Powering the AI Era: Strategic Energy Landscape Through 2030

Executive summary

Global electricity demand is undergoing its most violent inflection in three decades, and the company best positioned to capture it is the one that owns the picks and shovels of AI's physical buildout. After 15 years of near-flat growth, global power consumption surged 4.4% in 2024 and is now compounding at ~3.6% annually through 2030 per the IEA, equivalent to adding a Japan to the world every year. Data centers alone will consume 945 TWh by 2030 in the IEA base case (Goldman Sachs revised to 1,350 TWh in April 2026), up from 415 TWh in 2024, with ~50% of US electricity demand growth coming from AI infrastructure.

The supply side cannot keep pace. GE Vernova's gas turbine orderbook is sold out through 2030. Large power transformer lead times have stretched to 2.5 years. The US interconnection queue holds 2,300 GW of generation; only 13% of projects entering the queue from 2000–2019 ever reached commercial operation. PJM's 2027/28 capacity auction cleared at the price cap, $333.44/MW-day, after falling 6,623 MW short of reliability requirements. NERC now flags 13 of 23 regions at elevated or high resource adequacy risk.

The strategic prize for AES and its incoming private owners is enormous but bounded by execution. AES enters this period with structural advantages competitors cannot replicate quickly: 11.8 GW of signed PPAs with Amazon, Google and Microsoft; the #1 BloombergNEF Americas clean-energy seller ranking for five consecutive years; a 67 GW development pipeline; an Andes platform unique among US peers; and a March 2026 take-private by GIP/EQT/CalPERS/QIA at $33.4B enterprise value that removes quarterly EPS pressure precisely when the industry needs patient long-cycle capital.

Six strategic priorities anchor the 2030 thesis. First, lean fully into hyperscaler co-located and dedicated generation, where demand far outstrips supply and 20-year PPA terms are now standard. Second, scale battery storage aggressively while AES already leads, before NIMBY headwinds following Moss Landing harden permitting. Third, monetize the AES Next/AI Fund partnership as a software platform sellable to other utilities, the only Tesla-style optionality on the AES balance sheet. Fourth, defend and extend the Andes platform as European competitors retreat from Latin America. Fifth, build SMR optionality through partnership rather than balance-sheet bet. Sixth, position for the gas/LNG bridge that the Trump administration and hyperscaler urgency have made unavoidable through 2032.

What is overhyped: SMR commercial timelines, fusion before 2030, hydrogen as a near-term electricity demand driver, offshore wind in the US through 2028. What is underhyped: the gas turbine bottleneck as a binding constraint on AI buildout, transformer supply chain stress, the criticality of land/permits/queue position as the new alpha, and Latin American renewables as European majors retreat.

1. Global energy demand: the macro picture

The flat-to-surging inflection

For the first time in 30 years outside crises, global electricity demand growth in 2024 outpaced economic growth. The IEA's Global Energy Review 2025 records +4.4% YoY in 2024, adding more than 1,200 TWh, roughly twice the 2.6% average of the prior decade. Growth moderated to ~3% in 2025 on milder weather but remains structurally elevated; the IEA's Electricity 2026 forecasts 3.6% CAGR through 2030, 50% above the prior decade. Cumulative 2025–2027 demand growth alone is projected at +3,500 TWh.

The US tells the story most starkly. After 15 years of near-zero growth, EIA's Annual Energy Outlook 2026 documents 2.1% CAGR over the past five years and projects 0.9–1.6% through 2050. NERC's 2025 Long-Term Reliability Assessment, released January 29, 2026, raised its 10-year US summer peak forecast by +224 GW (+24%), 69% higher than the prior year's projection. Grid Strategies finds the 5-year peak load forecast has risen six-fold in just four years. The US built 54 GW of new utility-scale generation and storage in 2025, the most in over two decades (BNEF/BCSE).

Demand drivers, quantified

AI and data centers are the single largest swing factor. The IEA's Energy and AI report (April 2025) projects DC electricity rising from 415 TWh (2024) to 945 TWh (2030), with US DC demand more than doubling (+240 TWh, +130%), China rising 170%, Europe 70%. Goldman Sachs revised its forecast in April 2026 to +220% by 2030, or 1,350 TWh, with the US accounting for ~750 TWh. McKinsey's central case is 1,400 TWh by 2030. In the US, data centers will account for ~50% of all electricity demand growth through 2030.

Electrification of transport added 17M+ EV sales globally in 2024, hitting 20M+ in 2025 (more than 1 in 4 new cars). BNEF's NEO 2025 projects 42M EVs annually by 2030, lifting transport's share of global power to 11.2% by 2050.

Industrial reshoring is concentrated but powerful. US manufacturing construction spending in Q1 2025 ran at over twice Q1 2022 levels. China's "new energy products" (solar modules, batteries, EVs) consumed >300 TWh in 2024, equivalent to Italy's entire annual consumption, accounting for 35% of China's industrial electricity demand growth 2022–2024.

Building electrification and cooling drove 60% of 2024 global growth (+600 TWh), per IEA. AC penetration in Indonesia is set to triple from <15% to 50% by 2035. US air-source heat pump sales rose 14% in 2024.

Hydrogen remains a small near-term electricity driver: 2025 low-emissions production ~1 Mt with announced 2030 pipeline at 37 Mtpa, down from 49 Mtpa in 2024, the first decline ever as BP, Ørsted and others cancel projects.

Regional fault lines

Region	2025 demand growth	Key 2030 dynamic
United States	+2% retail; PJM peak +32 GW by 2030	Data centers = 50% of growth; 13/23 NERC regions at elevated risk
China	+5.1% YoY (>9,500 TWh); 58% of global demand growth	Renewables capacity passed thermal in 2025 (1,840 GW wind+solar)
EU	+1.5% (2024); colder winter (2025)	Eurelectric: +60% by 2030; REPowerEU 69% renewable electricity goal
India	+5.8% (2024); 6.3% CAGR forecast	Peak load 148→250 GW (2014–2024); 500 GW non-fossil target by 2030
Latin America	65% clean electricity	2.3% CAGR through 2050; 239 GW new capacity needed by 2030 (World Bank)
MENA	Saudi/UAE leading	Saudi Vision 2030: 50% renewables (~130 GW); Middle East renewables additions +28.9% in 2025 (record)

ERCOT is the canary. Texas demand growth ran 7% in 2025 and 14% (forecast) in 2026. The large-load interconnection queue swelled to 233 GW by December 2025, ~4x year-end 2024 levels, with 77% from data centers seeking pre-2030 connection. ERCOT's preliminary peak forecast: 367 GW by 2032, more than triple today.

Grid as the binding constraint

The IEA estimates 20% of planned data center projects globally are at risk of delay because of grid bottlenecks. Globally, ~3,000 GW of renewable projects sit in queues, with 1,500 GW in advanced stages, five times the capacity actually added in 2022. The DOE's National Transmission Needs Study concludes US regional transmission must more than double by 2035, with interregional capacity expanding more than 5x. The follow-on NTP Study (October 2024) sets the buildout requirement at 2.1–2.6x by 2050 under current policy.

2. State of data centers globally

Capacity and trajectory

Synergy Research counted 1,360 hyperscale data centers at end-Q4 2025, up from ~750 in 2018, with operational capacity doubling in less than four years and forecast to double again in roughly three years. Total global DC capacity has reached ~100 GW (IEA). McKinsey projects expansion from 82 GW (2025) to 219 GW by 2030, requiring $5.2–6.7 trillion in cumulative investment through 2030.

Forecaster	2030 DC electricity demand
IEA Energy and AI (April 2025)	945 TWh (~3% global)
Goldman Sachs (April 2026 update)	1,350 TWh (+220% vs 2023)
McKinsey	~1,400 TWh (~4% global)
451 Research / S&P	1,587 TWh
Wood Mackenzie (bull case)	up to 4,500 TWh by 2050

Geographic concentration and constraints

Northern Virginia remains the world's largest hub at 4,040 MW operational (CBRE H2 2025) and 5.6 GW per JLL methodology, with Dominion Energy reporting 47 GW of contracted data center load by late 2025, up from 21 GW in mid-2024. Dominion's 2025 IRP projects peak load rising 70% through 2045. Dallas-Fort Worth crossed 1 GW in 2025 and is projected to double by year-end 2026. Total US primary-market supply hit 9,432 MW with record-low 1.4% vacancy. JLL projects $1 trillion of North American data center development through 2030, with 100+ GW of colocation/hyperscale potentially breaking ground.

International hubs are constrained. Dublin data centers consumed 22% of all Irish electricity in 2024, up from 5% in 2015; the de facto moratorium ended in December 2025 only with stringent new on-site generation requirements. Frankfurt's data centers consume ~42% of city electricity. Singapore awarded just 300 MW under its 2024 Call for Applications. Amsterdam logged zero new supply in 2024.

AI compute and power density

Rack power density has exploded. Legacy enterprise racks ran 5–10 kW; modern AI clusters reach 50–100+ kW; NVIDIA's GB200 NVL72 mandates 120–140 kW per rack with liquid cooling required. Industry analysts expect next-generation NVIDIA/AMD configurations to push past 200 kW per rack by 2027. Liquid cooling is now table stakes. Industry-average PUE sits at 1.56, with hyperscale leaders at 1.09–1.20. AI workloads are projected to grow 30% annually versus 9% for conventional servers, lifting AI's share of DC power from ~5–15% historically to 35–50% by 2030 (IEA).

Water is becoming a binding constraint in stressed regions. US data centers consumed ~17B gallons directly in 2023 (Lawrence Berkeley); indirect water from electricity generation added another 211B gallons. Phoenix, Texas, and Northern Virginia face acute exposure.

The deals reshaping the industry

The signature transactions of the past 18 months establish the new template:

• Microsoft–Constellation Three Mile Island/Crane Clean Energy Center (Sept 2024): 835 MW, 20-year PPA, restart targeted 2027–2028, $1B DOE loan announced November 2025. Microsoft pays roughly twice wholesale rate.

• Amazon–Talen Susquehanna: original 960 MW behind-the-meter ISA blocked by FERC (Nov 2024, upheld April 2025); restructured June 2025 to front-of-meter, 1,680–1,920 MW by 2032, ~$18B contract value.

• Microsoft–Brookfield Renewable (May 2024): >10.5 GW global framework agreement, ~8x the largest prior corporate PPA.

• Meta–Constellation Clinton (June 2025): 1,121 MW, 20 years.

• Meta–Vistra (January 2026): 2.6 GW across Perry, Davis-Besse, Beaver Valley, 20-year terms.

• Google–Kairos Power: 500 MW SMR Master Plant Development Agreement; first SMR online 2030.

• Amazon–X-energy: anchored $500M Series C-1 (Oct 2024); up to 5 GW by 2039.

• Meta–Oklo: 1.2 GW, online from 2030; Meta–TerraPower: 2.8 GW + 1.2 GW storage across 8 units.

• Stargate (OpenAI/Oracle/SoftBank/MGX, Jan 2025): $500B / 10 GW four-year US commitment; ~7 GW announced by September 2025, though The Information reported in early 2026 that the JV remains largely unstaffed with Oracle as the de facto developer.

• Amazon–Dominion North Anna SMR MOU: expanded June 2025 to co-develop 5 GW of nuclear.

The gas-nuclear-renewables stack

The IEA forecasts data center power through 2035 will be sourced roughly 50% renewables, with ~+175 TWh of natural gas and ~+175 TWh of nuclear (the latter heavily SMR-dependent from 2030). Coal still represents ~30% globally, primarily in China. The data make clear that the historical "100% renewables" framing is being quietly abandoned in favor of an "any clean and firm power" pragmatism, with hyperscalers signing 4.7 GW of nuclear PPAs in 2025 alone (BNEF).

3. Hyperscaler energy strategies

The capex tsunami

Hyperscaler capital expenditure has gone vertical. The Big Four's combined 2026 guidance approaches $600–630B, up from $251B just two years prior:

Company	2024 capex	2025 capex	2026 guidance
Amazon	$77B	$125B	~$200B
Microsoft	~$56B	~$90B	$110–120B
Alphabet	$52.5B	$91–93B	$175–185B
Meta	$39B	$72B	$115–135B
Big Four total	$251B	~$378–388B	~$600–630B

Synergy Research recorded $142B in single-quarter hyperscale capex in Q3 2025, up 180% versus Q3 2022. Roughly 75% of 2026 capex is AI infrastructure. Free cash flow strain is mounting: Morgan Stanley/BofA project Amazon FCF at −$17B to −$28B in 2026; Pivotal projects Alphabet FCF down ~90% to $8.2B. Meta's Hyperion campus in Louisiana ($50B+, 2 GW) is being financed via a $27–30B JV with Blue Owl, signaling hyperscalers' growing reliance on infrastructure capital partners precisely the segment GIP and EQT now lead.

PPA volumes and structural shifts

BloombergNEF's 1H 2026 Corporate Energy Market Outlook shows global corporate clean PPA volumes fell 10% to 55.9 GW in 2025, the first decline since 2015, but the decline masks a profound concentration. Big Tech accounted for 49% of global activity, with US unique corporate buyers down 51% YoY to just 33. Meta led globally at 10.24 GW; Amazon followed at 10.22 GW. 4.7 GW of Meta+Amazon volume was nuclear, reflecting the firm-power pivot.

Engie was the top global supplier at 3.6 GW; AES was #1 in the Americas for the fifth consecutive year (excluding storage). PPA terms have lengthened from a 10–12 year norm to 15–20 years (Microsoft-Constellation TMI, Meta-Vistra, Meta-Constellation Clinton all 20). Microsoft's "100/100/100" by 2030 goal and Google's 24/7 carbon-free energy framework are driving demand for hourly-matched and dispatchable clean power, materially changing what a "good" PPA looks like.

The "additionality" debate has effectively been settled in favor of pragmatism: Amazon and Microsoft remain disciplined on incrementality, but the inclusion of nuclear restarts (TMI), license extensions (Clinton), and existing-asset uprates (Beaver Valley) shows the bar has shifted from "new wind/solar" to "any contracted clean MWh on the system."

4. The generation mix evolution

Solar dominance accelerates

IRENA's Renewable Capacity Statistics 2026 records 510–511 GW of solar additions globally in 2025, up from 452 GW in 2024 and ~75% of all renewables. Cumulative solar PV reached 2,391 GW. Solar generated ~2,700 TWh in 2025 (8% of global electricity) and was the single largest contributor to global energy supply growth for the first time. BNEF projects 700 GW (2025), 753 GW (2026), 780 GW (2027) annual additions. Fixed-axis solar global benchmark LCOE was $39/MWh in 2025 (BNEF), with a further 30% reduction expected by 2035.

Battery storage becomes infrastructure

BNEF's 2H 2025 Energy Storage Outlook records 92–94 GW / 247 GWh deployed globally in 2025, +23–35% YoY. The US added 15 GW (+35% YoY). Pricing hit historic lows: stationary storage packs fell 45% YoY to $70/kWh, becoming the cheapest segment for the first time. LFP packs averaged $81/kWh, with the lowest observed system at $50/kWh. BNEF projects cumulative deployment to reach 2 TW / 7.3 TWh by 2035 (~8x 2025), with 4-hour system LCOE already below $100/MWh in six markets. AES is one of a small number of global developers with operational scale here.

Natural gas's unexpected resurgence

The single most consequential supply-side fact for the AI buildout: gas turbines are sold out. GE Vernova's backlog ended 2025 at ~80 GW; Q1 2026 backlog hit 100 GW; CEO Scott Strazik says reservations will be sold out through 2030 by end of 2026. Q1 2026 data center electrification orders hit $2.4B, exceeding all of 2025. Wood Mackenzie projects gas turbine prices rising 195% by 2027. Siemens Energy's record FY2025 orders ($19.5B in Q3, +65% adjusted) tell the same story; Mitsubishi is doubling capacity. Wait times for large turbines have stretched from 2–3 years to 5–7. Gas power M&A valuations have doubled since 2024; Vistra's ~$5.5 GW Cogentrix acquisition (~$4B) reflects this revaluation.

NERC's queues hold 53 GW of new gas in development.

Nuclear renaissance, with caveats

Global nuclear generation hit a record in 2025 and approaches 3,000 TWh in 2026 (IEA). Twelve GW of new construction started in 2025. Two US restarts headline the renaissance: Holtec's Palisades (Michigan) became the first-ever US restart of a closed nuclear plant in 2025, backed by a $1.5B DOE loan. Constellation's Crane Clean Energy Center (former TMI Unit 1) is targeting 2027 restart for Microsoft.

SMR landscape (April 2026):

Developer	MWe	Status
GE Vernova-Hitachi BWRX-300	300	Construction started Darlington (Ontario) May 2025; first unit operational 2030; TVA Clinch River permit accepted July 2025; $400M DOE grant Dec 2025
Kairos Power	35 (test) / 150 commercial	Hermes 2 (50 MW) is first US utility PPA for Gen-IV reactor (TVA, Sept 2025); Google 500 MW
TerraPower Natrium	345 (peak 500)	Non-nuclear construction at Kemmerer WY; operation targeted 2030
X-energy Xe-100	80 (4-pack 320)	Amazon-anchored $500M Series C-1; Dow Seadrift TX FID by 2028
NuScale	77	Only NRC-certified design; 6 GW deployment program with TVA/ENTRA1 (Sept 2025)
Holtec SMR-300	300	First two units at Palisades site by mid-2030

Trump's May 23, 2025 nuclear executive orders set targets of 400 GW nuclear by 2050 (4x current) and 10 large reactors under construction by 2030, with DOE-NRC reform directives (18-month max licensing, Reactor Pilot Program criticality by July 4, 2026). Execution risk remains acute: NuScale's UAMPS project was cancelled in 2023 after costs rose 75% from $5.3B to $9.3B; Vogtle's two units came in at 140% over original budget.

Coal decline pauses; geothermal rises

US coal retirements collapsed to just 2.6 GW in 2025, the lowest since 2010, versus 13.7 GW in 2022. DOE Section 202(c) emergency orders are keeping J.H. Campbell, Centralia, Schahfer and others online; the EPA proposed rolling back the Biden-era power plant rule in November 2025; 47 utilities received MATS exemptions through 2029. Globally, coal grew 0.4% in 2025 with India and Southeast Asia still expanding capacity.

Fervo Energy's enhanced geothermal is the breakout next-gen play, raising $462M Series E in December 2025 with PPAs for SCE (320 MW), Shell, and Google/NV Energy (115 MW), targeting 500 MW total at Cape Station by 2028. Eavor (Germany), Sage Geosystems and others follow. Fusion remains pre-commercial: Helion announced first private D-T fusion in February 2026 with a Microsoft 50 MW PPA targeting 2028 (widely viewed by analysts as aggressive); Commonwealth Fusion Systems' ARC plant in Virginia targets early 2030s with a Google PPA. Industry funding hit $9.7B in 2025 across 53 companies, but no fusion plant will be on the grid before 2030.

5. Grid infrastructure and transmission

The transmission shortfall is the most underappreciated chokepoint in the energy transition. The IEA estimates global grid investment must nearly double from ~$300B/yr to >$600B/yr by 2030, with 80 million km of grid added or refurbished by 2040 (equivalent to today's entire global grid). The EU's 2023 Grid Action Plan projects €584B (~$633B) needed by 2030. China's State Grid budgeted a record $75B in 2023 alone and plans 15 additional ultra-high-voltage lines by 2030.

US queues and reform

Berkeley Lab's Queued Up: 2025 Edition (December 2025) tallies ~10,300 active US projects: 1,400 GW generation + 890 GW storage = ~2,300 GW total. The composition: 956 GW solar, 890 GW storage, 271 GW wind, 136 GW gas (+72% YoY). Median time from interconnection request to commercial operation has stretched from <2 years (2000–2007) to >4 years (2018–2024). Of projects entering queues from 2000–2019, only 13% reached COD by end-2024; 77% withdrew.

FERC Order 2023 (the cluster-study reform) is in compliance phase across all RTOs through 2026. PJM's Transition Cycle 1 produced 17.4 GW of executed interconnection agreements (September 2025), with the first cycle of the new process opening April 27, 2026 (today). CAISO's new Interconnection Prioritization Process took effect June 2025. ERCOT enacted SB 6 in May 2025 imposing standards on large loads ≥75 MW.

Equipment supply chain

Wood Mackenzie's Q2 2025 analysis quantifies the squeeze: US is short 30% on power transformers and 10% on distribution transformers in 2025. Power transformer demand has risen 116% since 2019; lead times stretched to 128 weeks (2.5 years) for power and 144 weeks (2.8 years) for GSUs. Imports cover 80% of US power transformer supply. Domestic capacity expansions (Eaton $340M South Carolina, Siemens Energy $150M North Carolina, Hitachi Energy expansions) total ~$1.8B since 2023 but won't close the gap before 2027–2028. Average US large power transformer in service is 38 years old; >50% of distribution units are past expected service life.

PJM capacity market: the canary

PJM's capacity auctions document the binding constraint better than any forecast:

Delivery year	Clearing price
2024/25	$28.92/MW-day
2025/26	$269.92/MW-day (+833%)
2026/27	$329.17/MW-day
2027/28	$333.44/MW-day (price cap)

The 2027/28 auction failed reliability targets by 6,623 MW, the first time the entire RTO including FRR fell short. Without the Shapiro-negotiated price cap, clearing would have hit ~$530/MW-day. Data centers were responsible for 40% of the $16.4B auction cost.

6. Competitive landscape

The four blocs competing for the same opportunity space

US merchant nuclear/gas IPPs have become the AI-baseload winners. Constellation's $26.6B acquisition of Calpine closed January 7, 2026, creating a ~75 GW national generator. Vistra reached ~44 GW post-Energy Harbor with a 2.6 GW Meta nuclear deal and 5,500 MW of Cogentrix gas. Talen has the AWS Susquehanna deal worth ~$18B over its life. These names trade at 7–15x EV/EBITDA on hyperscaler PPAs. AES has no nuclear, structurally a different value proposition.

US regulated utilities exposed to data-center load are capturing the franchise growth: Dominion 47 GW contracted, Duke 13 GW build plan, Southern 50+ GW pipeline, Entergy 7–12 GW signed with Amazon/Google/Meta. These compete directly with AES Indiana and AES Ohio.

Pure-play and European renewables developers are the most direct competitors. NextEra remains the giant, with ~72 GW total, ~30 GW NEER backlog and 13.5 GW added in 2025, including the Duane Arnold restart for Google. Brookfield Renewable (Microsoft 10.5 GW deal, Neoen acquisition) is the most directly comparable platform competitor at scale. Iberdrola (46.2 GW renewables, target 60 GW by 2028), Engie (#1 global PPA seller in 2025 at 3.6 GW; 95 GW target by 2030), and Enel (~66 GW renewables) are AES's closest international peers.

Oil-major retreat is the defining 2025 story. BP cut renewables annual spend from $7B to $1.5–2B and divested 1.3 GW US onshore wind to LS Power. Equinor cut its 2030 renewables target to 10–12 GW (from 12–16) and took a ~$1B writedown on Empire Wind. Shell scaled back. Ørsted raised DKK 60B (~$9.4B) in an emergency rights issue in late 2025, divested 50% of Hornsea 3 and is cutting ~2,000 jobs. The implication for AES and its private capital sponsors is direct: distressed asset supply should be abundant through 2027–2028.

Comparative positioning

Company	Generation	Renewables operational	Hyperscaler/tech PPAs	Geographic edge
NextEra Energy	~72 GW	~30 GW backlog	~10.5 GW operating + backlog	US
Constellation Energy	~75 GW (post-Calpine)	Mostly nuclear+gas	Microsoft Crane PPA	US merchant
Vistra	~44 GW	6.4 GW nuclear	Meta 2.6 GW	ERCOT, PJM
Talen Energy	~10.7 GW	Limited	AWS $18B contract	PJM
AES Corporation	34,740 MW	17.9 GW; 67 GW pipeline; 11.8 GW signed PPAs	5.6 GW Amazon, 1.8 GW Google, 1.8 GW Microsoft	US + Andes
Brookfield Renewable	~34 GW operating	157+ GW global pipeline	Microsoft 10.5 GW	Global
Iberdrola	58 GW	46.2 GW	Top European PPA seller	Iberia, UK, US, LatAm
Engie	~100 GW	49 GW	#1 global PPA seller 2025	Global
Masdar	Sovereign	65 GW portfolio (Jan 2026); 100 GW by 2030 target	EM-focused	MENA, Africa, Europe

Private equity infrastructure: the new center of gravity

The Infrastructure Investor 100 ranks Brookfield ($104B raised 2020–2024), GIP/BlackRock ($86B), KKR ($82B), and EQT in the top tier. Stonepeak ($48.6B) closed Castrol in December 2025 and is targeting Fund V at $15B. Blackstone Infrastructure Partners' $11.5B acquisition of TXNM Energy (May 2025) is the analogous take-private to AES. CalPERS and CDPQ are now direct co-underwriters; CDPQ took ~30% of AES Ohio in late 2025. Top 10 fund closes captured 46% of 2025 infrastructure capital raised, indicating extreme concentration in mega-managers, exactly the pool that just bought AES.

7. Regulatory and policy landscape

The Trump II energy agenda

The Trump administration's energy policy reorientation is the single largest regulatory variable through 2028. Key actions since January 2025:

• Day-one EOs (Jan 20, 2025): Lifted the LNG export pause; declared a national energy emergency; froze offshore wind leasing on the entire OCS; revoked NEPA implementing regulations.

• One Big Beautiful Bill Act (signed July 4, 2025): Wind/solar must begin construction within 12 months of enactment (by July 4, 2026) to keep full §45Y/§48E credits, or be placed in service by Dec 31, 2027. Storage co-located with wind/solar is exempt from the placed-in-service deadline. EV credits terminated September 30, 2025. Hydrogen §45V deadline pulled forward to January 1, 2028. FEOC "material assistance" rules apply to projects starting construction after 2025, with cost-ratio thresholds escalating annually. Treasury baseline raises ~$468B over 10 years from IRA reforms.

• Four nuclear EOs (May 23, 2025): 18-month max NRC licensing; target 400 GW nuclear by 2050; ≥3 pilot reactors to achieve criticality by July 4, 2026; 5 GW power uprates and 10 large reactors under construction by 2030.

• Tariffs: Section 201 solar tariffs expired February 6, 2026, but Section 301 China tariffs on poly/wafer/cells doubled to 50%; reciprocal tariffs added 10% baseline. AD/CVD on Cambodia/Vietnam/Thailand/Malaysia solar imports finalized April 2025 at up to 271%. Chinese BESS effective tariff rose to ~55% as of January 2026, peaking near 82% earlier in 2025. Section 232 expanded to 50% copper (August 2025).

• NEPA: CEQ rescinded all implementing regulations (effective April 2025; final January 2026). Seven County SCOTUS decision (May 2025) substantially narrowed scope. OBBBA §112 halves NEPA review periods for project sponsors who pay a fee.

• DOE LPO renamed Office of Energy Dominance Financing (October 2025); refocused on nuclear, oil/gas, coal, critical minerals. Largest LPO loan in history ($26.5B to Southern Company subsidiaries, October 2025). $25.8B BIL + $4.3B IRA undeployed; $6.8B BIL + $2.5B IRA already cancelled.

• Offshore wind has lost in court 5 of 5 times through April 2026, but DOI re-paused 6 GW (~Vineyard Wind 1, Revolution Wind, CVOW 2.6 GW, Sunrise Wind, Empire Wind 1) on December 22, 2025 citing national security.

EU and China

The EU is on track to deliver Fit for 55 with REPowerEU adding a 45% renewables aspiration. The 2040 target was provisionally agreed in late 2025 at -90% GHG vs 1990. CBAM entered its compliance phase January 1, 2026, with the first quarterly certificate price set at €75.36/t CO2 (April 7, 2026). EU ETS cleared at €83.79/t in December 2025 (+30% YoY); BNEF forecasts €149/t by 2030. ETS2 (buildings, road transport) launches January 2027.

China's 15th Five-Year Plan was adopted in March 2026 with a softer carbon-intensity target of -17% (vs -18% in 14th FYP, which itself missed at -12.4%). 2025 China wind+solar additions exceeded 430 GW (record); cumulative 1,840 GW now exceeds thermal capacity for the first time. China retains ~80% of global solar PV manufacturing and ~70% of wind turbine and EV battery manufacturing.

State-level diversity

Twenty-eight US states operate RPS or clean-energy mandates. California (100% clean by 2045) and New York (70% by 2030, 100% zero-emission by 2040) anchor the aggressive end. Virginia's Clean Economy Act remains the dominant signal in the Mid-Atlantic data center market. Texas has no RPS but added more solar and storage than any other state in 2025. The federal-state divergence is widening; 18 state AGs are litigating Trump offshore wind decisions.

8. Strategic opportunity analysis through 2030

The 2030 strategic opportunity for AES (and any global power company with comparable scale) decomposes into twelve vectors. Each is sized, time-stamped, and assessed on competitive intensity.

8.1 Hyperscaler co-location and dedicated generation

TAM: $300–500B in cumulative US PPA value through 2030. The 11.8 GW of AES PPAs already signed with Amazon, Google and Microsoft is the foundation; the 67 GW development pipeline represents the upside. With Big Tech now 49% of global PPAs and 4.7 GW of 2025 PPAs in nuclear (BNEF), the firm-power premium is real. Co-location remains regulatorily fraught after FERC's 2024–2025 Talen/AWS rulings, but hybrid front-of-meter structures with dedicated interconnection are the new standard (the restructured Amazon-Talen deal is the template). AES's existing relationships are its highest-value asset; competitive intensity is acute (NextEra, Brookfield, Engie, Constellation), so the moat is land + queue position + permit + transformer slot, not generation technology.

8.2 Behind-the-meter solutions, private wires, microgrids

TAM: $50–100B by 2030. With PJM's 2027/28 capacity at the price cap and 4-year average grid connection wait times, hyperscalers are increasingly willing to pay for behind-the-meter solutions that bypass the queue entirely. The Meta-Hyperion Louisiana model ($50B+, 2 GW, Blue Owl JV) is illustrative. Distinct from grid-connected PPAs, this is a build-own-operate utility business that AES Indiana and AES Ohio can extend with scaled solar+storage+gas microgrids. Competitive intensity: moderate; few players have BTM capability at GW scale.

8.3 Battery storage at scale

TAM: 220 GW / 972 GWh annual deployments by 2035 (BNEF); cumulative 2 TW / 7.3 TWh. US 2025 deployments hit 15 GW. Storage pack pricing has collapsed to $70/kWh (-45% YoY), making 4-hour LCOE sub-$100/MWh in six markets. AES is one of perhaps five Western developers with operational scale here. Risk: NIMBY post-Moss Landing (January 2025 fire damaged 55,000 NMC modules; California AB 303 introduced) plus FEOC rules that may exclude Chinese cell suppliers from credits in 2026 onward. Strategic priority: lock in non-Chinese cell supply (LG, Samsung SDI, Panasonic, Tesla 4680) to preserve credit eligibility.

8.4 Grid services and ancillary markets

TAM: $30–50B by 2030. PJM's capacity prices and ERCOT's ORDC pricing make ancillary services (frequency regulation, voltage support, black start) increasingly lucrative for storage and dispatchable assets. FERC Order 2222 implementation (PJM 2027–28 effective; MISO June 2029; SPP September 2025; NYISO end-2026) opens DER aggregation. Competitive intensity: high (Voltus, Enel X, Octopus, AutoGrid), but AES's owned-asset base is the differentiator.

8.5 SMR and nuclear partnerships

TAM: $200–400B in SMR capex through 2035 if commercialization holds; binary outcome. AES has no nuclear today. Recommended posture: partnership rather than balance-sheet exposure through a Kairos/X-energy/BWRX-300 supply agreement structure tied to specific hyperscaler customers. This preserves optionality without taking first-of-a-kind capex risk (NuScale UAMPS canceled at 75% cost overrun; Vogtle ran 140% over budget). Trump's 400 GW by 2050 target and DOE Reactor Pilot Program (criticality by July 4, 2026) create a forced timeline.

8.6 AI for energy operations

TAM: $20–40B/year by 2030 (predictive maintenance, dispatch optimization, asset health). AES's June 2024 partnership with Andrew Ng's AI Fund and AES Next venture (which produced Haven Safety AI, launched February 2026) is the foundational platform. Initial use cases include drone inspection, vegetation management, transformer health monitoring, predictive maintenance, and load forecasting. The opportunity is captive: deploy across the AES fleet first to prove ROI.

8.7 Building AI products to sell to other utilities

TAM: $10–25B by 2030 if AES productizes its software stack. This is the only Tesla-style optionality on the AES balance sheet. The path: validate Haven Safety AI and adjacent products on owned assets, then license to other utilities (Tesla's progression from auto OEM to FSD/Autopilot platform is the analogue, but in regulated power). Competitive intensity is moderate: GE Vernova GridOS, Hitachi Lumada, Schneider Electric EcoStruxure, Siemens Spectrum Power, AutoGrid (Schneider) all have plays, but no incumbent has both the AI fund pipeline and the operational scale of AES. This is the highest-multiple opportunity in the AES portfolio under private ownership.

8.8 Latin America renewable buildout (AES Andes platform)

TAM: $200B+ in LatAm renewables capex through 2030. With European majors retreating, LatAm renewables represents an underappreciated AES advantage. The region holds 65% clean electricity already (vs 43% global average); Brazil reached 86%+ renewable; Chile targets 90% by 2030; Mexico is opening its grid to IPPs. World Bank estimates 239 GW of new capacity needed by 2030. AES Andes (Chile) and AES Brasil are platforms few competitors can match in country knowledge or installed base. Risks: 2024 Brazilian drought (worst since 1950s) hit hydro hard; Colombia is losing gas self-sufficiency from 2030.

8.9 Virtual power plants and DERs

TAM: $50B+ globally by 2030; FERC Order 2222 catalyzes US. VPPs are the residential-and-C&I-distributed analog to grid-scale storage. AES's utility footprint (AES Indiana, AES Ohio, El Salvador) creates a captive testbed. Competitive intensity: high (Tesla, Sunrun, AutoGrid, OhmConnect, Span), but utility-anchored aggregators have the regulatory path of least resistance.

8.10 Industrial electrification partnerships

TAM: $100–200B by 2030. Beyond data centers, the next industrial electrification wave includes green hydrogen for steel (where it survives), EV manufacturing, critical mineral processing, and chip fabs. The IRA-era manufacturing buildout (Q1 2025 manufacturing construction at >2x Q1 2022) needs dedicated power. AES has done this template in Latin America (mining customers). Replicating in US Sun Belt manufacturing corridors is logical extension.

8.11 LNG infrastructure and gas as bridge fuel

TAM: $100B+ in US LNG export and gas-fired generation through 2030. Trump lifted the Biden LNG pause on day one; US liquefaction stands at ~15 Bcf/d operating, ~17 Bcf/d under construction, and ~19 Bcf/d FERC/DOE-approved. Gas turbines are sold out through 2030. Strategic posture: maintain optionality without overweighting; gas is a transition asset whose terminal value compresses post-2035 as renewables + storage + nuclear + geothermal stack improves. AES's existing thermal/LNG infrastructure positions it to participate without doubling down.

8.12 The Substrate Strategy: picks and shovels of AI

The unifying thesis. AES is best understood not as a power company competing on $/MWh but as the physical substrate of the AI economy, the company that delivers electrons, transformers, switchgear, software and grid services to the hyperscaler buildout. Under private ownership, AES can take longer-cycle bets (5–10 year payback), tolerate negative near-term FCF on growth capex, and partner aggressively with infrastructure capital pools (Blue Owl/Meta-style JVs) to scale beyond its own balance sheet. The substrate framing aligns the AES Next AI fund, the Andes platform, the storage scale, and the utility footprint into a single narrative: own the layer hyperscalers cannot vertically integrate.

9. Risks and headwinds

Tariffs and trade

Critical risk. The cumulative tariff stack on solar and storage is the single largest near-term project economics threat. Effective tariffs on Chinese BESS reached ~55–82% by early 2026; AD/CVD on Southeast Asian solar reached up to 3,500% on circumventers. Wood Mackenzie now ranks the US as the most expensive country in which to build solar. LevelTen Q1 2026 reported solar PPA prices up 13% YoY and wind up 24%.

Supply chain bottlenecks

Critical risk. Gas turbines are sold out through 2030; large power transformers run 2.5-year lead times; GSUs at 2.8 years. CEO of GE Vernova: backlog will be sold out through 2030 by end of 2026. Without secured turbine slots and transformer orders, projects are de facto blocked.

Interest rate and financing

Moderate-to-high risk. Fed funds at 3.50–3.75% (March 2026), with only one more cut signaled in 2026. OBBBA's accelerated tax credit phase-out forces a "race to safe harbor" by July 4, 2026. Tax equity supply tightening. PPA prices reflecting cost-plus inflation.

Labor

Moderate-to-high. USITC projects a US shortage of 360,000 welders by 2027; BLS projects 11% electrician growth needed but supply lagging. Hiring timelines for skilled trades doubled from 8 weeks to 4 months. Mass deportation policy further compresses available pool. A modern data center requires ~800 electricians; a 100 MW solar farm needs ~50.

Community opposition

High risk and worsening. Sabin Center (June 2025): 459 counties with severe restrictions (+16% YoY); 498 contested projects in 49 states (+32% YoY). Vistra's Moss Landing fire (January 16, 2025) damaged 55,000 of 100,000 NMC modules and triggered the largest BESS cleanup in EPA history; California AB 303 was introduced as a result. Loudoun County, Virginia, ended by-right zoning for data centers in March 2025 — a watershed in the world's largest DC market. $98B of data center projects have been blocked or delayed through 25 cancellations.

Cybersecurity

High and structural. Volt Typhoon, the Chinese MSS-affiliated APT, has had persistent access to US critical infrastructure for 5+ years. Dragos's 2025 report: "still very active...mapping out and embedding in US infrastructure." Salt Typhoon compromised 9+ US telecoms with 1–2 year dwell time. NERC's president called grid cyber risk a "five-alarm fire" in August 2025.

Climate and physical risks

High. Hurricanes Helene + Milton (Sept-Oct 2024) caused ~$115B combined damage; Helene knocked out power for 6M customers across 10 states. Eaton Fire (January 2025, LA) potential SCE liability $8–9B+. Latin American hydropower hit by 2024 drought, the worst Brazilian drought since the 1950s; Ecuador imposed nightly blackouts.

Renewables sector stress

Moderating but real. SunPower (August 2024), Sunnova ($8.9B debt, June 2025), and Solar Mosaic (June 2025) all filed Chapter 11. Residential solar installations fell 31% in 2024. Offshore wind has seen >$5B in impairments and >5 GW of cancelled or delayed capacity (Ørsted Ocean Wind, Avangrid Commonwealth Wind, Equinor Empire Wind 1, Sunrise Wind).

Stranded asset risk

Easing. US 2025 coal retirements were just 2.6 GW, the lowest since 2010. DOE Section 202(c) emergency orders are extending plant lives; EPA proposed rolling back the Clean Power Plan rule in November 2025. Trump policy is actively encouraging coal retention for AI/firm power needs. Long-term retirements may resume but on uncertain trajectory.

10. Synthesis: where the biggest opportunities lie

The 2028 and 2030 base case

By 2028, three structural facts will define the global energy landscape. First, data center electricity demand will have roughly doubled from 2024 levels, putting it on a trajectory to ~1,000 TWh by 2030 in the IEA central case and potentially 1,350 TWh in Goldman's revised view. Second, gas turbines will remain physically rationed, with delivery slots sold out and prices ~2x 2023 levels. Third, the first SMRs will be on the cusp of commercial operation (BWRX-300 Darlington in 2030; Kairos Hermes 2 in 2026; X-energy Cascade in early 2030s), but well below the volumes needed to materially displace gas.

By 2030, the picture sharpens further. Renewables plus nuclear together approach 50% of global generation (IEA Electricity 2026); solar PV is the single largest source of supply growth; battery storage is at 2 TW cumulative globally; the US has built more transmission than any decade since the 1970s, but still short of NERC reliability targets in multiple regions; PJM and ERCOT capacity prices remain elevated, sustaining merchant generation economics; and the take-private wave that swept AES, TXNM and others has substantially repriced the entire IPP space upward.

Top strategic priorities

1. Lock the hyperscaler relationships into 20-year platform agreements, not project-by-project PPAs. The Microsoft-Brookfield 10.5 GW framework and the Amazon-Dominion 5 GW MOU are the templates. AES has the customer base; private capital provides the multi-decade balance sheet. Aim to convert the 11.8 GW signed PPAs into 25+ GW framework agreements by 2028.

2. Build the AI software platform as a sellable product. Haven Safety AI is the proof of concept. The strategic ambition is a five-year path to a utility software business sold to peers globally, modeled on Tesla's AI optionality. The AES Next/AI Fund LP position is unique among utilities; lean in.

3. Scale battery storage aggressively in 2026–2027 while incumbents lead. Pack pricing is at historic lows, but FEOC rules and tariffs will harden after 2027, and NIMBY post-Moss Landing will compress permittable sites. Lock supply, lock land, lock interconnection.

4. Defend the Andes platform; opportunistically acquire from European retreaters. Iberdrola, Engie, BP, Shell and Equinor are all selectively divesting Latin American assets. AES's country knowledge and installed base is an underweighted moat.

5. Build SMR optionality through customer-anchored partnerships. Avoid first-of-a-kind balance-sheet exposure (NuScale's 75% overrun is the cautionary tale). Pursue Kairos, X-energy, BWRX-300 partnerships tied to specific hyperscaler customers, with private capital co-funding.

6. Position for the gas/LNG bridge through 2032 without overweighting. AES's existing thermal infrastructure participates without doubling down. Gas's terminal value compresses post-2035.

7. Treat transmission, transformers, queue position and skilled labor as the new alpha. The constraint is no longer cost of capital or LCOE; it is physical execution. Pre-purchase transformer slots; secure long-cycle EPC contracts; partner with vocational pipelines.

What's overhyped vs. underhyped

Overhyped. SMR commercial deployment before 2030 (only Darlington likely; first US SMRs probably 2031–2033). Fusion at grid scale before 2030 (Helion's 2028 Microsoft target is widely viewed as aggressive). Hydrogen as a 2030 electricity demand driver (project pipeline contracted in 2025). US offshore wind through 2028 (5 court losses for the Trump administration but DOI re-pausing under national security; investors should not underwrite to 2027 COD). Residential solar (post-NEM 3.0, post-tariff, post-bankruptcy wave).

Underhyped. Gas turbine bottleneck as a binding constraint on AI buildout. Transformer supply chain stress lasting through 2028. Battery storage as a generation-class asset rather than a balancing tool. The gravitational pull of infrastructure private equity in repricing the entire IPP sector. Latin American renewables as European competitors retreat. Geothermal (Fervo) as the next firm-clean-power technology to surprise to the upside. Cyber as a structural cost that compounds on every utility's P&L.

Conclusion: the substrate strategy

The most consequential observation in this report is that the hyperscalers are now investing more annually in physical infrastructure than the entire US utility industry combined. The Big Four's $600B+ 2026 capex is roughly 5x annual US utility capex. Hyperscalers are vertically integrating into power (Amazon-Talen, Microsoft-Constellation, Meta-Vistra) only because incumbent power supply cannot keep up. That vertical integration has natural limits: hyperscalers are consumers of electricity, not operators of grids. Someone has to own the substrate.

AES's strategic opportunity through 2030 is to be that someone, in markets where its customer relationships, geographic footprint, AI partnerships and battery storage scale uniquely qualify it. The take-private by GIP/EQT/CalPERS/QIA is precisely the capital structure required: long-duration, patient, infrastructure-mandated, well-aligned with 20-year customer contracts. The execution risks (tariffs, supply chain, labor, NIMBY, cyber, regulatory whiplash) are real and quantified. But the demand side has decoupled from any prior baseline. For the first time in three decades, the binding constraint on the global power system is not capital or technology but physical buildout, and that is exactly the constraint a private, capitalized, well-positioned platform like AES is built to relieve.