EQT's $2.6 Billion Copia Power Deal Shows Capital Rotating Into AI's Power Layer

I want you to notice what EQT is actually buying. Not chips. Not GPUs. Not a hyperscaler's balance sheet. It's buying the power. Copia has 2.6 gigawatts of generation and storage already operating or under construction, plus a development pipeline exceeding 9 gigawatts of grid-connected capacity aimed at data centers, according to EQT's own announcement. That same release puts Copia's broader solar and storage pipeline above 25 gigawatts, with another 7 gigawatts of natural gas capacity behind it. This is a bet on electrons, not silicon. And it tells you something about where the AI capital cycle is headed next.

The capital rotation nobody is pricing in yet

For two years, the AI investment story ran through semiconductors and hyperscaler capex. Nvidia's order book, Microsoft's data center spend, OpenAI's compute deals. That story is still real. But it has matured, and mature stories attract second-order capital. Sophisticated infrastructure managers are no longer betting on who wins the AI model race. They're betting that whoever wins will need more electricity than the grid currently has, and that owning the power generation, storage, and delivery assets underneath that demand is a better risk-adjusted trade than owning the compute layer itself.

Carlyle didn't build Copia as an AI play. It built Copia in 2021 as a power development platform, well before "AI data center power crunch" was a phrase anyone used. EQT is buying it in 2026 explicitly as "a leading integrated power and AI infrastructure platform," per its own press release. The underlying assets didn't change much. The framing did. That's the tell. Capital that used to chase power projects for utility-scale returns is now chasing the same projects for AI-scale returns, and it's paying AI-scale multiples to get there.

This matters for you as an accredited investor because it's a preview of where large infrastructure funds are putting new dollars, and those same themes are increasingly available in smaller-check vehicles built for individuals, not just sovereign wealth funds and pensions.

What makes infrastructure its own asset class, mechanically

Infrastructure investing isn't private equity with hard hats. The mechanics are different, and the differences are why institutions treat it as a distinct allocation bucket rather than a subset of buyout funds.

Three things define the asset class. First, long-duration, contracted cash flows. Power purchase agreements, regulated utility rate structures, and toll concessions often run 15 to 30 years, so an investor is underwriting a cash flow stream, not a growth story. Second, inflation linkage. Many contracts include explicit escalators tied to CPI or negotiated rate adjustments, which is part of why infrastructure held up better than growth equities during the 2022 rate shock. Third, an illiquidity premium. You are locking up capital in assets that can't be sold in a day, and in exchange, you're compensated with a return premium over liquid public infrastructure equivalents like utility stocks.

The fundraising numbers show how much capital now wants exposure to that combination. EQT set a €21 billion target for EQT Infrastructure VII in May 2026, roughly $24.5 billion, according to EQT's own disclosure. Its predecessor, EQT Infrastructure VI, closed in March 2025 at €21.5 billion, itself a 35% jump from Fund V's €15.7 billion close in 2021, according to Connect Money's reporting. That's roughly 37% cumulative growth in flagship fund size across two vintages in four years, in an asset class that used to be considered slow-moving. Fund VI was reportedly 75-80% invested by the time the Copia deal closed, which is why EQT needed a new vehicle in market. The Copia acquisition is expected to be one of the first deals for Fund VII, landing while the fund is still in its 0-5% invested stage.

Here's the comparison that matters for your own allocation decisions.

Notice the tradeoff column. Every step down in minimum investment and eligibility requirement costs you liquidity certainty. Interval funds promise quarterly repurchase windows, not guaranteed exits. That distinction is not a technicality. It's the entire risk premium you're being paid to accept.

The broader evergreen fund category, which includes interval funds and non-traded BDCs, still represents a small slice of private markets overall, roughly 5% of the roughly $700 billion evergreen segment relative to total private markets assets, according to Hamilton Lane research cited in JD Supra's July 2026 private markets update. That same research forecasts evergreen structures could reach 20% of private markets within a decade. If that forecast holds, the gap between what institutions can access through flagship funds like EQT Infrastructure VII and what individual accredited investors can access through interval funds and BDCs should narrow, not because minimums drop further, but because more sponsors build feeder structures specifically to capture individual investor demand for themes like AI power infrastructure.

Copia Power, and who EQT thinks the winners are

Copia Power's history is worth knowing before you evaluate the theme. Tenaska, a private power developer, built the pipeline that became Copia. Carlyle formed Copia as a standalone platform in 2021, backing management to develop and operate power generation with data centers as an anchor customer base. Five years later, per Capacity Media's reporting, Carlyle is exiting into a buyer with an explicit, multi-company AI infrastructure strategy. EQT isn't buying Copia in isolation. It already owns EdgeConneX, a data center operator. It owns Zayo, a fiber and connectivity backbone. It owns Cypress Creek Energy, a renewable power developer, and Scale, another infrastructure platform in its broader portfolio. Copia slots in as the power leg of a four-legged stool: power generation, data center real estate, fiber connectivity, and renewable development, all under one sponsor's ownership umbrella. EQT's stated intent, per its own release, is to have these portfolio companies collaborate directly. The goal is feeding power and connectivity into the data center buildout rather than each company sourcing those inputs from unrelated third parties.

Ray Henger and Alex Darden lead the deal team on EQT's side, with Pooja Goyal, who chairs EQT's infrastructure advisory business, publicly framing the acquisition as a continuation of the firm's thesis that AI infrastructure demand requires owning the full physical stack, not just one layer of it. That framing is convenient for a firm trying to raise $24.5 billion, but it also matches what utilities themselves are reporting. Dominion Energy, Georgia Power, and PJM Interconnection have all flagged data center load growth as the primary driver of new capacity requests in their most recent regulatory filings, which is the demand-side evidence that makes EQT's bet plausible rather than purely promotional.

That's a specific, falsifiable bet, and it's worth naming as such. EQT is betting that vertical integration across the AI infrastructure stack, power plus data centers plus fiber, produces both cost advantages and pricing power that a standalone power developer or standalone data center operator can't capture alone. If that's right, the Copia acquisition is cheap relative to the value EQT can extract by pairing it with EdgeConneX's real estate and Zayo's fiber network. If that's wrong, EQT overpaid for what is fundamentally still a power development company with commodity economics and long permitting timelines. You should watch how EQT actually cross-sells these assets over the next 24 months before assuming the integration thesis works.

The buildout-bubble, regulatory, and lockup risks you need to weigh

Three distinct risks sit inside this theme, and they don't move together, so don't let one reassurance paper over the other two.

The AI infrastructure buildout itself could be overbuilt relative to actual compute demand. Every hyperscaler is currently guiding to multi-year capex increases, and every infrastructure fund is underwriting power deals assuming that demand curve holds. If AI model training and inference demand growth decelerates, some of the 9 gigawatts in Copia's development pipeline may never get built, or may get built and sit underutilized. This isn't hypothetical. Data center vacancy and power curtailment have both happened in prior tech infrastructure cycles, most notably in the fiber buildout of the late 1990s, where capacity outpaced demand for the better part of a decade.

Power markets and regulation carry their own separate risk. Interconnection queues in most U.S. regions currently run three to seven years, according to Department of Energy grid studies, and utility commissions in states like Virginia and Georgia have already pushed back on rate structures that shift data center power costs onto residential customers. A regulatory decision that caps how fast new large loads can connect to the grid, or that changes who pays for grid upgrades, directly affects the revenue assumptions behind deals like this one. Natural gas capacity additions also carry emissions policy risk that can shift with a single administration change.

Then there's the structural risk specific to how you, as an individual accredited investor, actually access this theme. You are not buying into EQT Infrastructure VII directly unless you can write an eight-figure check. Your access point is an interval fund, a non-traded BDC, or a similar evergreen vehicle. Those vehicles promise quarterly liquidity through repurchase offers, but that liquidity is not guaranteed. Fund documents explicitly state that repurchase offers can be oversubscribed, in which case you get pro-rata partial liquidity, not full redemption, and boards can suspend repurchase programs entirely during periods of market stress. Treat the quarterly repurchase language as a best-effort mechanism, not a contractual exit right.

Pricing risk deserves its own mention too. Interval funds and non-traded BDCs price at net asset value, calculated by the fund's adviser, not by a public market. That NAV is only as good as the marks the adviser puts on illiquid power and data center assets that don't trade daily. If a fund's underlying power assets get marked using aggressive assumptions about future data center demand, and that demand doesn't materialize on schedule, you could see a markdown concentrated in a single quarter rather than a gradual price discovery process like you'd get with a public stock. Ask any fund sponsor how often independent third-party valuations inform their NAV, and how large the swings have been in prior quarters.

How to actually get exposure to this theme

If the infrastructure-for-AI thesis is compelling to you, there are three realistic paths as an accredited investor, and they sit at different points on the liquidity-versus-access spectrum.

• Interval funds with infrastructure mandates. Hamilton Lane converted its Private Infrastructure Fund to an interval fund structure in April 2026, with minimums as low as $2,500 in some share classes and quarterly repurchase offers covering at least 5% of outstanding shares, per Hamilton Lane's own announcement. The fund's stated focus includes power and energy alongside telecom and transportation.
• Non-traded BDCs and digital infrastructure vehicles. Products like Blue Owl's digital infrastructure fund and CION Grosvenor's infrastructure-focused vehicle target the data center and power buildout more narrowly. Read the underlying portfolio concentration before committing. A fund that's 40% exposed to a single hyperscaler's data center pipeline carries different risk than a diversified power and connectivity fund.
• Direct co-investment through your RIA or wealth platform. Some private wealth platforms now offer feeder access into slices of flagship infrastructure funds, at lower minimums than a direct LP commitment but still generally requiring $100,000 to $250,000 and multi-year lockups.

Whatever you choose, read the fee load and the repurchase mechanics in the prospectus before you look at the return projections. According to the Investment Company Institute's 2026 research, total assets across interval funds, tender offer funds, and BDCs reached $534 billion at year-end 2025, up from $140 billion five years earlier. That growth means more product choice, but it also means more products competing for your allocation with varying fee structures and varying quality of underlying deal access. Fast growth in a fund category is not the same as fast growth in fund manager skill.

Verify the sponsor's track record on actual exits, not just deployed capital. Carlyle's roughly fivefold return on Copia is a genuine data point in favor of the power infrastructure thesis. It is not proof that every power infrastructure deal from here performs the same way. Ask any fund you're considering how many realized exits it has, at what multiples, and over what holding periods, before you commit capital you can't easily get back.

Author Disclosure: Jeff Barnes, MBA has no personal position in any company, fund, or platform named in this article. Angel Investors Network has no current commercial relationship with any party mentioned. AIN provides marketing and education services, not investment advice. Past performance does not guarantee future results. All investments involve risk, including loss of principal.