The Biggest Private Market Deals Reveal Where AI Investing Is Headed

Five deals sum it up. Anthropic closed a $65 billion Series H, pricing the company at roughly $965 billion — essentially pre-IPO market-cap territory. Anduril pulled $5 billion for autonomous defense manufacturing. Reflection AI took $2.5 billion to automate knowledge work, already paired with SpaceX on compute. Long Lake secured $2.25 billion while executing an acquisition of American Express Global Business Travel. Baseten rounded at $1.5 billion for its inference deployment platform.

Total committed capital across these five rounds: $76.25 billion. The common thread is not "we're building a better chatbot." It's owning the pipes.

Where the money actually went

Strip out the AI label and examine the capex allocation. Anthropic is building cloud-scale model infrastructure and positioning for public markets. Anduril is pouring capital into manufacturing capacity and next-gen military hardware — physical stuff, not software margins. Baseten sells deployment tooling to enterprises that don't want to build their own ML ops stack. These are infrastructure and platform plays with recurring revenue mechanics, not viral consumer apps hoping for a $20/month subscription.

Long Lake's move is the clearest signal of convergence. It's not waiting for organic adoption — it's acquiring a mature enterprise services business (Amex GBT) and injecting AI workflows directly. That's a private equity playbook with an AI layer on top. VC returns, PE tactics, infrastructure scale.

The counter-signal from traditional PE

While venture capital accelerates into AI-native infrastructure, traditional private equity is slamming the brakes on sectors exposed to AI disruption. Wealth management deal volume in the UK fell from 193 transactions in 2023 to 157 in 2025, per Solve Partners data. Investment committees are running harsher screens on whether target businesses can survive AI-driven fee compression over a standard 3–5 year hold.

Nordic Capital, backing Ascot Lloyd, now explicitly scores "moat against AI competition." Co-head Emil Anderson put it bluntly: business models aren't fully disrupted yet, but that window is closing. The February tech sell-off, triggered in part by Altruist's AI tax planning tool, hammered shares in St. James's Place, Quilter, and Rathbones — proof that even listed incumbents price in displacement risk overnight.

One advisory source described the sentiment as a "blood bath in certain sectors" that happened "very, very quickly." Leveraged buyout funds now face a fundamental question: can you underwrite a decade of earnings stability when AI could collapse fee structures?

What to watch

Forge's Q2 readout makes one structural point worth stress-testing: large private rounds give companies the runway to fortify competitive moats ahead of liquidity events. The bet is that owning infrastructure — compute, inference, deployment, manufacturing — creates compound defensibility.

For investors, the practical filter is straightforward. If a company sells shovels to the miners, watch it. If it's a miner hoping the gold rush lasts, model the downside. That distinction applies to your public portfolio too — consumer tech pricing events like Best Buy's Apple shopping weekend move hardware at the margin, but the real alpha sits in the infrastructure layer underneath.

No guarantee these deals produce market leaders. Late-stage rounds have a mixed track record — capital alone doesn't build durable businesses. But the direction of private capital flow is a data point, not a narrative. And the data says: the biggest fortunes in AI will belong to those who own the rails, not the trains running on them.