Steve Cohen's Point72 Just Dropped $2.8 Billion on AI Infrastructure

Steve Cohen built his reputation as one of the sharpest stock pickers alive. But his latest 13F filing tells a story that goes well beyond stock picking — it's a structural bet on the companies building the backbone of the artificial intelligence era.

Point72 Asset Management's Q1 2026 13F, filed May 15, shows a portfolio worth $78.1 billion as of March 31. Buried inside that filing is a cluster of new positions totaling roughly $2.8 billion, all pointed at a single theme: AI infrastructure.

What the Filing Actually Shows

The five headline buys from Q1:

• Taiwan Semiconductor (TSM) — $870 million new position
• Nvidia (NVDA) — $510 million
• Amazon (AMZN) — $492 million
• Equinix (EQIX) — $487 million (position grew ~467%)
• Broadcom (AVGO) — $439 million

Each of these is a pillar of the AI supply chain. TSMC manufactures the chips Nvidia designs. Nvidia makes the GPUs that run model training and inference. Amazon's AWS is one of the dominant buyers of GPU compute. Equinix operates the interconnected data centers where cloud traffic flows. Broadcom supplies the networking silicon inside those data centers.

Cohen isn't betting on any single AI company. He's betting on the entire hardware and infrastructure layer underneath them.

Why Infrastructure Over Models?

There's been fierce debate in the investment world about where AI value will ultimately accrue: at the model layer (OpenAI, Anthropic, Google DeepMind), the application layer, or the infrastructure layer. Cohen's portfolio makes a clear statement.

Infrastructure is the safer bet for public market investors because it captures revenue regardless of which AI models or applications win. Whoever trains the next frontier model will need TSMC's fabs, Nvidia's GPUs, and Amazon's cloud capacity. Whoever deploys it at scale will flow traffic through Equinix. The picks-and-shovels logic that drove the first dot-com infrastructure boom is playing out again — except this time the volumes are far larger and the capital requirements more intense.

Point72's internally managed Turion Fund — its dedicated AI investment vehicle — reportedly returned 30% on similar positioning. Cohen appears to be scaling that thesis into the broader flagship portfolio.

META and the Advertising Angle

Beyond the infrastructure cluster, Point72 held a substantial Meta Platforms (META) position worth roughly $518.8 million. Meta is a different kind of AI beneficiary: its AI investments flow directly into ad targeting, Reels recommendations, and AI-powered creative tools. Advertisers are seeing measurably better returns, and Meta's revenue per user is climbing as a result.

Owning both Nvidia (the chip) and Meta (the application) is a hedged way to play AI — if the infrastructure buildout slows, consumer AI adoption may still drive Meta's numbers. If model scaling continues, Nvidia wins big.

What Cohen Exited

Not every position was a winner worth holding. Point72 completely exited TE Connectivity, Texas Roadhouse, and AstraZeneca during the quarter, while opening new positions in Chevron (CVX) and PPL Corp. The exits from Texas Roadhouse (restaurants) and AstraZeneca (pharma) suggest a deliberate rotation away from defensive-yield plays toward higher-conviction growth themes.

The entry into Chevron is interesting: it may reflect a macro hedge against energy prices, or a view that energy demand from data centers — which now consume enormous quantities of power — will keep commodity prices elevated longer than the market expects.

Reading Point72's Moves Alongside Other Funds

Cohen isn't alone in this positioning. The Q1 13F season showed a broad pivot among major hedge funds toward the AI infrastructure stack. InvestorLens's overlap tool lets you see exactly where top funds are converging — and the data shows unusual consensus around semiconductors and data-center REITs like Equinix.

When a fund with Point72's research depth and trading sophistication makes a $2.8 billion cluster bet on a single theme, it's worth paying attention to what the pattern says about where institutional conviction is running strongest. You can track all of Point72's latest disclosed positions, and compare them to other major funds, on InvestorLens's investor pages.

The Bigger Picture

One quarter's 13F is a snapshot, not a forecast. Cohen's positions reflect holdings as of March 31, and the portfolio may have already shifted. Point72 runs a high-turnover multi-strategy book — individual positions can appear or disappear in a single quarter.

But the scale of this cluster is hard to dismiss. Nearly $3 billion in coordinated new buys across five closely related AI infrastructure names suggests this isn't a casual tactical trade. It looks more like a conviction call that the infrastructure buildout cycle has years, not quarters, left to run.

The macro consensus signals on InvestorLens show how this kind of institutional positioning is shaping the broader smart-money outlook. And if you want to track new capital flows as the Q2 13Fs hit this summer, the flow tracker is worth bookmarking.

---

Data sourced from public SEC 13F filings. Educational research only — not investment advice.