Corning GLW

Reference table

1. What the business is, and the segment structure that changed this year

Corning describes itself in the 10-Q as "one of the world's leading innovators in materials science, with a 175-year track record of life-changing inventions", applying "expertise in glass science, ceramic science, and optical physics". Chief executive, chairman and president Wendell P. Weeks; approximately 67,200 full-time employees; headquartered in Corning, New York.

Effective the first quarter of 2026, Corning reorganised into four reportable segments, and this is the single most important disclosure-structure fact in the file. Per the 10-Q:

> "Effective in the first quarter of fiscal 2026, the Company revised its segment structure... the Company began managing its Display and Specialty Materials businesses as a single operating segment, referred to as Glass Innovations, and its Hemlock Semiconductor Group, solar wafer, and solar module businesses as a single operating segment, referred to as Solar. In addition, the Company's Life Sciences business does not meet the quantitative threshold for separate reporting... Optical Communications and Automotive remain unchanged."

The four segments, plus a residual, with the recast comparatives from the 10-Q (core net sales, $M):

Read that table and the company splits cleanly in two. Optical Communications and Solar together are 52.9% of core sales and grew 40.8% year on year. Glass Innovations, Automotive and Life Sciences together are 47.1% and grew 0.9% — Life Sciences actually shrank. Corning is one fast business and one flat business bolted together, and the fast one has just become the bigger one. That is the whole growth story and it is the whole valuation problem: you are paying a 37x forward multiple for a blend in which nearly half is growing at less than the rate of inflation.

Segment net income tells the same story with more force. Reportable-segment net income was $867M in the June quarter against $652M a year earlier, +33.0%, with the 10-Q attributing the increase to "higher segment net income of $191 million in Optical Communications, partially offset by increased variable compensation and higher costs to ramp up capacity in Solar." Optical is $191M of a $215M increase.

Geography (FY2025, from the 10-K's own table, on the segment/core basis totalling $16,408M):

The United States grew from $4,439M to $6,760M in two years — 52.3% — while China grew 12.0%. That is the AI data-centre build showing up geographically, and it is why the NVIDIA partnership is framed around US-based capacity. China at 30.3% of the total remains the largest single-country concentration outside the US and sits in display glass and polysilicon, the businesses growing 1-2%.

Note the vendor's geographic block omits a line. The 10-K's table includes "All Other" net sales of $245M for FY2025; the vendor's seg_geo reports only North America, Asia Pacific and Europe, summing to $16,163M against the filing's $16,408M. A $245M, 1.5% omission — small, but it means the vendor's geographic block does not foot to any published total. Details in Section 8.

2. The transaction the vendor payload does not know about

On 6 May 2026 Corning entered a securities purchase agreement in connection with what the 10-Q calls "the long-term partnership with a customer to strengthen U.S. manufacturing for AI infrastructure." The earnings release names the customer: NVIDIA.

The terms, verbatim from the 10-Q's Note 12:

> "the Company issued and sold (i) a warrant (the "Traditional Warrant") to purchase up to 15 million shares of common stock of the Company at an exercise price of $180.00 per share, and (ii) a pre-funded warrant... to purchase up to 3 million shares of common stock at an exercise price of $0.0001 per share."

And the accounting, which matters:

> "The Traditional Warrant was issued without separate consideration and does not convey rights to a distinct good or service. Accordingly, it was accounted for as consideration payable to a customer and, as an equity-classified award, was measured at its grant-date fair value of $296 million..."

> "The Pre-Funded Warrant was issued for cash consideration of $500 million... In addition, as a result of the nominal exercise price, the underlying common shares are considered outstanding for purposes of calculating earnings per common share."

Alongside it, from Note 2:

> "During the second quarter of 2026, the Company recorded an approximate $0.7 billion net contract liability, consisting of a $1.0 billion customer deposit under a long-term supply agreement extending through December 31, 2029, partially offset by $296 million for a warrant issued to the customer... The fair value of the warrant will be recognized as a reduction of revenue as sales occur under the agreement."

Four consequences, none of which is in any vendor field.

First, there is $500 million of cash on the balance sheet that came from selling equity, not from operations, and 3 million shares that are already counted as outstanding for earnings-per-share purposes.

Second, there is a $1.0 billion customer deposit sitting inside operating cash flow. The 10-Q is explicit: first-half operating cash flow of $2,079M against $859M a year earlier was "primarily driven by higher net income and the receipt of an upfront customer deposit of $1.0 billion, related to a long-term supply agreement." Trailing free cash flow of roughly $2.40 billion therefore contains approximately $1.0 billion of prepayment. Adjusted, it is nearer $1.4 billion — and the reported price-to-free-cash-flow of 57.5x becomes roughly 98x. The contract-liability balance rose from $2.3 billion to $2.7 billion across the half.

Third, $296 million of future revenue has already been spent. The warrant's grant-date fair value will be recognised as a reduction of revenue as sales occur under the agreement. That is a direct, disclosed, scheduled headwind to reported optical revenue over the life of a contract running to end-2029, and consensus revenue estimates may or may not carry it.

Fourth, there are 15 million warrant shares at $180 outstanding — 1.7% of the share count, struck 12.6% above today's price, exercisable at any time and expiring on the third anniversary. The $137.74B market capitalisation does not include them. At $180 they represent $2.70 billion of proceeds and 15 million new shares; on a fully-exercised basis today's enterprise value would rise and the per-share arithmetic would change.

What this transaction actually signals is more interesting than the accounting. A customer that pays $500 million for a warrant, grants a further warrant on 15 million shares, and puts down a $1.0 billion deposit is not placing an order. It is co-financing capacity because it does not believe the capacity would otherwise exist. From the supplier's side, that is the strongest possible evidence that optical interconnect is a genuine physical bottleneck — and simultaneously the strongest possible evidence that Corning's growth is now tied to a small number of identifiable counterparties. Both readings are correct and both are load-bearing.

The other partnership disclosed in the same release: "Amazon announced a multiyear, multibillion-dollar agreement under which Corning will supply the optical fiber, cable, and connectivity solutions that power Amazon's expanding data center infrastructure across the United States."

3. Growth, margins and the two sets of books

Corning reports two revenue lines and two earnings lines, and the difference is not trivial. GAAP net sales for the June quarter were $4,505M; core net sales were $4,738M — a $233M, 5.2% gap. GAAP EPS was $0.64; core EPS was $0.78 — a 21.9% gap. The 10-Q explains the adjustment: core measures apply the company's own hedge rates to the hedged portion of foreign-currency exposure and exclude "realized and unrealized gains and losses on our... translated earnings contracts", restructuring and impairment charges, discrete tax items, pension mark-to-market and acquisition costs.

This matters because the analyst estimates in this file are on the CORE basis and the income statement is on GAAP. We verified it: the vendor's earn_cal reports a June-quarter revenue "actual" of $4,738M, which is exactly the 10-Q's core net sales, not the $4,505M GAAP figure in inc_q; and the March-quarter "actual" of $4,345M is exactly the six-month core figure of $9,083M less $4,738M. Every forward multiple in this dive is therefore struck on core EPS, and the trailing multiple is shown on both bases (72.4x GAAP, 55.9x core). A reader who computes a forward multiple using GAAP trailing earnings and a core forward estimate will get a nonsense number.

The GAAP annual series:

FY2023 is the base the Springboard plan was launched from — the 10-Q says it began in the third quarter of 2023 "with an annualized sales run rate of $13 billion." From there, GAAP revenue has risen 24.2% and GAAP operating income has risen 162%. Operating margin doubled from 7.1% to 14.9%. This is not a story that depends on forward estimates; it has already happened.

The quarterly trajectory, GAAP, with the core figures where disclosed:

Five consecutive quarters of 16-21% GAAP revenue growth, against a company that shrank 11.3% in FY2023. The 10-Q's own summary is that this was "our ninth consecutive quarter of year-over-year growth."

And the core margin expansion is the part management emphasises, from the earnings release: "We expanded core gross margin 120 basis points to 39.6% and core operating margin 190 basis points to 20.9%. We also grew core ROIC 180 basis points to 14.9%." Note the gap between core and GAAP margins — 39.6% versus 36.1% gross, 20.9% versus 15.5% operating. That is a persistent 350-540 basis point difference, driven mostly by the currency-hedge adjustment. Both are real; neither is the whole picture; we quote both throughout and label which is which.

The tax rate is doing some work and should be named. The June quarter's effective tax rate was 6.2%, against 14.4% a year earlier; the six-month rate was 13.7% against 16.9%. The vendor's effectiveTaxRateTTM is 13.8%. A June-quarter provision of $40M on $649M of pre-tax income is not a durable run-rate, and the 10-Q attributes the core/GAAP EPS difference partly to "largely non-cash discrete tax items." Any normalisation toward a low-20s rate is a mid-to-high-single-digit percentage headwind to GAAP EPS that is not visible in the growth rates above.

Research, development and engineering was $299M in the June quarter, 6.6% of GAAP revenue, up 8% — growing well below revenue, which is itself a source of the margin expansion and worth watching if the capacity build requires more process engineering than it currently gets.

4. Balance sheet and enterprise value — and a 21% debt overstatement in the vendor file

From the 10-Q at 2026-06-30, filing-verified:

The company states its own leverage in the 10-Q's liquidity table, which settles the question:

> Total debt: $8,424M (2026-06-30) and $8,434M (2025-12-31). Total debt to total capital: 39% and 41%. Working capital $4,896M. Current ratio 1.8:1. Days sales outstanding 59. Inventory turns 3.3.

The vendor's bal_a reports FY2025 totalDebt of $10,223M. The filing says $8,434M. The vendor overstates by $1,789M, or 21.2%. The mechanism is a double count: the vendor's shortTermDebt of $901M and longTermDebt of $8,476M already include the current and non-current finance-lease components (the filing's $804M and $7,630M plus roughly $97M and $846M of leases), and the vendor then adds capitalLeaseObligations of $846M on top. netDebt inherits the error: the vendor prints $8,697M where the correct figure including leases once is $7,851M and the company's own definition (excluding leases) gives $6,908M.

The enterprise-value rebuild:

> Market cap $137.740B

> + total debt per the 10-Q $8.424B

> − cash $2.504B

> = enterprise value $143.660B on the company's own debt definition

> (adding roughly $0.95B of finance leases gives $144.61B, which is what the vendor reports)

The vendor's enterpriseValueTTM of $144.615B is, unusually, defensible — it is built on the current June 2026 balance sheet and includes finance leases once. The defect is confined to the annual bal_a block, which double-counts them. We use $144.6B for enterprise-value multiples, on the lease-inclusive basis, and state that the company's own total-debt definition gives $143.7B.

Leverage is comfortable but not trivial. Net debt of $5.92B (company definition) against our recomputed trailing EBITDA of $4.06B is 1.46x; on the lease-inclusive basis, 1.69x. The vendor's netDebtToEBITDATTM of 1.959x is wrong in both directions — an overstated net debt divided by an understated EBITDA. Interest coverage is 7.4x. There is a $1.5 billion undrawn revolving credit facility.

Working capital and receivables carry one disclosure worth flagging. The 10-Q states: "During the six months ended June 30, 2026 and 2025, we accelerated the collection of $611 million and $802 million, respectively, in accounts receivable" through factoring arrangements. Receivables sold are removed from the balance sheet and the proceeds run through operating cash flow. Days sales outstanding of 59 is therefore a flattered figure, and operating cash flow contains factoring proceeds alongside the $1.0 billion customer deposit. Neither is improper; both mean reported operating cash flow overstates cash earned from selling product.

Approximately 49% of consolidated cash is held outside the United States, and the 10-K notes roughly $1.9 billion of indefinitely reinvested foreign earnings at 2025-12-31.

Book value. bookValuePerShareTTM is $15.264 and tangibleBookValuePerShareTTM is $11.676 — the vendor correctly distinguishes them here, a 23.5% difference reflecting $2,476M of goodwill and $610M of intangibles. That defect class (book equal to tangible book) does not occur in this file and it is worth recording, given how often it has.

5. Cash flow and capital returns — and why the free-cash-flow line is not what it looks like

First half of 2026, from the 10-Q: operating cash flow $2,079M against $859M a year earlier. That is a $1,220M improvement of which the 10-Q attributes $1.0 billion to the upfront customer deposit. Investing outflow was $562M, "primarily driven by increased capital expenditures of $238 million."

So the trailing free-cash-flow figure needs a health warning printed on it. The vendor's freeCashFlowPerShareTTM of $2.785 implies roughly $2.40 billion of trailing free cash flow, against $1.41 billion for full-year 2025. Approximately $1.0 billion of the improvement is a prepayment, not earnings. Adjusted, trailing free cash flow is roughly $1.4 billion, and the price-to-free-cash-flow multiple moves from the vendor's 57.5x to approximately 98x. We use the adjusted figure in every conclusion and flag the reported one as inflated.

Capital intensity is rising and that is the point of the plan, not a failure of it. Capital expenditure was $1,282M in FY2025 (8.2% of revenue) and is running at roughly $1,520M trailing (9.0%). The 10-Q says Corning "expect[s] to invest in capacity expansion and technology capabilities where appropriate to support customer demand" — a tenfold optical-connectivity expansion is not free. The Springboard plan promises "significantly higher returns on invested capital and substantially more free cash flow"; those are management claims and are half-weighted here. The observable version is core ROIC at 14.9%, up 180 basis points year on year.

Capital return is a dividend and not much else. The trailing dividend is $1.12 per share, a 0.70% yield, with $999M paid in FY2025 — a 52.2% payout of GAAP earnings. Buybacks are token: $163M in FY2025, 0.12% of market capitalisation, and treasury shares actually rose only 2 million across the half. Total shareholder yield is roughly 0.82% and shares are drifting slightly higher, not lower — diluted weighted-average shares went from 865M (Q2 2025) to 875M (Q2 2026), +1.2%, before any warrant exercise. This is not a capital-return story and should not be sold as one; the cash is going into plant.

6. Valuation — priced in or room?

At $159.905 (market cap $137.740B, enterprise value ~$144.6B):

Estimate coverage is thinner than the analyst count suggests. Eight to ten analysts on FY2026 and FY2027 is adequate; six on FY2028 EPS and four on FY2029 is not, and the FY2028 range makes the point — epsLow of $3.541 against epsHigh of $7.831, a 121% spread. The FY2027 row is much tighter ($3.888 to $5.052, a 30% spread) and is therefore the anchor for all three fair values. FY2029 supports no conclusion here.

est.ebitdaAvg and est.ebitAvg are rejected outright, with the same signature found on other names in this batch. In every forward year, ebitdaAvg is exactly 23.54% of revenueAvg and ebitAvg is exactly 13.43%, to four decimals — a fixed ratio applied to the revenue line, not a forecast. It is also wrong: the company's core operating margin is 20.9%, not 13.43%. Corroborating defect: netIncomeAvg exceeds ebitAvg in FY2027 ($3,850.7M against $3,063.0M) and FY2028 ($5,251.2M against $3,704.9M) — arithmetically impossible. All forward valuation runs on epsAvg and revenueAvg.

Peer context is unusable. The vendor peer set is Autodesk, Celestica, Fortinet, Infosys, Keysight, Motorola Solutions, Strategy Inc, Roblox and others — a software-and-services list with one contract manufacturer in it, against a capital-intensive materials-science company. Strategy Inc, a bitcoin treasury vehicle, is in the peer set for a glass manufacturer. No peer-multiple comparison is drawn. The company's own filings name no comparable public competitor set.

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

At $159.905 — 48.9x FY2026 consensus core EPS, 37.2x FY2027 and 27.2x FY2028 — the price embeds:

• Core revenue reaches $22.803B in 2027 and $27.582B in 2028. (Consensus-derived; 9 and 10 analysts.) That is +18.7% then +21.0% — growth accelerating in the outer year, which requires Optical Communications to sustain 30%-plus growth for three consecutive years while roughly half the company grows at 1-2%. The nearest test is the September quarter: management guided core sales to $4.9-5.0B, approximately +16% year on year — a DECELERATION from the June quarter's +17.1%.
• Core EPS reaches $4.303 by 2027 and $5.875 by 2028 — a 105% increase on the trailing $2.86 in two and a half years. (Consensus.) With share count drifting up 1.2% and no meaningful buyback, all of that has to come from revenue and margin.
• Core operating margin expands materially beyond 20.9%. (Our derivation.) To get $5.875 of core EPS on $27.582B of revenue with roughly 890M shares requires net income near $5.2B — an 18.9% net margin, against a trailing core figure closer to 14%. This is the most fragile assumption in the price, because it must be delivered while capital expenditure runs at 9% of revenue building capacity, while $296M of warrant value is being charged against revenue, and while the effective tax rate normalises off a 6.2% June quarter.
• The market keeps paying roughly 30x or better on the forward year. (Our number.) At 25x FY2028E the stock is $147; at 35x it is $206. The multiple, not the earnings, is the widest source of variance — and this stock has already been re-rated twice inside twelve months, up 157.7% and then down 37.5% from the high.
• The Springboard targets are approximately met. (Management's own claim, half-weighted.) Consensus FY2029 revenue of $30.981B is consistent with the "$30B annualised run rate by end of 2028" target. At $159.905 you are paying 25.0x the 2029 consensus — meaning three years of the plan are already in the price at a normal multiple. That single arithmetic fact is the core of the valuation case against.

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

Expected return over the next twelve months decomposes as: core EPS growth (+31.7%, from FY2026E $3.268 to FY2027E $4.303) + multiple drift (COMPRESSION, from 48.9x on the forward year to roughly 38x on the then-forward year, about −22%) + shareholder yield (+0.82%, of which some is offset by 1.2% share creep) ≈ +3% to +10%.

Our base assumes compression, and the reason is arithmetic rather than scepticism. Consensus EPS growth runs +31.7% (FY2027) then +36.5% (FY2028) then +8.9% (FY2029). The multiple the market pays has to fall as it approaches the year where growth halves, and it normally begins doing so a year early. Assuming today's 48.9x forward multiple simply persists would be assuming the market never notices the deceleration built into its own estimates.

Almost all of the expected return is earnings growth, and the multiple is working hard against it. If the multiple held at 48.9x on FY2027E the price would be $210 (+31.7%) — which is exactly the street's high target. If it compressed to 30x the price would be $129 (−19.3%) — which is exactly the street's low target. The street's own $129-to-$210 range is a pure multiple range on a shared earnings estimate, and that is the most honest description of the uncertainty here.

The bear case at $110 requires no earnings collapse — 28.3x the FY2027 consensus low of $3.888. The bull case at $240 requires no re-rating beyond today's level — 47.5x the FY2027 consensus high of $5.052, roughly today's FY2026 multiple applied a year forward.

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

• We do not differ from the street on direction, and only slightly on level — our $165 against consensus $173.75, a 5.0% gap. With 21 buy, 14 hold and 2 sell across 37 analysts, this is the most divided coverage in the batch, and the division is entirely about multiple, not about the business. When our number lands inside the consensus range and the disagreement in the market is about what to pay rather than what will happen, we have no analytical edge — and no-edge names default toward Watch. We say that plainly.
• We differ from the vendor data on leverage and cash flow, and both corrections cut against the stock. Total debt is overstated 21.2% in the annual block (a lease double-count, which we correct downward in Corning's favour), but trailing free cash flow is overstated by roughly $1.0 billion (a customer deposit, which we correct downward against it). The net effect of doing the work properly is a company with less debt and much worse cash conversion than the screens show, and the second matters more at a 37x forward multiple.
• We think the NVIDIA warrant is under-analysed in both directions. The bullish reading — a counterparty co-financing capacity is the strongest evidence of a real bottleneck — is well understood. The bearish half is not: 15 million warrant shares at $180 are 1.7% dilution not in the market capitalisation, and $296 million of warrant value is queued to be charged AGAINST REVENUE as the very sales the thesis depends on are recognised. Watchable number: reported Optical Communications revenue against segment growth commentary in the 2026-10-27 print — a visible gap between the two would be that charge appearing.
• Where the knowledge base and the price disagree is the interesting part. The lane is eighteen bullish claims and zero bearish, from the highest-skill voice in the file, several with explicit multi-year tests ('compounds over next six years', 'moves meaningfully higher over the next 2 years', 'outperforms Micron over the next year'). The knowledge base thinks this is early; the multiple thinks it is priced. Our resolution is that both can be right on different horizons, which is precisely what a Watch with a named entry expresses.
• Positive surprise that would force a re-rate: the September quarter beating the top of management's own $5.0B core sales guide with core operating margin above 22%, which would validate the margin leg of 6a; or a third named hyperscale partnership on Amazon/NVIDIA terms, which would convert customer concentration from a risk into a moat; or Glass Innovations returning to growth, which would fix the blend problem at a stroke.
• Negative surprise that would break the thesis: September core sales below $4.9B, i.e. missing management's own floor seven days after issuing it; Solar profitability failing to improve after the disclosed shutdown and equipment upgrade; the contract-liability balance falling without corresponding revenue; a capital-expenditure guide materially above 9% of revenue, which would push the free-cash-flow inflection out again; or any deterioration in the 30.3% of core sales sold into China.

Synthos fair values

All three anchors are multiples of the FY2027 consensus core EPS distribution (mean $4.303, low $3.888, high $5.052, 10 analysts) — the best-covered forward row with a defensible range — cross-checked against FY2026E of $3.268 and FY2028E of $5.875.

• Bear ~$110 — 28.3x the FY2027 consensus LOW of $3.888. Cross-check: 33.7x FY2026E; 18.7x FY2028E; roughly 19% below the rising 200-day moving average of $135.92. The scenario: hyperscale optical orders normalise after the initial build, Solar's ramp costs persist, Glass Innovations stays flat, the tax rate normalises, and the multiple compresses toward what a capital-intensive materials company usually carries. −31.2%.
• Base ~$165 — 38.3x the FY2027 consensus MEAN of $4.303. Cross-check: 50.5x FY2026E; 28.1x FY2028E; enterprise value of $149.0B on FY2027E revenue of $22.803B is 6.5x sales. Sensitivity, stated openly: 34x gives $146 — 8.6% BELOW spot — and 42x gives $181, 13.2% above. The entire answer lives in a 34-42x band, and at 37.2x the market is already inside it. The scenario: consensus roughly lands, margins expand less than the outer years require, and the multiple compresses in line with the deceleration built into the estimates. +3.2%.
• Bull ~$240 — 47.5x the FY2027 consensus HIGH of $5.052, approximately today's FY2026 multiple carried a year forward. Cross-check: 40.9x FY2028E; 6.1% below the 52-week high of $255.69. This bull case requires no re-rating above where the stock has already traded this year — only that the Springboard plan runs at the top of the analyst range. +50.1%.

Base is 3.2% above spot; asymmetry roughly 1.61:1 to the upside (31.2% down, 50.1% up). That ratio is acceptable but the base case is not: a fair value 3.2% above the price, sitting below the street, is the definition of a fully-valued name. What makes this a Watch rather than a Hold is the quality of the knowledge-base lane and the possibility of a fill nearer the 200-day.

7. Knowledge base — twenty-seven hits, eighteen bullish claims, zero bearish, and a concentration test it passes

Raw hits: 27. Entity matches: 18. Text-only matches: 9. Used as name-level conviction: 16. Used as basket/sector colour: 9. Discarded: 2.

The sweep ran entity terms GLW, Corning, Corning Inc and Corning Incorporated, plus a free-text sweep on corning, optical fiber, fiber optic and Springboard, across all 51,928 distilled claims.

Discards first, because the discipline matters. Two claims from arthur_hayes, both dated 2023-09-05, matched on the word springboard: "ongoing regional bank collapses... trigger macro turmoil and money printing that springboards Bitcoin toward $1 million." The verb, not the plan. Discarded in full. No other false positive survived the sweep — "Corning" is a rare enough token that it does not collide, unlike the short tickers elsewhere in this batch.

Every remaining claim is bullish. There is no bearish claim about Corning in the knowledge base. Conviction runs 58 to 85.

The name-level claims, in date order (all bullish):

> 2025-12-06 · conviction 78 · thesis · speaker jordi_visser (role: independent) · skill 2.0

> "Corning hit a 'springboard' — optical fiber for data-center buildout plus a coming auto/phone/computer AI-chip upgrade cycle — a long-dormant physical-AI beneficiary where you'll see doublings."

> 2025-12-07 · conviction 82 · thesis · channel jordi_visser, no named speaker · skill 2.0

> "Corning, dead 17 years, now moving like early Nvidia — driven by data-center optical fiber, upgrade-cycle glass/sensors and humanoids; a compound beneficiary with multiple exponential capex cycles ahead."

> Test: "GLW compounds over next six years like a Qualcomm-4G/Broadcom-hyperscaler beneficiary"

> 2025-12-13 · conviction 78 · entities GLW, TER, COHR · speaker jordi_visser · skill 2.0

> "2026 is the embodied-AI hardware year; the Blackwell gateway lifts component names — Corning, Teradyne, Coherent, lidar, batteries, sensors."

> 2025-12-14 · conviction 80 · channel jordi_visser, no named speaker · skill 2.0

> "Focus on hardware, hardware, hardware — semiconductors and physical-AI names geared to embodied AI (e.g. Corning) beat software as the investment theme."

> 2025-12-20 · conviction 58 · speaker jordi_visser · skill 2.0

> "Corning benefits from the expansion into glass and is an AI company via data-center buildout."

> 2026-01-31 · conviction 68 · entities GLW, MU · speaker jordi_visser · skill 2.0

> "Corning will outperform Micron over the next year — optical-fiber shortage (Meta's $6B deal) plus glass buildout for phones, computers and autos; still early stage."

> Test: "Corning outperforms Micron over the next year"

> 2026-02-01 · conviction 85 · entities Corning, Meta · channel jordi_visser, no named speaker · skill 2.0

> "Optical fiber is the emerging physical bottleneck; the $6B Meta–Corning deal is early, before the glass buildout for autos, phones and computers."

> 2026-02-05 · conviction 80 · entities Corning, GLW, Meta · channel jordi_visser, no named speaker · skill 2.0

> "Corning is early-stage; optical fiber is running out (Meta $6B deal) and it benefits from glass buildout for phones, computers and autos; outperforms Micron."

> 2026-02-27 · conviction 75 · entities Corning, EOS Energy, Enersys · channel compound_and_friends, no named speaker · skill 1.0 (two near-identical claims from the same session)

> "Electrical/power infrastructure for data centers is the leadership trade — Corning is the year's hottest stock; small-cap electrical suppliers charts look like 2021 cryptos."

> 2026-04-26 · conviction 75 · entities COHR, GLW, MRVL, LITE · speaker Jordi Visser · skill 2.0

> "Optical is necessary for inference and power efficiency — Coherent, Lumentum, Corning, plus Marvell (silicon-photonics leader...); transceiver market in shortage."

> 2026-05-03 · conviction 62 · speaker Josh Brown · channel compound_and_friends · skill 1.0

> "Corning, a 100+ year-old glassmaker, is Halo — makes fiber optics for data centers, heavily involved in GPU buildout."

> 2026-05-03 · conviction 60 · entities GLW, MRVL, COHR · channel jordi_visser, no named speaker · skill 2.0

> "Optical fiber is early-cycle AI and ripping to all-time highs — Meta buying $6B of Corning fiber, Jensen doing deals across optical fiber, coherent light and Marvell."

> 2026-05-09 · conviction 85 · entities MU, MRVL, GLW, NVDA · speaker jordi_visser · skill 2.0

> "Parabolas in compute names (Micron, Marvell, Corning, Nvidia) are fundamentally justified — inference/agent demand is running ahead of supply; this is not the dot-com bubble."

> 2026-05-23 · conviction 65 · speaker jordi_visser · skill 2.0

> "Corning has a massive order and earnings intact; expects it meaningfully higher over the next 2 years, question is acceleration in other business lines."

> Test: "Corning moves meaningfully higher over the next 2 years"

> 2026-07-29 · conviction 78 · speaker Josh Brown · skill 1.0 — the most recent claim in the lane, dated the day after the second-quarter print

> "Corning is on his list of best stocks in the market, a buy — wants to own it and be long it (frustrated the momentum sleeve sold it)."

And, quoted separately at half weight because it is counterparty management voice:

> 2026-05-21 · conviction 62 · horizon: FACT · speaker jensen_huang · channel jensen_huang · skill 1.0

> "Because of the NVIDIA partnership, Corning will build 10 times as much fiberoptics capability in the United States as previously."

> Test: "Corning scales US fiberoptics capacity 10x"

This is the chief executive of the counterparty describing his own transaction, and under the house rule on management voice it does not enter the conviction pool. What makes it useful anyway is that it is a FACT-horizon claim and Corning's own 10-Q and earnings release corroborate it independently — the release states Corning "will expand its U.S.-based optical connectivity manufacturing capacity by 10x." A claim that a filing confirms is evidence regardless of who said it, and we treat it as corroboration rather than conviction.

The concentration test

The jordi_visser channel supplies 20 of 27 raw hits (74%) and 13 of 18 entity matches (72%). That is well past the one-third threshold at which house rules require a sensitivity test, and the test is the one that moved SCHW from 7-bullish/7-bearish to 2/6.

Stripping jordi_visser entirely leaves five entity claims from three independent sources:

All five remain bullish. The lane survives its own stress test. That is an uncommon result in this programme and it is stated as a finding rather than assumed.

Two qualifications that keep the rating from being higher.

First, six of the twenty jordi_visser claims carry neither a speaker nor a role and are attributed only to the channel (2025-12-07, 2025-12-14, 2026-02-01, 2026-02-05, 2026-03-29, 2026-05-03). Under the 4-lane attribution policy that is the weakest sourcing we accept, and given the documented failure mode where a channel slug carries a guest's voice rather than the host's, those six cannot be confirmed as that speaker's own view. They include the highest-conviction claims in the lane (85 and 82).

Second, one factual assertion recurs across four claims and we cannot verify it: a "$6B Meta–Corning deal." The filings in this archive name Amazon and NVIDIA as the second-quarter partnerships and do not mention Meta. The claim may be true and simply predate or postdate the disclosure window in this archive, but it is not corroborated by any document we hold, and we flag it rather than repeating it as fact.

Conclusion. Breadth 4, net conviction positive-high, claim count 18. This is the deepest and most one-sided knowledge-base lane in the batch and it is a genuine input to the growth and exponential scores — but a conviction lane is not a valuation argument, and eighteen bullish claims do not make 37.2x FY2027 consensus cheap. The lane is why this is a Watch with a named entry rather than an Avoid, and it is why the flip conditions in Section 11 are specific and short.

8. Data integrity — what we rejected and why

Nine findings. Corning's vendor file is middling: the current-quarter enterprise value is right, the book-versus-tangible-book distinction is correct, and the FY2025 product table foots exactly — but the annual debt block, the trailing EBITDA, the forward EBIT/EBITDA rows and the segment history are all wrong in different ways.

1. bal_a.totalDebt double-counts finance leases — REJECTED, 21.2% overstatement. The vendor reports FY2025 total debt of $10,223M. The 10-Q's own liquidity table states "Total debt | 8,424 | 8,434" for 2026-06-30 and 2025-12-31. The mechanism: the vendor's shortTermDebt ($901M) and longTermDebt ($8,476M) already contain the current and non-current finance-lease components, and capitalLeaseObligations ($846M) is then added again. The vendor reports total debt of $10,223M and net debt of $8,697M; the filing shows $8,434M and (against cash of $1,526M) $6,908M; the lease-inclusive figure counted once is $9,377M and $7,851M. We use the filing. The same pattern appears in FY2024 ($8,091M reported against $880M of leases) and FY2023.

2. Trailing EBITDA is internally contradictory — REJECTED and recomputed. ratios_ttm.ebitdaMarginTTM of 20.685% implies trailing EBITDA of $3,509M, which against trailing operating income of $2,624M implies depreciation and amortisation of only $885M. But FY2025 cash-flow D&A was $1,347M, and the vendor's own capexToDepreciationTTM of 1.0585 against trailing capex of $1,520M independently implies $1,436M. The vendor contradicts itself. We compute EBITDA as operating income $2,624M + D&A $1,436M = $4,060M (23.9% margin) and use that. Consequence: evToEBITDATTM/enterpriseValueMultipleTTM of 41.21x becomes ~35.6x, and netDebtToEBITDATTM of 1.959x becomes 1.46x on the company's own net-debt definition.

3. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature — REJECTED in full. In FY2026 through FY2029, ebitdaAvg is exactly 23.54% of revenueAvg and ebitAvg is exactly 13.43%, to four decimal places. The company's actual core operating margin is 20.9%. Corroborating defect: netIncomeAvg exceeds ebitAvg in FY2027 ($3,850.7M against $3,063.0M) and FY2028 ($5,251.2M against $3,704.9M), which is impossible. All forward valuation uses epsAvg and revenueAvg.

4. seg_prod does not reflect the current segment structure and is on inconsistent bases across years — FY2025 used, all earlier years REJECTED. The FY2025 entry has seven lines summing to $15,629M, which matches GAAP revenue exactly and is therefore a legitimate GAAP-basis disaggregation — but it is the OLD segment vocabulary (Display Technologies, Specialty Materials, Polycrystalline Silicon, Life Sciences), not the four-segment structure (Optical Communications, Glass Innovations, Automotive, Solar) effective from the first quarter of 2026. The FY2024 entry sums to $14,056M, which matches neither GAAP revenue ($13,118M) nor the 10-K's segment total ($14,469M), and uses a third vocabulary again ("Environmental Technologies", which no longer exists). FY2021, FY2022 and FY2023 are absent entirely; the series jumps from 2024 to 2020, then to 2019, 2018, 2012 and 2011. We use the FY2025 lines for GAAP mix and take all current segment analysis from the 10-Q directly.

5. seg_geo omits the 10-K's "All Other" geography line — corrected. The vendor reports North America $7,105M, Asia Pacific $7,646M and Europe $1,412M for FY2025, summing to $16,163M. The 10-K's table adds "All Other | 245" and totals $16,408M. The vendor's block therefore foots to no published total. We use the 10-K's table throughout, including the country-level detail the vendor does not carry (United States $6,760M, China $4,971M). Note also that the geographic table is on the segment/core basis, not GAAP.

6. The NVIDIA warrant transaction is absent from every vendor field — supplied from the filings. A May 2026 securities purchase agreement issuing a warrant on 15 million shares at $180 and a pre-funded warrant on 3 million shares for $500 million cash, with a $296 million grant-date fair value recorded as consideration payable to a customer and a $1.0 billion customer deposit under a supply agreement to December 2029. Nothing in quote, profile, bal_a, cf_a, km_ttm or est reflects any of it. The 3 million pre-funded shares are already counted as outstanding for EPS; the 15 million are not. This is the "whole transaction absent" defect class and it is material to both the share count and the forward revenue line.

7. Trailing free cash flow contains a $1.0 billion customer prepayment — flagged, and adjusted in every conclusion. The vendor's freeCashFlowPerShareTTM of $2.785 implies roughly $2.40B; the 10-Q states that first-half operating cash flow of $2,079M (against $859M) was "primarily driven by higher net income and the receipt of an upfront customer deposit of $1.0 billion." Adjusted trailing free cash flow is approximately $1.4B, and priceToFreeCashFlowRatioTTM of 57.5x becomes roughly 98x. Separately, the 10-Q discloses $611M of accounts receivable collections accelerated through factoring in the half, also inside operating cash flow.

8. earn_cal mixes bases — the 2026 entries are CORE, and the 2025 entries reconcile to neither. The June-quarter "revenueActual" of $4,738M is exactly the 10-Q's core net sales, not the $4,505M GAAP figure in inc_q; the March entry of $4,345M is the six-month core total less the June quarter. But the 2025 entries ($4,412M for Q4, $4,100M for Q3) do not sum with the filing's H1 2025 core figure of $7,724M to the 10-K's FY2025 core total of $16,408M — they fall $172M short. We treat earn_cal as core-basis for 2026 and as unreliable for 2025, and take all 2025 quarterly figures from inc_q on a GAAP basis, labelled as such.

9. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech. The quote block reports $271.78 and $63.07; the computed technical block reports $255.69 and $63.17, a 5.9% discrepancy on the high. We use tech throughout for consistency with the moving averages, drawdown and relative-return series.

Not defects, correctly reported and independently confirmed — worth recording: the current enterprise value of $144.615B, which reconciles to the June 2026 balance sheet including finance leases once; the share count (10-Q common stock of $927M at $0.50 par implies 1,854M issued, less 994M treasury = 860M outstanding, against 861.4M implied by market capitalisation ÷ price, a 0.2% match); bookValuePerShareTTM of $15.264 correctly distinct from tangibleBookValuePerShareTTM of $11.676 — the book-equals-tangible-book defect does not occur here; the FY2025 seg_prod lines footing exactly to GAAP revenue; the FY2025 revenue, gross profit, operating income and net income lines; and interestCoverageRatioTTM of 7.35x.

Vendor composite rating — REJECTED with the mechanism. The rating block scores Corning C+ / 2 overall, built on 1 out of 5 for price-to-earnings, 1 out of 5 for price-to-book and 1 out of 5 for debt-to-equity. The price-to-earnings score uses the 72.4x trailing GAAP figure without acknowledging the core basis on which the company and every analyst estimate is struck; the debt-to-equity score uses the 21.2%-overstated total debt documented in finding 1; and the price-to-book score penalises a company with $14.9B of net property, plant and equipment carried at historical cost. A composite rating built on a debt figure the company's own filing contradicts is not a usable lens and we discard it.

Non-equity tripwire — checked and passed. GLW is common stock, $0.50 par value, NYSE-listed, per the 10-Q cover. Price of $159.905 is not par-like; beta is 1.086; the dividend is variable and has been raised over time; volume was 16.64M shares (~$2.7B of turnover); and the 52-week band of $63.17 to $255.69 is a 305% range. This is common equity.

9. Technicals

• Price $159.905. −37.5% from the 52-week high of $255.69, but +153.1% above the 52-week low of $63.17. Position within the annual range: 50th percentile — precisely the middle of an enormous range.
• Moving averages, split decision: −11.8% below a 50-day average of $181.16, which is declining; +17.6% above a 200-day average of $135.92, which is rising. The long-term uptrend is intact; the intermediate trend is broken.
• RSI 43.3 — neutral-to-weak, nowhere near oversold. MACD −12.90 — negative.
• Maximum drawdown from peak over the trailing year: −37.5%.
• Relative performance, and the twelve-month figure is the headline of the whole file: 3-month −0.03% against SPY +7.6% and QQQ +7.7%; 6-month +44.9% against SPY +11.1% and QQQ +15.7%; 12-month +157.7% against SPY +24.3% and QQQ +30.8%. Corning has outperformed the S&P 500 by 133 points over twelve months and underperformed it by 7.6 points over three. That is a parabola and its correction inside the same year.
• Sentiment is the most divided in this batch: 21 buy, 14 hold, 2 sell across 37 analysts; consensus target $173.75 (+8.7%), median $177.50, high $210 (+31.3%), low $129 — 19.3% BELOW spot. Thirty-eight percent of the coverage is not at buy on a stock that just raised its multi-year plan, and the $81 spread between the high and low targets is entirely a multiple disagreement.

Today's move and what it does to the entry

GLW closed 2026-08-04 at $159.905, up 9.05% or $13.265 from a previous close of $146.64. It opened at $155.00 — a 5.70% gap — traded $150.85 to $162.49, and closed 1.6% below the day high on 16.64M shares against a 14.48M average. No company-specific news dated 2026-08-04 appears in the filing archive; the last company event was the 2026-07-28 earnings release, seven days earlier. 2026-08-04 was a broad technology and semiconductor up-day across this batch, and a 9.05% move on a 1.086-beta stock with no filing behind it is best read as sector beta amplified by a name that had fallen 37.5%.

The honest read. This is a bounce inside a downtrend, not a breakout: the 50-day average sits at $181.16, still 13.3% overhead, and RSI at 43.3 and MACD at −12.90 are not confirming anything. One strong session after a 37.5% drawdown resolves nothing.

So the gap is not the reason to decline the entry. The valuation is. Even 37.5% below its high, Corning trades at 72.4x trailing GAAP earnings, 55.9x trailing core, 37.2x FY2027 consensus and 8.5x sales, with adjusted trailing free cash flow near $1.4 billion against a $137.7 billion market capitalisation. The drawdown removed the parabola; it did not create a discount.

What would change this, and these are the falsifiers:

• A price near $135-140 — roughly 31-33x FY2027E, on or just above the rising 200-day at $135.92 — which would put our base $165 at an 18-22% premium and make the risk/reward genuinely asymmetric. That is 12-16% below today's close and inside the range this stock has traversed in a month.
• Or the 2026-10-27 print landing above the top of management's own $5.0B core sales guide with core operating margin above 22%, which would validate both legs of 6a at once. At that point paying up is defensible, because the margin expansion the price requires would have been observed rather than assumed.

10. Insiders — a chief executive exercising and selling 100,000 shares, and a $186 print

The reading, stated carefully. The four director entries are routine restricted-stock-unit grants of 137-161 shares at zero cost — noise. The CHRO's 18,378 shares are coded F-InKind, shares withheld by the issuer to satisfy tax on vesting; not an open-market sale, and it carries no signal.

The chief executive's transactions are the only ones with content, and they are a genuine exercise-and-sell. On 9 June 2026 Wendell Weeks exercised options at $27.03 and sold 100,000 shares at $186.46, realising roughly $18.6 million on an exercise cost of $2.7 million. A second exercise of 100,000 shares at the same strike is recorded on a separate ownership line. This is not withholding; it is a discretionary disposal.

Two things temper the reading and one sharpens it. Tempering: the shares came from options with a $27.03 strike, which is the kind of position that has to be monetised before expiry, and Weeks retains 908,353 shares — approximately $145.3 million at today's close — so the remaining stake dwarfs the sale. Sharpening: the sale printed at $186.46, which is 16.6% ABOVE today's close of $159.905, and the CHRO's withholding priced at $209.83, 31.2% above. The insider transactions in this file were struck in a price band the stock has since left. That is a fact about the drawdown rather than about anyone's judgement, but it is the only concrete evidence in the file of anyone connected to the company transacting, and it was a sale.

What the file does not contain: any open-market purchase by any insider. With the stock 37.5% below its high and the multi-year plan just upgraded, the absence of a single discretionary buy is worth one line — not bearish, but not the corroboration a bullish knowledge-base lane would ideally have.

11. Verdict, kill-criteria and flip conditions

Watch.

What is genuinely excellent, and none of it is in dispute: a business that shrank 11.3% in FY2023 and has since grown GAAP revenue 24.2% and GAAP operating income 162%, with operating margin doubling from 7.1% to 14.9%; nine consecutive quarters of year-on-year growth; a June quarter with core sales +17%, core EPS +30%, core gross margin +120 basis points to 39.6%, core operating margin +190 basis points to 20.9% and core return on invested capital +180 basis points to 14.9%; Optical Communications +32.3% and Solar +89.6%; two named hyperscale partnerships disclosed in a single quarter, one of them backed by $500 million of warrant proceeds and a $1.0 billion deposit in exchange for a tenfold US capacity expansion; total debt of $8.42B against $2.50B of cash and 39% debt-to-capital; and the deepest, most uniformly bullish knowledge-base lane in this batch — eighteen entity claims, zero bearish, surviving a concentration test that strips the channel supplying 74% of them.

What we are declining to pay for: 72.4x trailing GAAP earnings, 55.9x trailing core earnings, 48.9x FY2026 consensus, 37.2x FY2027, 8.5x sales, roughly 35.6x recomputed trailing EBITDA, and roughly 98x adjusted free cash flow once the $1.0 billion customer deposit is removed from it. A base fair value of $165 that is 3.2% above spot and 5.0% below the street's $173.75, in a range where 34x gives $146 — below the price. A forward case that needs core operating margin to move materially above 20.9% while capital expenditure runs at 9% of revenue and $296 million of warrant value is charged against revenue. A twelve-month return of +157.7% already in the price. And a September guide, issued by management seven days ago, that decelerates to +16% core sales.

The distinction that matters. We are not saying Corning is a poor business or that the Springboard plan is fantasy — the growth score of 9 and the exponential score of 8 are the highest pair in this batch, and the knowledge base agrees. We are saying that at 25.0x the FY2029 consensus, three years of the plan are already priced at a normal multiple, and there is no cushion left for the plan to slip. A name where our fair value sits 3.2% above the price and below the crowd's, with a conviction lane this strong, is a Watch with a short list of specific triggers — not a Hold, because we want the exposure; not a Buy, because we want it 15% cheaper.

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

• September-quarter core sales below $4.9B, i.e. missing the floor of management's own guidance issued 2026-07-28. That is the growth story breaking within one quarter of being upgraded.
• Core operating margin below 20.0% in any quarter, which would say the capacity ramp is consuming the operating leverage the price requires.
• The contract-liability balance falling from $2.7 billion without corresponding revenue recognition, which would mean the NVIDIA-linked prepayment is being returned or renegotiated rather than converted.
• Any material reduction in a named hyperscale agreement, or a disclosed change to the NVIDIA partnership terms. At a tenfold capacity commitment, one counterparty's change of mind is a company event.
• Capital expenditure guided above roughly 11% of revenue without a corresponding revenue commitment, which would push the free-cash-flow inflection beyond the estimate window entirely.
• China revenue, 30.3% of FY2025 core sales, deteriorating materially on tariff or export policy.

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

• A price in the $135-140 band with no deterioration in the fundamentals — roughly 31-33x FY2027E, on or just above the rising 200-day moving average of $135.92 — putting the base case at +18% to +22%. This is the single most likely path to a Buy and it is only 12-16% below today's close.
• The 2026-10-27 print above $5.0B of core sales with core operating margin above 22%, which validates the margin leg of the price at the same moment it validates the volume leg.
• A third named hyperscale partnership on Amazon/NVIDIA terms, which would convert the customer-concentration risk into evidence of a structural position and would be worth roughly a tier on its own.
• Glass Innovations returning to mid-single-digit growth, which would fix the blend problem — nearly half the company currently grows at 1-2% and dilutes everything the optical business does.
• A reclaim of the 50-day moving average at $181.16, converting today's 9.05% session from a bounce into a trend change — a technical flip only, and it would need one of the fundamental legs with it.

Where GLW fits in the Synthos Framework Portfolio. The AI physical-infrastructure sleeve, at 0% today with a 2-3% target on a fill in the $135-140 band or a confirming October print. On the batch and cross-sleeve overlap question: Corning is an optical-interconnect expression of the same AI capital-expenditure cycle that drives the semiconductor-equipment sleeve, and holding both without acknowledging the shared driver is concentration dressed as diversification. Within the optical complex specifically, the knowledge base names Coherent, Lumentum and Marvell alongside Corning in three separate claims — Corning is the incumbent with the manufacturing base, and also the most expensive way to own the theme relative to its blended growth rate, because half the company is not in the theme at all. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $159.905.

Single biggest risk: the optical order book is a capital-expenditure cycle, not an annuity. Corning is expanding US optical-connectivity manufacturing capacity tenfold and US fibre production by more than 50% for a named counterparty, funded partly by that counterparty. If hyperscale interconnect spending normalises — not collapses, merely normalises — Corning owns the plant, the depreciation and the fixed cost, and the market owns a 37.2x forward multiple with a 1-2%-growth business underneath nearly half of it. The FY2023 experience is the precedent that matters: revenue fell 11.3% and operating margin fell to 7.1% in a single year the last time an end market turned. There is no valuation cushion to absorb a repeat. That is the $110 bear case, and it requires only that the outer years of consensus — +21.0% revenue growth in FY2028, accelerating — turn out to be an extrapolation rather than a forecast.

Provenance & disclosures

• Traceability: 18 tagged knowledge-base claims name Corning out of 51,928 distilled claims (raw hits 27, entity matches 18, text-only 9, used as name-level conviction 16, used as sector/basket colour 9, discarded 2; breadth 4, net conviction positive-high). Every entity-matched claim is bullish; there is no bearish claim on Corning in the knowledge base. Sources: jordi_visser (20 raw, 13 entity; skill 2.0 — the highest weight in the file; conviction 58-85), compound_and_friends (3, including two 2026-02-27 claims at conviction 75 and Josh Brown's 2026-05-03 claim at 62), Josh Brown as a named channel (1, 2026-07-29, conviction 78 — the most recent claim in the lane, dated the day after the second-quarter print), and jensen_huang (1, 2026-05-21, conviction 62). Concentration sensitivity test RUN, as required when one channel supplies a third or more of a lane: jordi_visser supplies 74% of raw hits and 72% of entity matches; stripping it leaves five claims from three sources and ALL FIVE REMAIN BULLISH. The lane survives. Attribution qualifications: six of the twenty jordi_visser claims carry neither speaker nor role and are attributed only to the channel (2025-12-07, 2025-12-14, 2026-02-01, 2026-02-05, 2026-03-29, 2026-05-03), which under the 4-lane policy is the weakest sourcing we accept and includes the two highest-conviction claims (85 and 82). The jensen_huang claim is COUNTERPARTY MANAGEMENT VOICE — the chief executive of NVIDIA describing his own partnership — and is quoted separately at half weight, excluded from the conviction pool; it is retained as corroboration only because Corning's own earnings release independently confirms the tenfold capacity claim. Discards: 2, both arthur_hayes 2023-09-05, matched by a text sweep on "Springboard" against the verb "springboards Bitcoin toward $1 million." One unverified factual assertion recurs in four claims — a "$6B Meta–Corning deal" — which no document in this archive corroborates; it is flagged rather than repeated as fact. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — balance sheet, cash flow and segment data through 2026-06-30, filing-verified from the 10-Q filed 2026-07-29; annual figures for FY2025, FY2024 and FY2023 from the 10-K filed 2026-02-12; second-quarter results and third-quarter guidance from the 8-K filed 2026-07-28, which carries the full earnings release · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873599 = 2026-08-04T19:59:59Z ($159.905, +9.05%; 50-DMA $181.16; 200-DMA $135.92; RSI 43.3; MACD −12.90) · knowledge-base claims 2026-08-04. Corning's fiscal year is the calendar year. All figures come from the Synthos vendor data file for GLW or from the SEC filings in the GLW archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-12 (fiscal 2025); 10-Q filed 2026-05-01 (March 2026 quarter); 10-Q filed 2026-07-29 (June 2026 quarter); 8-K filed 2026-05-06; 8-K filed 2026-05-15 (a J.P. Morgan conference appearance notice, Regulation FD, no financial content); 8-K filed 2026-07-28, which carries the FULL second-quarter earnings release including third-quarter guidance, the Springboard targets and the Amazon and NVIDIA partnership announcements — an unusually complete extraction and the source of several figures unavailable elsewhere. All carry preserved [TABLE] statement data.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): total debt, where the vendor's annual block reports $10,223M against the 10-Q's stated $8,424M / $8,434M — a 21.2% overstatement from double-counting finance leases, with net debt correspondingly rebuilt from $8,697M to $6,908M on the company's own definition; trailing EBITDA, recomputed from the vendor's internally contradictory $3,509M to $4,060M (operating income $2,624M plus D&A of $1,436M, which the vendor's own capexToDepreciationTTM independently implies), moving EV/EBITDA from 41.21x to ~35.6x and net-debt/EBITDA from 1.959x to 1.46x; est.ebitdaAvg/ebitAvg, rejected in full for a fixed-ratio signature (exactly 23.54% and 13.43% of revenue in every forward year) against an actual core operating margin of 20.9%, with netIncomeAvg exceeding ebitAvg in two years; seg_prod, which uses three different segment vocabularies across three years, reflects none of the four-segment structure effective Q1 2026, and is missing FY2021-FY2023 entirely; seg_geo, which omits the 10-K's $245M "All Other" line and therefore foots to no published total; the entire NVIDIA warrant transaction (15M shares at $180, 3M pre-funded for $500M cash, $296M charged as consideration payable to a customer, $1.0B customer deposit), absent from every vendor field; trailing free cash flow, inflated by that $1.0B deposit — adjusted from roughly $2.40B to $1.4B, moving price-to-free-cash-flow from 57.5x to roughly 98x; earn_cal, which reports CORE net sales for 2026 quarters against GAAP in inc_q and reconciles to neither basis for 2025; and quote.yearHigh/yearLow ($271.78/$63.07) rejected for tech.hi52/lo52 ($255.69/$63.17). Where vendor and filing AGREED — recorded because these defect classes usually bite: the current enterprise value of $144.615B, which reconciles to the June 2026 balance sheet with finance leases counted once; the share count (10-Q common stock of $927M at $0.50 par = 1,854M issued, less 994M treasury = 860M, against 861.4M implied by market cap ÷ price); bookValuePerShareTTM of $15.264 correctly distinct from tangibleBookValuePerShareTTM of $11.676; the FY2025 seg_prod lines footing exactly to GAAP revenue of $15,629M; and the FY2025, FY2024 and FY2023 income-statement lines.
• Basis note — this is the most important reading instruction in the dive. Corning reports GAAP net sales and EPS and, separately, core net sales and EPS, which apply the company's own hedge rates to hedged currency exposure and exclude translated-earnings-contract gains and losses, restructuring and impairment, discrete tax items and pension mark-to-market. The gap is large: June-quarter GAAP sales $4,505M against core $4,738M (5.2%), and GAAP EPS $0.64 against core $0.78 (21.9%). Analyst estimates in this file are on the CORE basis — verified because earn_cal's June "actual" of $4,738M is exactly the 10-Q's core net sales. All forward multiples in this dive use core EPS; the trailing multiple is shown on both bases (72.4x GAAP, 55.9x core) and labelled every time. A reader mixing a GAAP trailing figure with a core forward estimate will get a meaningless number.
• Estimate coverage: 8-9 analysts on FY2026 revenue and EPS, 9-10 on FY2027 — the anchor for all three fair values; 6 on FY2028 EPS, with a range of $3.541 to $7.831 (a 121% spread), and 4 on FY2029 — both are shown for context and neither carries a conclusion.
• Peer note: the vendor peer set — Autodesk, Celestica, Fortinet, Infosys, Keysight, Motorola Solutions, Strategy Inc, Roblox and others — is a software-and-services list containing a bitcoin treasury vehicle, offered as comparables for a capital-intensive materials-science manufacturer. Rejected in full; no peer-multiple comparison is drawn. The knowledge base names Coherent, Lumentum and Marvell as thematic co-beneficiaries in the optical complex, and those are the relevant comparison set, though no multiples for them are available in this file.
• Fair-value caveat: the $110 / $165 / $240 anchors are multiples of the FY2027 consensus core EPS distribution — 28.3x the low of $3.888, 38.3x the mean of $4.303, and 47.5x the high of $5.052. Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 34x gives $146, which is BELOW spot; 42x gives $181. The whole answer lives inside a 34-42x band and the market at 37.2x is already inside it — which is the clearest statement available that this name is fairly valued rather than mispriced.
• Management-voice note: the Springboard targets ($20B annualised by end-2026, $30B by end-2028, $40B by end-2030, a 19% sales CAGR from Q4 2026 to Q4 2030 with "earnings growing faster than sales"), the third-quarter guidance ($4.9-5.0B core sales, $0.85-0.89 core EPS), the statement that Solar profitability will improve from the third quarter, and the "more than $3 billion" Solar revenue ambition are all management's own claims and are half-weighted throughout this dive per house rule. Where they are quoted, they are labelled. The observable facts underneath them — the reported segment growth rates, margins and the disclosed partnership terms — are not discounted.
• Timing: second-quarter 2026 results were released 2026-07-28, seven days before this dive, so the full half-year on the new four-segment structure is reflected. The next print is 2026-10-27, 84 days away — no near-term company-specific catalyst, which is an explicit input to the Watch verdict. 2026-08-04 was a broad technology up-day across this batch; GLW's 9.05% move carried no company-specific filing and is treated as sector beta.
• 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.