Brad Gerstner’s $6.7B Altimeter Capital Cuts China, Adds CoreWeave and Bloom Energy

Altimeter Capital Management, the Boston hedge fund run by tech investor and BG2 podcast co-host Brad Gerstner, disclosed a $6.66 billion equity portfolio in its February 13F filing, built around concentrated bets on US AI mega-caps and the picks-and-shovels of AI infrastructure.

On February 17, 2026, Altimeter Capital Management filed its quarterly 13F-HR (the SEC form U.S. institutional managers use to disclose their stock holdings 45 days after each quarter end) reporting $6.66 billion in equity positions across 18 names as of December 31, 2025. The Boston-based fund, founded by Brad Gerstner in 2008 and run as a concentrated long-only book on technology winners, made its biggest portfolio shift in two years during the fourth quarter. The Q4 book is a consolidation story.

That story has two parts. The first is a quiet exit from China, with Alibaba, PDD Holdings, and ARM Holdings all sold during the quarter alongside several smaller names. The second is a tilt into the same AI-power infrastructure trade that has anchored other institutional books this winter. Both parts read as a single decision. Pull capital out of names where Gerstner has lower conviction, and route it into a smaller set of US AI plays he can sit with for years.

From $7.6B to $6.7B in One Quarter

Altimeter’s reported equity book stood at roughly $7.58 billion at end-September 2025 and fell to about $6.66 billion at end-December, a 12 percent reduction. The decline came almost entirely from position exits rather than price drops in the names that stayed. Gerstner closed eight positions during the quarter. Alibaba ($173 million), Arm Holdings ($155 million), PDD Holdings ($181 million), Synopsys ($127 million), Maplebear ($122 million), Gemini Space Station ($38 million), Pattern Group ($10 million), and Netskope ($7 million) all left the book. He also cut Broadcom by about 95 percent and Robinhood by roughly 60 percent. The largest US AI mega-cap positions Nvidia, Meta, Microsoft, and Amazon were each held or modestly added to, leaving the surviving book more concentrated than the one that came before it.

The Top 5 Positions and the Thesis Behind Them

Five names account for about 65 percent of the disclosed portfolio. Nvidia is the largest at $1.51 billion. Nvidia designs the GPU compute (the specialized chips AI training runs on) that every frontier-AI lab is racing to buy. Meta Platforms is second at $1.22 billion. Meta is both an AI customer through its Llama model line and a hyperscaler (Amazon, Microsoft, Google, Oracle, Meta and a handful of frontier-AI labs) building its own data-center capacity. Microsoft is third at $618 million, the cloud arm hosting OpenAI workloads. Amazon is fourth at $511 million, almost entirely a bet on AWS. Uber is fifth at $457 million, the lone non-AI name in the top tier and a position Altimeter has held since the company’s pre-IPO rounds. New positions added in Q4 are smaller but tell the same story. CoreWeave entered at $230 million, a bet on the GPU-rental capacity that lets frontier labs lease compute by the hour. Shopify entered at $92 million. Bloom Energy entered at $23 million for on-site fuel cells (devices that generate electricity on-site without burning fuel) that let data centers bypass the grid entirely.

Why This 13F Is Trending Now

One day ago, Aschenbrenner’s $5.5 billion Situational Awareness Fund disclosed the same two anchor positions in CoreWeave and Bloom Energy that Gerstner added in Q4. Read together, the filings show two unrelated managers reaching the same picks-and-shovels conclusion in the same quarter. AI compute needs physical power and rentable GPU capacity faster than the grid or hyperscalers can deliver. Gerstner is the more visible of the two on retail finance Twitter and through his BG2 podcast with Bill Gurley, where he has spent the past year calling the AI infrastructure build-out the trade of the decade. The Q4 13F is the document version of those views.

What to Watch

The May 15 deadline: Altimeter’s Q1 2026 13F is due by mid-May. Watch whether Gerstner held the new CoreWeave and Bloom Energy stakes after their Q1 rallies, or trimmed them as Aschenbrenner-aligned holdings ran higher.

New position discipline: The Q4 book added three names and exited eight. If Q1 reverses that pattern and adds more than it exits, that signals Gerstner is rotating capital back into single-name bets rather than consolidating around mega-caps.

The power supply story: The Bloom Energy stake is small at $23 million but symbolically large. If Gerstner doubles or triples it in Q1, the read is that he sees the on-site power thesis lasting beyond the current AI capex cycle.