The Southern SO

Reference table

1. What the business is, and the one number that has changed

Southern Company is a Delaware holding company headquartered in Atlanta, NYSE-listed, with approximately 29,800 employees, led by Chairman, President and Chief Executive Christopher C. Womack and Chief Financial Officer David P. Poroch. It describes itself in the July earnings release as "a leading energy provider serving 9 million customers across the Southeast and beyond... electric operating companies in three states, natural gas distribution companies in four states, a competitive generation company, a leading distributed energy solutions provider with national capabilities, a fiber optics network and telecommunications services."

The demand picture, from the June 10-Q, and it is the whole story:

One customer class is growing at double digits and the other two are not growing at all. That concentration is the growth story and the risk in a single line.

The Large Load Contracts — quoted in full because the terms are the thesis

From the 10-Q, verbatim:

> "Since 2023, the traditional electric operating companies have contracts with new data centers and other large load customers covering approximately 11 gigawatts of electric load, with each contract individually representing a maximum electric load greater than 100 MWs, that have been reviewed by the applicable state PSC. Additional contracts totaling approximately five gigawatts of electric load have been agreed with customers and are subject to review by the applicable state PSC. The approximately 16 gigawatts of new contracts (Large Load Contracts) fully ramp up over several years after commencement of service. Some of the Large Load Contracts are currently in effect and have commenced service. Service under the remaining Large Load Contracts is expected to begin through 2028."

> "All of the Large Load Contracts contain minimum bill provisions. Large Load Contracts totaling approximately 13 gigawatts of electric load (including substantially all signed after 2024) also contain minimum duration, termination payment requirements, and financial security requiring customers to post collateral or provide an acceptable guarantee based on the customer's credit ratings. Certain Large Load Contracts requiring construction of specifically dedicated assets contain provisions for contribution by the customer to construction costs."

Read that carefully, because it answers the standard objection to every data-centre-utility thesis in the market. The objection is that announced load is not contracted load, and that contracted load can be cancelled when an AI capital-expenditure cycle turns. Southern's answer is four specific protections: a minimum bill on every contract; minimum duration on 13 of 16 gigawatts; termination payments; and posted collateral or a credit-rated guarantee. "Including substantially all signed after 2024" is the most informative clause — it says the terms got tighter as the counterparties got more numerous, which is what a supplier with pricing power does.

Two honest caveats. The 5 gigawatts still "subject to review by the applicable state PSC" is not yet approved, and state commission approval is what converts a contract into recoverable rates. And the filing does not name the counterparties, disclose the concentration among them, or quantify the revenue the 16 gigawatts represents. The economic size is therefore not derivable from this archive and no attempt is made to derive it.

The regulatory model, stated plainly

From the 10-Q: "The traditional electric operating companies are engaged in continuous construction programs to accommodate existing and estimated future loads on their respective systems. Major generation construction projects are subject to state PSC approval in order to be included in retail rates, through which the traditional electric operating companies recover their approved investment and a return on investment."

That sentence is the entire earnings mechanism. Southern invests capital, a commission approves it into the rate base, and Southern earns an allowed return on it. Earnings grow with rate base, not with revenue — which is why adjusted EPS rose 22.8% in a quarter revenue rose 0.1%. It is also why free cash flow is deeply negative and why the share count rises every year: the model requires permanent external financing.

2. The event dated today — $2.375 billion of convertible notes

This is the highest-yield finding in this file and it postdates every statement in the payload by seven months.

Two 8-Ks, one day apart:

2026-08-03: Southern "announced offerings of $650 million in aggregate principal amount of its convertible senior notes due December 15, 2027... and $1.5 billion in aggregate principal amount of its convertible senior notes due September 15, 2029... in private placements to persons reasonably believed to be qualified institutional buyers pursuant to Rule 144A."

2026-08-04 — the date of this dive: "Southern Company announces upsize and pricing of $725 million in aggregate principal amount of Series 2026A 2.125% Convertible Senior Notes due December 15, 2027 and $1.65 billion in aggregate principal amount of Series 2026B 3.50% Convertible Senior Notes due September 15, 2029."

A cross-check that validates the whole file. A 12.5% premium to a $104.56 conversion price implies a reference price of $92.94; a 27.5% premium to $118.50 implies $92.94. The vendor's quote.previousClose is $92.94, to the cent, twice over. That is the cleanest independent confirmation of a payload price this programme has recorded, and it is worth stating because it means the quote block is struck correctly even where other fields are not.

Use of proceeds, from the pricing release: "approximately $403 million of the net proceeds... to repurchase approximately $369 million aggregate principal amount of its Series 2024A 4.50% Convertible Senior Notes due June 15, 2027... Southern Company intends to use the remaining net proceeds to repay all or a portion of its outstanding short-term debt and for other general corporate purposes, which may include investment in its subsidiaries."

Four things follow and each is a judgement, stated as one.

This is a refinancing at less than half the coupon, achieved by selling equity optionality. The 2024A notes pay 4.50%; the 2026A pay 2.125%. Southern is buying a 237-basis-point saving with a call option struck 12.5% above the market. On $725 million, that is roughly $17 million a year of interest saved against 6.93 million shares of contingent dilution. Whether that is a good trade depends entirely on where the stock goes, and it is an explicit management bet that it does not go above $104.56 quickly.

Southern is now a serial convertible issuer, which is unusual for a regulated utility. The releases disclose three separate series: Series 2024A 4.50% due June 2027, Series 2025A 3.25% due June 2028, and now Series 2026A/2026B. A utility reaching the convertible market three years running is telling you something about the cost of straight equity and the size of the funding requirement.

The contingent dilution is real but capped and conditional. "Upon conversion, Southern Company will pay cash up to the aggregate principal amount... and pay or deliver, as the case may be, cash, shares of Southern Company's common stock, or a combination... at Southern Company's election, in respect of the remainder." So the principal is settled in cash and only the in-the-money excess can become shares, at Southern's option. Maximum contingent dilution on the base deal is 20.86 million shares, 1.81% of the current count; with both greenshoes, 23.99 million, 2.08%. Neither is catastrophic; both are additive to a share count already rising 3.3% a year.

And there is a specific, disclosed, dated near-term price mechanism. The release states Southern "expects that many holders of the Series 2024A Convertible Notes employ a convertible arbitrage strategy with respect to the Series 2024A Convertible Notes and have a short position with respect to Southern Company's common stock that they would close, through purchases..." — that is a stated expectation of buying pressure from the notes being repurchased — while the new notes will be bought by arbitrageurs who typically establish fresh short positions. The offering closes 2026-08-06. The stock traded down to $91.25 intraday on 10.4 million shares on the day of pricing before closing at $93.25.

None of this is in the payload. bal_a ends 2025-12-31; cf_a ends FY2025; there is no 2026 row anywhere in the balance-sheet or cash-flow blocks. Anyone valuing Southern on this file is valuing a company with $2.4 billion less debt and no convertible optionality outstanding.

3. The June quarter and the first half

From the 8-K earnings release filed 2026-07-30, five days before this dive:

Four observations.

First, the revenue line is nearly static and the earnings line is not. +0.1% revenue against +22.8% adjusted EPS in a single quarter. That gap is the regulated model working exactly as designed: earnings come from the allowed return on a growing rate base, and from mix and rate structure, not from selling more units at the same price. Management's stated drivers are "investment in state-regulated utilities, customer usage and growth, higher earnings from equity method investments and lower income taxes, partially offset by higher interest expense" — three of the four are not volume.

Second, the adjustments are transparent and mostly non-recurring, with one exception that is scheduled. The accelerated depreciation from repowering — $143M pre-tax in the quarter, $296M in the half — relates to Southern Power's wind repowering projects, and the 10-Q sizes what remains: "At June 30, 2026, the remaining pre-tax accelerated depreciation is projected to total approximately $200 million in 2026 and $100 million in 2027." So this adjustment has a disclosed end date and a disclosed remaining size, which is better disclosure than most companies in this programme provide. The $129M loss on extinguishment of debt in the June quarter is a one-off and is consistent with the refinancing activity described in Section 2.

Third, the share count is a material drag that the EPS growth partially conceals. Average shares rose 3.3% year on year. Net income excluding items rose 27.1% while adjusted EPS rose 22.8% — the 4.3-point gap is dilution. Over a decade, at 3% a year, that is the difference between a good outcome and an ordinary one, and it is the permanent cost of the funding model.

Fourth, the beat record is consistent but the forward bar has been set oddly low:

And here is the arithmetic that Section 6 turns into a variant perception. First-half adjusted EPS was $2.46. The FY2026 consensus is $4.582. That requires second-half adjusted EPS of $2.12 against an actual $2.15 in the second half of 2025 — a 1.4% DECLINE — immediately after a first half that grew 14.4%. The Q3 consensus of $1.66 against $1.60 a year earlier says the same thing at quarterly resolution: +3.75% after +9.1% and +11.9%. Either the estimates have not caught up, or the second half carries a headwind the archive does not disclose.

4. What the growth costs — and the vendor field that hides it

This is the section where the payload is most misleading and the filing most necessary.

Now the vendor's version. km_ttm reports capexPerShareTTM of $6.750 and capexToRevenueTTM of 25.43%, which on 1,150.4M shares implies trailing capital expenditure of approximately $7,765M. The filings say $13,920M. That is a 44% understatement. It flows straight into freeCashFlowPerShareTTM of $2.263 — implying positive trailing free cash flow of roughly $2,603M — and into freeCashFlowYieldTTM of 2.40% and priceToFreeCashFlowRatioTTM of 41.7x. Southern's actual trailing free cash flow is approximately MINUS $3,269 million. The vendor reports a positive free-cash-flow yield on a company with a $3.3 billion cash deficit, and the sign is the part that matters.

Note that the payload contradicts itself here too: cf_a for FY2025 correctly reports capitalExpenditure of −$12,737M and freeCashFlow of −$2,935M, both matching the filing. So the annual cash-flow block is right and the trailing key-metrics block is wrong, in the same file, by $6.2 billion. capexToDepreciationTTM of 1.209 is the same error viewed from another angle — the true ratio is approximately 2.17x.

Why this matters more here than it would elsewhere. For most companies, a mis-stated free-cash-flow yield is a screening annoyance. For a regulated utility in a construction supercycle it is the entire financing question, because the gap between operating cash flow and capital expenditure is exactly what must be raised in the debt and equity markets each year — and it is why $2.375 billion of convertible notes were priced this afternoon.

The financing, from the filings:

H1 2026 financing flows: long-term debt issued $4,790M, repaid $3,085M, common stock issued under the dividend reinvestment plan $109M. Equity rose $3,475M against $2,531M of net income, so roughly $944M came from issuance and other movements. Add the $2,375M of convertible notes closing 2026-08-06 and the trajectory is unambiguous.

Net debt to EBITDA is approximately 5.2x on the vendor's calculation, which for a regulated utility is elevated but not unusual. The relevant question is not the level but the direction, and the direction is up while free cash flow is negative and the construction programme is accelerating — property additions of $6,639M in a half-year against $12,737M for all of 2025.

The dividend is covered by earnings and not by free cash flow, and this should be said plainly. dividendPerShareTTM is $2.98, a 3.20% yield, at a payout of roughly 65% of FY2026 consensus adjusted EPS. On free cash flow the coverage is negative, and the vendor's own dividendPaidAndCapexCoverageRatioTTM of 0.951 — computed on the understated capex — is comfortably below 1.0 even before correction. This is normal for a growing regulated utility, it is how the model has always worked, and it is the reason the share count rises.

5. Balance sheet and the lease overstatement

The vendor's totalDebt includes $1,484M of operating lease obligations, overstating debt by 2.0%. This is the T/TJX lease-inclusion defect class, and at 2.0% it is the mildest instance this programme has recorded — it does not change any conclusion, and it is recorded because the check was run and this is what it found. All debt figures in this dive use the filing's pure-debt basis and say so.

Enterprise value reconciles acceptably. enterpriseValueTTM of $181,374M against market capitalisation of $107,271M plus net debt of $72,436M gives an unexplained residual of $1,667M, 0.9% of enterprise value — well inside the 15% withholding threshold, and roughly consistent with partial inclusion of the $2,851M minority interest. evToEBITDATTM of 12.79x is therefore usable, with the caveat that it excludes the $2,375M of convertible notes priced this afternoon (which would take it to approximately 13.0x).

Two other balance-sheet facts worth recording. Property, plant and equipment net was $116,441M at 2025-12-31 against total assets of $155,720M — 74.8%. This is what a utility is: a pile of physical assets with an allowed return attached. And goodwill of $5,161M has been unchanged for three consecutive years, which is the correct behaviour for goodwill that is not being tested to impairment.

6. Valuation — priced in or room?

At $93.25 (market cap $107.27B, ~1,150.4M implied shares):

Estimate coverage is good where it matters and collapses immediately after. 14 analysts on FY2027 EPS — the anchor — 13 on FY2026, and only 4 on FY2028, 2 on FY2029 and 2 on FY2030. The FY2029 and FY2030 rows are excluded from every conclusion. The FY2027 EPS range is $4.843 to $4.964, a 2.5% spread — very tight, and typical of a rate-regulated business where the earnings are largely a formula.

est.ebitdaAvg and est.ebitAvg are REJECTED for the fixed-ratio fabrication signature. ebitdaAvg is exactly 43.23% of revenueAvg and ebitAvg exactly 25.05% in every forward year from FY2026 through FY2030. They are also wrong in level: FY2025's ebitdaAvg of $12,322.6M against a reported $14,334M understates by 14.0%. All forward valuation uses epsAvg.

A note on the trailing EPS basis. inc_a reports FY2025 EPS of $3.94 and earn_cal reports quarterly adjusted actuals summing to $4.30, against a FY2025 consensus of $4.289. The difference is the adjusting items — principally accelerated repowering depreciation — and every forward multiple in this dive is struck on the adjusted basis, matching the estimates. The vendor's priceToEarningsRatioTTM of 22.36x is on a partly-GAAP trailing basis and is not comparable to the forward multiples.

Peer context is unusable and worth naming. The vendor peer set — AEP, CEG, D, DUK, ETR, EXC, NEE, NGG, XEL — is genuinely the right comparison universe, the best peer set in this batch. But the payload supplies only their market capitalisations and no multiples, earnings or growth rates. No peer-multiple comparison is drawn, because inventing the inputs is not permitted. (Recorded for completeness: NEE $181.9B, DUK $96.9B, CEG $96.0B, NGG $80.9B, AEP $69.9B, D $60.9B, ETR $50.8B, XEL $48.6B, EXC $47.2B.)

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

At $93.25 — 20.4x FY2026 consensus, 18.9x FY2027, 3.20% yield — the price embeds:

• Adjusted EPS reaches $4.582 in 2026 and $4.924 in 2027. (Consensus; 13 and 14 analysts.) The 2026 figure is demonstrably conservative: first-half adjusted EPS was $2.46, so it requires second-half EPS of $2.12 against $2.15 a year earlier — a 1.4% decline after a first half up 14.4%. Even flat year-on-year second-half earnings produce $4.61.
• The 16 gigawatts of Large Load Contracts arrive on schedule and are recovered in rates. (Filing-sourced quantity; our derivation of the consequence.) Service under the remaining contracts is "expected to begin through 2028", and roughly 5 of the 16 gigawatts are still "subject to review by the applicable state PSC." This is the most fragile assumption in the price, not because the contracts look weak — the minimum-bill, minimum-duration, termination and collateral terms are stronger than any comparable disclosure in this batch — but because the entire earnings acceleration from FY2027 to FY2028 (+9.2%) depends on regulatory approval and construction execution, and neither is under management's control.
• The construction programme is financed without a step-change in the cost of capital. (Our number, from the filings.) Property additions of $13.9 billion trailing against $10.7 billion of operating cash flow means roughly $3.3 billion a year must be raised externally, indefinitely. The price assumes that this continues to be available at rates like 2.125% and 3.50% on convertibles and at an average share count growing 3.3%. It has been. That is not a guarantee.
• The market keeps paying roughly 19-20x forward earnings for a 7-8% grower with a 3.2% yield. (Our number.) At 17x FY2027E the stock is $84 — within a dollar of the 52-week low; at 22x it is $108. The band is narrow because utility multiples are anchored by the dividend yield and the ten-year yield, neither of which this archive contains.
• Regulators do not respond to affordability pressure. (Our number; the filings name the risk.) Southern's own forward-looking-statement list places "customer affordability matters" among its named uncertainties, and the chief executive's quoted remarks emphasise "keeping reliability and rate stability at the center of our work" and "when growth is done right, everyone benefits." When a company pre-empts a criticism in its earnings release, the criticism is live.

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

Expected return over the next twelve months decomposes as: adjusted EPS growth (+7.5%, from FY2026E $4.582 to FY2027E $4.924) + multiple drift (COMPRESSION of roughly 2-3%, from today's 20.4x on the forward year toward about 19.9x) + dividend yield (+3.20%) ≈ +8%.

Our base assumes a very small amount of multiple compression, and the reason is dilution rather than any view about the business. A company issuing 3.3% more shares a year while running negative free cash flow does not usually re-rate upward, and the addition of $2.375 billion of convertible notes with contingent dilution above $104.56 puts a soft ceiling under any multiple expansion in the near term — the conversion price is itself a statement about where sophisticated buyers of that optionality think the stock is capped.

Note what this makes the return: roughly 60% earnings growth and 40% dividend, with essentially nothing from the multiple. That is a defensible, low-volatility, bond-like return — and at 1.42:1 asymmetry it is not enough to earn a Buy.

If the multiple held at 20.4x FY2027E the price would be $100 (+7.4%). If it compressed to 17x, $84 (−10.0%) — which is $84.08, the 52-week low, to the cent.

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

• On the level we have essentially NO variant perception and our base is BELOW the street's. $98 against $100.08, and $3.50 below the median target of $101.50, with 10 of 34 analysts at buy and three at sell. We are marginally more cautious than a consensus that is already at Hold.
• Where we differ from consensus is the FY2026 number, and the difference is arithmetic. The FY2026 consensus of $4.582 implies second-half adjusted EPS of $2.12 against $2.15 a year earlier, a decline, after a first half that grew 14.4% and two consecutive beats of 9.1% and 11.9%. Watchable number: the 2026-10-29 print against $1.66. A print at $1.75 or better would put FY2026 near $4.70 and pull FY2027 with it. But note what this is worth: on a 19x multiple, a 2.5% estimate revision is $2.35 of price. It is not a thesis.
• We think the market is underweighting the financing side and overweighting the load headline. The 16 gigawatts is genuinely excellent and genuinely contracted. The $3.3 billion annual external funding requirement, the 3.3% annual share issuance and the third consecutive year of convertible issuance are the price of it, and none of them appears in an EPS estimate. Watchable number: average shares outstanding in the October print, at 1,137 million in June against 1,101 million a year earlier.
• We think the knowledge base's utility-model criticism deserves more weight than its zero name-level status implies. no_priors (2025-10-15, bearish 78): "Regulated utility monopolies earn a set rate of return on capex, creating a perverse incentive to build rather than innovate, pushing transmission costs up." That is a description of Southern's earnings mechanism written as a criticism, and it identifies the exact channel — retail affordability — through which a state commission would cap the returns this thesis needs. It is a structural claim about the industry, it names no company, it carries no conviction weight, and it is the most useful sentence in the discarded lane.
• Positive surprise that would force a re-rate: state commission approval of the remaining ~5 gigawatts of Large Load Contracts on the signed terms, plus a quantification of what the 16 gigawatts contributes to revenue or rate base — which the filings have never provided and which would convert a qualitative growth story into a modelable one; or a September quarter above $1.75 with commercial kilowatt-hour growth still in double digits.
• Negative surprise that would break the thesis: a state commission denying recovery on a material portion of the construction programme, or reducing an allowed return on equity; commercial kilowatt-hour growth falling below 5%; a Large Load Contract counterparty invoking termination and testing the collateral provisions; property additions rising further while operating cash flow does not, forcing an equity issuance at a discount rather than convertibles at a premium; or a major storm or wildfire event, which the filings name first among catastrophic risks and which no multiple in this section prices.

Synthos fair values

All three anchors are multiples of the FY2027 consensus adjusted EPS distribution (mean $4.924, low $4.843, high $4.964, 14 analysts), each cross-checked against FY2028E and the dividend.

• Bear ~$78 — 16.1x the FY2027 consensus LOW of $4.843, and 14.5x FY2028E. Cross-check: 7.2% below the 52-week low of $84.08; 1.3% below the lowest street target of $79; implied dividend yield 3.82%. The scenario: a regulatory setback on cost recovery or allowed return, commercial load growth decelerating, the financing requirement forcing dilutive equity, and the multiple compressing to where a slow-growth utility with 5.2x leverage would trade. −16.4%.
• Base ~$98 — 19.9x the FY2027 consensus MEAN of $4.924, and 18.2x FY2028E. Cross-check: 2.1% below the street's $100.08; 1.7% below the 52-week high of $99.72; implied dividend yield 3.04%. Sensitivity, stated openly: 18x gives $89 — below spot — and 22x gives $108. The whole answer lives in an 18-22x band on an estimate whose own dispersion is 2.5%. The scenario: the Large Load Contracts ramp as scheduled, regulators approve the remaining 5 gigawatts, earnings compound at 7-9%, the dividend grows, and the multiple stays where it is while the share count keeps rising. +5.1%.
• Bull ~$115 — 23.2x the FY2027 consensus HIGH of $4.964, and 21.4x FY2028E. Cross-check: 2.7% above the highest street target of $112; 15.3% above the 52-week high; conversion price of the Series 2026B notes is $118.50, so the bull case sits just below the level at which the new convertibles are in the money. The scenario: US electricity demand growth accelerates toward the pace the knowledge base's independent lane describes, additional gigawatts are contracted at improving terms, and the market re-rates regulated utilities in the Southeast as growth assets. +23.3%.

Base is 5.1% above spot; asymmetry roughly 1.42:1 to the upside (16.4% down, 23.3% up), before a dividend yield of 3.20%. A base under 6%, a payoff ratio under 1.5:1, a base BELOW the street's consensus, an empty name-level conviction lane and negative free cash flow is a Hold.

7. Knowledge base — zero on the name, twenty-seven on the theme

Raw hits: 27. Case-sensitive entity hits on SO, Southern Company, Southern Co, Georgia Power or Vogtle: ZERO. Name-level claims: 0. Discarded: 27.

The case-sensitive sweep was run FIRST, before any free-text search, because "SO" is among the most collision-prone tickers in this store and the brief flagged it as such. It returned nothing — not a single claim in 51,928 carries Southern Company, Georgia Power or Vogtle as an entity. No entity lane exists, and none is manufactured. That is the finding.

The free-text sweep on "Southern Company", "Georgia Power", "Vogtle", "regulated utilit", "data center power" and "electricity demand" returned 27 hits across 15 channels — and this is the unusual part: the lane is not noise. It is a multi-sourced, independent, strongly bullish thesis on precisely the variable that determines Southern's growth, and it names no company.

The demand side, verbatim:

> 2025-01-23 · bullish · conviction 88 · horizon: thesis · channel: macrovoices · skill 0.9

> "Electricity demand will grow dramatically driven by AI, data centers, EVs, crypto, power-to-x and rising global prosperity, reversing decades of slow growth."

> 2025-10-15 · bullish · conviction 78 · horizon: fact · channel: no_priors · skill 1.0

> "US electricity demand growth may accelerate from ~2% to ~10% CAGR, driven largely by AI and data centers; projections may even be understated."

> 2025-05-06 · bullish · conviction 90 · horizon: principle · channel: all_in · skill 1.0

> "AI has flipped a flat electricity demand curve into explosive growth; the five biggest tech firms are now power's largest customers ($300B capex)."

> 2025-06-04 · bullish · conviction 75 · horizon: thesis · channel: luke_gromen · skill 1.0

> "AI electricity demand is exploding while the US grid and supply chain can't keep pace — a looming bottleneck requiring rapid nuclear buildout."

> 2025-11-08 · neutral · conviction 70 · horizon: thesis · channel: jordi_visser · skill 2.0

> "Hyperscalers have run on grid excess capacity; next year hits full capacity, making the power buildout critical — AI/LLM training is where electricity demand grows."

> 2026-02-03 · bullish · conviction 62 · horizon: fact · speaker: Dylan Patel · role: independent · channel: dylan_patel · skill 1.0

> "Power was the biggest constraint in 2024-25 but people have woken up; ~15-18GW added this year, ~30GW added next year — almost a doubling — and 'the power is there for it.'"

A redaction, disclosed under house rules. One further claim (all_in, 2025-08-09, bullish 70) reports a named artificial-intelligence laboratory's request for 50 gigawatts of US power over three years — "roughly a full year's added capacity" — as a signal of "structurally new, price-insensitive electricity demand." The laboratory's name is redacted here in line with the house prohibition on naming AI models or labs; the quantity, the framing and the source are reported in full. For scale: Southern's entire contracted Large Load book is 16 gigawatts.

The counterweight, and it comes from the same channel as the most bullish demand claim:

> 2025-10-15 · BEARISH · conviction 78 · horizon: principle · channel: no_priors · skill 1.0

> "Regulated utility monopolies earn a set rate of return on capex, creating a perverse incentive to build rather than innovate, pushing transmission costs up."

> 2025-01-23 · neutral · conviction 70 · horizon: principle · entities: PG&E · channel: no_priors · skill 1.0

> "Regulated utilities love CapEx (marked up ~15% to ratepayers); underinvestment in undergrounding/upgrades comes from regulators capping rates, not utilities."

These two are the most useful claims in the lane and they are directly about Southern's business model even though they never name it. They describe the earnings mechanism set out in Section 1 — invest capital, earn an allowed return — and characterise it as a distortion that pushes costs onto ratepayers. That is the precise channel through which the affordability backlash the filings name would arrive. They are structural claims about an industry, they carry no name-level conviction, and they are reported because the honest reading of a 27-claim bullish demand lane is incomplete without the two claims inside it that describe how the returns get capped.

Attribution and hygiene.

• Channel distribution: jordi_visser 7, all_in 3, no_priors 3, lex_fridman 2, money_of_mine 2, and ten channels with one each (andreas_steno, forward_guidance, luke_gromen, macrovoices, a16z, odd_lots, dylan_patel, elon_musk_mgmt, interviewers, quarantine_misattributed). jordi_visser supplies 26% of the lane — under the one-third threshold, so no concentration sensitivity test is triggered. (Recorded: had it exceeded a third, the test would have been run, because jordi_visser carries a skill weight of 2.0 and would dominate any weighted aggregate.)
• One hit comes from the quarantine_misattributed channel and is EXCLUDED on that basis alone, in line with the attribution policy. Its content — that AI becomes a state-regulated utility and hyperscaler valuations compress — is irrelevant to Southern in any case.
• One hit comes from the elon_musk_mgmt channel, which is a management-voice channel for a different company entirely and is excluded from the pool.
• Four claims carry named speakers — Drew Baglino (a16z), Daniel Yergin (odd_lots), Dylan Patel (dylan_patel), Matthew Fist (money_of_mine) — none of whom is Southern Company management. No claim carries a management speaker_role for this registrant. The remaining 23 carry speaker: null and are channel-attributed.

Conclusion. Breadth 0, claim count 0, net conviction none. The Synthos knowledge base has no view on Southern Company. What it has is a strong, independent, fifteen-channel conviction that the thing Southern sells is about to be in structurally short supply — and, inside that same lane, the sharpest available statement of why a regulated monopoly may not be allowed to keep the resulting profits. Neither enters the score. Both are in the analysis.

8. Data integrity — what we rejected and why

Seven findings. Southern's vendor file has one severe and consequential error — a trailing capital-expenditure figure 44% below the filing's, which inverts the sign of free cash flow — plus a segment block covering a sixth of revenue, an empty geography block, and a whole corporate financing event that postdates every statement.

1. km_ttm understates trailing capital expenditure by 44% and INVERTS the sign of free cash flow — REJECTED. capexPerShareTTM of $6.750 and capexToRevenueTTM of 25.43% imply trailing capital expenditure of approximately $7,765M. The filings give property additions of $12,737M for FY2025 and $6,639M for H1 2026 against $5,456M for H1 2025, so trailing capital expenditure is $13,920M. Consequential rejections: freeCashFlowPerShareTTM $2.263 (implying positive trailing free cash flow of ~$2,603M against an actual of approximately −$3,269M), freeCashFlowYieldTTM 2.40%, priceToFreeCashFlowRatioTTM 41.7x, evToFreeCashFlowTTM 70.5x, capexToDepreciationTTM 1.209 (true ≈ 2.17x), capexToOperatingCashFlowTTM 0.749 (true ≈ 1.31x), freeCashFlowOperatingCashFlowRatioTTM 0.251, freeCashFlowToEquityTTM $8,370M and dividendPaidAndCapexCoverageRatioTTM 0.951. This is the COP defect class — a single capital-expenditure line materially understated — and here it does not merely change a yield, it changes the sign. Note the internal contradiction: cf_a for FY2025 reports capital expenditure of −$12,737M and free cash flow of −$2,935M, both matching the filing exactly. The annual block is right and the trailing block is wrong, in the same file, by $6.2 billion.

2. A $2,375 million convertible-note pricing dated the day of this dive is absent from every vendor field. Priced 2026-08-03, announced in an 8-K filed 2026-08-04, closing 2026-08-06: $725M of Series 2026A 2.125% notes due December 2027 convertible at ~$104.56, and $1,650M of Series 2026B 3.50% notes due September 2029 convertible at ~$118.50, with greenshoes of $108.75M and $247.5M. bal_a ends 2025-12-31, cf_a ends FY2025, and there is no 2026 row anywhere in either block. The payload also contains no trace of the Series 2024A 4.50% notes due June 2027 or the Series 2025A 3.25% notes due June 2028, both of which the pricing release confirms are outstanding. This is the "whole transaction absent" defect class, and Southern has three convertible series the file has never heard of.

3. totalDebt includes $1,484M of operating lease obligations — a 2.0% overstatement. The vendor's FY2025 totalDebt of $74,075M equals the filing's pure debt of $72,591M (securities due within one year $6,220M + notes payable $722M + long-term debt $65,649M) plus operating lease obligations of $1,484M (current $197M + deferred $1,287M). This is the T/TJX lease-inclusion defect class and at 2.0% it is the mildest instance recorded in this programme. It changes no conclusion. All debt figures in this dive use the filing's pure-debt basis.

4. seg_prod covers 17% of revenue and omits the electric business entirely for the two most recent years — REJECTED. The FY2025 entry contains only Gas Distribution Operations $4,428M, Gas Marketing Services $582M, Gas Pipeline Investments $32M and Southern Company Gas $5,044M — which double-counts its own subtotal and, taken at the segment level, sums to $5,042M against reported revenue of $29,553M, 17.1%. The electric utilities — which are the company — appear nowhere in the FY2024 or FY2025 entries. FY2023 is missing entirely (the series runs 2025, 2024, 2022, 2021, 2020...), and the FY2022 entry does include Electric Utilities at $22,873M and Southern Power at $3,369M. A reader taking seg_prod at face value would conclude Southern Company is a gas distributor. All segment and demand analysis in this dive comes from the 10-Q.

5. seg_geo is EMPTY. The block returns []. For a company whose entire business is three electric states and four gas states, this is a total omission — though a benign one, since Southern is entirely domestic and the geography is disclosed narratively in the filings. Recorded because the check was run.

6. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature — REJECTED. ebitdaAvg is exactly 43.23% of revenueAvg and ebitAvg exactly 25.05% in every forward year FY2026 through FY2030. The historical rows carry different ratios (FY2023: 36.0% and 14.3%), confirming the forward rows are generated. They are also wrong in level: FY2025 ebitdaAvg of $12,322.6M against a reported $14,334M, a 14.0% understatement. All forward valuation uses epsAvg. The FY2029 and FY2030 rows rest on 2 EPS analysts each and are excluded from every conclusion.

7. tech.max_dd_from_peak of −6.49% is IDENTICAL to pct_from_hi — the ADP-class signature — but is NOT flagged here, because the return series is consistent with it. With hi52 of $99.72 and lo52 of $84.08, a 15.7% peak-to-trough range exists inside the year, so a −6.49% maximum drawdown is only possible if the low preceded the high. The return series supports that: ret_12m of −1.83% implies a price near $95.00 a year ago, ret_6m of +5.74% implies $88.16 six months ago, and ret_3m of −2.85% implies $95.99 three months ago — a path in which the $84.08 low falls between twelve and six months back and the $99.72 high is recent. Consistent, therefore not flagged. (Recorded: quote.yearHigh/yearLow of $100.84/$83.80 sit 1.1% and 0.3% from tech.hi52/lo52; tech is used throughout for internal consistency.)

Where vendor and filing AGREED — recorded, and one of these is the best cross-check in the batch:

• quote.previousClose of $92.94 is confirmed to the cent, twice, by the convertible pricing release. A 12.5% conversion premium on a $104.56 conversion price implies $92.94; a 27.5% premium on $118.50 implies $92.94. An independent SEC filing validating a vendor price field to four significant figures is the strongest price verification this programme has recorded.
• cf_a FY2025 capitalExpenditure of −$12,737M and freeCashFlow of −$2,935M both match the filings — which is what makes finding 1 an error in the trailing block specifically rather than a systematic one.
• bal_a FY2025 balance-sheet lines reconcile to the 10-Q's December comparatives exactly: long-term debt $65,649M, accounts payable $3,710M, total liabilities $116,853M, total equity $38,867M, total assets $155,720M.
• earn_cal reports adjusted EPS and matches the earnings release — Q2 2026 epsActual of $1.13 and revenueActual of $6,977,000,000 both match.
• The enterprise-value rebuild has an unexplained residual of $1,667M, 0.9% — well inside tolerance, and roughly consistent with partial inclusion of the $2,851M minority interest.
• The implied share count of 1,150.4M against a June-quarter average of 1,137M is a 1.2% difference in the right direction for a company issuing equity continuously through a dividend reinvestment plan and at-the-market programme. Checked, and not an error.

Vendor composite rating — accepted with one caveat that inverts it. The rating block gives B+ / 3 overall, with 5/5 on discounted cash flow and 4/5 on returns. The 5/5 discounted-cash-flow score is built on the free-cash-flow figure corrected in finding 1 — that is, on a positive $2.6 billion where the company actually consumed $3.3 billion — and should carry NEGATIVE weight, not zero. The 2/5 on debt-to-equity and price-to-earnings and 1/5 on price-to-book are unobjectionable descriptions of a leveraged, capital-intensive, regulated business.

Non-equity tripwire — checked and passed. SO is common stock, NYSE-listed. Price of $93.25 is not par-like; beta is 0.332 — low, and characteristic of a regulated utility rather than of a fixed-income instrument; the dividend is a variable declared amount at a 3.20% yield; volume was 10.41M shares (~$971M of turnover); the 52-week band of $84.08 to $99.72 is a 19% range. Note the three separate series of convertible senior notes, which are NOT this security and which are Rule 144A private placements not listed anywhere. This is common equity.

9. Technicals

• Price $93.25. −6.5% from the 52-week high of $99.72; +10.9% above the 52-week low of $84.08. Position within the annual range: 59th percentile.
• The stock is essentially on its long-run average and below its short one. 1.3% BELOW a 50-day moving average of $94.49; only 0.5% above a 200-day average of $92.77. The 50-day sits above the 200-day but the price is below both crossings — a stalled uptrend, not a downtrend.
• RSI 44.7 — neutral, slightly weak. MACD −0.22 — the only negative MACD reading in this batch.
• Relative performance is uniformly poor and that is the most interesting fact in this section. 3-month −2.9% against SPY +7.6% and QQQ +7.7% — underperforming by 10 points; 6-month +5.7% against SPY +11.1%; 12-month −1.8% against SPY +24.3% and QQQ +30.8% — a 26-point deficit against the index. In a twelve-month period in which Southern grew adjusted earnings, signed gigawatts of contracted new load and raised its dividend, the stock went nowhere. That is either a valuation that started too high, a rate-driven de-rating of the whole utility complex, or a market that does not believe the load story — and this archive cannot distinguish between them.
• Maximum drawdown from peak over the trailing year: −6.49%, verified consistent with the return series (Section 8, finding 7). Combined with a 0.332 beta, this is by a wide margin the lowest-volatility security in this batch.
• Sentiment: 10 buy, 21 hold, 3 sell across 34 analysts; consensus target $100.08 (+7.3%), median $101.50 (+8.8%), high $112 (+20.1%), low $79 (−15.3%). A clear Hold consensus, with three sell ratings on a regulated utility — unusual, and worth recording.

Today's move and what it does to the entry

SO closed 2026-08-04 at $93.25, up 0.33% or $0.31 from $92.94. It opened at $92.09 — 0.9% BELOW the previous close — traded down to $91.25, and recovered to close at $93.25 on 10.41 million shares, roughly $971 million of turnover.

That intraday shape has an explanation dated the same day. Southern priced $2.375 billion of convertible senior notes off the 3 August close of $92.94, and the 8-K announcing it was filed 2026-08-04. The company's own release states that holders of such notes typically "employ a convertible arbitrage strategy... and have a short position with respect to Southern Company's common stock." A 0.9% gap down and a 1.8% intraday drawdown on elevated volume on the day of a convertible pricing is exactly what that mechanism produces. The offering closes 2026-08-06, two days after this dive.

The honest read: this is a neutral entry into a stock that has gone nowhere for a year, with a specific, dated, disclosed source of near-term supply arriving in forty-eight hours. That is not a reason to sell and it is a reason not to hurry. The level at which the arithmetic changes is $84 — the 52-week low, 17x FY2027 consensus, and a 3.55% dividend yield — and it is named in Section 11.

10. Insiders — eight transactions, two trivial sales and six director awards

The reading, and it is close to nothing.

Six of the eight are director deferred-stock-unit and phantom-stock awards granted on the same date at a zero price. These are quarterly board compensation credits, not decisions. They carry no signal whatsoever. They do show that Southern's non-executive directors hold meaningful equity — William G. Smith Jr holds 163,013 units, David J. Grain 86,881, Shantella E. Cooper 32,606, Anthony F. Earley Jr 31,592 — which is a modest governance positive and is recorded as such.

The other two are 100-share sales by the Comptroller, one month apart, at $95.77 and $94.28. A hundred shares is approximately $9,400. Two of them across two months is approximately $19,000. This is not a signal about anything. It is worth one observation only: the 2026-08-03 sale was executed at $94.28, on the day the convertible notes were priced off a $92.94 close, and one day before this dive.

What the file does not contain: any transaction of any kind by the chief executive, the chief financial officer, or any executive officer other than the Comptroller — and not a single open-market purchase. On a stock flat for twelve months at a 3.2% yield, the absence of insider buying is mildly disappointing and carries little weight.

11. Verdict, kill-criteria and flip conditions

Hold.

What is genuinely good here, and it is more than the price suggests. Southern has converted the single biggest structural change in its industry in forty years into roughly 16 gigawatts of signed Large Load Contracts with terms that are stronger than any comparable disclosure in this batch: minimum-bill provisions on all of them, and minimum duration, termination payments and posted collateral on approximately 13 gigawatts. Weather-adjusted commercial kilowatt-hour sales grew 10.9% in the June quarter. Adjusted earnings per share grew 22.8% on revenue that grew 0.1%. The dividend yields 3.20% at a 65% payout, the beta is 0.332, and the maximum twelve-month drawdown was 6.5%. The knowledge base's independent electricity-demand lane — fifteen channels, none of them naming a company — is as one-directional as any thematic lane this programme has surfaced.

What stops it being a Buy, stated as four specific reservations.

First, the base case is 5.1% and the payoff ratio is 1.42:1, and our base is BELOW the street's. $98 against a consensus of $100.08 and a median target of $101.50, with the street at Hold by 21 to 10 and three sells. There is no edge here and we are not going to pretend there is.

Second, the free cash flow is negative and the sign matters. Property additions of $13.9 billion trailing against $10.7 billion of operating cash flow is a $3.3 billion annual external funding requirement, permanently. The average share count rose 3.3% year on year, which consumed 4.3 points of the June quarter's 27.1% net income growth before it reached a per-share number. And the vendor file reports this company as generating a positive 2.40% free-cash-flow yield, which is wrong by $6.2 billion and by a sign.

Third, the financing has reached the convertible market for the third consecutive year, and $2,375 million of it was priced on the day of this dive. The Series 2026A notes convert at $104.56 — 12.5% above the reference price — which is a market-set statement about where the option value of this equity sits over eighteen months. It is 12% above where we are and 6.6% above our base case.

Fourth, and this is the reservation the knowledge base supplies: the earnings mechanism is an allowed return on invested capital, granted by state commissions, at a moment when the invested capital is growing faster than at any time in the company's recent history and the beneficiaries are data centres rather than households. no_priors: "Regulated utility monopolies earn a set rate of return on capex, creating a perverse incentive to build rather than innovate, pushing transmission costs up." Southern's own filings name "customer affordability matters" among their principal uncertainties, and its chief executive's quoted remarks pre-empt the charge. When a company answers a criticism before it is made, the criticism is live.

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

• A price near $84, which is 17.1x the FY2027 consensus of $4.924, a 3.55% dividend yield, and the 52-week low to the cent. At that level the base case is +16.7%, the asymmetry improves past 2:1, and nothing needs to change at the company. This is the most likely route and it requires a 10% decline — which this security experienced within the last twelve months.
• State commission approval of the remaining ~5 gigawatts of Large Load Contracts on the signed terms, which would take the fully-reviewed book from 11 to 16 gigawatts and remove the principal regulatory contingency in the growth case.
• A quantified disclosure of what the Large Load Contracts contribute — revenue, rate base, or contracted minimum bills. Southern has told us the gigawatts and the contract terms and has never told us the dollars. A number would convert this from a qualitative story into a modelable one, and it is the single most valuable disclosure the company could make.
• A September quarter at or above $1.75 adjusted EPS (consensus $1.66) with commercial kilowatt-hour growth still in double digits, which would confirm the FY2026 consensus is too low and pull FY2027 with it.
• Property additions plateauing while operating cash flow rises, closing the external funding gap and slowing the share issuance.

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

• A state commission denying recovery on a material portion of the construction programme, or reducing an allowed return on equity. This is the only event that breaks the model rather than delaying it.
• Weather-adjusted commercial kilowatt-hour growth below 5%, against 10.9% in the June quarter — the load is the thesis and this is the number that reports it.
• A Large Load Contract counterparty terminating, testing the minimum-duration, termination-payment and collateral provisions in public for the first time.
• An equity issuance at a discount rather than convertibles at a premium, which would say the funding model has become more expensive than management's own convertible pricing implies.
• Net debt to EBITDA above 5.75%, from approximately 5.2x, or a credit-rating action — the filings name "changes in Southern Company's credit ratings" explicitly among their risk factors.
• A major storm or wildfire event. The filings list "catastrophic events such as fires, including wildfires, land movement, earthquakes, explosions, floods, high winds, tornadoes, hurricanes and other storms" first among their catastrophic risks, and Southern operates across the US Southeast hurricane corridor. No multiple in this dive prices it.

Where SO fits in the Synthos Framework Portfolio. No position today. It goes on the watch list at $84 (17.1x FY2027 consensus, 3.55% yield) for a 2% initial sleeve in the utilities/infrastructure bucket, with a 3% target on state commission approval of the remaining 5 gigawatts. Sizing note: this is the highest-quality expression of the data-centre power thesis available in a regulated wrapper — contracted, collateralised, and yielding 3.2% — and the mistake is paying 20x for it when the same asset traded at 17x within the year with the contracts already signed. Logged as a tracked Synthos call (Hold) as of 2026-08-04 at $93.25, with the fair-value anchors, the kill criteria and the $84 upgrade trigger all gradeable.

Single biggest risk: that the financing gets harder before the earnings arrive. Southern is spending $13.9 billion a year on property additions against $10.7 billion of operating cash flow, has grown total debt $2,991 million in six months to $75,582 million, has issued 3.3% more shares in a year, and reached the convertible market for the third consecutive year on the afternoon of this dive. The earnings from the 16 gigawatts of contracted load ramp "through 2028"; the capital to serve it is being raised now. That timing mismatch is the structural feature of every regulated utility construction cycle, and it is normally resolved by regulators granting recovery — which is why every kill criterion above except the storm risk is ultimately a regulatory one. At 18.9x FY2027 consensus with a 5.1% base case, the security is not being paid to bear it.

Provenance & disclosures

• Traceability: 0 knowledge-base claims name Southern Company out of 51,928 distilled claims (raw hits 27, case-sensitive entity hits ZERO, name-level 0, discarded 27; breadth 0, net conviction none). The case-sensitive entity sweep on SO, Southern Company, Southern Co, Georgia Power and Vogtle was run FIRST, precisely because "SO" is among the most collision-prone tickers in the store, and it returned nothing. The free-text sweep on "Southern Company", "Georgia Power", "Vogtle", "regulated utilit", "data center power" and "electricity demand" returned 27 hits across 15 channels which are NOT noise but a multi-sourced independent thesis on US electricity demand that names no company: macrovoices 2025-01-23 (bullish 88), all_in 2025-05-06 (bullish 90), no_priors 2025-10-15 (bullish 78, "US electricity demand growth may accelerate from ~2% to ~10% CAGR"), luke_gromen 2025-06-04 (bullish 75), jordi_visser 2025-11-08 (neutral 70, skill 2.0), dylan_patel 2026-02-03 (bullish 62, named speaker), lex_fridman, a16z (Drew Baglino), odd_lots (Daniel Yergin), money_of_mine (Matthew Fist), andreas_steno and forward_guidance. The counterweight, from the same channel as the most bullish demand claim: no_priors 2025-10-15 (BEARISH 78) — "Regulated utility monopolies earn a set rate of return on capex, creating a perverse incentive to build rather than innovate, pushing transmission costs up" — and no_priors 2025-01-23 (neutral 70, entities PG&E) — "Regulated utilities love CapEx (marked up ~15% to ratepayers)." These two describe Southern's exact earnings mechanism as a distortion and identify the affordability channel through which regulators would cap it; they carry no name-level conviction and are the most useful claims in the lane. A REDACTION is disclosed: one further claim (all_in, 2025-08-09, bullish 70) reports a named artificial-intelligence laboratory requesting 50 gigawatts of US power over three years, "roughly a full year's added capacity"; the laboratory's name is redacted under the house prohibition on naming AI models or labs, and the quantity, framing and source are reported in full. Channel concentration: jordi_visser 7 of 27 (26%) — below the one-third threshold, so no concentration sensitivity test is triggered; had it exceeded a third the test would have been run, because that channel carries a skill weight of 2.0. One hit from quarantine_misattributed is EXCLUDED on attribution grounds alone; one from elon_musk_mgmt is excluded as a management channel for a different registrant. No claim carries a management speaker_role for Southern Company. All quotes are verbatim from the stored claim text.
• Data as-of: all operating figures — segment demand, weather-adjusted sales, Large Load Contract terms, property additions, operating cash flow, debt, equity and total assets — through 2026-06-30, from the 10-Q filed 2026-07-30 · quarterly earnings, adjusting items and share counts from the 8-K filed 2026-07-30 · the $2,375 million convertible-note pricing from the 8-Ks filed 2026-08-03 and 2026-08-04 · annual figures for FY2025 and earlier from the vendor payload, cross-checked against the 10-Q's December comparatives · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873602 = 2026-08-04T20:00:02Z ($93.25, +0.33%; 50-DMA $94.49; 200-DMA $92.77; RSI 44.7; MACD −0.22) · knowledge-base claims 2026-08-04. Southern's fiscal year is the calendar year. All figures come from the Synthos vendor data file for SO or from the SEC filings in the SO archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-19 (fiscal 2025); 10-Q filed 2026-04-30 (March 2026 quarter); 10-Q filed 2026-07-30 (June 2026 quarter — the source of the Large Load Contract disclosure, the weather-adjusted sales data, the cash-flow statement and the balance sheet); 8-K filed 2026-07-30 (the complete second-quarter earnings release with the GAAP-to-adjusted bridge); 8-K filed 2026-08-03 (announcement of the convertible offerings); 8-K filed 2026-08-04 — the date of this dive — (upsize and pricing of $725M of Series 2026A 2.125% notes and $1,650M of Series 2026B 3.50% notes). All carry preserved [TABLE] statement data. The 8-K corporate-action scan was run and found the largest single financing event in this batch, dated the day of the dive.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): the trailing capital-expenditure figure — km_ttm implies approximately $7,765M against the filings' $13,920M, a 44% understatement that turns an actual trailing free cash OUTFLOW of approximately −$3,269M into a reported freeCashFlowYieldTTM of +2.40% and a freeCashFlowPerShareTTM of +$2.263, invalidating priceToFreeCashFlowRatioTTM (41.7x), evToFreeCashFlowTTM (70.5x), capexToDepreciationTTM (1.209 against a true ~2.17x), capexToOperatingCashFlowTTM (0.749 against ~1.31x) and dividendPaidAndCapexCoverageRatioTTM (0.951), and contradicting the SAME FILE's cf_a block, which reports FY2025 capital expenditure of −$12,737M and free cash flow of −$2,935M correctly; the $2,375 million convertible-note pricing of 2026-08-03, absent from every vendor field, along with the Series 2024A 4.50% and Series 2025A 3.25% convertible notes the pricing release confirms are outstanding; totalDebt of $74,075M, which includes $1,484M of operating lease obligations against the filing's pure debt of $72,591M, a 2.0% overstatement of the T/TJX lease-inclusion class; seg_prod, whose FY2024 and FY2025 entries contain only the natural-gas businesses — $5,042M against $29,553M of reported revenue, 17.1% — omitting the electric utilities that are 83% of the company, with FY2023 missing entirely; seg_geo, which is completely EMPTY; and est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio signature (exactly 43.23% and 25.05% of revenue in every forward year FY2026–FY2030) and for understating FY2025 EBITDA by 14.0% against the reported figure. Where vendor and filing AGREED — recorded, and one is exceptional: quote.previousClose of $92.94 is independently confirmed to the cent, twice over, by the convertible pricing release's stated 12.5% and 27.5% premiums to conversion prices of $104.56 and $118.50 — the strongest external price verification recorded in this programme; cf_a FY2025 capital expenditure and free cash flow both match the filings exactly; bal_a FY2025 reconciles to the 10-Q's December comparatives line by line (long-term debt $65,649M, accounts payable $3,710M, total liabilities $116,853M, total equity $38,867M, total assets $155,720M); earn_cal reports adjusted EPS and revenue matching the earnings release ($1.13 and $6,977,000,000 for Q2 2026); the enterprise-value rebuild leaves a residual of $1,667M, 0.9%, well inside tolerance; and tech.max_dd_from_peak of −6.49%, which carries the ADP-class signature of equalling pct_from_hi, was CHECKED against the three return windows and found CONSISTENT with a recent peak and an earlier trough — verified, not flagged.
• Basis note: analyst estimates are on adjusted EPS, which excludes accelerated depreciation from repowering, losses on extinguishment of debt, estimated losses on Nicor Gas capital investments and on plants under construction, and disposition impacts. The June-quarter bridge is disclosed cleanly: GAAP net income $1,174M plus accelerated repowering depreciation of $143M pre-tax ($111M after tax), the $129M loss on debt extinguishment ($97M after tax) and smaller items equals $1,289M excluding items, or $1.13 against $1.03 GAAP — a 9.7% adjustment. The largest recurring component has a disclosed end date and remaining size: "At June 30, 2026, the remaining pre-tax accelerated depreciation is projected to total approximately $200 million in 2026 and $100 million in 2027," relating to Southern Power's Kay, Grant, Grant Plains, Wake and Bethel wind repowering projects completing through the third quarter of 2027. Trailing adjusted EPS of $4.30 is the sum of the four quarterly actuals; the vendor's inc_a EPS of $3.94 is the GAAP figure, and its priceToEarningsRatioTTM of 22.36x is on a partly-GAAP basis not comparable to the forward multiples used here.
• Estimate coverage: 13 analysts on FY2026 EPS, 14 on FY2027 — the anchor for all three fair values — with an FY2027 range of $4.843 to $4.964, a 2.5% spread; only 4 on FY2028, and 2 each on FY2029 and FY2030, all of which are excluded from every conclusion. Revenue coverage: 12, 13, 14, 9 and 13 analysts respectively.
• Peer note: the vendor peer set — AEP, CEG, D, DUK, ETR, EXC, NEE, NGG and XEL — is the correct comparison universe and the best peer set in this batch. But the payload supplies only market capitalisations and no multiples, earnings or growth rates, so no peer-multiple comparison is drawn. Recorded for completeness: NEE $181.9B, DUK $96.9B, CEG $96.0B, NGG $80.9B, AEP $69.9B, D $60.9B, ETR $50.8B, XEL $48.6B, EXC $47.2B.
• Fair-value caveat: the $78 / $98 / $115 anchors are multiples of the FY2027 consensus adjusted EPS distribution — 16.1x the low of $4.843, 19.9x the mean of $4.924, and 23.2x the high of $4.964 — each cross-checked against FY2028E (14.5x, 18.2x, 21.4x) and against the implied dividend yield (3.82%, 3.04%, 2.59%). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 18x gives $89, below spot; 22x gives $108. The base case assumes a small amount of forward-multiple compression, driven by a share count rising 3.3% a year and by the contingent dilution embedded in the convertible notes priced on 2026-08-03 at conversion prices of $104.56 and $118.50.
• Utility-metric note: free-cash-flow-based measures are corrected rather than discarded here, because for a regulated utility the free-cash-flow deficit is the financing requirement and is therefore the most important number in the file. Trailing free cash flow is approximately MINUS $3,269 million (operating cash flow ≈$10,651M less property additions ≈$13,920M, both derived from the filings), against the vendor's implied PLUS $2,603M. The measures used throughout are adjusted EPS, weather-adjusted kilowatt-hour sales by customer class, contracted gigawatts of Large Load, property additions, total debt on a pure-debt basis, average shares outstanding, and the dividend yield and payout — all filing-sourced except the estimates.
• Half-weight management framing: the chief executive's quoted remarks — "extraordinary economic development momentum," "investing responsibly," "keeping reliability and rate stability at the center of our work," "when growth is done right, everyone benefits" — are company framing and are half-weighted under house rules, identified as such at every use. The Large Load Contract quantities and terms are NOT management framing: they are quantified disclosures in the 10-Q and are treated as filing facts.
• Timing: second-quarter 2026 results were released 2026-07-30, five days before this dive, and beat the consensus adjusted EPS estimate by 11.9% ($1.13 against $1.01). The $2,375 million convertible-note offering was priced 2026-08-03, announced in an 8-K filed 2026-08-04 — the date of this dive — and is expected to close 2026-08-06, two days after it. The next earnings print is 2026-10-29, 86 days away (consensus adjusted EPS $1.66, revenue $8,199M). The most recent insider filing is dated 2026-08-04 (a 100-share sale by the Comptroller executed 2026-08-03).
• 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.