Aes Corp AES

Thesis

AES Corp is a globally diversified power company with four strategic business units: US regulated utilities (AES Indiana, AES Ohio), US Renewables (a top-three developer with a multi-GW signed backlog tied to hyperscaler PPAs), Energy Infrastructure (legacy thermal/IPP assets in Chile, Dominican Republic, Argentina, Panama, Mexico, Vietnam, Bulgaria), and New Energy Technologies (Fluence battery storage stake, hydrogen JVs). The bull thesis is straightforward: data-center power demand is exploding, AES has signed ~12 GW of corporate PPAs largely with hyperscalers, the regulated utilities provide a rate-base growth platform at allowed ROEs, and the sprawling EM IPP fleet is being harvested and recycled into US renewables. Buy the data-center and rate-base narrative, ignore the leverage, get re-rated to a US-utility multiple.

The scorecard is harsher. Composite is 62/100. Profitability scores 13 — 10-year average ROIC of effectively 0.0 per the deterministic scorer, which captures that two decades of capital deployment into emerging-market IPPs has produced cumulative excess returns of approximately nothing. FCF conversion of 7.91x looks anomalous and reflects that owner-earnings TTM is only $1.8B against a multi-billion capex run-rate. Net debt / EBITDA reads -1.231 (a sign artifact / classification quirk in the scorecard), but the actual recourse + non-recourse debt stack is well over $30B. Interest coverage scored 0.0. Share count has barely moved (+0.83% in 10 years — equity-light at parent, but they raise project debt and tax-equity instead).

The price-vs-IV math: IV base of $74.48, low $41.86, high $112.21 — a 2.7x spread between low and high, flagged by the scorer because maintenance capex is uncertain by more than 50%. With no current_price input, the actionable rule is: only interesting below the low-IV floor of ~$42, where you are essentially being given the regulated utilities for free and renewables backlog as a call option. Above $74 base IV you are paying for execution that the 10-year ROIC record does not support. This is a price-driven, not quality-driven, situation.

Moat

AES has the form of a moat — long-lived physical assets, regulatory licenses, contracted cash flows — without the substance of one. Walk the five moat types:

1. Pricing power. Effectively none in merchant generation, where AES is a price-taker on regional power curves (PJM, Chilean SIC, Mexican wholesale). In renewables, AES locks in 15-20 year fixed-price PPAs with hyperscalers and corporates — that's not pricing power, that's pricing absence (you trade upside for bankability). The regulated US utilities (AES Indiana, AES Ohio) have the standard regulated-monopoly construct Damodaran flags as 'a mixed blessing': legal monopoly, but the regulator 'preserves the right to control the prices charged and margins earned' to ensure firms 'did not earn excess returns' [2]. AES Ohio's allowed ROE is in the ~9.7% range; AES Indiana similar. That is cost-of-capital pricing, not pricing power.

2. Switching costs. Inside the regulated franchises, switching costs are infinite — captive ratepayers in Indianapolis and Dayton cannot choose another wire. But the regulator extracts the surplus, so switching costs do not translate into excess returns. In merchant and contracted generation there are no customer switching costs once a PPA expires; recontracting risk is real. In Fluence (battery storage software/hardware), Damodaran's [4] caution about Yahoo-style search engines applies: 'there is little cost to an end-user from switching from one engine to another.' Battery storage integration software has weak switching costs versus Tesla Megapack, Wartsila, Sungrow.

3. Network effects. Zero. Power generation is not a network business. The grid is the network, and AES doesn't own it.

4. Intangibles (brand, license, IP). The licenses are the asset in the regulated and contracted portions of the book — Damodaran calls these 'exclusive licensing rights to service a market' [2] — but as he explicitly notes, the regulator captures the rents. Brand value in B2B power sales is near-zero; hyperscalers pick on price/availability/ESG. Fluence has some IP and integration know-how but is racing against far better-capitalized incumbents.

5. Cost advantages. This is where the bull case has to live. AES claims scale advantages in renewables development (top-3 US developer, ~12 GW signed backlog), interconnect-queue position, and a 'first-mover' advantage on hyperscaler PPAs. Damodaran [4] lists three cost-advantage flavors: economies of scale, exclusive distribution, and lower-cost inputs. AES has modest development scale, no exclusive distribution (transmission is independent), and no structural input-cost edge — solar panels, batteries, turbines are commodity-priced and the IRA tax-credit transferability has commoditized financing too. Run the $10B / 5-year stress test [3]: NextEra, Brookfield Renewable, Constellation, Vistra, and infrastructure funds (Brookfield, KKR, EQT, Blackstone) can all write $10B checks. None of them face barriers to replicating AES's renewables pipeline. Excess returns in this segment will erode toward cost of capital — exactly what Damodaran [3] predicts in competitive sectors.

Erosion risk. The regulated utility piece is durable but capped at allowed ROE. The contracted renewables piece is bankable but rapidly commoditizing as every infrastructure fund crowds in. The legacy EM IPP piece has currency, expropriation, and recontracting risk (Argentina peso, Chilean tariff disputes, Dominican Republic FX controls). The 0.0 ten-year ROIC is the moat-erosion verdict already delivered: AES has been running this business for 30 years and has produced zero cumulative excess return on capital. That is the most damning moat evidence imaginable.

One nuance worth pricing in: the replacement cost of AES's installed base — particularly transmission-connected sites with grandfathered interconnection rights in tight markets like ERCOT-North, PJM-Mid-Atlantic, and CAISO — is structurally above book value because new interconnection studies are now 4-7 years to completion. That gives AES a small, decaying option-value moat on its existing fleet that does not show up in 10-year ROIC. It is not durable enough to change the verdict, but it does justify the IV-low of $42 rather than a deeper distressed mark. Erosion timeline: 5-10 years as transmission build-out catches up.

Moat verdict: NARROW (and arguably NONE in economic terms — physical-asset durability without excess returns).

Management & Capital Allocation

Management here is Andrés Gluski (CEO since 2011) and a deep bench of long-tenured operators. The capital-allocation track record is the relevant lens, not the personalities. Run the five-choice framework.

1. Reinvestment in the business. AES has been a heavy reinvestor — the renewables build-out has consumed $5-7B of annual capex in recent years, financed primarily with non-recourse project debt and tax equity. The reinvestment quality is the central question, and the 10-year average ROIC of 0.0 per the scorecard is the answer: cumulative reinvestment has not earned the cost of capital. You can quibble that the historical EM IPP investments dragged down the pooled ROIC and the new renewables vintages will earn 8-10% unlevered, but that is precisely the promise — 30 years of similar promises preceded it, and the result is in the data. The scorer also flags 'NOPAT declined; ROIIC not meaningful,' which means even on a five-year incremental basis, the new dollars are not visibly earning more than the old ones.

2. Acquisitions. Mixed-to-bad. AES has spent decades acquiring distressed EM utilities (Brazil, Argentina, Venezuela exit, Eastern Europe exit). The discipline has improved — Gluski has been a divestor, exiting non-core geographies (Brazil distribution, Cameroon, Jordan, Bulgaria coal). The Fluence creation (with Siemens) and IPO was creative; the fall in Fluence stock from peak has eaten most of the unrealized value.

3. Debt. This is the structural feature of the business. AES carries massive non-recourse project debt at the subsidiary level and meaningful recourse debt at the parent. Interest coverage scored 0.0 in the scorecard, which by itself should make a Buffett-Munger investor pause. The defense — most debt is non-recourse and matched to long-dated PPAs — is real but does not eliminate the equity risk: in any rising-rate or refinancing-stress scenario, project equity gets pinched first. Recourse debt has been deleveraged modestly; the parent leverage ratio target is ~3x EBITDA, which is high for a 'utility.'

4. Buybacks. Negligible. Share count is up 0.83% over 10 years per the scorecard — essentially flat, which is the best you can say. AES has not been a serial diluter at the parent, which deserves credit. There is no record of opportunistic buybacks at deep discount to IV (no 'avg P/IV when buying' to assess because there is no buying).

5. Dividends. AES has paid and grown a dividend, currently yielding mid-single-digits at most price points. Dividend coverage by parent free cash flow has been tight; the dividend has at times been funded partly by asset-sale proceeds rather than recurring FCF. That's a yellow flag — financing distributions out of capital recycling rather than pure earnings.

Communication quality. AES discloses parent-level FCF, contracted backlog (in MW and dollars), and segment EBITDA cleanly. The 'Adjusted EBITDA with Tax Attributes' metric the company emphasizes capitalizes IRA tax credits into operating earnings — an aggressive presentation Buffett would mock as managerial creativity. They guide to long-term EPS growth of 7-9% which has been missed multiple times; 2023-2024 saw downward revisions and an embarrassing earnings reset.

Verdict. The team is competent operators in a structurally hard business. They have not destroyed value at the parent (share count flat, dividends paid), but they also have not created excess returns over a full cycle. The aggressive non-GAAP presentation and history of guidance resets dock them.

Capital allocator: C

Industry Structure

Porter's Five Forces on AES's blended business — segment-weighted because the four SBUs face very different structures.

1. Threat of new entrants — HIGH (in renewables), LOW (in regulated utilities). US regulated utilities have near-perfect entry barriers: legal monopolies, certificate-of-need requirements, infrastructure sunk costs. US renewables development has low entry barriers: capital is the only requirement, and capital is abundant. NextEra, Brookfield, Invenergy, EDP Renewables, plus every infrastructure fund (Brookfield IP, KKR, EQT, Blackstone Infrastructure, BlackRock GIP, Stonepeak) can and do enter. Hyperscalers (Microsoft, Google, Amazon, Meta) increasingly contract directly or build their own. Damodaran [3]: 'the presence of these excess returns will undoubtedly draw in new competitors over time, putting downward pressure on these returns.' This is happening live in renewables.

2. Bargaining power of buyers — HIGH and rising. Hyperscaler offtakers are the most sophisticated power buyers in history. They run multi-developer auctions, demand 24/7 carbon-free terms, push for fixed-price + indexation hybrids, and require high-credit security packages. The 'top-3 US renewables developer' position AES touts is real but is not a moat — it's table stakes. Regulated retail customers have zero bargaining power individually but operate via PUCs, which represent them aggressively in rate cases.

3. Bargaining power of suppliers — MODERATE to HIGH. Solar modules, wind turbines, transformers, batteries, GE/Siemens turbines: most of the supply chain has had pricing power episodes (battery cells in 2022, transformers and high-voltage switchgear in 2023-2025). Lead times for transformers stretched to 2-3 years. Lithium, polysilicon, and copper expose AES to commodity supplier power. Skilled construction labor (IBEW electricians) is increasingly scarce.

4. Threat of substitutes — MODERATE. Electricity has no consumer substitute, so demand is structurally inelastic — that's the enduring tailwind. But within power, gas turbines, nuclear, geothermal, hydro, behind-the-meter solar/storage, and demand response all substitute for utility-scale renewables. Microsoft signing nuclear PPAs (Three Mile Island restart, SMR pre-orders) is a direct substitute for AES's renewables PPA pitch.

5. Rivalry among existing competitors — HIGH. Renewables development is hyper-competitive. Regulated utilities compete in regulatory forums (rate cases, capital plans), not in product markets, so rivalry is low there but is mediated by the regulator's appetite for rate increases.

Value pool. Where do excess returns sit in this industry? Empirically: in transmission (rate-based, allowed ROE plus growth in plant), nuclear at scarcity premiums (Constellation, Vistra), and in vertically integrated regulated utilities with constructive regulatory regimes. Not in merchant generation, not in EM IPP, not in cost-plus contracted renewables. AES's portfolio is over-indexed to the lower-return sub-sectors. Trajectory: data-center demand creates a temporary sellers' market in renewables, but the option value is being competed away in real time.

Industry Verdict: Average. The regulated utility piece is Good; the contracted renewables piece is Average and trending down; the EM IPP and merchant pieces are Poor.

Inversion (Bear Case)

I am short AES. Here is how this ends badly.

1. The single event that kills this. A material refinancing-cliff event meets a hyperscaler PPA renegotiation cycle. Roughly $4-6B of project and parent debt rolls in any given 18-month window. If the 10-year Treasury parks at 5%+ for an extended period and credit spreads widen, AES's marginal project IRRs collapse below the cost of capital. Simultaneously, one or more hyperscalers — having overbuilt power contracts during the AI hype cycle of 2024-2025 — exercises termination-for-convenience clauses, force-majeure claims, or simply renegotiates for lower indexation. This is exactly what happened to Ørsted in 2023 (canceling Ocean Wind 1 & 2 with a $4B writedown) and to multiple US offshore developers. AES takes a $3-5B impairment, the dividend is cut, and recourse-debt covenants are tested. The equity gets repriced to a stressed merchant-IPP multiple.

2. Why the moat is narrower than bulls think. Bulls hear 'top-3 US renewables developer with 12 GW backlog' and infer durability. The moat is interconnection-queue position and signed PPAs — both are point-in-time advantages that decay. The IRA's tax-credit transferability has commoditized renewables financing: any infrastructure fund with a Cayman vehicle can now monetize PTC/ITC, eliminating the prior advantage of large balance-sheet sponsors. Hyperscalers are pivoting hard to nuclear (Microsoft-Constellation, Amazon-X-Energy, Google-Kairos), gas (Meta is buying gas, full stop), and behind-the-meter generation. The 'AI demand for renewables' narrative is six months out of date. The 0.0 ten-year ROIC tells you AES has never had a moat — bulls are extrapolating a few good 2023-2024 quarters into a permanent re-rating.

3. Why management is worse than it appears. The 'Adjusted EBITDA with Tax Attributes' metric capitalizes IRA tax credits into operating earnings. This is the kind of non-GAAP construction Buffett identified in 1984 as the 'walking dead' tactic — businesses 'simply preferred to take an extraordinarily optimistic view' of their economics [Buffett 1984]. Guidance has been reset downward multiple times in 2023-2024. The dividend has been partially funded out of asset sale proceeds — financing distributions out of capital recycling, not earnings. The company has spent 30 years building a global IPP and produced zero cumulative excess return; that is the resume. A new strategy ('we are a US data-center power company now') has been articulated maybe four times over those 30 years; each time the next regime collapsed.

4. What bulls are extrapolating that won't hold. (a) Hyperscaler PPA pricing — currently elevated 30-50% above 2022 levels — will normalize as supply catches up by 2027-2028. (b) Backlog conversion rates of 90%+ assume no PPA cancellations or repricing; historical IPP backlog conversion is 60-75%. (c) The regulated utility ROEs are not a growth engine — Indiana and Ohio rate bases grow 4-6%/yr, the allowed ROE caps the upside, and rate cases periodically claw back. (d) EM IPP cash flows are extrapolated as stable; Argentina, Dominican Republic, and Chile have all had recent tariff-renegotiation episodes. (e) Fluence value is treated as a free option; Fluence stock is down ~70% from peak and the equity stake mark is materially impaired.

5. Valuation trap (multiple compression / regime change). Bulls anchor on 'data-center utility' multiples — 18-22x EPS, similar to NextEra. Bear case: the correct comp set is merchant IPP + EM exposure, which trades at 6-9x EPS (Vistra, NRG ex-AI hype, Brazilian Eneva, Engie). At 8x normalized parent FCF (~$1.2-1.5B sustainable), equity value is $9-12B — versus mid-2025 market cap above $9B at $13/sh and >$15B at higher 2024 prices. The IV-low of $41.86 is itself optimistic in a hard regime change; a true bear case ROIC-to-cost-of-capital re-rating implies stress-case fair value of $8-12/sh, below the IV-low. Multiple compression is the dominant risk factor — the operating model 'works,' the equity simply isn't worth what bulls model.

If I am right, the stock could be worth $9-14 within 2-3 years.

Lollapalooza Bias Check

Active biases in me, the analyst, right now:

1. Anchoring (high). I am anchoring hard on the IV-base of $74.48 because the deterministic scorer produced it. The scorer also explicitly flags that maintenance capex is uncertain by more than 50% and that 'no historical P/FCF available; using neutral 12/17/22 multiples.' The IV is a rough triangulation, not a precise number. I should anchor less on $74.48 and more on the range — and within the range, on the low-IV of $41.86, which embeds a more realistic view of normalized economics for a 0-ROIC business.

2. Confirmation bias (medium-high). I came into this analysis suspicious of levered EM IPPs (a Buffett-Munger prior) and the ROIC = 0.0% number confirmed exactly that prior. I should ask: what would I need to see to reverse the verdict? Answer: a clean 5-year US-renewables vintage demonstrating 8%+ unlevered IRRs, parent-level FCF coverage of dividend by 2x, and recourse-debt deleveraging below 2x EBITDA. None of those are in the scorecard data; my prior may be doing more work than the data warrants in either direction.

3. Recency bias (medium). I am over-weighting Ørsted-2023 and the offshore-wind-cancellation cycle as a template for what could happen to AES. AES's onshore-wind, solar, and storage book is meaningfully different from offshore-wind risk profile. I should not lazily map one cancellation cycle onto another.

4. Authority bias (low-medium). Buffett famously avoids capital-intensive utilities except via Berkshire Hathaway Energy, which has strong regulatory franchises. I am defaulting to 'Buffett wouldn't own this' as a substitute for original analysis.

5. Deprival super-reaction (low). Not strongly active — I do not own AES.

6. Social proof (medium). AES is widely held by 'data-center power' theme funds and is a frequent ESG-momentum name. The crowded narrative makes me reflexively contrarian, possibly too much so.

7. Incentive bias (relevant for the company). Management's compensation is tied to Adjusted EBITDA with Tax Attributes — a metric they themselves engineered. Munger: 'Show me the incentive and I'll show you the outcome.' This biases reported results upward.

Net. The biggest risk is anchoring on $74 IV and confirmation bias on 'levered IPP = bad.' The biggest protection against my biases is the scorer's deterministic output: 0.0 ROIC, 13/30 profitability, 0.0 interest coverage. Those are not opinions.

10-Year Outlook

Will AES in 2036 be the same fundamental business model? Mostly yes — owning long-lived power-generation and distribution assets, contracted or rate-based, financed with project and parent debt. The geography mix will be more US-tilted (continued EM divestitures), the technology mix more renewables/storage and less coal/gas (the coal exit is essentially done; some gas remains).

Will the customer base be larger? Likely yes, marginally. US power demand is finally growing again at 2-3% after two decades of flat consumption, driven by data centers, electrification, and reshored manufacturing. AES's regulated franchise grows with rate base. Renewables MW deployed will likely double from current ~37 GW to 60-75 GW.

Will profit per customer be higher? Uncertain. The hyperscaler PPA pricing tailwind of 2024-2025 likely fades. Allowed utility ROEs are stable to slightly down. EM IPP economics are structurally pressured. Real profit per kWh sold is more likely flat-to-down than up.

Will the moat be wider? Almost certainly not. The renewables development space is more crowded each year, not less. Hyperscalers continue to build internal energy teams. Battery storage software (Fluence) faces ferocious competition. The only widening moat is the regulated utility franchise — but that is a small slice of the total business.

Single biggest threat. A sustained higher-rate environment (5%+ 10-year Treasury) combined with hyperscaler over-contracting unwind. AES's leverage and refinancing schedule are vulnerable on both flanks. A secondary threat: a debt-financed acquisition or a transformative capital-recycling transaction that subtracts rather than adds value (a Gluski legacy risk).

My ability to forecast which of these scenarios plays out is genuinely limited. The scorer's wide IV range ($42 to $112) is itself a confession that the 10-year forward is hard to pin down. The Munger 4-test was borderline at the 'requires predicting commodity prices / regulatory outcomes' check — power prices, interest rates, hyperscaler demand, and Latin American politics are all required forecasts.

CONFIDENCE: low