Private equity investment in US data centers hit a record $45.7 billion in 2025. That was the highest total in at least five years and made up 72 percent of all data-center deals.
The momentum has carried straight into 2026 as hyperscalers and sponsors rely on SPVs and joint ventures to fund the AI buildout.
The AI buildout is reshaping US infrastructure spending.
At the same time, public spending on transportation came in at $49.9 billion.
This means data center construction spending has outpaced government
The deal everyone still points to is Meta's Hyperion data-center campus in Louisiana. In October 2025 Meta and funds managed by Blue Owl Capital set up a joint venture through a special purpose vehicle. The SPV raised roughly $27 billion in high-grade debt plus equity. PIMCO and BlackRock anchored the debt. Meta owns just 20 percent of the project. Blue Owl owns the rest. Meta builds and runs the campus and pays a long-term lease.
In plain terms, the big financing and debt stay inside the SPV instead of landing on Meta's main balance sheet. Meta skips a giant one-time cash hit and pays steady rent over time. The risks stay locked inside the SPV, and ownership can change hands cleanly without touching Meta's books.
Meta followed the same playbook again this spring. In early May reports surfaced on a new roughly $13 billion SPV financing for its El Paso, Texas campus using Morgan Stanley and JPMorgan. The project has already been scaled up significantly, with Meta targeting 1 gigawatt of capacity by 2028.
The model is spreading fast, and SPVs are also opening up innovative new uses. Private-credit giants like Apollo and Blackstone recently teamed with Broadcom on a $36 billion SPV to finance Anthropic's purchase of Google chips — a structure that lets AI companies access critical hardware without traditional balance-sheet strain. Similar vehicles are being explored for on-chain liquidity and tokenized exposure to private AI assets, giving accredited investors cleaner, faster ways to participate in the buildout.
Venture teams keep spinning up co-invest SPVs for opportunistic AI deals outside their main fund mandates. These let limited partners jump into high-conviction opportunities without changing fund economics. Volume stays strong because speed and simplicity win when infrastructure and application deals move fast.
Here is why SPVs keep delivering conviction right now:
Risk stays isolated to that one asset.
Access opens to deals that once required full fund commitments.
Founders see just one new line on the cap table.
Closes happen faster with custom terms and tax-efficient setups.
Everything stays transparent with no commingled assets.
Bottom line: SPVs are the professional-grade tool powering the AI infrastructure supercycle. They give precise exposure with built-in protections, real capital at work, and scalable efficiency.
The public markets just delivered another reminder that AI infrastructure is the defining trade of the cycle.
This week AI stocks exploded higher. The clear winners were the companies supplying the actual hardware, memory, and networking backbone.
Hewlett Packard Enterprise (HPE) rocketed more than 25 percent after crushing fiscal Q2 earnings. Revenue hit a record $10.7 billion (up 40 percent year-over-year), server revenue alone jumped about 33 percent to $5.5 billion, and the AI infrastructure business is now the clear growth engine. HPE raised full-year guidance, pulled forward its long-term financial targets by two years, and reported a record backlog with orders more than doubling year-over-year. Investors are finally treating HPE as a pure AI play.
That's +$90 BILLION in market cap since Jensen Huang's comment.
Marvell Technology (MRVL) soared more than 45 percent, its biggest one-day gain ever, after Nvidia CEO Jensen Huang stood onstage at Computex and called it "the next trillion-dollar company." Huang highlighted Marvell's custom AI chips and connectivity semiconductors that hyperscale data centers need to link thousands of chips without bottlenecks. The endorsement sent the stock to record highs and confirmed its key role in the AI stack.
This stock was worth $60 billion just 13 months ago.
Memory was the real headline. Micron (MU) broke $1,000 a share and crossed a $1 trillion market cap, now around $1.14 trillion, the fastest company in history to reach that mark after a more than 300 percent run over the past year. Samsung hit a record high and SanDisk an all-time high on the same wave. The driver is a genuine supply crunch: AI data centers are on track to consume roughly 70 percent of global memory supply this year, with high-bandwidth memory sold out well into next year.
Nvidia held near highs, while Super Micro Computer and Dell also climbed on the same wave. Broader indices pushed higher on this concentrated strength in AI infrastructure.
'The Pricking Is Coming': Dalio Warns AI Bubble Will Burst Like Dot-Com, But Tech Will Endure
ZeroHedge - On a long enough timeline, the survival rate for everyone drops to zero
This moment echoes the dot-com boom. Back then the biggest gains came from the companies building the internet's plumbing before the killer apps fully arrived. Today the capital is flooding into the picks-and-shovels providers of AI servers, memory, networking chips, and data-center hardware. Real earnings and guidance are backing the move. No hype. Just surging demand meeting real supply constraints.
Whether you lead a syndicate, run a fund, or operate a business, an SPV gives you a clean single-asset vehicle to move fast and stay protected. Our top 10 uses, from the most common to the most frontier:
Co-invest vehicles. Your fund finds a great deal but wants to put in more than the fund alone can. You spin up an SPV so trusted LPs can write extra checks into that one company, without changing how your main fund works.
Cross-border stacks. Your investors sit in different countries with different tax and legal needs. The SPV is set up in the jurisdiction that fits them best, whether that's Delaware for US investors or BVI for a global, crypto-native group.
Real estate and RWA. Hold a single property or income-producing real-world asset in its own SPV, so investors get clean exposure to that one deal and the risk stays sealed off from everything else.
Pro-rata defense. One of your early bets is taking off and raising a big new round. A standalone SPV lets you put in fresh money to keep your percentage from shrinking, without having to raise a whole new fund to do it.
Employee liquidity. Startup employees and early shareholders often hold stock they can't sell. An SPV pools buyers together to purchase those shares, giving sellers cash and giving your investors a position in a proven company.
Continuation and secondary plays. You've held a position for years and some backers want their money back, but you still believe in the company. You move that stake into a new SPV: early investors can cash out, and those who want to stay roll forward.
Emerging-manager warehousing. You want to start a fund but don't have a track record yet. Running a few deals one at a time through SPVs lets you build a real, provable history before you ask LPs to commit to a blind-pool fund.
Energy and infrastructure. Big physical projects like data centers or power carry real risk. Holding each one in its own SPV means a problem with that single asset stays contained and can't spill over onto everything else you own.
Retirement capital access. There's over $15 trillion sitting in US retirement accounts that mostly can't touch private deals. A Self-Directed Roth IRA can invest into an SPV, so that long-term money can back startups and secondaries, with gains compounding tax-free.
Compute access SPVs. AI companies need expensive hardware like GPUs and TPUs but don't want to buy it all outright. An SPV finances the hardware separately, so the company gets access without a giant hit to its own balance sheet.
Isolated risk, faster closes, custom terms, no commingled assets.
Looking to set up an SPV for AI, infrastructure, or any high-conviction private deal? Contact us at GCRx and we'll get you the details.
Invest together.
Best,
Arthur and the GCRx Team