Six Gigawatts Signed in a Quarter, One Cent of It in the Earnings Bridge

Results vs. Consensus

Q2 2026 Scorecard

The consensus figure requires a footnote. Three providers carried the quarter at $0.99, $1.00 and $1.01; we use $1.00. That number is also, to the cent, the estimate Southern itself published for the quarter on the April call. So the thirteen-cent beat is a beat against the company's own conservatism, which the Street had adopted wholesale. It is the second consecutive quarter in which Southern has cleared its own estimate by twelve or thirteen cents, and that pattern matters more for the guidance discussion below than the beat itself does.

Year-Over-Year Comparisons

The Twenty-One Cent Bridge

Southern publishes a cent-by-cent reconciliation of the change in adjusted EPS. It is the most useful table the company produces and it is where this quarter is decided. The right-hand column is ours.

Quality of Beat: Revenue

Total operating revenue rose $4 million, or 0.1%, and came in 3.5% below the $7.23 billion the Street had modelled. Both facts are less informative than they look, and in opposite directions.

The miss is almost entirely commodity pass-through. Retail fuel revenue fell $76 million and natural gas revenue fell $13 million; on the cost side, fuel and purchased power fell $34 million and the cost of natural gas fell $78 million. Fuel revenues under the traditional electric operating companies' recovery mechanisms generally equal fuel expense and do not affect net income. Strip retail fuel and natural gas from both years and the remaining $4,948 million of revenue grew 1.9% against $4,855 million. That is the honest top-line growth rate for a company whose reported revenue was flat.

Which is the part that should worry you. Non-fuel retail electric revenue, the line on which the regulated return is actually earned, rose $63 million to $3,682 million, or 1.7%. That is a genuine improvement on the first quarter, when the same line was unchanged to the dollar at $3,384 million in both years, and it satisfies the letter of the test we set at initiation. It does not satisfy the substance. Retail kilowatt-hour sales rose 2.1%, faster than the 1.7% revenue growth, so non-fuel revenue per retail kilowatt-hour fell about 0.3%. The company's own retail revenue bridge shows why: rates and pricing subtracted $10 million, "primarily due to lower contributions from commercial and industrial customers with variable demand-driven pricing at Georgia Power." Volume growth is arriving in the lowest-priced class, and the pricing mechanism attached to it is contributing less, not more.

Quality of Beat: Margins

Operating margin was 25.5% against 25.3%, sixteen basis points better, which reverses the 182-basis-point compression of the first quarter. Read carefully, the improvement is arithmetic rather than operational: revenue was flat, so the margin ratio simply reflects that expenses were also flat. Operating income rose $12 million on a base of $1,764 million.

Inside that flat expense line, two regulatory mechanisms are doing heavy lifting. Depreciation rose $111 million in total, but the Form 10-Q decomposes it as $98 million of accelerated wind repowering charges at Southern Power plus $102 million from additional plant in service, less $99 million "resulting from the extension of Georgia Power's 2022 ARP." At the regulated utilities, in other words, essentially all of the depreciation on newly in-service plant was offset by a rate-plan accounting benefit. Separately, non-fuel operations and maintenance rose only 1.2% partly because Alabama Power drew $19 million from its reliability reserve to offset transmission, distribution and generation expense, and taxes other than income taxes fell $36 million on a $37 million property-tax true-up of prior-year assessments at Georgia Power.

Assessment: None of this is improper and all of it is disclosed. But a reader who takes 25.5% as evidence that the cost structure is holding up under a construction program of this scale is reading a number that has been smoothed by a depreciation deferral, a reserve drawdown and a prior-year tax actualization, each of which has a finite life.

Quality of Beat: EPS

The effective tax rate was 13.6% against 25.3%, a decline of 1,171 basis points that reduced the tax provision by $102 million on pre-tax earnings that rose $233 million. The Form 10-Q is specific about the cause: a $93 million decrease in charges to a valuation allowance on certain state tax credit carryforwards at Georgia Power, plus $35 million of higher wind production tax credits from Southern Power's purchase of the noncontrolling interest in the SP Wind tax equity partnership, partly offset by $17 million less flowback of excess state deferred income taxes.

Divide the $93 million by the 1,137 million average shares and you get 8.2 cents, which is the eight-cent income-tax line in the company's own bridge almost exactly. The single largest contributor to the quarter is therefore not a lower tax rate. It is the absence of a charge Georgia Power took a year ago. It cannot recur, because it has already stopped.

A second way to size it: hold the effective rate at last year's 25.3% and the quarter produces GAAP EPS of roughly $0.89 rather than $1.03, and adjusted EPS of roughly $0.99 rather than $1.13. That is a hair below the consensus the company beat by 13%.

Two further items belong in the same paragraph. Earnings from equity method investments swung from $10 million to $86 million; the 10-Q attributes $58 million of that to "gains and losses associated with investments in energy-related venture capital funds" at Southern Holdings, whose entire equity-method book is $156 million. A $59 million swing on a $156 million venture portfolio is a mark, and marks reverse. And reported interest expense fell $78 million, but $129 million of the decline is the non-repeat of last year's convertible-note extinguishment loss, which the company itself excludes from adjusted earnings; on the ongoing basis, interest expense rose on $65 million of higher average borrowings, $11 million of higher rates and $7 million of power-purchase finance-lease interest, offset by $17 million more capitalized interest.

The GAAP-to-Adjusted Wedge

Adjusted EPS exceeded GAAP EPS by ten cents this quarter against twelve cents a year ago, which looks like narrowing. It is not. Last year's twelve-cent wedge was nine cents of convertible-note extinguishment plus roughly three cents of wind repowering. This year's ten cents is almost entirely repowering: $143 million pre-tax, $111 million after tax. On a like-for-like repowering basis the wedge went from about three cents to about ten. For the first half the wedge widened outright, from fourteen cents to twenty-two.

The repowering charges also got bigger, quietly. At March 31 the company projected roughly $335 million remaining in 2026 and $100 million in 2027. It incurred $143 million in the June quarter, which should leave about $192 million for 2026. It now projects approximately $205 million in 2026 and $120 million in 2027. Total expected remaining charges rose about $33 million during the quarter, with the 2027 estimate specifically raised by a fifth. The program still runs through the third quarter of 2027.

Segment Performance

Georgia Power

Revenue rose 0.7% and net income rose 28.3%. Earnings before income taxes rose 10.3%, so roughly two-thirds of the net income growth came below the pre-tax line. Georgia Power's own effective tax rate was 14.3% for the first half against 21.7% a year ago, on the same valuation-allowance release and increased amortization of federal production and investment tax credits that drives the consolidated rate. This is the subsidiary where the quarter was made, and it was made in the tax provision.

Georgia Power is also where the capital is going. Its property additions were $4,233 million in the first half against $3,090 million, up 37%, and its construction work in progress rose from $6,764 million at year end to $8,215 million at June 30, up 21.5% in six months. That is what produced the 60% increase in the allowance for equity funds used during construction, which the 10-Q attributes specifically to "an increase in capital expenditures subject to AFUDC at Georgia Power."

"The OpenAI contract in Georgia pushes us beyond our recently approved capacity by right around 1 GW. So just making sure that was real clear terms of where we are. And what the upside opportunities are."
— Chris Womack, Chairman, President and CEO

Assessment: Georgia Power is simultaneously the best asset in the portfolio and the clearest illustration of the problem with this quarter. Its load book is full, its capital program is compounding at better than 20% a year, and its reported earnings growth this quarter came from a tax item and a depreciation deferral, neither of which is a return on any of that capital.

Alabama Power

Revenue fell 0.3% while pre-tax earnings rose 15.3% and net income rose 14.7%. Unlike Georgia, the earnings growth here is broadly pre-tax, which makes it the cleaner result of the two big utilities. Roughly 3 GW of new large-load contracts were signed across three Alabama projects during the quarter, the first time the state has carried the headline in this story. Management framed it as the westward migration of demand it has been describing for several quarters. Cost discipline helped: the reliability reserve absorbed $19 million of transmission, distribution and generation expense in the quarter and $40 million in the first half, a regulatory smoothing mechanism that will eventually be exhausted.

Assessment: Alabama is the quarter's genuine operating bright spot and the place where the large-load funnel visibly widened beyond Georgia. Track whether the pre-tax earnings growth persists once the reliability reserve stops contributing.

Southern Power

Revenue fell 2.0% and the business posted a GAAP net loss of $25 million against $51 million of income, with a pre-tax loss of $88 million. That is entirely the wind repowering program, which took $143 million pre-tax this quarter and runs through the third quarter of 2027. On the adjusted basis the segment contributed nothing to the year-over-year change in the quarter and two cents in the first half.

Two structural items sit inside Southern Power's numbers. The purchase of the noncontrolling membership interests in the SP Wind tax equity partnership added $35 million of federal production tax credits, and it simultaneously removed $38 million of loss allocations that previously flowed to tax equity partners. Net income attributable to noncontrolling interests therefore swung from a $27 million loss to $14 million of income, a $41 million reduction in what is attributable to Southern. The transaction moved earnings from the noncontrolling-interest line to the tax line and was roughly a wash.

The more interesting Southern Power story is forward-looking. The existing tolling agreements begin rolling off into a power market that has repriced substantially, and management says the recontracting conversations are underway.

"But you look at in terms of where they were contracted from a pricing standpoint, in terms of where the market is today, we do see upside opportunities in those pricing opportunities. That will contribute to the durability and the length of our long-term plan"
— Chris Womack, Chairman, President and CEO

Assessment: A promised Southern Power update slipped to "later this year" for a second consecutive quarter. The recontracting optionality is real and probably underpriced, but it has now been described three ways in three quarters without a number attached to it.

Southern Company Gas

Revenue fell 1.3% on lower gas commodity prices flowing through, while net income rose 18.9% and pre-tax earnings rose 13.7%. Roughly half the improvement is identifiable: earnings from the equity investment in Southern Natural Gas rose $9 million in the quarter and $16 million in the first half. The segment also carries this quarter's new excluded item, an $8 million estimated loss related to Nicor Gas capital investments disallowed by the Illinois Commerce Commission in November 2025, with the release noting that further charges may occur.

Assessment: A solid quarter operationally and the least discussed part of the company. The Nicor disallowance is small, but it is now a recurring line in the exclusions table with no total exposure disclosed, and it is the only place in the portfolio where a regulator has taken capital away.

Mississippi Power

Revenue rose 0.8% while pre-tax earnings fell 10.5% and net income fell 11.9%. It is the only operating company going backwards, and it was not mentioned in prepared remarks. Management's only comment on Mississippi came in response to a question about the state's receptiveness to data centers, where the answer pointed to 100 MW projects and to activity in the pipeline rather than to anything in the current numbers.

Assessment: Immaterial to the consolidated result at roughly 4% of net income, but a negative print that goes unexplained on a call that found time for the World Cup is a small tell about disclosure priorities.

Load and Customer Metrics

The volume table repays a minute of arithmetic. Total retail sales grew 773 GWh year over year. Commercial contributed 934 GWh, residential subtracted 177 GWh, industrial added 14 GWh and the residual class added 2 GWh. All of the retail growth, and then some, is commercial. The Form 10-Q attributes the 7.4% weather-adjusted commercial increase to "increased customer usage, largely driven by data centers at Georgia Power." Management sized the system data center load at more than 1.2 GW, an increase of more than 500 MW. At any plausible load factor, 500 MW of incremental data center capacity running through a 91-day quarter produces more than the entire 773 GWh of retail growth.

That is the thesis working exactly as advertised, and it deserves to be said plainly: the load is real, it is showing up in the meter data, and it is now the whole of retail growth. It also means the other two classes are contributing nothing. Weather-adjusted residential sales fell 0.7% on "decreased customer usage, partially offset by customer growth," even though roughly 11,000 residential electric customers were added in the quarter and more than 40,000 in the year. Weather-adjusted industrial sales were flat, with primary metals and miscellaneous manufacturing offset by paper, textiles and chemicals. Southern's non-data-center demand base is not growing.

Key Topics & Management Commentary

Overall Management Tone: Management was confident and almost entirely forward-looking, spending the prepared remarks on contracted gigawatts and the pipeline rather than on how the quarter's earnings were produced. The composition of the increase went undiscussed: the tax valuation allowance, the rate-plan depreciation offset and the venture capital marks that together account for most of the twenty-one cents were not mentioned by management and not raised by any of the ten analysts who asked questions. Tone was more assured than at the first quarter, and the exchange was the least searching we have seen on this name.

1. Eight of the Twenty-One Cents Is a Charge That Did Not Repeat

The CFO opened with the number and the beat, and characterized the drivers in general terms.

"For the second quarter of 2026, our adjusted EPS was $1.13 per share, $0.21 higher than the second quarter of 2025 and $0.13 above our estimate. The primary drivers of our performance for the current quarter compared to last year included increased usage and customer growth, along with higher AFUDC from ongoing construction projects, higher earnings from equity method investments, and tax related impacts at our state-regulated electric utilities."
— David Poroch, Chief Financial Officer

"Tax related impacts at our state-regulated electric utilities" is doing a great deal of work in that sentence. The Form 10-Q names the impact: a $93 million decrease in charges to a valuation allowance on certain state tax credit carryforwards at Georgia Power. On 1,137 million average shares that is 8.2 cents, which reconciles to the eight-cent income-tax line in the company's own bridge. The other tax movements are smaller and go in both directions: $35 million more of federal wind production tax credits at Southern Power, $17 million less flowback of excess state deferred income taxes at Georgia Power.

The distinction that matters is between a rate and a charge. A structurally lower tax rate, driven by production and investment tax credit amortization on a growing renewable and storage fleet, is durable and worth capitalizing. Southern is in fact monetizing credits at scale: Alabama Power, Georgia Power and Southern Power received $39 million, $91 million and $49 million of cash respectively from credits transferred to third parties in the first half. But the largest single piece of this quarter's tax contribution is the non-repeat of a prior-year charge, and that is a comparison effect with a one-year life.

Assessment: Neither the release nor the call identified the valuation allowance. An investor who read only the earnings package would conclude the tax line reflects an improving structural rate. It mostly reflects an item that ends here.

2. The Contracted Book Went From Eleven Gigawatts to Seventeen

This is the quarter's real news and it is a large number. Three Alabama projects added roughly 3 GW, and Georgia Power signed a 3.2 GW, 25-year agreement with OpenAI at a site near Savannah that includes 1 GW of flexible demand response and begins taking service in phases in 2028.

"Combined, these four projects representing 6 GW of newly contracted customer load along with agreements previously signed brings our total contract to large load agreements across our electric subsidiaries to over 17 GW by the mid-2030s."
— Chris Womack, Chairman, President and CEO

Three months ago the contracted book was "over 11 GW" with 12 GW in late-stage discussion. It is now over 17 GW with 8 GW late-stage, of which 3 GW is described as near finalization. The prospective pipeline is unchanged at "well above 75 GW." Read as a funnel, roughly 6 GW converted from late stage to contracted and about 2 GW of fresh late-stage entered, which is a genuinely high conversion rate.

"Beyond the 17 GW already contracted there are an additional 8 GW of projects in late stages including 3 GW projected to be finalized in the near-term."
— Chris Womack, Chairman, President and CEO

Assessment: A 55% increase in the contracted book in one quarter is the single most thesis-relevant thing that happened, and it broadens the story beyond Georgia for the first time. It is also entirely a mid-2030s revenue event. Nothing about it appears in 2026 or 2027 earnings.

3. Twenty-One Billion Dollars of Collateral, Finally Sized

The unverifiable part of the bull case at initiation was the credit and cancellation protection behind the contracts. Management put a number on it this quarter, though it took two attempts. In answering a question about non-investment-grade counterparties, the CFO first described roughly $20.8 billion of collateral in a way that appeared to attribute it to the OpenAI contract alone, then returned unprompted to correct the record.

"You know, the whole portfolio of the 17 GW, we got about $21 billion of collateral. So, and I recognize you were speaking specifically to the OpenAI contract. So I just wanna clarify that for you."
— David Poroch, Chief Financial Officer

Twenty-one billion dollars against 17 GW is roughly $1.2 billion per gigawatt of contracted load, against a company rule of thumb of about $2 billion of capital per gigawatt of new company-owned generation. The collateral is described as a portfolio of parent guarantees, letters of credit and surety bonds assembled to a credit standard of "about an A- or better position," and it backs termination payments rather than the capital itself.

Assessment: This is a material improvement in disclosure and it deserves credit. The protection is real, it is sized, and the correction was voluntary. What remains undisclosed, for a second consecutive quarter, is the minimum bill itself: no dollar level, no rate, and no statement of what share of the contracted revenue is contractually floored.

4. "Full Share of the Cost to Serve" Became "Incremental Cost to Serve"

The pricing standard management describes for these contracts changed wording between April and July, and the two formulations are not equivalent in ratemaking. In April the framing was full embedded cost:

"These bilaterally negotiated agreements are structured so that customers driving incremental demand cover the full share of the cost to serve them, helping to assure this growth benefits all customers."
— Chris Womack, Chairman, President and CEO, first-quarter 2026 call

In July, twice, it was incremental cost:

"Recall, the framework under which we approach contracting with large load customers includes pricing with minimum bills, to cover at least 100% of the incremental cost to serve. Large load customers are paying their full share."
— Chris Womack, Chairman, President and CEO

The CFO used the same construction: minimum bills that "cover 100% of the incremental cost to serve." Incremental or marginal cost of service is the cost of adding the new load. Full or embedded cost of service also allocates a share of the existing system to the new customer. A large load paying only its incremental cost is being served on terms that are accretive to existing customers only to the extent the incremental cost is above the average, which for a system building new dispatchable generation at $2 billion per gigawatt it may well be, but it is a weaker claim than the one made in April.

Assessment: Possibly nothing more than a speechwriter's word choice, and we do not treat it as evidence of a change in the contracts. But it is the second consecutive quarter in which the strongest claim in the bull case is supported by adjectives rather than by a disclosed number, and the adjectives got weaker.

5. Georgia Is Oversubscribed, Which Turns the Procurement Into a Requirement

The OpenAI contract pushed Georgia Power past the generation capacity approved in its most recent resource plan. That converts the all-source procurement from optional upside into a need.

"And, like we have talked about in the past, we are probably about 1 GW or 2 away from if you will, selling out the capacity. That we had approved in Georgia last year. So now we are gonna work through that process"
— David Poroch, Chief Financial Officer

The timeline management gave: selection through the rest of the summer and into the fall, with clarity on which projects were chosen by year end; certification proceedings through much of 2027; and company spending beginning to "feather in" around 2028 for resources in service in 2030 and 2031. Alabama runs on a similar schedule. The capital arithmetic is the CFO's own: about $2 billion, "a little bit above maybe," per gigawatt of company-owned generation. On a 2 to 6 GW Georgia procurement plus whatever Alabama adds, the potential increment to the capital plan is measured in the high single-digit to low double-digit billions.

"there is not placeholders in our capital plan. We do not get ahead of our regulators. And so, obviously, you can see how the upside that we have talked about is not in our capital plan right now."
— David Poroch, Chief Financial Officer

Assessment: The absence of placeholders is disciplined and we would rather have it than the alternative. It also means the capital plan an investor is valuing today excludes the most likely source of growth, and will keep excluding it until late 2027. That is a long time to hold a 20-times multiple on faith.

6. Data Centers Are the Entire Retail Growth Story

The demand data was the strongest part of the quarter and management led with it.

"Notably, data center usage was 55% higher compared to the second quarter of 2025 and is now up 49% year-to-date. Primarily due to accelerating load ramps from our large load customers. System wide, our data center load now exceeds 1.2 GW, an increase of more than 500 megawatts over the prior year and we expect this trend to continue accelerating as our 17 GW of contracted demand comes online."
— David Poroch, Chief Financial Officer

"Year to date, weather-normal retail electricity sales were 2.3% higher than the first half of 2025, consistent with the trends observed earlier this year. This represents the highest retail sales growth through June we have seen in nearly two decades."
— David Poroch, Chief Financial Officer

Data center usage growth accelerated from 42% at the first quarter to 55%, and the installed data center load crossed a gigawatt during the quarter. Set against the volume table, the 500 MW of incremental data center load accounts for the whole of the 773 GWh increase in retail sales, because residential volumes fell and industrial volumes were flat.

Assessment: Confirming for the demand thesis and quietly concerning for the diversification of it. Southern's retail growth is now a single-customer-class phenomenon concentrated in one state, and the residential and industrial base underneath it is static.

7. Non-Fuel Retail Revenue Grew, But Revenue Per Kilowatt-Hour Fell

At initiation we flagged the flat non-fuel retail electric revenue line as the cleanest single test of whether the load is converting into earned return. It grew this quarter, by $63 million to $3,682 million. On the first-half view the entire increase came in this quarter; the line was unchanged at $3,384 million in both first quarters.

The composition is less encouraging than the direction. The company's own retail revenue bridge shows sales growth adding $77 million while rates and pricing subtracted $10 million, weather subtracted $10 million and fuel and other cost recovery subtracted $70 million, for a net $13 million decline in total retail electric revenue. The rates-and-pricing decline is attributed to "lower contributions from commercial and industrial customers with variable demand-driven pricing at Georgia Power," which is to say the demand-driven tariff attached to the growth is contributing less than it did. Retail volumes grew 2.1% and non-fuel retail revenue grew 1.7%, so revenue per retail kilowatt-hour on that line fell about 0.3%.

Assessment: Growth resumed, which is genuine progress against the test we set. But it is volume growth at a slightly lower realized price, arriving in the cheapest customer class, while base rates are frozen in Georgia and Alabama, whose two utilities between them book revenue equal to 73% of the consolidated total. The quantity is moving; the price is not.

8. The Capital Is in Construction Work in Progress, Not in Rate Base

Four disclosures in this quarter are the same fact seen from four angles. Property additions were $6,639 million in the first half against $5,456 million, up 21.7%. Construction work in progress rose from $10,534 million at year end to $11,874 million at June 30, up 12.7% in six months, with Georgia Power's balance up 21.5%. The allowance for equity funds used during construction rose 60% to $128 million, which the 10-Q attributes to "an increase in capital expenditures subject to AFUDC at Georgia Power." And depreciation on newly in-service plant added $102 million while a Georgia rate-plan extension removed $99 million of it.

Put together: the money is going into construction, the construction is accruing a regulatory return rather than earning a cash one, the assets are not yet depreciating in a way that shows up in the income statement, and the revenue requirement that would eventually pay for all of it cannot reset in Georgia until after 2028 or in Alabama until after 2029. That is a coherent and entirely legitimate structure. It is also why operating income has now grown 0.4% and 0.7% in consecutive quarters while adjusted EPS grew 7.3% and 22.8%.

Assessment: The construction allowance is a genuine economic return and it converts to rate base at commercial operation. The question this raises for 2029 and beyond is what happens to the income statement at the crossover, when the accrual stops, the depreciation starts and the rate-plan offsets have run their course. Nobody is modelling that yet because it is outside the guidance horizon.

9. The Full-Year Guide Implies a Flat-to-Down Second Half

Management characterized the outlook as continuing momentum and tilted the existing range upward without changing it.

"Looking towards the second half of the year, we anticipate this momentum continuing and now project our full-year 2026 adjusted earnings to be near or at the top of our 2026 adjusted EPS guidance range of $4.50 to $4.60. Our adjusted EPS estimate for the third quarter is $1.50 per share."
— David Poroch, Chief Financial Officer

The arithmetic runs the other way. Adjusted EPS was $2.46 in the first half against $2.15, growth of 14.4%. Southern delivered $2.15 in the second half of 2025 as well, $1.60 in the third quarter and $0.55 in the fourth. Holding the full year at the top of the range therefore implies second-half adjusted EPS of $2.14, which is 0.5% below last year. At the bottom of the range it implies $2.04, down 5.1%. And the third-quarter estimate of $1.50 is 6.3% below the $1.60 the company earned a year ago.

Assessment: One of two things is true. Either the second half genuinely decelerates sharply, in which case "momentum continuing" is the wrong description, or the quarterly estimates are as conservative as the last two prints suggest and the full-year range is stale. We think the second, but it was not tested. This was the single most answerable question on the call and no one asked it.

10. Equity: $2.5 Billion Settled, the Remaining Need Cut to $1.1 Billion

The financing story continues to be the least contentious part of the case, and it improved again.

"In the second quarter, we sourced an additional $700 million of equity through our at-the-market or ATM program with forward contracts to settle at our discretion through 2028. Together with the significant amount of equity previously sourced, we have reduced our projected remaining equity need by 2030 to $1.1 billion."
— David Poroch, Chief Financial Officer

The projected residual need fell from $1.8 billion at the first quarter to $1.1 billion, a 39% reduction, and the 17% funds-from-operations-to-debt objective for 2029 was reaffirmed. The 10-Q fills in the mechanics: Southern established a new at-the-market program in June for up to 50 million shares and terminated the 2024 program; 27.7 million shares settled during the first half for net proceeds of approximately $2.5 billion; and 11.5 million shares sold under forward contracts remain to be settled at the company's discretion.

The near-term cost of this is visible. Average shares rose 3.3% year over year against 2.2% at the first quarter, and dilution took four cents out of the twenty-one, up from three of nine last quarter.

Assessment: A genuinely good outcome on the terminal number and a worse one on the near-term run rate. Roughly 23 million further shares are implied between the outstanding forwards and the residual $1.1 billion, about 2% more dilution spread through 2030, which is a manageable figure for a plan of this size.

11. Rate Stability as Political Strategy

Base rates are held stable in Georgia through 2028 and in Alabama through 2029, and Southern joined the National Ratepayer Protection Pledge in the week before the call. The CEO tied the two together in unusually direct language.

"I mean, I was listening to a lot of commentary yesterday after the Fed decision. Lot of commentary about inflation and electric rates going up across the country. Electric rates are not going up. In our in our territory. We are delivering rate stability to our customers, and that is something that we are thrilled and privileged to make available and provide to our customers."
— Chris Womack, Chairman, President and CEO

On the political noise around data centers specifically, the answer pointed at the OpenAI announcement as the template: a project communicated alongside the rate-stability and community-benefit case rather than on its own. Management twice said the industry needs to explain the benefits better, and noted that while there may be moratoria in individual counties there are none at the state level.

Assessment: The rate freeze is the licence to build, and defending it is the right priority. Investors should also note its accounting counterpart: the extension of Georgia Power's 2022 rate plan is what removed $99 million of depreciation from this quarter's income statement. Rate stability for customers and earnings stability for shareholders are being produced by the same mechanism.

12. New Nuclear: Enthusiasm, and an Explicit Disclaimer

Asked whether Southern would build the next AP1000, the CEO gave the clearest answer he has given on the subject.

"I mean, Southern Company is not going to be next. Let me be clear about that. But we are going to continue to work constructively and very I would say, pretty aggressively with this administration. And with a lot of other parties to see how we can get this done."
— Chris Womack, Chairman, President and CEO

Pressed on whether hyperscalers are willing to absorb cost-overrun risk above budgeted amounts on new AP1000s, the answer was that conversations are happening "in terms of what role they can play in this conversation, in this equation," without a firm conclusion.

Assessment: Useful and slightly deflationary. Southern's Vogtle experience makes it the natural first mover, and the market has periodically priced some option value for that. Management has now said plainly that it will not be first. The optionality is in licensing and advisory rather than in owning the next unit, which is a much smaller number.

Guidance & Outlook

Implied second-half ramp. The guidance range, the third-quarter estimate and the first-half actual together pin down the shape of the year more tightly than management's language does.

Street at: The sell-side carries roughly $1.65 for the third quarter against the company's $1.50 estimate, a fifteen-cent gap that is best understood as analysts adjusting for a pattern rather than disagreeing with management. Southern cleared its own quarterly estimate by twelve cents in the March quarter and thirteen cents in the June quarter. Apply half of that to the third quarter and the year lands at $4.62; apply all of it and it lands above the range.

Guidance style: Conservative and, on the evidence of two consecutive quarters, systematically so. The company beat its own estimate twice and still declined to move the range, choosing instead to describe the outcome as "near or at the top." That formulation raises the expected value without creating a new number to be held to. It is a defensible way to guide and it is also why the reported full-year range now implies a second half that shrinks.

The longer-dated arithmetic. Full-year 2025 adjusted EPS was $4.30. The 2026 range of $4.50 to $4.60 represents growth of 4.7% to 7.0%, so even the top of the range only reaches the bottom of the 7% to 8% long-term objective management describes, and the midpoint falls short of it. The company's stated goal is to run in the top half of that trajectory. Nothing in the 2026 guide does that. The gap between the long-term objective and the guided year is the reason the second-half arithmetic deserves more scrutiny than it received.

Analyst Q&A Highlights

Whether the Contracted Book Buys a Longer Rate Stay-Out

The dominant line of questioning on the call was whether the visibility created by contracted load lets Southern extend rate stability further, and whether the customer credits that accompanied the last extension can be repeated on a larger scale. Management would not be drawn on either, and repeated the "we do not get ahead of our regulators" formulation three separate times across three different questioners.

Q: "But is there an opportunity to come back with bigger rate credit here? I mean, obviously, the number, I think, from last year at this point was north of half a billion. Is there a way to, again, do what you did before and effectively sidestep the revenue process here?"
— Julien Dumoulin-Smith, Jefferies

A: "Julien, let me say, as you know, we do not get ahead of our regulators. We work very constructively with them. Kinda hand in hand. I mean, our goal at all times is to keep rates as low as possible for our customers. And so we are thrilled with how we are delivering rate stability to our customers through 2028"
— Chris Womack, Chairman, President and CEO

Assessment: The refusal is procedurally correct and analytically unhelpful. The unstated logic is that a larger contracted book generates enough revenue to fund larger customer credits, which in turn buys a longer stay-out, which in turn defers the revenue-requirement reset that shareholders eventually need. Everyone on the call understood this; nobody made management say it.

How Large the Generation Procurement Could Actually Get

Two questioners tried to bound the incremental capital opportunity by working from the load additions back to megawatts of generation and then to dollars. Management confirmed the direction and supplied the capital multiple, while declining to confirm a gigawatt range.

Q: "But is the range beyond that as well. Like, can you just frame is this 4 to 5 plus gigawatts of gen that we could put some type of, you know, capacity multiple on and try and, you know, estimate what the size is?"
— Multiple analysts incl. Stephen D'Ambrisi, RBC Capital Markets, Carly Davenport, Goldman Sachs

A: "And keep in mind, I think we have shared this in the past. Maybe a decent rule of thumb to think about capital opportunities going forward is about $2 billion or so, a little bit above maybe related to 1 GW of new generating capacity."
— David Poroch, Chief Financial Officer

Assessment: The most useful exchange on the call. Management confirmed the framing was "directionally correct," confirmed roughly 3 GW of newly signed Alabama load and roughly 1 GW of Georgia oversubscription, and handed over the $2 billion per gigawatt multiplier. It stopped short of a total because the procurement processes are live, which is the correct place to stop.

Credit Protection Behind a Non-Investment-Grade Counterparty

The sharpest question of the day went at the obvious vulnerability in a 25-year contract with a private, cash-consuming counterparty. The answer produced the first sizing of the collateral behind the whole book, and then a voluntary correction when management realized the number had been attached to the wrong thing.

Q: "So just could you just remind, in Georgia, for OpenAI and other customers who maybe are not investment-grade credits? Or strong investment-grade. How do the tariffs work for from a credit standpoint? Both the large load tariffs?"
— Steven Fleishman, Wolfe Research

A: "But at the end of the day, the collateral portfolio that we are going to take to back up these contracts is going to put us at about an A- or better position"
— David Poroch, Chief Financial Officer

Assessment: Management engaged fully and the four-pillar structure it described (long-term contract, minimum bill, default provisions, collateral) is coherent. The correction that followed, restating $21 billion of collateral as covering all 17 GW rather than the single contract, was volunteered and reflects well on the disclosure culture. Note the arithmetic implication: roughly $1.2 billion of collateral per contracted gigawatt.

Demand Response as a Contract Feature

The gigawatt of flexible demand response embedded in the Georgia agreement was raised twice, once as a strategic question about whether it becomes standard and once as a valuation question about what it is worth to existing customers. The strategic answer was substantive. The valuation answer was not given.

Q: "I want to follow-up on the gigawatt of demand response. I agree that is a great resource. I know we something we have seen something similar in Michigan, but am I right that this is the first time you are doing something like that? And I know that the press release talks about long-term savings for customers, Have you quantified that?"
— Andrew Weisel, Scotiabank

A: "I mean, the rest of the customer base will see benefits across the entire project. I mean, what we are doing is doing peak load periods, during high-demand periods, we are able to shave the peak. And so as we operate this economic system on a minute-by-minute, second-by-second basis, we are always looking to put online the most economical resource to meet and serve that load."
— Chris Womack, Chairman, President and CEO

Assessment: Management confirmed this is the first demand-response structure it has done for a data center and described the operating benefit clearly, but never answered the quantification half of the question. A gigawatt of dispatchable curtailment materially changes the reserve margin arithmetic and therefore how much generation the procurement needs to buy. That is a number the company can compute and chose not to give.

Recontracting the Southern Power Fleet

The competitive generation fleet's tolling agreements begin expiring into a market that has repriced upward since they were signed, which is arguably the largest near-term earnings lever in the portfolio and the one management has said least about.

Q: "So, just, Christopher, on Southern Power, the existing tolling agreements are going to start to roll off. I know there is obviously an opportunity to repurpose the capacity towards serving the hyperscalers. Can you maybe discuss how you are thinking about that opportunity set Have conversations with the hyperscalers begun with these assets? And would any opportunities be captured within your existing 75 gigawatt pipeline?"
— Shahriar Pourreza, Wells Fargo

A: "And so the team is now in the midst of having those conversations with a host of different counterparties that you understand our risk profile there in terms of making sure that they are creditworthy counterparties. But you look at in terms of where they were contracted from a pricing standpoint, in terms of where the market is today, we do see upside opportunities in those pricing opportunities."
— Chris Womack, Chairman, President and CEO

Assessment: Three quarters of "we see upside" without a megawatt count, an expiry schedule or a price differential. A Southern Power update was again promised for later this year. This is the cheapest disclosure improvement available to the company and the one most likely to move the stock, and it keeps slipping.

The Shape of Future Equity Issuance

With the residual need cut to $1.1 billion and 11.5 million forward-sold shares still to settle, the question was less about how much equity and more about when the dilution actually lands.

Q: "Are you able to give any guidance on how to think about the pace of equity and when it will actually hit over the next few years? I know a lot of moving parts there, and I do not expect you to get too specific. But how should we think about the year-by-year dilution and whether it will be ratable or be maybe accelerated?"
— Andrew Weisel, Scotiabank

A: "But, you know, we generally wanna shape that in terms of kind of mirroring, if you will, our capital outlays. For the construction effort that we are gonna have. But it also you know, dovetails into the plan to get near 17% FFO to debt by 2029."
— David Poroch, Chief Financial Officer

Assessment: The answer is honest and unhelpful in equal measure. Settlement timing is discretionary and available at a few days' notice, so the company genuinely does not know. For modelling purposes the practical implication is that dilution front-loads with capital outlays, and capital outlays rose 21.7% in the first half.

Whether Contracted Ramps Match Contracted Load

A quieter question asked what the company has learned from data centers already energized, and whether actual ramp rates track what was assumed at contract signing. The answer was the most candid thing management said all day.

Q: "Any learnings you would highlight from some of that ramp-up and the sales trends? And I guess, in particular, how that might apply to your outlook for all this load growth and running that through to the EPS growth guidance as well."
— Richard Sunderland, Truist Securities

A: "No. I would start by saying, I mean, one of the things we see is that, you know, we have to work very closely with these projects in terms of what their ramp rates are. They may not be what was predicted when the when the projects were initially approved."
— Chris Womack, Chairman, President and CEO

Assessment: An admission that realized ramps differ from contracted ramps, immediately followed by the argument that it does not matter because the minimum bill decouples revenue from volume. That argument is the whole bull case for the contract structure, and it remains unverifiable because no minimum bill level has been disclosed. This exchange is precisely why that disclosure matters.

Whether Southern Builds the Next AP1000

New nuclear surfaced as it does every quarter, this time with a question about whether federal backstopping of cost-overrun risk would change Southern's willingness to build. The answer separated advocacy from participation more sharply than before.

Q: "Should there be, like, any financial backing from the government in helping taking on some of the cost overrun risk Is that something Southern would be interested in building, i.e., through the back end? Just any color on potentially participating in this consortium or you just looking to license the blueprints?"
— Shahriar Pourreza, Wells Fargo

A: "I mean, Southern Company is not going to be next. Let me be clear about that. But we are going to continue to work constructively and very I would say, pretty aggressively with this administration."
— Chris Womack, Chairman, President and CEO

Assessment: Clear, and worth pricing. Any residual option value in the equity for Southern building the next AP1000 should now be written to zero. What is left is a licensing and advisory role plus the possibility that hyperscalers underwrite overrun risk for somebody else, which management confirmed is under discussion but not concluded.

What They're NOT Saying

1. The $93 million valuation allowance: the largest single contributor to the quarter appears in the release only as "tax related impacts at our state-regulated electric utilities" and on the call only as "lower income taxes." Its nature as the non-repeat of a prior-year charge is disclosed exclusively in the Form 10-Q, filed the same day, and was not raised by anyone.
2. The $99 million depreciation offset: the extension of Georgia Power's 2022 rate plan removed almost exactly as much depreciation as newly in-service plant added. This is the reason a company that spent $6.6 billion on property additions in six months shows benign depreciation growth at its regulated utilities, and it was never mentioned.
3. Venture capital marks inside adjusted earnings: "higher earnings from equity method investments" resolves, in the 10-Q, to a $58 million increase at Southern Holdings from gains on energy-related venture capital funds, on a portfolio of $156 million. Roughly three cents of adjusted EPS is a mark that can reverse.
4. Why the second half decelerates: a third-quarter estimate 6.3% below last year and a full-year range implying flat-to-down second-half earnings, described as "momentum continuing," went entirely unchallenged.
5. The minimum bill, for a second consecutive quarter: no dollar level, no rate, no term structure, and no disclosure of what share of contracted revenue is floored. The claim that these contracts transfer ramp risk to the customer cannot be verified from any public document.
6. What "incremental cost to serve" means in dollars: the pricing standard was restated this quarter in weaker terms than last quarter, and the difference between incremental and full embedded cost was neither explained nor questioned.
7. Earned return on equity by subsidiary: still not disclosed. With base rates frozen through 2028 and 2029 and construction work in progress up 12.7% in six months, whether the utilities are over-earning or under-earning against their authorized returns is unobservable until a rate case reopens.
8. The value of the demand-response gigawatt: the press release cites long-term customer savings; the question asking for the number was answered qualitatively.
9. Nicor Gas disallowance exposure: a new excluded item at $8 million in the quarter, with the release warning that further charges may occur and no total quantified.
10. Mississippi Power's 11.9% earnings decline: the only operating company going backwards, unmentioned in prepared remarks and unasked in Q&A.
11. Residential usage per customer: weather-adjusted residential sales fell 0.7% on "decreased customer usage" despite roughly 11,000 customer additions in the quarter. Management cited the customer additions and not the usage decline.
12. The Southern Power update: promised "later this year" for a second consecutive quarter, still without a date, a megawatt schedule or a price.
13. When the capital plan gets refreshed: with Georgia oversubscribed by about a gigawatt and two live procurements, no date was given for an updated capital forecast.

Market Reaction

• Pre-print setup: Closed at $96.05 on 29 July, up 10.1% year to date against 6.9% for the S&P 500, up 1.2% over the trailing twelve months and up 0.4% over the trailing thirty days, inside a 52-week closing range of $84.08 to $99.72.
• Reaction session (30 July, before-open print): Opened at $97.20, up 1.2%; traded between $93.72 and $97.31; closed at $94.34, down 1.8% or $1.71. The stock gave up the entire opening gain and closed in the bottom quarter of the day's range.
• Volume: 6.4 million shares against a 5.8 million thirty-day average, 1.1 times normal. Unremarkable.
• Sector: The regulated utility complex fell on a session when the S&P 500 rose 1.7%. The utilities sector exchange-traded fund fell 0.6%, Duke 2.2%, American Electric Power 1.3%, Dominion 1.2% and NextEra 0.6%.
• Run-up: The OpenAI agreement was announced on 25 July, five days before the print, not with it. The stock rose 3.6% over the three sessions into that announcement and gave back 1.2% over the three sessions after it.

The instinct is to read a 1.8% decline on a 13% earnings beat as a verdict on quality. It is mostly not. On a day when the utilities sector ETF fell 0.6% and two large-cap peers that were not reporting fell 2.2% and 1.3%, the idiosyncratic component of Southern's move is roughly 1.2 percentage points. A Federal Reserve decision landed the previous afternoon and the S&P rose 1.7% while bond proxies sold off. This was a rotation day for the group, and Southern participated in it.

What the intraday shape does say is more interesting than the close. The stock opened up 1.2% on the press release, which contained the headline beat, the 17 GW figure and the guidance tilt, and then declined through the 1:00 p.m. call to finish 1.8% down and near the low. The identical structure ran in the opposite direction three months ago, when the stock opened up 1.1% and closed up 3.4% within a penny of the high as the call added to the release. The release was received well both times. The difference is what the call added, and this quarter it added very little that was not already in the July 25 announcement.

The pre-print setup also matters less than it did in April. Southern entered this print up 10.1% year to date, roughly 320 basis points ahead of the index, but up only 1.2% over twelve months and essentially flat over the prior month. It sits mid-range in its 52-week band. This is not a stock that had run into the number, and it is not one that was priced for disappointment either. The move is small and the interpretation should be correspondingly modest.

Street Perspective

Debate: Is a Beat Built on Tax and Accrual Items a Beat?

Bull view: The composition question is a distraction. Tax credits, the construction allowance and equity-method income are ordinary features of a regulated utility deploying capital at scale; the construction allowance in particular is an authorized return that converts to rate base at commercial operation, and the production tax credits are being converted to cash through third-party transfers. Adjusted EPS is up 22.8% and the first half is up 14.4%, and management has beaten its own estimate twice running.

Bear view: Fifteen of twenty-one cents came from a valuation allowance that cannot repeat, a non-cash accrual and a mark on a venture portfolio. Operating income has grown 0.4% and 0.7% in consecutive quarters. Strip the tax rate back to last year's level and the quarter is roughly in line rather than 13% ahead. A business whose reported growth requires this much explanation is not compounding, it is being presented as compounding.

Our take: The bear has the better of the specifics and the bull has the better of the framing. The construction allowance genuinely is an economic return and we do not discount it to zero. The valuation allowance and the venture marks are a different matter, and together they are roughly eleven cents of a twenty-one cent increase. The right conclusion is narrower than either camp's: this is a quarter in which the asset improved a great deal and the earnings did not improve at all, and the multiple is being paid for the earnings.

Debate: Does Seventeen Gigawatts Justify Twenty Times Before Any of It Is in the Plan?

Bull view: The contracted book grew 55% in a single quarter, Georgia is already oversubscribed, two live procurements could add high single-digit billions of capital at roughly $2 billion per gigawatt, and none of it is in the plan. The market is being offered a call option on the largest load-growth franchise in the sector at a sector-average multiple, with a 3.2% dividend and an equity need that just fell by 39%.

Bear view: Procurement selection lands at the end of 2026, certification runs through 2027, spending begins around 2028 and the resources come online in 2030 and 2031. Contracted load reaches 17 GW "by the mid-2030s." Every dollar of this is beyond any reasonable discounted horizon for a utility, and in the meantime the investor holds a security priced at 20.5 times whose operating income is flat and whose revenue requirement cannot reset until 2029.

Our take: This is the actual debate on the name and the honest answer is that both are describing the same fact with different discount rates. Our position is that the option is real, is worth something, and is already in a 20.5 times multiple on a business whose regulated peers trade lower. We would pay up for it on a pullback or on the first evidence that contracted load converts to earned return. We are not willing to pay a premium multiple for a mid-2030s revenue event that the company itself refuses to put in its capital plan.

Debate: Is the Full-Year Guide Conservative or Is the Second Half Genuinely Slowing?

Bull view: Southern has beaten its own quarterly estimate by twelve and thirteen cents in consecutive quarters. The third-quarter estimate of $1.50 will be beaten the same way, the full year will land at or above the top of the range, and "near or at the top" is management being characteristically careful. The Street's $1.65 third-quarter number reflects exactly this understanding.

Bear view: The company had every opportunity to raise the range after clearing its own first-half estimates by twenty-five cents and did not. The stated range implies second-half earnings of $2.04 to $2.14 against $2.15 delivered a year ago. The prior-year third quarter was $1.60 and this year's estimate is $1.50. Management knows something about the second half that the "momentum continuing" language obscures, most obviously that the tax comparison stops helping.

Our take: We side with the bulls on the number and with the bears on the disclosure. The pattern of conservatism is well established and we model $4.62 for the year, above the top of the range. But the tax comparison does get harder, the venture marks may not repeat, and management chose a formulation that raises expectations without creating an obligation. That the arithmetic was not challenged on the call is a failure of the process, not evidence that it does not matter.

Model Update

Valuation framework. Applying 20 to 21 times to a 2027 adjusted EPS estimate of roughly $4.90 produces a fair-value range of approximately $98 to $103 with a midpoint near $100, unchanged from our initiation. We have deliberately not raised it despite a contracted book that grew 55% in the quarter, because none of the associated capital is in the plan, spending does not begin until roughly 2028, certification does not conclude until late 2027, and the contracted load is described as arriving "by the mid-2030s." Against the $94.34 reaction close, a $100 fair value implies about 6.0% of price return plus a 3.2% dividend yield, for a twelve-month total return around 9%. At the top of our range it is closer to 12%; at the bottom, closer to 7%.

Thesis Scorecard Post-Earnings

The pillars below are the ones established at initiation three months ago, graded against what this quarter's filings and call revealed. Status tags follow the standing convention: bull pillars run on track, at risk or broken; risk points run contained, emerging or materializing.

Overall: The thesis is unchanged in shape and more polarized in content. All four bull pillars are on track and two of them strengthened materially, with the contracted book up 55% and the residual equity need down 39%. All four bear points were confirmed, and the central one escalated: the share of the earnings increase coming from outside operations went from 56% to 71%, and the mechanism is now fully visible in the Form 10-Q rather than merely suspected. The asset improved and the earnings did not. Those two things are not in tension, because the capital that will produce the earnings is sitting in construction work in progress accruing a return it cannot yet collect, and the revenue requirement that would collect it is frozen until 2028 in Georgia and 2029 in Alabama. That is the whole story of this company for the next three years.

Action: Maintaining Hold, and we are closer to an upgrade than we were in April without being at one. At initiation we said we would move to Outperform on two consecutive quarters of growing non-fuel retail electric revenue with the construction allowance falling below a quarter of the year-over-year earnings increase, on a disclosed procurement outcome that adds materially to the plan before the end of 2026, or on a pullback toward the mid-$80s. One and a half of those four tests were met. Non-fuel retail revenue grew, but this is the first such quarter, not the second, and the price per kilowatt-hour on that line fell. The construction allowance did drop to 19% of the earnings increase from 56%, clearing the stated threshold, but only because a tax item took its place, which meets the letter of the test and defeats its purpose. The procurement outcome is not due until year end. The stock is at $94.34, not the mid-$80s. Our fair value stays near $100, which at 20.5 times the top of the guidance range and a 3.2% yield implies a twelve-month total return around 9%: market-like, for a franchise we continue to regard as best-in-class in its sector. We would upgrade on a third-quarter print in which non-fuel retail revenue grows again with pricing flat or better and the beat comes from the operating lines of the bridge, on a procurement outcome disclosed with size and terms before year end, or on a move toward the mid-$80s where the yield approaches 3.6%. We would move toward Underperform on a Georgia commission outcome that reopens the rate freeze, on a fourth-quarter miss that reveals the second-half deceleration to be real rather than conservative, or on a third consecutive quarter of flat operating income while the non-GAAP wedge widens.