Altimeter Capital Management Lp

The biggest buys chart reads like a blueprint for owning every profitable bottleneck in AI compute. New positions and adds cluster tightly around semiconductors and specialized infrastructure, indicating Altimeter wants the toll booths on AI rather than just the applications.

• NVDA: Increasing Nvidia by 15.3% and lifting it to 28.57% of the book is a blunt statement that the fund still sees upside despite a 264.5% gain vs its average cost. They are not trading around a winner; they are upgrading it to the portfolio’s defining asset.

• ARM: The new 4.55% position in Arm, worth $259.5M, plugs a clear gap in Altimeter’s AI architecture exposure. Arm’s CPU and IP footprint gives them a royalty-style claim on edge and data-center designs that Nvidia alone cannot cover.

• TSM and AVGO: Adds of 12.0% to Taiwan Semi and +108.8% to Broadcom round out the semiconductor complex. Together, these moves say Altimeter is betting not just on GPUs, but on the broader explosion in high‑end, AI‑grade compute and networking.

• CRWV and META: The 40.0% add to CoreWeave and a fresh 5.9% bump to Meta expand the downstream side of the thesis. CoreWeave is an AI‑first cloud, and Meta is a hyperscale buyer and builder of AI infrastructure — both extend the same bet on GPU scarcity and model proliferation.

• UBER and AXON: Away from chips, the 42.6% increase in Uber and a new 1.11% stake in Axon show a taste for software‑levered, real‑world networks that can monetize AI in transportation, logistics, and public safety rather than in generic SaaS.

The sells this quarter are not a repudiation of tech, but a re‑ranking of which tech assets deserve to sit next to Nvidia and Meta. Altimeter is clearly harvesting liquidity from large, profitable platforms and one underperforming software name to fund a narrower, more leveraged AI stack.

• MSFT and AMZN: Microsoft and Amazon remain large at 7.69% and 7.64%, but share counts are down 7.3% and 5.7%, respectively. With gains vs average cost of 66.0% for Microsoft and 45.6% for Amazon, these trims look like classic funding trades: recycle capital from relatively “de‑risked” hyperscale platforms into higher‑beta AI infrastructure.

• SNOW: The 5.0% reduction in Snowflake, which is sitting at a -39.0% loss vs Altimeter’s average buy, is more telling. This is not profit‑taking; it is patience wearing thin on a high‑multiple data platform that has not lived up to the rest of the AI complex.

• HOOD: A 30.4% cut to Robinhood, despite an 81.2% gain vs cost, signals that consumer trading volume and brokerage monetization are now second‑tier ideas. Capital is shifting away from cyclical retail flow and toward structural AI demand.

In aggregate, these trims show Altimeter using liquid, mega‑cap and financials exposure as an internal capital market. If a name does not sit at the heart of AI compute or a dominant network, it is fair game as a funding source.

On the sector chart, the story is rotation within tech, not a step away from it. Technology’s share is roughly flat at just over 80%, but that headline hides a decisive tilt from broad software into semiconductors and AI‑tied infrastructure.

Within tech, incremental dollars flowed to Nvidia, Arm, Taiwan Semi, Broadcom, CoreWeave, and Meta, while Snowflake and Microsoft were net sources of cash. The fact that the largest software trim is also the biggest loser underscores a discipline: Altimeter will protect capital for names where the AI payoff is clearer and nearer.

Outside tech, the real estate label on Uber is a data quirk — functionally, Uber is a global, two‑sided logistics and mobility network. Its move from 8.25% to 10.06% weight gives Altimeter a non‑chip expression of AI in the physical world. Consumer exposure via Amazon is nudged down, and financials via Robinhood are cut from 1.84% to 1.09%, freeing capacity to launch a new 1.11% position in Axon.

Net-net, sector rotation is about culling peripheral themes. Consumer discretionary and finance weights shrink, while an industrial name that is effectively a software and data business (Axon) enters alongside a denser cluster of chip and infra plays.

Put together, this 13F paints Altimeter as an investor that wants to own the rails of AI more than the apps. The combination of a 28.57% Nvidia position, a new 4.55% Arm stake, and increased exposure to Taiwan Semi, Broadcom, CoreWeave, and Meta signals a view that the non‑linear profits from AI will accrue first to those who manufacture, design, and deploy compute at scale.

The trims in Microsoft, Amazon, Snowflake, and Robinhood show a willingness to treat even beloved growth platforms as disposable funding when they are not at the sharpest edge of that thesis. Altimeter is not exiting these names; it is insisting they earn their place against a surging pipeline of AI infrastructure opportunities.

Uber and Axon hint at the second act of the strategy: once the chips and data centers are in place, demand will flow into real‑world networks that can embed AI into transportation, commerce, and public safety. These are still smaller positions, but the direction is clear.

Investors reading this 13F should not see a generic “tech hedge fund.” They should see a concentrated, high‑conviction vehicle that is comfortable being wrong in public for a few quarters if it means owning what it believes will become the enduring profit pools of the AI era.