GE Vernova GEV

Reference table

1. What the business is, and how the disclosure actually works

GE Vernova describes itself in the 10-K as "a global leader in the electric power industry, with products and services that generate, transfer, orchestrate, convert, and store electricity." Its installed base generates approximately 25% of the world's electricity. Delaware corporation, headquarters in Cambridge, Massachusetts; approximately 75,000 employees, of whom roughly 70% work in manufacturing, engineering or services — approximately 24,000 in Europe, 21,000 in the United States, 19,000 in Asia and 6,000 in Latin America. Chief executive Scott L. Strazik (vendor profile; the extracted 10-K text refers to the CEO by role rather than by name). The employee count predates the Prolec GE acquisition, which added approximately 10,000.

The single most important disclosure-structure point: this company is two years old, and half the history is not its own

On 2 April 2024, General Electric — which now operates as GE Aerospace — completed the spin-off of GE Vernova, distributing all of the shares of its common stock to GE stockholders. That is the date given in the 10-K, and we use it. (The vendor profile reports ipoDate: 2024-03-27, which is the start of when-issued trading, and says the company was "established in 2023," which is the incorporation date. Neither is the spin date. Where they disagree with the filing, the filing wins.)

This makes the pre-2024 columns of inc_a, bal_a and cf_a a different kind of number, and reading them as a continuous series is the largest single error available on this name. The 10-K's Note 1 Basis of Presentation says it directly:

> "For periods prior to the Spin-Off, the combined financial statements have been derived from the consolidated financial statements and accounting records of GE, including the historical cost basis of assets and liabilities comprising the Company, as well as the historical revenues, direct costs, and allocations of indirect costs attributable to the operations of the Company, using the historical accounting policies applied by GE. The combined financial statements do not purport to reflect what the results of operations, comprehensive income, financial position, or cash flows would have been had the Company operated as a separate, stand-alone entity during the periods prior to the Spin-Off."

And on the cost allocations specifically:

> "The Consolidated and Combined Statement of Income (Loss) includes expense allocations for certain corporate, infrastructure, and shared services expenses that were provided by GE on a centralized basis (GE Corporate Costs), including, but not limited to, finance, supply chain, human resources, IT, insurance, employee benefits, and other expenses… The GE Corporate Costs allocations may not be indicative of the actual expense that would have been incurred had the Company operated as an independent, stand-alone public entity."

Three concrete consequences, each visible in the vendor file:

1. The share count is fictional before the spin. inc_a reports 272,083,576 weighted average shares for fiscal 2021, 2022 and 2023 — the identical figure in all three years, which no real company produces. The 10-K explains why: "The computation of earnings (loss) per share for all periods through April 1, 2024 was calculated using 274 million common shares that were issued upon Spin-Off." Every pre-2024 EPS figure in this file is a retrospective per-share allocation, not an earnings-per-share history.

2. There was no standalone capital structure. The 10-K: "Historically, we participated in cash pooling and other financing arrangements with GE to manage liquidity and fund our operations. As a result of completing the Spin-Off, we no longer participate in these arrangements… Our capital structure, long-term commitments, and sources of liquidity have changed significantly from our historical practices." The 2023 balance sheet shows commonStock of $8.051 billion and retainedEarnings of $0 — that is GE's net parent investment line, not equity in any ordinary sense.

3. The 2021-2023 operating losses are not comparable to the 2024-2025 profits. The pre-spin years carry allocated GE corporate costs; the post-spin years carry the actual cost of running a standalone public company, which the 10-K names as one of the drags on 2025 operating income ("higher corporate costs required to operate as a stand-alone public company"). The direction of the bias is not even consistent, so we do not attempt an adjustment. We simply refuse to compute any growth rate or trend that spans 1 April 2024.

Every table in this dive that includes a pre-spin year is labelled as such, and every conclusion in this dive rests on fiscal 2024, fiscal 2025 and the first half of 2026 only.

The three reportable segments (10-K, Item 1)

• Power — gas, nuclear, hydro and steam. Gas Power sells heavy-duty and aeroderivative turbines plus lifecycle maintenance; Nuclear Power covers boiling-water reactors, fuel and small modular reactors through a joint venture with Hitachi; Hydro and Steam round it out. Power RPO was $94.4 billion at 2025-12-31 with a gas-turbine installed base of approximately 7,000 units, of which approximately 1,800 are under long-term service agreements with an average remaining life of approximately ten years. As of the same date there were 51 HA-Turbines in RPO, 43 being installed and commissioned, and 126 in the installed base with approximately 3.6 million operating hours.
• Wind — Onshore Wind (workhorse 2.8-127m, 3.6-154m, 6.1-158m and 6.0-164m units), Offshore Wind (the Haliade-X 220m), and LM Wind Power (blades). Workhorse products are approximately 75% of equipment RPO. Services agreements cover approximately 24,000 onshore turbines from an installed base of approximately 59,000 units. The U.S. is approximately 60% of Onshore Wind's equipment RPO.
• Electrification — Grid Solutions (high-voltage direct current, alternating-current substations, transformers, switchgear, grid automation), Power Conversion & Storage, and Electrification Software.

The vendor supplies no segment revenue table at all. seg_prod contains exactly two lines — Product and Service — which is the equipment-versus-services disaggregation, not the segment disaggregation. A reader who takes seg_prod as the segment table would conclude that GE Vernova is a two-segment company. It has three, and one of them is losing money. This is defect class 7 in a form we have not seen before in this batch: not a wrong tie, but a wrong table.

What seg_prod actually is, and it does tie:

Tie test, run line by line as the contract requires. FY2025 sums to $38,068M against reported revenue of $38,068M — an exact match. FY2024 sums to $34,935M against the vendor's own reported revenue of $34,943M, an $8 million (0.02%) shortfall. We report the gap rather than hide it; it is immaterial and is most likely a reclassification, since the FY2025 10-K states that "we have reclassified certain prior year amounts to conform to the current year's presentation." No double-counting, no missing line. The table is used.

Geography (seg_geo) — and here there IS a double-count trap.

The four regional lines sum to exactly $20,728M, which is the "Non-US" line. Non-US is therefore a subtotal, not a peer line, and a naive sum of all six vendor entries gives $58,797M — 1.545x actual revenue. This is the XOM defect. We have de-duplicated it: United States plus the four regions gives $38,069M against reported revenue of $38,068M, a $1 million rounding difference. Verified. The 10-K confirms the taxonomy: "Revenues are classified according to the region to which equipment and services are sold. For purposes of this analysis, the U.S. is presented separately from the remainder of the Americas."

The United States is the largest single market and is gaining share fast — 42.0% of revenue in 2024, 45.6% in 2025. That matters for the AI-datacentre thesis, for tariff exposure in both directions, and for the wind policy risk, since the U.S. is approximately 60% of Onshore Wind's equipment RPO.

Customers and competitors

Customer concentration is not disclosed as a risk in this filing — there is no "one customer accounted for X% of revenue" statement in the extracted 10-K text, which is consistent with a business selling to hundreds of utilities, independent power producers, governments and industrials worldwide. We note the absence rather than infer safety from it.

Competitors, named by the 10-K segment by segment:

• Power: Siemens Energy, Mitsubishi Power, Westinghouse, Framatome, Rolls-Royce.
• Wind: Vestas, Siemens-Gamesa, Nordex, Envision, Goldwind.
• Electrification: Hitachi Energy, Siemens Energy, Siemens, Schneider Electric, Mitsubishi Electric, ABB.

The vendor's peer set contains none of them. peers supplies AQN, CEG, CWEN, DUK, ENLT, NEE, ORA and SO — every one of which is a utility or a renewable power producer, i.e. a customer archetype rather than a competitor. The vendor also classifies GE Vernova under sector "Utilities" and industry "Renewable Utilities." GE Vernova is an industrial capital-goods manufacturer whose renewable segment is its smallest and its only unprofitable one. The classification is wrong, and it is wrong in a way that would corrupt any sector-relative screen, any utility-style yield comparison and any beta-adjusted peer multiple. No peer-multiple comparison is drawn in this dive, because the file contains no valid peer set and no peer multiples.

2. The single most important thing to understand about these numbers: not one dollar of the reported profit is operating profit

Read the reported income statement of this company for the last three periods and you will get the wrong answer three times.

The arithmetic that follows from this is the central fact of the dive.

Trailing twelve months (September 2025 through June 2026 quarters) reported net income is $9.529 billion on revenue of $41.368 billion — a 23.0% net margin, which the vendor duly reports as netProfitMarginTTM: 0.2303 and converts into a headline P/E of 29.24x.

Now remove the items the filings identify:

> Reported TTM net income $9.529B

> − the Q4 2025 tax valuation-allowance release $2.900B (10-K, verbatim)

> − the Q1 2026 Prolec remeasurement gain $3.992B (10-Q Note 8, verbatim; the 10-Q states this gain was non-taxable)

> − the Q1 2026 Proficy, equity-method and China XD gains $0.667B

> − the Q2 2026 China XD gain $0.013B

> = clean trailing net income of approximately $1.957B

On 270 million diluted shares that is approximately $7.25 of clean trailing earnings per share, and at $1,018.53 the stock trades at approximately 140x it.

Two independent checks say this is roughly right rather than a hostile construction.

First, the company's own non-GAAP measure agrees. The 10-K reports Adjusted EBITDA of $3.2 billion and an Adjusted EBITDA margin of 8.4% for fiscal 2025. A company with $3.2 billion of adjusted EBITDA, $853 million of depreciation and amortisation and a normal tax rate cannot produce $4.9 billion of net income from operations. It produced it from the tax line.

Second, the consensus estimate for fiscal 2025 agrees. est shows epsAvg of $7.276 for the year ended 2025-12-31 against reported GAAP diluted EPS of $17.69. Our independently derived clean figure of $7.19 (net income of $4.884B less the $2.9B release, over 276 million diluted shares) lands within 1.2% of the analyst consensus for the same year. Two different methods, one filing-sourced and one consensus-sourced, converge on the same number. That is the honest earnings power of this company as at the end of 2025: roughly $7 per share, not $17.69.

And the cash flow tells the same story from a third direction. The 10-Q reports operating cash flow of $10.7 billion for the six months ended 2026-06-30 against $1.5 billion a year earlier, and free cash flow of $9.9 billion against $1.2 billion. Spectacular — until the same document explains it:

> "Cash from operating activities of $10.7 billion for the six months ended June 30, 2026 included a $11.7 billion inflow from changes in working capital. The cash inflow from changes in working capital was primarily driven by: contract liabilities and current deferred income of $13.7 billion, driven by down payments on orders and slot reservation agreements at Power, and down payments at Electrification…"

The working-capital inflow is larger than the operating cash flow. On our arithmetic, trailing twelve-month operating cash flow of approximately $14.19 billion less an approximate trailing working-capital inflow of $13.0 billion leaves roughly $1.2 billion of operating cash flow before working capital, against trailing capital expenditure of approximately $1.68 billion — i.e. working-capital-neutral free cash flow is approximately zero to slightly negative. (Our derivation, from cf_a full-year figures netted against the 10-Q's stated half-year figures; labelled as such and not filing-stated.)

None of this means the business is bad. Customer prepayments against a four-year order book are exactly what you would expect from a supplier with the pricing power to demand slot-reservation deposits, and collecting $14.1 billion of customer cash in six months is a strength of the commercial position. What it means is that the trailing profit and cash figures on every screen are unusable as a valuation anchor, and that the entire investment case has to be made on forward margin — a thing this company has not yet demonstrated. Section 6 does exactly that, and that is where the disagreement with the street lives.

3. Growth and margin trajectory

Annual, with the pre-spin years fenced off:

Gross margin is the one series in this file that is unambiguously and consistently improving, and the 10-K states it verbatim: "gross margin was 19.8%, 17.4%, and 14.5% for the years ended December 31, 2025, 2024, and 2023." That is 530 basis points in two years, and it continued into 2026: 21.3% in the June quarter against 20.3% a year earlier, and 20.3% for the first half against 19.3% — all four figures confirmed verbatim in the 10-Q.

(Note one vendor-versus-filing gap: inc_a reports FY2024 gross profit of $6.309B, which is 18.05% of $34.943B, against the 10-K's stated 17.4%. The difference is approximately $229 million. The vendor carries the figures as originally filed on 2025-02-06; the FY2025 10-K restates them, saying "we have reclassified certain prior year amounts to conform to the current year's presentation." The filing wins: we use 17.4% for FY2024.)

Quarterly, cross-checked line by line against earn_cal:

Three readings, in order of importance.

First, the revenue acceleration is real and it is the strongest part of the case. The June 2026 quarter did $11.104 billion, up 21.9% year on year — the 10-Q confirms "total revenues increased $2.0 billion (22%)" — and the first half did $20.443 billion, up 19%. But 12 of those 22 points were organic: the 10-Q states organic revenue rose $1.1 billion (12%), with organic equipment up 14% and organic services up 10%. The remainder is Prolec GE, which the same document says contributed $859 million of revenue in the quarter. Twelve percent organic on an $11 billion quarterly base is very good. It is not 22%.

Second, seasonality is severe and the March quarter is a trap. Operating margin ran 0.9% in Q1 2025 and 1.9% in Q1 2026 against 5.5-5.9% in the fourth and second quarters. Any reading of this company off a single first quarter is wrong in both directions — which is precisely why the $17.44 headline from March 2026 is so misleading: it sat on top of the weakest operating quarter of the year, with $179 million of operating income.

Third, the "miss" in the June quarter is a basis artefact and we will not report it as a miss. earn_cal shows epsActual: 2.47 against epsEstimated: 3.17 — a 22% shortfall that would ordinarily be a serious event. But epsActual is GAAP diluted EPS (it matches inc_q exactly, $2.47) and epsEstimated is a non-GAAP consensus (the same field for Q1 2026 read $1.95 against a GAAP actual of $17.44, an apparent 794% "beat"). These are different measurements and comparing them produces noise, not information. GE Vernova's own adjusted earnings measure for the quarter is not available: the 8-K of 2026-07-22 is in the archive as the cover page only, with the results carried in an Exhibit 99 that was not extracted. We therefore do not know whether the June quarter beat or missed on the basis the street was struck on, and we say so rather than guess.

Research and development was $1.197 billion in 2025 (3.1% of revenue), up from $982 million, and ran $334 million in the June 2026 quarter. The 10-K describes the split as roughly half industrialising existing products and supporting the installed base for this decade, half long-term innovation. Selling, general and administrative costs are falling as a share of revenue — 12.4% of revenue in the June 2026 quarter against 13.0%, and 13.1% for the half against 13.8% — which is the operating leverage the forward case needs, arriving in small increments.

The segments, reconstructed — and why the reconstruction is necessary

The vendor supplies no segment figures and the filing's segment tables were stripped by the text extraction. What survives is the MD&A prose, which gives changes and percentages but not levels. We reconstruct the levels from those changes, label the reconstruction as ours, and check it against a disclosed total.

The reconstruction ties to within approximately 0.4%, which is what one should expect from figures the filing rounds to the nearest $0.1 billion. We use it for structure and proportion only, never for a multiple.

And here is the Wind problem, stated as precisely as the disclosure permits. GE Vernova measures segments on "segment EBITDA," which the 10-K says its CEO computes after excluding "certain non-cash charges, such as depreciation and amortization, impairments and other matters, major restructuring programs, and certain gains and losses." Even on that generous basis:

• Fiscal 2025: Wind segment EBITDA "decreased slightly (2%)" while revenue fell 6%. The MD&A separately describes the contribution to consolidated operating income as "a slight decrease in segment results at Wind of less than $0.1 billion." On roughly $9.4 billion of revenue, Wind's contribution to consolidated operating income moves by less than $100 million in either direction — the segment is approximately breakeven.
• June 2026 quarter: Wind segment revenue "down $0.2 billion (10%)" and segment EBITDA "down $0.1 billion (67%)" — and organically down 79%.
• First half 2026: revenue down $0.6 billion (16%), segment EBITDA down $0.3 billion, attributed "primarily at Onshore Wind due to lower equipment deliveries and the impact of tariffs, and at Offshore Wind due to higher contract losses."
• Wind RPO fell 6% in the half and 9% year on year. It is the only segment whose backlog is shrinking.

Underneath it, the Offshore Wind contract losses, stated exactly as the 10-K does:

> "During the years ended December 31, 2025, 2024, and 2023, primarily as a result of changes in product and project cost estimates, we recorded incremental contract losses for certain Offshore Wind contracts of $637 million, $1,005 million, and $379 million, respectively. The incremental contract losses primarily relate to the estimated impact of changes in execution timelines, project-related commercial liabilities, costs to remediate quality issues including the removal of previously installed blades at the Vineyard Wind project, and additional project-related supply chain and manufacturing costs. Further changes in our execution timelines or other adverse developments could result in further losses beyond the amounts that we currently estimate."

Two billion dollars of incremental contract losses in three years, with the company explicitly declining to say the number is final. The 10-K also records that on 22 December 2025 the U.S. Department of the Interior paused leases for all large-scale offshore wind projects under construction in the United States, directly affecting the Vineyard Wind completion timeline; the 10-Q reports the pause was lifted on 27 January and that all remaining turbines were installed in the first quarter, with the company now "working with our customer to resolve outstanding claims and counterclaims." That resolution is a genuinely two-sided future event and we score it as such.

The honest summary of the segment picture: Power and Electrification are carrying the entire company, Wind is a roughly $9.4 billion revenue business contributing approximately nothing, and the 45% of consolidated revenue that is services is what keeps gross margin rising while equipment mix shifts.

4. Balance sheet and the enterprise-value rebuild

The starting point is a defect, and it is the first one the brief asked us to check. bal_a[0] is dated 2025-12-31 and filed 2026-01-29. It is seven months stale and three material events have happened since:

1. 2026-02-02 — the $5.254 billion acquisition of the remaining 50% of Prolec GE, cash consideration, $4.9 billion net of cash acquired.

2. 2026-02-04 — a $2.6 billion senior-note issuance, the company's first debt of any size.

3. The first half of 2026 — $3.6 billion of share repurchases and a $14.088 billion increase in contract liabilities and deferred income.

We therefore build the balance sheet from the 10-Q filed 2026-07-22, which reports as of 2026-06-30, and use bal_a only for the prior-year comparison. The 10-Q's figures are prose-stated and filing-verified.

The netDebt sign test — the one that broke AAPL and BRK-B — PASSES here, and we say so. The vendor reports netDebt: −$8,848,000,000 for 2025-12-31, i.e. a net cash position, and enterpriseValueTTM of $262.109 billion against a market capitalisation of $271.269 billion — enterprise value correctly BELOW market cap. The sign is right, the direction is right, and the magnitude is close. This is worth recording as a pass because the same field has been wrong on two mega-caps in this programme.

What the vendor gets wrong on the balance sheet is different and smaller:

• shortTermInvestments is $0 in all three years while longTermInvestments is $1.834B (2025), $2.149B (2024) and $2.341B (2023). The 10-K refers to "interest earned on cash balances and short-term investments", so short-term investments exist; the vendor has classified them as long-term or netted them elsewhere. We do not add them to liquidity — we cannot verify they are liquid, and the 10-K's own liquidity statement cites cash only. Net cash of $10.1 billion is therefore conservative, not aggressive.
• totalDebt is $0 for 2025-12-31, along with shortTermDebt, longTermDebt and capitalLeaseObligations, all zero. The 10-Q says total debt excluding finance leases was "less than $0.1 billion" at that date, so the senior-debt figure is nearly right — but the 2024 balance sheet carried $810 million of finance leases and the 2025 one carries zero. GE Vernova did not extinguish its leases in one year. The totalDebt series is not comparable across 2024 and 2025 and we do not use it.
• The vendor contradicts itself on whether debt exists at all. ratios_ttm reports debtToMarketCapTTM: 0 and longTermDebtToCapitalRatioTTM: 0 alongside debtToEquityRatioTTM: 0.331, debtToAssetsRatioTTM: 0.049 and interestDebtPerShareTTM: 14.776 — the last of which implies roughly $3.96 billion of debt on 268 million shares. Three mutually inconsistent debt figures in one payload. The filing settles it: $2.6 billion at 2026-06-30, all of it the February 2026 notes.

The debt itself, from the 10-Q verbatim: on 2026-02-04 the company issued $2,600 million of senior notes — $600 million of 4.250% notes due February 2031, $1,000 million of 4.875% notes due February 2036, and $1,000 million of 5.500% notes due February 2056 — with net proceeds of $2,543 million used for general corporate purposes including financing part of the Prolec acquisition. There is no maturity before 2031. Liquidity is backed by $6.0 billion of credit facilities: a $3.0 billion unsecured revolver and a $3.0 billion standby letter-of-credit and bank-guarantee facility, both maturing 2029-04-02, with no borrowings outstanding. The company states it "estimated an insignificant liquidity impact of a ratings downgrade below investment grade."

The rebuilt enterprise value:

> Market cap $271.269B (266,333,581 shares × $1,018.53)

> + total debt $2.600B

> − unrestricted cash $12.700B

> = rebuilt enterprise value ≈ $261.17B

The vendor reports $262.109B, implying net cash of $9.160B against our $10.100B. The gap is $0.94 billion, or 0.36% — immaterial, and we say so. All EV multiples in this dive use $261.17B.

The rejection of evToEBITDATTM is the most consequential correction in this dive. The vendor's 30.25x implies trailing EBITDA of $8.664 billion, which reconciles exactly to inc_q EBITDA of $959M + $1,335M + $5,445M + $925M. The $5,445 million from the March 2026 quarter is a quarter in which operating income was $179 million. The vendor computes EBITDA as pre-tax income plus depreciation and amortisation, so the $3.992 billion Prolec remeasurement gain — a non-cash accounting entry — is sitting inside the EBITDA denominator of a valuation multiple. Our figure, built from trailing operating income of $1.801 billion plus approximately $1.135 billion of depreciation and amortisation, is roughly $2.94 billion, giving EV/EBITDA of about 88.9x. Against the company's own Adjusted EBITDA of $3.2 billion for fiscal 2025 it is 81.6x. Either way the honest number is three times what the vendor prints, and no reader should be shown 30x for this business.

The contract-liability balance, and why it is not a fortress

Contract liabilities and deferred income were $25.774 billion at 2025-12-31 and rose $14.088 billion in the first half of 2026 — to approximately $39.9 billion. (The 2025 figure is the vendor's deferredRevenue; the $14,088 million increase is stated verbatim in the 10-Q; the sum is our arithmetic and is labelled as such.)

This is the largest single item on the liability side of this company and it is routinely mistaken for strength. The 10-Q defines it plainly: "Contract liabilities primarily represent cash received from customers under ordinary commercial payment terms in advance of delivery of equipment orders or servicing of customers' installed base."

It is a genuine commercial strength — a supplier who can demand slot-reservation deposits on turbines that will not ship for three years has real pricing power, and it is the reason the company holds $13.1 billion of cash while carrying $2.6 billion of debt.

It is not earned money. Every dollar of it must be discharged by delivering equipment at a contracted price, in an environment where the 10-K discloses $250 million of 2025 tariff costs, the 10-Q discloses a $100-200 million estimate for 2026, and the same company has taken $2.0 billion of incremental contract losses on offshore projects in three years. A contract liability is an obligation to perform at a price fixed in the past. In an inflationary supply chain that is a risk, not an asset — and it is precisely the mechanism that produced the Offshore Wind losses in the first place.

Working capital is negative $8.397 billion and the current ratio is 0.850. For most companies that is a distress signal. Here it is the customer-financed model working as designed. We report it as neither, and we do not use currentRatioTTM, quickRatioTTM or netCurrentAssetValueTTM as lenses on this business.

5. Cash flow and capital returns

The 10-K confirms the fiscal 2025 figures verbatim: "Cash flows from operating activities were $5.0 billion and $2.6 billion" and "Free cash flow was $3.7 billion and $1.7 billion for the years ended December 31, 2025 and 2024."* Vendor and filing agree on the cash-flow statement to the rounding, which is a clean reconciliation and is recorded as one.

Read the working-capital column against the free-cash-flow column and the whole story is there. In fiscal 2025, $2.910 billion of a $3.711 billion free cash flow was working capital. In the first half of 2026, $11.700 billion of a $9.900 billion free cash flow was working capital — more than 100% of it. The company is transparent about this: the 10-Q attributes the first-half inflow to "contract liabilities and current deferred income of $13.7 billion, driven by down payments on orders and slot reservation agreements at Power, and down payments at Electrification."

GE Vernova's own free-cash-flow definition is management's, and the 10-K says as much: "We measure ourselves on a free cash flow basis… however, free cash flow does not delineate funds available for discretionary uses as it does not deduct the payments required for certain investing and financing activities." Per house rule, management's own framing is carried at half weight. We report the company's $9.9 billion and we report our working-capital-neutral estimate of approximately zero, and we let the reader see both.

Capital returns are large, real, and running down the authorisation.

• Buyback. The board raised the repurchase authorisation to $10.0 billion on 2025-12-09, from $6.0 billion authorised 2024-12-10. The company repurchased 2.5 million shares for $2.3 billion in the June quarter and 4.3 million shares for $3.6 billion in the first half, and states that cumulatively $7.0 billion has been repurchased over the life of the programme. That leaves approximately $3.0 billion of authorisation, or 1.1% of the market capitalisation.
• The execution price is worth noting and cuts both ways. $3.6 billion for 4.3 million shares is an average of approximately $837, and $2.3 billion for 2.5 million shares in the June quarter is approximately $920. Today's close of $1,018.53 is 21.7% above the first-half average and 10.7% above the second-quarter average. The company has been a large buyer, but not at this price.
• Share count is genuinely shrinking: the 10-Q covers show 268,719,995 shares at 2026-03-31 and 266,333,581 at 2026-06-30 — a 0.9% reduction in one quarter, roughly 3.5% annualised. That is real accretion, not dilution-offset.
• Dividend. profile.lastDividend and ratios_ttm.dividendPerShareTTM both read $1.75, yielding 0.17%. Neither figure is confirmed anywhere in the filings in this archive — no per-share dividend rate appears in the extracted 10-K or 10-Q text. What the filings do give is fiscal 2025 dividends paid of $275 million (cf_a, corroborated by the 10-K's capital-allocation discussion) which on approximately 276 million weighted shares is roughly $1.00 per share for calendar 2025, and a 10-Q statement of "higher dividends paid of $0.1 billion" in the first half of 2026. We state the $1.75 as vendor-sourced and unverified, and we draw no conclusion from it. The payout ratio of 4.3% is meaningless against earnings that are one-offs.

Total shareholder yield is therefore roughly 1.3% to 2.8% depending on whether the first-half buyback pace annualises — 0.17% of dividend plus 1.22% (fiscal 2025 basis) to 2.65% (first-half annualised) of repurchase. With $3.0 billion of authorisation remaining, the higher figure cannot persist past roughly two more quarters without a new authorisation.

6. Valuation — priced in or room?

At $1,018.53 (market capitalisation $271.269B, 266,333,581 shares, rebuilt enterprise value $261.17B):

Estimate coverage is genuinely good on the years that matter — 22 analysts on FY2027 EPS and 23 on FY2028 — but the dispersion is extraordinary. The FY2028 EPS range runs from $21.962 to $47.408: the high is 2.16 times the low. For comparison, the FY2027 high is 1.40x the low. The street agrees on next year and has no idea about the year after, which is itself the most useful thing the estimate block tells us. The FY2029 and FY2030 rows rest on 10 analysts each and are excluded from every conclusion.

Three estimate-block rejections, tested rather than assumed:

First, the FY2026 EPS row of $30.734 is rejected as contaminated. It is arithmetically inconsistent with everything around it — higher than the FY2027 estimate of $24.892 on a company whose revenue is growing 15% a year. The explanation is mechanical: $17.44 (Q1 GAAP actual) + $2.47 (Q2 GAAP actual) + $4.09 (Q3 consensus) leaves a residual of $6.73 for the fourth quarter. The row is a hybrid of realised GAAP earnings that include a $4.0 billion non-cash gain and forward non-GAAP consensus. It is not a forecast of anything and no conclusion in this dive uses it. Our own clean estimate for 2026, stripping the $4.672 billion of first-half gains, is roughly $13.5-14.5 of EPS — which makes the FY2027 consensus of $24.892 a step of roughly 75%, and that is the number the price is really paying for.

Second, est.ebitdaAvg and est.ebitAvg are rejected in full — this is the worst example of defect class 8 seen in this programme. The FY2025 row is testable against a reported actual and it fails catastrophically:

The last line is the fabrication signature. Across every one of the eight estimate years, ebitdaAvg minus ebitAvg — which must equal depreciation and amortisation — ranges from $32.4 million to $76.5 million, on a company reporting $853 million to $1,797 million of annual depreciation and amortisation. The row is not a consensus estimate that happens to be wrong; it is a margin assumption applied mechanically with a depreciation figure that has no relationship to this company. The 2023 row is the reductio: ebitdaAvg of $7.480 billion for a year in which the company posted an operating loss of $923 million. All forward valuation in this dive runs on epsAvg and revenueAvg, per the data contract.

Third, epsAvg passes the same test and is used. The FY2025 epsAvg of $7.276 lands within 1.2% of our independently derived clean actual of $7.19 (reported $4.884B net income less the filing-stated $2.9B tax release, over 276 million diluted shares). Two different methods converge. epsAvg is the reliable field on this name as it is elsewhere, with the single exception of the contaminated FY2026 row.

Which basis is consensus struck on? Non-GAAP, and the gap here is enormous rather than trivial. Reported fiscal 2025 GAAP diluted EPS was $17.69; consensus for the same year was $7.276. That is not a 3% adjustment as at KLA — it is a 59% haircut, because GAAP includes a $2.9 billion tax release that no analyst modelled. Every forward multiple in this dive is on the non-GAAP consensus basis, and the trailing multiple is shown on three bases — 29.2x on reported GAAP, ~140x on our clean operating derivation, and the vendor's 30.3x EV/EBITDA which we reject entirely. A reader who takes 29.2x from a screen and thinks the stock is reasonably priced has been misled by a tax election.

6a. What today's price assumes (the inversion)

At $1,018.53 — 40.92x the FY2027 consensus and 29.16x FY2028 — the price embeds the following falsifiable claims, each with a number and a date:

• Revenue reaches $52.798 billion in calendar 2027 and $60.861 billion in 2028 — compound growth of about 15% a year off a $41.368 billion trailing base. (Consensus-derived; 27 and 31 analysts.) This is the assumption we are most comfortable with, because it is the one the backlog directly supports: equipment RPO of $87.821 billion with 36% scheduled inside one year and 65% inside two, plus services RPO of $88.463 billion at 16% and 54%, implies roughly $45.8 billion of contracted conversion in the next twelve months alone. Nearest check: the 2026-10-28 print against a consensus of $11.995 billion, +20.3% year on year.
• Net margin expands from approximately 4.7% (our clean trailing figure) to 12.6% in 2027 and 15.6% in 2028. (Derived by us from the consensus revenue and net-income rows, which cross-check against epsAvg × share count to within 1%.) This is the most fragile assumption in the price and it is not close. GE Vernova's own Adjusted EBITDA margin was 8.4% in fiscal 2025, its trailing operating margin is 4.35%, and a 15.6% net margin would require an EBITDA margin somewhere in the low twenties — roughly a tripling of the 2025 level in three years, while one segment is breakeven, tariffs cost $100-200 million a year, and $39.9 billion of contract liabilities lock in prices agreed in the past.
• EPS grows from roughly $14 (our clean 2026 estimate) to $24.892 in 2027 — about 75% — and then a further 40.3% to $34.924 in 2028. (Our clean base; consensus for the forward years.) Two consecutive years of 40%-plus earnings growth are in this price. The FY2028 consensus low of $21.962 implies a net margin of about 9.8% and, at today's price, a multiple of 46.4x. The bear case is not a crash scenario; it is simply the low end of the analysts' own range.
• The market keeps paying roughly 29x or better on the forward year at every date between now and then. (Our number.) At 25x FY2028E the stock is $873; at 35x it is $1,222. Given the FY2028 earnings range is itself 2.16x wide, the earnings variable dominates the multiple variable here — the reverse of most names in this batch.
• Wind does not get worse. (Our assumption; filing-sourced inputs.) The price contains no visible allowance for a further Offshore Wind charge, and the 10-K explicitly declines to cap the exposure: "Further changes in our execution timelines or other adverse developments could result in further losses beyond the amounts that we currently estimate." Three years, $2.021 billion of incremental contract losses, and an unresolved claims-and-counterclaims negotiation on Vineyard Wind.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (+40.3%, from FY2027E $24.892 to FY2028E $34.924) + multiple drift (COMPRESSION, from 40.9x on the forward year today to roughly 28x on the forward year in twelve months, −31%) + shareholder yield (+1.3% to +2.8%) ≈ roughly flat to modestly negative.

Our base case assumes multiple compression, and we are explicit that this is not a rolldown artefact. Today the market pays 40.92x on the next calendar year's estimate. In twelve months, if it pays 28x on the then-next year's estimate, that is a genuine de-rating, not the mechanical effect of the calendar advancing. We assume it for three reasons, in order of weight:

1. The earnings being capitalised are not yet demonstrated. Every trailing profit metric on this company is an accounting event. A multiple of 41x forward earnings on a business whose clean trailing earnings are $2 billion is a multiple paid on faith, and faith multiples compress as the evidence arrives — in either direction.

2. Growth decelerates on the consensus path itself: +40.3% (FY27→FY28), +34.6% (FY28→FY29), +20.5% (FY29→FY30). A multiple appropriate to 40% growth is not appropriate to 20%, and the market normally begins discounting that a year ahead.

3. The dispersion is a risk premium waiting to be charged. When 23 analysts produce a range from $21.96 to $47.41 on the same year, the market is not being asked to pay for a forecast; it is being asked to pay for a distribution. Distributions get discounted.

Almost none of the expected return in our base case is multiple expansion, and that is deliberate. If the multiple simply HELD at 29.16x on FY2028E the price would be $1,018 — exactly today. The entire question is whether the FY2028 earnings land near $34.92, near $21.96, or near $47.41, and our fair-value construction below is built to make that explicit rather than to hide it inside a multiple assumption.

6c. Variant perception (where we differ, what would surprise)

• We differ from the street on the quality of the trailing earnings, and it is not a matter of opinion — it is a matter of reading the tax footnote. The screen number is 29.2x. The clean number is roughly 140x. A 22-buy, 7-hold, ZERO-sell rating distribution with a $1,274 median target on a stock at 40.9x forward earnings, whose trailing profits are entirely a valuation-allowance release and an acquisition remeasurement, is a consensus that has not stress-tested its own denominator. Our base of $978 is 22.1% below the street's $1,256.29, and the street's own low target of $949 is 6.8% below spot — so the disagreement is not only with us.
• We differ on what the free-cash-flow line means, and this is the cleanest single divergence. GE Vernova reported $9.9 billion of free cash flow in six months. The same 10-Q says $13.7 billion of it came from contract liabilities and deferred income — customer down payments and slot-reservation deposits. On a working-capital-neutral basis our arithmetic puts trailing free cash flow at approximately zero. A price/free-cash-flow ratio of 21.8x is being shown on screens for a company that, absent customer prepayments, is roughly cash-neutral after capital expenditure. Watchable number: the contract-liability line in the 2026-10-28 report. If the inflow decelerates sharply while revenue grows, the free-cash-flow figure collapses without anything operational changing.
• We agree with the street on the backlog, and we want to say so plainly, because it is the strongest evidence in this file. Equipment RPO growing 36.7% in six months to $87.821 billion is not a narrative; it is a disclosed number, and it is corroborated independently by four knowledge-base sources describing multi-year turbine queues. We are not making a demand-side argument against this stock. We are making a margin and price argument.
• We are aligned with the knowledge base's most analytically serious voice, and we did not start there. Jordi Visser's claim of 2026-01-11 — "Data-center infrastructure names at 30-50x already price in three years of demand; bottlenecks make upside surprises unlikely over next two years" — is a valuation claim, not a demand claim, and GE Vernova trades at 40.9x the 2027 consensus, inside his stated band. The same speaker was 85-conviction bullish on this exact trade thirteen weeks earlier, which is itself a caution about how much weight to put on any single reading.
• Positive surprise that would force a re-rate: the 2026-10-28 print showing third-quarter operating margin above 7% — which would be the first quarter in the file above 6.4% and would begin validating the margin path in 6a — combined with equipment RPO holding above $85 billion. Or a resolution of the Vineyard Wind claims in GE Vernova's favour, which would remove the open-ended tail the 10-K refuses to bound. Or the Wind segment reporting positive and growing segment EBITDA, which would convert a $9.4 billion drag into an earnings contributor.
• Negative surprise that would break the thesis: a further Offshore Wind contract loss of any size, which after $2.021 billion in three years would establish the charges as structural rather than terminal; equipment RPO growth stalling, which would say the order surge was a one-time pull-forward of slot reservations rather than a durable demand shift; a contract-liability inflow below roughly $2 billion in the quarter, which would strip the free-cash-flow story; or the FY2028 consensus migrating toward its own low of $21.962, at which point today's price is 46.4x.

Synthos fair values

All three anchors use ONE multiple — 28x — applied to the three points of the FY2028 consensus EPS distribution (low $21.962, mean $34.924, high $47.408; 23 analysts). This construction is deliberate and it is the honest one for this name: on GE Vernova the uncertainty is overwhelmingly in the earnings, not in the multiple. The FY2028 high is 2.16 times the low. Varying the multiple as well would manufacture false precision on top of a distribution that is already doing all the work.

Why 28x? Today the market pays 29.16x FY2028E and 40.92x FY2027E. 28x on FY2028E in twelve months is a modest de-rating of about 4% relative to today's forward multiple, justified by the three reasons in 6b: undemonstrated earnings, decelerating consensus growth, and a dispersion premium. It is not a punitive multiple — it is a premium industrial multiple that still assumes the growth story is broadly intact.

• Bear ~$615 — 28x the FY2028 consensus LOW of $21.962. Cross-check: 31.3x the FY2027 consensus low of $19.655; 12.2% above the 52-week low of $547.96; 26.8% below the 200-day moving average of $840.32. The scenario: the margin expansion under-delivers, Offshore Wind takes another charge, the contract-liability inflow normalises and the free-cash-flow optics deteriorate, and the equipment order surge proves to have been a pull-forward. Note this is not a catastrophe scenario — it is the low end of the analysts' own published range at a reasonable multiple. −39.6%.
• Base ~$978 — 28x the FY2028 consensus MEAN of $34.924. Cross-check: 39.3x FY2027E; rebuilt EV of $250.6B on FY2028E revenue of $60.861B is 4.12x sales; 1.42x current RPO. Sensitivity, stated openly: 26x gives $908 (−10.9%) and 30x gives $1,048 (+2.9%). The whole answer lives in a 26-30x band applied to a mean estimate that has a 2.16x dispersion around it, and we claim no more precision than that. The scenario: revenue lands broadly on the consensus path because the backlog delivers it; margin expands materially but less than modelled; Wind stays roughly breakeven; and the multiple gives back a little as growth decelerates. −4.0%.
• Bull ~$1,327 — 28x the FY2028 consensus HIGH of $47.408. Cross-check: 53.3x FY2027E. This bull case requires no re-rating at all — only that GE Vernova delivers the top of the analysts' own range. The scenario: gas-turbine pricing runs ahead of plan as the queue lengthens, Prolec GE turns profitable and the transformer shortage persists, Offshore Wind is resolved without further charge, and services margin compounds off the growing installed base. +30.3%.

Base is 4.0% BELOW spot; asymmetry is roughly 0.77:1 to the upside (39.6% down, 30.3% up) — negative. That ratio, not the growth story, is the verdict. We are not being paid to take the earnings distribution risk at this price. At $840, roughly the 200-day moving average and 24x the FY2028 mean, the base case is +16% and the asymmetry inverts to favourable. That is the number to want.

7. Knowledge base — eleven claims, one dominant voice, and a reversal

Raw hits: 11. Entity matches: 11. Text-only matches: 0. Used: 9. Discarded for collision: 2.

The sweep covered GEV, GE Vernova, Vernova, gas turbine, grid equipment, offshore wind and datacenter power across all 51,928 distilled claims. This is the deepest name-level lane in this batch and we want to record that up front — it is not a padded sector lane. Nine of the eleven claims name GE Vernova or GEV in the entity list as a subject, not as a basket member.

The two discards, and why

Both discarded claims are dated 2023-05-25 — ten months BEFORE GE Vernova existed as a public company. The spin-off completed on 2 April 2024. A claim tagged "GE Vernova" and dated 2023-05-25 is a claim about a group of divisions inside the General Electric conglomerate, discussed under a name that had been announced but not yet separated, with no standalone capital structure, no standalone cost base and no share price. Under the brief's collision rule — a claim about the old GE conglomerate is not a claim about GEV — both are discarded from the conviction pool. They are:

> 2023-05-25 · bearish · conviction 60 · horizon: thesis · entities: GE Vernova, Renewables · channel: business_breakdowns · NO NAMED SPEAKER

> "Skeptical of management's long-term renewables targets; they underappreciate competitive risks and it won't break even until mid-decade."

> 2023-05-25 · bullish · conviction 55 · horizon: thesis · entities: GE Vernova, Power · channel: business_breakdowns · NO NAMED SPEAKER

> "Gas power benefits from energy transition as a bridge/backup technology given renewable intermittency that batteries can't yet correct."

We note without weighting them that both have aged well — renewables has indeed not broken even, and gas power has indeed been the beneficiary — but a pre-spin claim about a GE division cannot be scored as conviction on a standalone security, and we are not going to launder a correct forecast into a source of edge it did not earn.

The nine used claims, verbatim

> 2025-07-27 · bullish · conviction 60 · skill 2.0 · horizon: thesis · entities: GEV · channel: jordi_visser · SPEAKER NULL

> "Gas turbines were 'dead' in 2022-23 under ESG; now makers can't meet data-center surge, backlogs beyond 2029 — Crusoe ordered 19 GE Vernova units. Supply-demand favors makers."

> Test: "gas turbine order backlogs extend beyond 2029"

> 2025-09-20 · bullish · conviction 85 · skill 2.0 · horizon: thesis · entities: GEV, Siemens Energy · speaker: jordi_visser (independent) · appears in an anthony_pompliano document

> "AI data-center demand outstrips power supply; gas turbines don't exist and face multi-year delays, so a tide lifts solar, batteries and hybrid power solutions — a field day."

> Test: "gas turbine shortage pushes PMIs up"

> 2025-09-21 · bullish · conviction 75 · skill 2.0 · horizon: thesis · entities: GEV · channel: jordi_visser · SPEAKER NULL

> "Gas-turbine shortage stretches supply out ~5 years while AI demand grows faster than expected; turbines fill only ~half the power gap, so the shortage worsens."

> Test: "turbine supply crunch persisting ~5 years"

> 2025-10-04 · BEARISH · conviction 55 · skill 2.0 · horizon: thesis · entities: ORCL, GEV · speaker: jordi_visser (independent) · appears in an anthony_pompliano document

> "Oracle and GE Vernova have insatiable demand but are sold out ~5 years with high PEs; the buildout is already priced in, so where does incremental growth come from?"

> 2025-10-05 · BEARISH · conviction 52 · skill 2.0 · horizon: thesis · entities: GEV, Siemens Energy · channel: jordi_visser · SPEAKER NULL

> "Big AI-power names like GEV and Siemens Energy face gas-turbine/production bottlenecks and can't justify multiples if they can't grow into demand; the crowded, correlated concentration is a risk."

> 2026-01-11 · BEARISH · conviction 60 · skill 2.0 · horizon: thesis · entities: GE Vernova · channel: jordi_visser · SPEAKER NULL

> "Data-center infrastructure names at 30-50x already price in three years of demand; bottlenecks make upside surprises unlikely over next two years."

> 2026-02-12 · bullish · conviction 65 · horizon: thesis · entities: GEV · channel: anthony_pompliano_show · SPEAKER "uncertain" (guest)

> "Building your own generation is a multi-year long game — Vernova/GE Power/Siemens orders up 260% with 36-month equipment queues — so operators chase existing powered land, pushing its value up sharply."

> Quote: "So building your own generation capacity is a long game."

> 2026-03-27 · bullish · conviction 70 · horizon: thesis · entities: GE Vernova, Caterpillar · channel: jensen_huang · NO NAMED SPEAKER

> "Power and generation equipment (incl. small modular nuclear) are key downstream beneficiaries of AI datacenter buildout."

> 2026-06-10 · bullish · conviction 80 · horizon: thesis · entities: GEV, Siemens, Mitsubishi · speaker: Jensen Huang · channel: jensen_huang

> "Energy is the bottom layer of AI and the single greatest opportunity in several generations; any form — nuclear, wind, solar, hydrogen — will get funded."

What this lane is, stated without inflation

Per-speaker concentration is the first finding and it is severe. Six of the nine used claims (66.7%) — and six of eleven raw (54.5%) — trace to a single voice: Jordi Visser. He appears under two channel labels (jordi_visser and inside anthony_pompliano documents), which is exactly the failure mode the attribution policy exists to catch: attributed by channel he would look like two sources; attributed by speaker he is one. We attribute by speaker.

The second finding is a clean position reversal inside that concentration, and it is visible in the dates.

Thirteen days separate an 85-conviction bull call from the first bear call, and the underlying facts did not change in thirteen days. What changed was the price. That is worth reading carefully in both directions. It undercuts any claim that this speaker's July-September enthusiasm was a considered long-term view — but it also means the three bearish claims are not a reflexive bear's default position; they are the same person, on the same trade, changing his mind on valuation specifically after having been maximally bullish on the fundamentals. In our frame that makes the bearish half of this lane more informative than the bullish half, not less, because it is the only place in the lane where a voice has updated against his own prior.

The third finding is the identity of the two most recent bullish claims, and their frame must be disclosed. Both the 2026-03-27 and 2026-06-10 bullish claims are attributed to Jensen Huang, whose company's economics depend directly on the AI build-out being funded. This is promoter-class voice under house rule and it is weighted low. The 2026-06-10 claim — "any form — nuclear, wind, solar, hydrogen — will get funded" — is a claim about capital availability for the entire energy complex, not a view on GE Vernova's margin or multiple, and the 2026-03-27 claim carries no named speaker at all. Neither is company-level analysis and neither is used as conviction.

Sourcing quality, reported as the policy requires. Six of the eleven raw claims carry speaker: null or no speaker field and are attributed only to a CHANNEL — the weakest sourcing the 4-lane policy admits. A seventh records the speaker as "uncertain" with attribution: "guest". Only three of eleven carry a genuinely named speaker, and two of those three are the promoter-class voice. No claim in this lane carries a management speaker_role — GE Vernova's own executives do not appear anywhere in the lane, which is a meaningful positive: this is not the LLY failure mode where a company's own CEO supplied two-thirds of the "independent" view.

Net conviction: mixed. Seven bullish (conviction 55-85) against four bearish (52-60) on the raw count; five bullish against three bearish among the nine used. But the bullish claims are systematically about DEMAND and the bearish claims are systematically about PRICE, and they are not in conflict. Every bullish claim in this lane says turbines are scarce and the queue is long — which the filings confirm in detail, with equipment RPO up 36.7% in six months. Every bearish claim says the scarcity is already in the multiple. Both can be true, and on our arithmetic both are.

Conclusion, stated as a finding. The knowledge base's contribution to this dive is not a directional signal; it is a corroboration of the demand mechanism and an independent challenge to the price. Its most serious voice, at skill 2.0, having been maximally bullish on the fundamentals, now says these names at 30-50x price in three years of demand. GE Vernova trades at 40.92x the 2027 consensus, squarely inside that band, and our own independent fair-value work arrives 4.0% below spot without reference to it. Two methods, one qualitative and one arithmetic, agreeing that the price is ahead of the evidence. Breadth 4, conviction rating Medium-low, and the concentration and reversal are disclosed rather than smoothed.

8. Data integrity — what we rejected from the vendor file and why

Fourteen findings. GE Vernova is the most defect-dense file in this batch, and the defects are concentrated in exactly the places a two-year-old spin-off with lumpy one-off gains would break a data pipeline.

1. bal_a[0] is SEVEN MONTHS STALE and three material post-dating events sit between it and today — REJECTED as the balance-sheet basis. The most recent balance sheet in the vendor file is 2025-12-31, filed 2026-01-29. Since then: the $5.254 billion Prolec GE acquisition closed 2026-02-02 ($4.9 billion net of cash acquired); $2.6 billion of senior notes were issued 2026-02-04; $3.6 billion of shares were repurchased; and contract liabilities and deferred income rose $14.088 billion. We build the balance sheet from the 10-Q filed 2026-07-22 (as of 2026-06-30) throughout, and use bal_a only for prior-year comparison. This is the GOOGL/LLY defect class and it is present here in full force.

2. est.ebitdaAvg and est.ebitAvg are arithmetically impossible across all eight years — REJECTED in full. The FY2025 ebitAvg of $8.257B is 595% above the reported operating income of $1.388B; ebitdaAvg of $8.293B is 159% above the company's own Adjusted EBITDA of $3.2 billion. The signature: ebitdaAvg minus ebitAvg ranges from $32.4M to $76.5M across every year, implying depreciation and amortisation of under $77 million for a company reporting $853M-$1,797M. The 2023 row asserts $7.480 billion of EBITDA for a year with a $923 million operating loss. This is not a noisy estimate; it is a fabricated margin structure. No forward valuation in this dive touches these rows.

3. The FY2026 epsAvg of $30.734 is a GAAP/non-GAAP hybrid contaminated by a $4.0 billion non-cash gain — REJECTED. It exceeds the FY2027 estimate of $24.892 on a company growing revenue 15% annually. The decomposition is mechanical: $17.44 (Q1 GAAP actual) + $2.47 (Q2 GAAP actual) + $4.09 (Q3 consensus) + $6.73 residual. No conclusion in this dive uses the FY2026 estimate row. (By contrast, epsAvg for FY2025 — $7.276 — lands within 1.2% of our independently derived clean actual of $7.19 and IS used. The field is reliable; this one row is not.)

4. Trailing EBITDA and every multiple built on it capitalise a one-off accounting gain — REJECTED and rebuilt. km_ttm.evToEBITDATTM of 30.25x implies trailing EBITDA of $8.664 billion, which reconciles exactly to the sum of inc_q.ebitda across the four trailing quarters — including $5.445 billion from the March 2026 quarter, a quarter with $179 million of operating income. The vendor computes EBITDA as pre-tax income plus D&A, so the $3.992 billion Prolec remeasurement gain is inside the denominator of a valuation multiple. We compute trailing EBITDA as operating income $1.801B + approximately $1.135B of D&A = approximately $2.94B, giving EV/EBITDA of roughly 88.9x — or 81.6x against the company's own $3.2 billion Adjusted EBITDA for fiscal 2025. netDebtToEBITDATTM of −1.057x and enterpriseValueMultipleTTM of 30.25x are rejected with it. This is the GOOGL/JPM/NFLX/ORCL defect class and it is the most consequential single correction in this dive.

5. Reported trailing net income is 79% non-operating — a clean operating EPS is derived and stated. TTM net income of $9.529 billion comprises the $2.9 billion Q4 2025 tax valuation-allowance release (10-K verbatim), the $3.992 billion Prolec remeasurement gain (10-Q Note 8 verbatim), $330M on Proficy, $86M on an equity-method disposal and $264M on China XD Electric. Clean trailing net income is approximately $1.957 billion, approximately $7.25 per share, and the honest trailing multiple is approximately 140x, not the 29.24x the vendor prints. Two independent checks corroborate: the company's own Adjusted EBITDA of $3.2 billion, and the FY2025 consensus epsAvg of $7.276.

6. GAAP and non-GAAP are mixed without labels in earn_cal, producing a fictitious 794% beat and a fictitious 22% miss. epsActual is GAAP diluted EPS (it matches inc_q exactly in four of five quarters); epsEstimated is non-GAAP consensus. Q1 2026 reads $17.44 actual against $1.95 estimated; Q2 2026 reads $2.47 against $3.17. Neither comparison means anything. (One genuine internal discrepancy: earn_cal reports epsActual: 13.39 for the 2026-01-28 report while inc_q reports diluted EPS of $13.28 for the December 2025 quarter — an $0.11 gap between two vendor fields describing the same quarter. Unresolved; we use inc_q.) GE Vernova's own adjusted EPS is NOT available: the 8-Ks of 2026-04-22 and 2026-07-22 are in the archive as cover pages only, with the results in an Exhibit 99 that was not extracted. We therefore do not state whether the June quarter beat or missed on the street's basis.

7. seg_prod is not a segment table at all — corrected in presentation, accepted in substance. The vendor supplies exactly two lines, Product $20.934B and Service $17.134B, which is the 10-K's revenue-by-type disaggregation. GE Vernova has THREE reportable segments — Power, Wind and Electrification — and none of them appears anywhere in the vendor payload. The tie test passes: FY2025 sums to $38,068M against reported revenue of $38,068M, an exact match; FY2024 sums to $34,935M against $34,943M, an $8 million (0.02%) shortfall attributable to the 10-K's disclosed prior-year reclassification. The table is used, correctly labelled as equipment-versus-services. Segment figures in this dive are reconstructed from MD&A prose and labelled as our derivation.

8. seg_geo contains a subtotal masquerading as a peer line — a naive sum overstates revenue by 54.5%. The four regional lines (Europe $7.594B, Middle East and Africa $5.389B, Asia $4.629B, Americas excluding U.S. $3.116B) sum to exactly $20,728M, which is the "Non-US" entry. Summing all six vendor entries gives $58,797M against $38,068M of revenue — 1.545x. De-duplicated (United States + four regions) it ties to $38,069M against $38,068M, a $1 million rounding difference. This is the XOM defect and it is present.

9. totalDebt is $0 for 2025-12-31 while three other vendor fields say debt exists — the debt block is internally inconsistent and is REJECTED. shortTermDebt, longTermDebt, capitalLeaseObligations and totalDebt all read zero for fiscal 2025, while ratios_ttm simultaneously reports debtToMarketCapTTM: 0, debtToEquityRatioTTM: 0.331, debtToAssetsRatioTTM: 0.049 and interestDebtPerShareTTM: 14.776 (implying roughly $3.96 billion). Additionally, capitalLeaseObligations falls from $810M (2024) to $0 (2025) — GE Vernova did not extinguish its finance leases in a year, and the 2024-to-2025 totalDebt comparison is therefore invalid. The filings settle it: debt excluding finance leases was "less than $0.1 billion" at 2025-12-31 and $2.6 billion at 2026-06-30, all of it the February 2026 notes. This is the MSFT/ACN lease defect class combined with a zeroed field.

10. depreciationAndAmortization is ZERO for the June 2026 quarter in inc_q — a zeroed field that breaks the trailing D&A series. The quarterly series reads $205M, $205M, $213M, $230M, $342M, then $0. A company that has just consolidated a $5.254 billion acquisition with recognised intangibles does not report zero amortisation. The trailing D&A figure used in this dive (~$1.135 billion) substitutes an estimate of approximately $350 million for the missing quarter, labelled as our estimate. (interestIncome is likewise $0 for the March 2026 quarter against $28M-$135M in every other quarter — the same class of defect.)

11. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech and say why. The quote block reports $1,195.94 / $530.16; the computed technical block reports $1,174.86 / $547.96 — discrepancies of 1.8% and 3.3%. We use the tech figures throughout, because they are computed from the same close series that produces the moving averages, drawdown and relative-return figures, and mixing sources within the technical section would make the percentages inconsistent. (tech.last of $1,018.26 also differs from quote.price of $1,018.53 by 0.03%; we use the quote price for all valuation and the tech series for all technical percentages, and say so.)

12. Sector and industry classification are materially wrong, and no peer comparison is drawn. profile classifies GE Vernova as sector "Utilities", industry "Renewable Utilities", and the peers list is eight utilities and independent power producers (AQN, CEG, CWEN, DUK, ENLT, NEE, ORA, SO). GE Vernova is an industrial capital-goods manufacturer that SELLS to utilities; its renewable segment is its smallest and only unprofitable one. The 10-K's own competitor list — Siemens Energy, Mitsubishi Power, Westinghouse, Framatome, Rolls-Royce, Vestas, Siemens-Gamesa, Nordex, Envision, Goldwind, Hitachi Energy, Schneider Electric, Mitsubishi Electric, ABB — contains not one name from the vendor peer set. The classification would corrupt any sector screen or yield comparison, and no peer multiple appears anywhere in this dive. The profile description also states that "The Wind division concentrates on the fabrication and sale of wind turbine blades," which describes LM Wind Power, one of three Wind business units — the filing wins.

13. Pre-spin financial history is a different measurement basis and is fenced off, not adjusted. Fiscal 2021, 2022 and 2023 in inc_a and cf_a, and fiscal 2023 in bal_a, are carve-out combined financials derived from GE's records with allocated corporate costs, which the 10-K says "do not purport to reflect what the results of operations… would have been had the Company operated as a separate, stand-alone entity." Three tells in the vendor data: the weighted-share count is identical at 272,083,576 in all three years (the 10-K: EPS for all periods through 1 April 2024 was computed using the 274 million shares issued at spin); commonStock of $8.051 billion with retainedEarnings of $0 in 2023, which is GE's net-parent-investment line; and totalDebt of $1.709 billion in 2023 against a company that had no standalone capital structure and participated in GE cash pooling. No growth rate, trend or multiple in this dive spans 1 April 2024. We label rather than estimate around it.

14. profile reports the wrong incorporation and listing dates for the separation. It gives ipoDate: 2024-03-27 and says the company "was established in 2023." The 10-K states the spin-off completed on 2 April 2024; 27 March is when-issued trading and 2023 is the Delaware incorporation date. We use 2 April 2024 throughout. Minor, but on a name where the spin date defines which half of the history is usable, it is worth being exact about.

Not defects, correctly reported and independently confirmed — and there is a great deal of this, which is why the file is usable at all:

• The share count is exactly right and the four-figure price is not hiding a split. $271,268,761,608 ÷ $1,018.53 implies 266,333,600 shares; the 10-Q cover page states 266,333,581 shares outstanding at 2026-06-30. The difference is 19 shares, or 0.000007%. The prior 10-Q cover gives 268,719,995 at 2026-03-31, consistent with the disclosed 2.5 million shares repurchased in the June quarter. The three 8-Ks in the archive (2026-04-22 results, 2026-05-22 annual meeting, 2026-07-22 results) contain no split, no stock dividend and no recapitalisation. The diluted EPS denominator of 270 million for the June quarter sits appropriately above the 266.3 million outstanding, as a weighted average through a quarter of repurchases plus dilution should. This check was demanded by the price level and it passes cleanly.
• netDebt has the correct SIGN and the correct direction. bal_a reports −$8.848 billion (net cash) and km_ttm.enterpriseValueTTM of $262.109 billion is correctly BELOW the market capitalisation of $271.269 billion. Our rebuilt figure of $10.10 billion of net cash differs by $0.94 billion (0.36%), which is immaterial. This is the field that was sign-inverted on AAPL and understated by $321.4 billion on BRK-B; on GEV it is right and we record the pass.
• Currency is clean. inc_a[0].reportedCurrency = USD, profile.currency = USD, exchange NYSE, quote in USD. No mixing, no ADS ratio, no conversion. Every multiple in this dive is computed on a single currency.
• The cash-flow statement reconciles verbatim to the 10-K: operating cash flow of "$5.0 billion and $2.6 billion" and free cash flow of "$3.7 billion and $1.7 billion" for 2025 and 2024 match cf_a's $4.988B/$2.583B and $3.711B/$1.700B exactly.
• Revenue, gross margin and SG&A for the current periods match the filings to the decimal: gross margin 21.3% (June quarter) and 20.3% (half) against the 10-Q's stated figures; 19.8%, 17.4% and 14.5% for 2025, 2024 and 2023 against the 10-K verbatim; SG&A at 12.4% of revenue in the June quarter against the 10-Q's 12.4%. Vendor and filing agree on every current-period operating line.
• earn_cal revenue actuals match inc_q exactly in all five reported quarters ($11,104M, $9,339M, $10,956M, $9,969M, $9,111M).
• operatingProfitMarginTTM of 4.354% and netProfitMarginTTM of 23.035% both reconcile exactly to our independently summed trailing operating income of $1,801M and net income of $9,529M on $41,368M of revenue. The vendor's arithmetic is correct; the inputs are what is contaminated.

Non-equity tripwire — checked and passed. GEV is common stock, NYSE-listed, and the 10-Q cover page states "shares of common stock with a par value of $0.01 per share." The price of $1,018.53 is not par-like ($25, $100 or $1,000 with a fixed coupon); beta is 0.935, well above the sub-0.3 typical of a baby bond, though it is the lowest in this batch; the dividend is variable (lastDividend $1.75 against fiscal 2025 dividends paid implying roughly $1.00 per share) rather than a fixed coupon; volume was 2.10 million shares (roughly $2.1 billion of turnover); and the 52-week band of $547.96 to $1,174.86 is a 114% range, the opposite of the narrow band a fixed-income-like instrument shows. This is common equity. The four-figure price is simply a high-priced share with no split.

9. Technicals

• Price $1,018.53. −13.3% from the 52-week high of $1,174.86, and +85.8% above the 52-week low of $547.96. Position within the annual range: 75th percentile — the upper quarter, not the middle.
• Moving averages, and this is the whole picture: −0.5% below a 50-day average of $1,023.95, which is flat; +21.2% above a 200-day average of $840.32, which is rising. The long-term uptrend is intact and the intermediate trend has gone flat. The stock has been oscillating around its own 50-day line rather than trending.
• RSI 45.9 — neutral, neither overbought nor oversold. MACD −17.57 — negative.
• Maximum drawdown from peak over the trailing year: −13.3%, i.e. the current price is the worst point of a shallow drawdown. This stock has not had a serious correction.
• Relative performance, and this is the single most informative technical fact in the file: 3-month −5.19% against SPY +7.59% and QQQ +7.67%; 6-month +34.87% against SPY +11.09% and QQQ +15.71%; 12-month +55.10% against SPY +24.26% and QQQ +30.80%. GE Vernova outperformed the S&P by 31 points over twelve months and UNDERPERFORMED it by 12.8 points over the last three. The leadership stopped in roughly May and has not resumed.
• Sentiment: 22 buy, 7 hold, zero sell across 29 analysts; consensus target $1,256.29 (+23.3%), median $1,274 (+25.1%), high $1,450 (+42.4%), low $949 — 6.8% BELOW spot. A 22-0 buy-to-sell ratio on a stock at 40.9x forward earnings is a crowded book, and the crowding is itself a risk that one knowledge-base claim names directly ("the crowded, correlated concentration is a risk", 2025-10-05).

Today's move and what it does to the entry

GEV closed 2026-08-04 at $1,018.53, up 1.17% or $11.77 from a previous close of $1,006.76. But it OPENED at $1,046.00 — a 3.90% gap up — printed a high of $1,046.93 in the first minutes and then declined all session, closing $27.47 (2.6%) BELOW the open and only 1.85% above the day's low of $1,000.01, on 2.10 million shares against average volume of 2.91 million.

That is a failed rally, and the distinction matters. In the KLAC dive from the same session, a 6.95% gap closed within 0.7% of the day's high on above-average volume — buyers in control into the bell. Here the reverse: a large gap open, sold all day, closing near the low on 28% below-average volume. The broad tape was strongly higher (ARM +17.36%, PANW +5.53%, TXN +5.42%, IBM +3.91%, ANET +3.06%), and GE Vernova's +1.17% was among the weakest moves in the batch despite the most obvious thematic link to the AI trade that was driving it. No company-specific news is in this file for 2026-08-04; the last company event was the 2026-07-22 earnings release, thirteen days earlier. We treat the move as beta and the fade as information.

The honest read. The stock is 13.3% below its high, sitting exactly on a flat 50-day moving average, having underperformed the index for three months, with RSI at 45.9 and MACD negative. That is a stock in a range, not a stock in a trend. What it is not is cheap: 40.92x the 2027 consensus, 6.31x trailing sales on rebuilt enterprise value, approximately 140x clean trailing earnings, and 1.54x the entire $176.3 billion order book.

So the technical setup is not the reason to decline the entry — the technical setup is neutral. The valuation and the earnings quality are the reasons. Two things would change it, and they are the falsifiers:

• A price near $840, roughly the 200-day moving average and about 24x the FY2028 consensus mean, which would put our base of $978 at a +16% premium and invert the asymmetry. That is 17.5% below today's close and well inside the range this stock has traversed in a quarter.
• Or the 2026-10-28 print showing operating margin above 7% with equipment RPO holding above $85 billion, which would begin converting the backlog into the margin the price already assumes. At that point paying up is defensible, because the uncertainty being paid for has been partly resolved.

Buying today means paying 40.9x forward earnings, on a stock that has underperformed for three months, that faded 2.6% from its own open on the day, for a base case that is 4% BELOW the price.

10. Insiders — seven director grants and one real sale, by the head of the loss-making segment

Seven of the eight transactions are the annual non-employee director equity grant, all dated 2026-05-20 — the day of the annual meeting confirmed by the 8-K filed 2026-05-22 — all coded A-Award at a price of $0. An A-Award is a grant of restricted stock units, not an open-market purchase. It carries no signal about the recipients' view of the price, and we score it as neutral. Total: 1,536 units across six directors, worth approximately $1.56 million at today's close — a rounding error and an ordinary board-compensation event. Note also that none of these are F-InKind tax-withholding events, so the reading is different from the KLAC file: there is no vesting-related withholding here at all.

The eighth transaction is the only one with content, and it deserves to be read carefully rather than dramatised. Victor Abate, Chief Executive Officer of the Wind segment, sold 4,819 shares on 2026-06-01 at $948.08 — approximately $4.57 million — leaving 1,835 shares. That is a disposal of 72.4% of the position, by the officer running the one segment that is approximately breakeven and losing backlog, seven weeks before the second-quarter print in which Wind segment EBITDA fell 67%.

What we can and cannot say about that. The file contains no Rule 10b5-1 plan disclosure, so we do not know whether it was scheduled. A single sale by a divisional chief executive is a weak signal in isolation, executives sell for many reasons, and $4.57 million is not a large sum in absolute terms. What is notable is the residual: 1,835 shares, worth about $1.87 million. The head of a segment generating roughly $9.4 billion of revenue retaining under $2 million of stock is a small alignment position by any measure. We report it as a mild negative and no more than that, and we note it is the only open-market transaction of any kind in this file — zero insider purchases, one insider sale.

11. Verdict, kill-criteria and flip conditions

Watch.

This is the most interesting business in this batch and the one where the evidence and the price are furthest apart in opposite directions at the same time.

What is genuinely excellent, and none of it is in dispute: remaining performance obligations of $176.284 billion — 4.26 years of trailing revenue — up 37% year on year, with equipment RPO up 36.7% in six months to $87.821 billion; an installed base generating approximately 25% of the world's electricity, including approximately 7,000 gas turbines with 1,800 under long-term service agreements averaging ten years of remaining life; gross margin that has risen 530 basis points in two years (14.5% → 17.4% → 19.8%, and 21.3% in the June quarter); revenue accelerating to +21.9% year on year in the June quarter with 12 points of it organic; net cash of approximately $10.1 billion with no maturity before 2031 and $6.0 billion of undrawn facilities; $7.0 billion of buyback executed including $3.6 billion in the first half at an average roughly 18% below today's price; and a shrinking share count. Four independent knowledge-base sources corroborate the demand mechanism, with turbine queues described as running beyond 2029.

What we are declining to pay for: 40.92x the 2027 consensus and 29.16x 2028, against trailing earnings that are 79% non-operating — a $2.9 billion tax valuation-allowance release and a $3.992 billion acquisition remeasurement gain — and a clean trailing multiple of approximately 140x. Trailing free cash flow of $12.4 billion that is entirely explained by $13.7 billion of customer down payments, with working-capital-neutral free cash flow at approximately zero. A margin assumption in the price requiring net margin to go from about 4.7% to 15.6% by 2028, against a company whose own Adjusted EBITDA margin was 8.4% in 2025. A segment doing $9.4 billion of revenue at approximately breakeven after $2.021 billion of incremental Offshore Wind contract losses in three years, with the 10-K explicitly declining to bound the exposure. A FY2028 consensus EPS range whose high is 2.16x its low. A 22-buy, zero-sell street. And a base fair value of $978 that is 4.0% below spot and 22.1% below the consensus target.

The distinction that matters. We are not saying GE Vernova is a bad business or that the demand is not real — the demand is the best-evidenced thing in this entire batch, and it is in a filing rather than a narrative. We are saying that the backlog answers the volume question and is silent on the margin question, that the margin question is the entire valuation, and that the reported financials currently provide no evidence on it whatsoever because they are dominated by accounting events. In the Synthos frame, a name whose fair value sits below spot and below the street, where the trailing earnings cannot be used and the forward earnings carry a 2.16x dispersion, is a Watch by construction. We want this exposure. We want it 15-20% lower, or after the margin shows up.

Pre-registered KILL criteria — what would take this to Avoid:

• Any further Offshore Wind contract loss disclosed at the 2026-10-28 or subsequent print. After $637M, $1,005M and $379M in three years, a fourth consecutive charge establishes the losses as structural rather than terminal, and the 10-K's refusal to bound them becomes the operative fact.
• Equipment RPO falling below approximately $80 billion, or growing less than 3% in a half. That would say the 36.7% six-month surge was a one-time pull-forward of slot reservations rather than a durable demand shift, and it removes the only leg of the case we currently rate as proven.
• The contract-liability inflow falling below roughly $2 billion in a quarter while revenue grows. The free-cash-flow line collapses without anything operational changing, and the market discovers what §5 already shows.
• Operating margin failing to exceed 6.0% in any quarter through mid-2027. The 2028 consensus needs a step change; a company still printing 5-6% two years out is not on the path, and at that point the FY2028 low of $21.962 becomes the central case, putting today's price at 46.4x.
• A FY2028 consensus EPS migration below roughly $28, which would take our base fair value under $790 — 22% below spot.
• A break below the 200-day moving average of $840.32 on company-specific news rather than market beta.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price below approximately $840 (roughly 24x the FY2028 mean, 34x FY2027E) with no deterioration in the backlog. This is the single most likely path to a Buy and it is 17.5% below today's close — at $840 the base case is +16% and the asymmetry inverts.
• The 2026-10-28 print showing operating margin above 7.0% with equipment RPO above $85 billion. That is the margin thesis beginning to validate, and it is the specific evidence 6a says is missing.
• Wind reporting positive and growing segment EBITDA for two consecutive quarters, converting a $9.4 billion drag into a contributor and removing the single largest quality discount we apply.
• Resolution of the Vineyard Wind claims and counterclaims without a further charge, which would close the open-ended tail the 10-K refuses to bound.
• A new buyback authorisation materially above the remaining $3.0 billion, executed at prices near current levels — the company bought at an average of $837 in the first half and its willingness to buy at $1,000-plus is a live test of management's own view of value.

Where GEV fits in the Synthos Framework Portfolio. The electrification and power-infrastructure sleeve, at 0% today with a 2-3% target on a fill near $840 or a confirmed margin inflection in October. On the batch overlap question: GEV is thematically adjacent to the AI-infrastructure complex but is not another expression of the semiconductor trade — it is the physical-constraint leg of the same demand, with a genuinely different cycle, a 0.935 beta and a customer base of utilities rather than hyperscalers. That diversification is real and it is the strongest structural argument for owning this name rather than a fifth semiconductor. It does not, however, change the price. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $1,018.53.

Single biggest risk: the margin never arrives. Consensus requires net margin to go from approximately 4.7% (our clean trailing figure) to 12.6% in 2027 and 15.6% in 2028, on a company whose own Adjusted EBITDA margin was 8.4% last year, whose trailing operating margin is 4.35%, and which is carrying $39.9 billion of contract liabilities representing prices agreed in the past against costs that include $100-200 million of 2026 tariffs. The backlog guarantees the revenue and guarantees nothing about the profit. At 40.9x forward earnings with clean trailing earnings of $7.25 a share, a margin disappointment does not produce a proportionate price decline; it produces a de-rating on top of it, which is what the $615 bear case is — and that bear case is nothing more exotic than the analysts' own low estimate at a reasonable multiple.

Most fragile assumption in the price: that the FY2028 consensus mean of $34.924 is the right centre of the distribution. Twenty-three analysts produce a range from $21.962 to $47.408. The mean is not a forecast; it is the midpoint of a disagreement about whether this company earns roughly $6 billion or roughly $13 billion in 2028. The entire fair-value spread in this dive — $615 to $1,327 — comes from that one distribution at a single constant multiple. Everything else, including the multiple, is second order. The first data point that will narrow it is the 2026-10-28 operating margin, and until it prints, there is no responsible way to claim precision about what this business earns three years out.

Provenance & disclosures

• Traceability: 11 tagged knowledge-base claims name GE Vernova (raw hits 11, entity matches 11, text-only matches 0, 2 discarded for collision; breadth 4, net conviction mixed, 9 used). The two discards are both dated 2023-05-25, ten months before the 2 April 2024 spin-off, and are therefore claims about divisions inside the General Electric conglomerate rather than about the standalone security — discarded under the brief's explicit collision rule and not scored, notwithstanding that both have aged well. Sources and per-speaker concentration, reported as the attribution policy requires: Jordi Visser supplies 6 of 11 raw claims (54.5%) and 6 of 9 used (66.7%), appearing under two channel labels — jordi_visser and inside anthony_pompliano documents — which by channel would look like two sources and by speaker is one. Within those six there is a clean position reversal visible in the dates: bullish 2025-07-27 (conviction 60), 2025-09-20 (85), 2025-09-21 (75), then bearish 2025-10-04 (55), 2025-10-05 (52), 2026-01-11 (60) — thirteen days between the maximum bull call and the first bear call, on a valuation rather than a demand argument. The remaining used claims: Jensen Huang, 2 claims (2026-03-27 bullish 70 with no named speaker, and 2026-06-10 bullish 80) — promoter-class voice, weighted low with the frame disclosed, since his company's economics depend on the AI build-out being funded; and anthony_pompliano_show, 1 claim (2026-02-12 bullish 65, speaker recorded as "uncertain", attribution "guest"). Sourcing quality: 6 of 11 claims carry no named speaker and are attributed only to a CHANNEL — the weakest sourcing the 4-lane policy admits — and a 7th records the speaker as "uncertain." Only 3 of 11 carry a genuinely named speaker, and 2 of those 3 are the promoter-class voice. No claim in this lane carries a management speaker_role; GE Vernova's own executives appear nowhere in it, which is a meaningful positive and the opposite of the LLY failure mode. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — income statement and cash flow through the quarter ended 2026-06-30 and the fiscal year ended 2025-12-31, with the balance sheet taken from the 10-Q filed 2026-07-22 (as of 2026-06-30) rather than from bal_a, whose most recent entry is 2025-12-31 and is seven months and three material corporate events stale · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 = 2026-08-04T20:00:02Z — one second after the batch reference stamp of 1785873601 = 2026-08-04T20:00:01Z carried in the front matter, an immaterial difference recorded for completeness — ($1,018.53, +1.17%; 50-DMA $1,023.95; 200-DMA $840.32; RSI 45.9; MACD −17.57) · knowledge-base claims 2026-08-04. Fiscal/calendar note: GE Vernova's fiscal year ends 31 December, so fiscal and calendar years coincide throughout. Spin-off note: GE Vernova was separated from General Electric on 2 April 2024. Fiscal 2021, 2022 and 2023 are CARVE-OUT combined financials with allocated GE corporate costs and no standalone capital structure, and fiscal 2024 is a part-year. No growth rate, trend or multiple in this dive spans 1 April 2024. All figures come from the Synthos vendor data file for GEV or from the SEC filings in the GEV archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and its gaps: 10-K filed 2026-01-29 (fiscal 2025); 10-Q filed 2026-04-22 (March 2026 quarter); 10-Q filed 2026-07-22 (June 2026 quarter); 8-K filed 2026-04-22 (first-quarter results); 8-K filed 2026-05-22 (annual meeting voting results, 2026-05-20); 8-K filed 2026-07-22 (second-quarter results). Both earnings 8-Ks are present as the cover document ONLY — Exhibit 99, the release carrying the financials and the company's own adjusted measures, was not extracted, and neither file contains a single number. Numeric tables are stripped from the current filing-text extraction across all documents (assume PROSE-ONLY; a fix is pending), which is why the segment revenue and segment EBITDA levels in §3 are RECONSTRUCTED from MD&A prose changes and labelled as our derivation rather than quoted. The freshest hard numbers on segment EBITDA, the contract-liability balance and the June-quarter balance sheet detail are therefore prose-stated approximations from the 10-Q, and this is said explicitly so no reader mistakes a derived figure for a tabulated one.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the balance sheet basis was moved from bal_a[0] (2025-12-31) to the 10-Q (2026-06-30) because the $5.254B Prolec GE acquisition (2026-02-02), the $2.6B senior-note issuance (2026-02-04), $3.6B of buybacks and a $14.088B contract-liability increase all post-date the vendor's most recent balance sheet; trailing EBITDA rebuilt from the vendor's implied $8.664B to approximately $2.94B after establishing that the vendor's figure includes the $3.992B non-cash Prolec remeasurement gain inside a quarter with $179M of operating income, moving EV/EBITDA from the vendor's 30.25x to approximately 88.9x; trailing net income restated from $9.529B to a clean approximately $1.957B on the strength of the 10-K's verbatim "$2.9 billion benefit primarily from a U.S. tax valuation allowance release in the fourth quarter of 2025" and the 10-Q's verbatim $3.992B/$330M/$86M/$264M gain disclosures, moving the trailing multiple from 29.24x to approximately 140x; est.ebitdaAvg and est.ebitAvg rejected across all eight years after the FY2025 rows tested 159% and 595% above the company's own Adjusted EBITDA and reported operating income respectively, with an implied depreciation charge of $32-77M against actuals of $853M-$1,797M; the FY2026 epsAvg row of $30.734 rejected as a GAAP/non-GAAP hybrid contaminated by the Prolec gain; totalDebt of $0 rejected in favour of the 10-Q's "less than $0.1 billion" at 2025-12-31 and $2.6 billion at 2026-06-30, with capitalLeaseObligations dropping from $810M to $0 invalidating the year-on-year comparison; seg_geo de-duplicated after the four regional lines were found to sum exactly to the "Non-US" subtotal, a naive sum overstating revenue by 54.5%; seg_prod re-identified as the equipment/services disaggregation rather than the segment table, with the three reportable segments absent from the payload entirely; FY2024 gross margin corrected from the vendor's implied 18.05% to the 10-K's stated 17.4%, the vendor carrying the originally filed figures and the 10-K the reclassified ones; inc_q D&A of $0 for the June 2026 quarter and interestIncome of $0 for the March 2026 quarter identified as zeroed fields; quote.yearHigh/yearLow ($1,195.94/$530.16) rejected in favour of tech.hi52/lo52 ($1,174.86/$547.96); and the sector/industry classification ("Utilities" / "Renewable Utilities") and the entire eight-name utility peer set rejected against the 10-K's own competitor list, with no peer multiple used anywhere. Where vendor and filing AGREED — worth recording, because it is substantial: the share count (10-Q cover 266,333,581 at 2026-06-30 against a market-cap-implied 266,333,600 — a 19-share, 0.000007% difference, with no split in any 8-K); the netDebt SIGN and direction, correctly showing net CASH and an enterprise value below market capitalisation — the field that was inverted on AAPL and understated by $321.4B on BRK-B is right here; the currency (USD throughout, no ADS ratio); the fiscal 2025 and 2024 cash-flow statement to the rounding ("$5.0 billion and $2.6 billion" operating, "$3.7 billion and $1.7 billion" free cash flow); gross margin of 19.8%/17.4%/14.5% for 2025/24/23 and 21.3%/20.3% for the June quarter and half; SG&A at 12.4% of June-quarter revenue; seg_prod FY2025 summing exactly to $38,068M of reported revenue; all five earn_cal revenue actuals matching inc_q exactly; and operatingProfitMarginTTM and netProfitMarginTTM reconciling precisely to our independently summed trailing figures.
• Basis note: analyst estimates are non-GAAP; reported diluted EPS is GAAP; and on this name the gap is enormous rather than cosmetic. Fiscal 2025 GAAP diluted EPS was $17.69 against a consensus epsAvg of $7.276 for the same year — a 59% difference, because GAAP includes a $2.9 billion tax valuation-allowance release that no analyst modelled. Forward multiples in this dive use consensus non-GAAP EPS on the FY2027 and FY2028 rows only. The trailing multiple is shown on three bases and the reader should see all three: 29.24x on reported GAAP diluted EPS of $34.83; approximately 140x on our clean operating derivation of approximately $7.25; and the vendor's 30.25x EV/EBITDA, which is rejected outright. GE Vernova's own guided measures — "Adjusted EBITDA" and "free cash flow" — are defined by the company itself, and the 10-K concedes that free cash flow "does not delineate funds available for discretionary uses." Both are carried at half weight per house rule, reported alongside our own derivations rather than instead of them.
• Estimate coverage: 27 analysts on FY2027 revenue and 22 on FY2027 EPS; 31 and 23 on FY2028. The FY2029 and FY2030 EPS rows rest on 10 analysts each and are excluded from every conclusion. The FY2026 EPS row is discarded outright as a GAAP/non-GAAP hybrid. The FY2028 EPS range of $21.962 to $47.408 — a high 2.16 times the low — is reported as the central uncertainty of the name rather than smoothed into a point estimate.
• Peer note: the vendor peer set (AQN, CEG, CWEN, DUK, ENLT, NEE, ORA, SO) contains no comparables — every name is a utility or an independent power producer, i.e. a customer archetype rather than a competitor, and the vendor's sector and industry classification ("Utilities" / "Renewable Utilities") is materially wrong for an industrial capital-goods manufacturer. The 10-K's own competitor lists are used instead: Siemens Energy, Mitsubishi Power, Westinghouse, Framatome and Rolls-Royce (Power); Vestas, Siemens-Gamesa, Nordex, Envision and Goldwind (Wind); Hitachi Energy, Siemens Energy, Siemens, Schneider Electric, Mitsubishi Electric and ABB (Electrification). No peer multiples are supplied in the file for any of them, so no peer-multiple comparison is drawn anywhere in this dive.
• Fair-value caveat: the $615 / $978 / $1,327 anchors are a single constant multiple of 28x applied to the three points of the FY2028 consensus non-GAAP EPS distribution — the low of $21.962, the mean of $34.924 and the high of $47.408, from 23 analysts. The construction is deliberately unusual: on this name the uncertainty is overwhelmingly in the earnings rather than in the multiple, since the FY2028 high is 2.16x the low, and varying the multiple as well would manufacture false precision on top of an already enormous distribution. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 26x gives $908 (−10.9%) and 30x gives $1,048 (+2.9%). The bear case is not a crash scenario — it is the analysts' own low estimate at the same reasonable multiple; the bull case requires no re-rating at all, only delivery of the top of their own range.
• Timing: second-quarter 2026 results were released 2026-07-22, thirteen days before this dive, so the June quarter and the 2026-06-30 balance sheet are both filing-verified. The next print is 2026-10-28, 85 days away (consensus revenue $11.995B, EPS $4.09) — there is no near-term company-specific catalyst, which is an explicit input to the Watch verdict. 2026-08-04 was a broad technology up-day (ARM +17.36%, PANW +5.53%, TXN +5.42%, IBM +3.91%, ANET +3.06%); GEV's +1.17% was among the weakest in the batch despite the most direct thematic link to the AI trade driving it, and the session's shape — a 3.90% gap open at $1,046.00 followed by an all-day fade to close at $1,018.53, 2.6% below the open and 1.85% off the low, on 28% below-average volume — is treated as information rather than noise.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.