Gerstner Slashed Broadcom 96% and Dumped Alibaba to Build a $6.7B AI Infrastructure Portfolio: Inside Altimeter's Q4 2025 Conviction Rotation

Brad Gerstner just executed one of the most aggressive portfolio rotations in recent 13F history. Altimeter Capital's Q4 2025 filing reveals he slashed Broadcom (AVGO) by 96% — from $245.7 million to just $11.1 million — while simultaneously exiting Alibaba (BABA) entirely and deploying over $2.4 billion into eight brand-new positions. The result: a portfolio that grew 42% to $6.66 billion and expanded from 14 to 18 holdings, all while pivoting decisively toward AI infrastructure and U.S.-domiciled software platforms.

This isn't a trim-around-the-edges filing. Gerstner liquidated his entire China exposure, abandoned his semiconductor diversification thesis in Broadcom, and rebuilt a third of his portfolio from scratch in a single quarter. The message is unmistakable: he's betting everything on the AI infrastructure super cycle — and he wants that bet concentrated in American companies he can touch.

TL;DR — Altimeter's Q4 2025 Moves at a Glance

• AUM surged 42% from $4.70B to $6.66B, driven by new capital inflows and position additions
• Broadcom slashed 96% — sold 712,647 of 744,776 shares, reducing from $245.7M (5.23%) to $11.1M (0.17%)
• Alibaba fully exited — all 969,100 shares ($173.2M) sold. Also exited ARM Holdings, PDD Holdings, Instacart, and Synopsys
• 8 new positions built totaling $2.44B: Microsoft ($618M), Amazon ($511M), Snowflake ($445M), TSMC ($371M), CoreWeave ($230M), Robinhood ($146M), Shopify ($92M), Bloom Energy ($23M)
• NVIDIA remains #1 at $1.51B (22.68%) with 455K shares added (+6%)
• Coupang nearly doubled — added 5.65M shares (+56%) to reach $370M position
• Meta trimmed minimally — just 20.6K shares sold (-1.1%), weight fell from 29% to 18% as portfolio diversified
• Portfolio diversified from 59.5% top-2 concentration (Q3) to 40.96% (Q4)
• China exposure: zero. Alibaba, PDD, and all Chinese-adjacent positions eliminated
• Q4 fund return was -4.92% despite AUM growth, suggesting growth came from inflows rather than performance

Filing Snapshot

Metric	Q3 2025	Q4 2025	Change
Total AUM	$4.70B	$6.66B	+42%
Number of Holdings	14	18	+4
Top Holding	NVDA (30.35%)	NVDA (22.68%)	Weight reduced via diversification
Top-2 Concentration	59.5%	40.96%	-18.5pp
New Positions	—	8	MSFT, AMZN, SNOW, TSM, CRWV, HOOD, SHOP, BE
Full Exits	—	4+	BABA, GEMI, PTRN, NTSK (+ ARM, PDD, CART, SNPS)

What He Bought: $2.4 Billion in New Conviction

Gerstner didn't nibble. He deployed over $2.4 billion into eight new positions in a single quarter — an extraordinary build-out that reshaped the entire portfolio.

Microsoft ($618M) — The AI Platform Anchor

Microsoft (MSFT) entered as Altimeter's third-largest holding at $617.8 million (9.27% weight). This isn't a momentum trade — it's a thesis on enterprise AI adoption. Azure is the delivery mechanism for OpenAI's models, GitHub Copilot is rewriting how software gets built, and Microsoft 365 Copilot is the Trojan horse for AI in every enterprise. Gerstner is betting that the company collecting the toll on enterprise AI deployment is as important as the company making the chips.

Amazon ($511M) — Cloud + Commerce Convergence

Amazon (AMZN) came in at $511.4 million (7.68%). AWS remains the largest cloud platform globally, and Amazon's logistics infrastructure — same-day delivery, robotic fulfillment — creates a moat that's measured in physical assets, not just code. For Gerstner, this is the intersection of AI compute (AWS Bedrock, custom Trainium chips) and real-world commerce logistics.

Snowflake ($445M) — The Data Layer for AI

Snowflake entered at $444.8 million (6.68%), marking a significant re-entry for Gerstner who had previously exited the stock. The timing isn't coincidental: Snowflake announced a $200 million partnership with OpenAI in February 2026, positioning its data platform as the connective tissue between enterprise data warehouses and large language models. Revenue hit $1.14 billion in Q2 at +32% year-over-year. Gerstner is betting that AI without structured data is useless — and Snowflake owns the data layer.

Taiwan Semiconductor ($371M) — The Irreplaceable Foundry

TSMC entered at $370.5 million (5.56%). If NVIDIA designs the brains of AI, TSMC manufactures them. There is no alternative at the leading edge of chip fabrication. This position completes Gerstner's AI infrastructure stack: NVIDIA for design, TSMC for fabrication, CoreWeave for deployment.

CoreWeave ($230M) — GPU-as-a-Service for the AI Boom

CoreWeave entered at $230.1 million (3.45%). This is Gerstner's most distinctive bet — a GPU rental company that has built 33 data centers housing over 250,000 NVIDIA GPUs, with a revenue backlog exceeding $55 billion in multi-year contracts from major AI labs. Microsoft is 62-71% of CoreWeave's revenue, which creates both opportunity and concentration risk. Gerstner backed CoreWeave since its private stage, and this public position reflects conviction, not discovery.

Robinhood ($146M), Shopify ($92M), Bloom Energy ($23M)

The smaller new positions reveal additional thesis threads. Robinhood at $146.3 million (2.20%) is a fintech platform bet riding the crypto and retail trading wave. Shopify at $91.8 million (1.38%) is a founder-led commerce platform with a durable SMB moat — fitting Gerstner's global commerce bucket. Bloom Energy at $22.7 million (0.34%) is the portfolio's lone clean energy position, a fuel cell company that could benefit from data center power demand.

What He Sold: The China Exit and Broadcom Demolition

Alibaba — Complete Exit ($173M → $0)

Gerstner sold all 969,100 shares of Alibaba, eliminating a $173.2 million position (3.69% of Q3 portfolio). He also exited PDD Holdings, effectively zeroing out his entire China tech exposure. The reasoning is straightforward: US-China tech decoupling has made these positions unquantifiable risks. Tariff tensions, regulatory unpredictability, and the potential for sudden delisting or sanctions make China tech uninvestable for a fund manager who needs to explain risk to LPs. This wasn't a valuation call — it was a regime-change call.

Broadcom — 96% Slash ($245.7M → $11.1M)

This is the headline move. Gerstner sold 712,647 of his 744,776 Broadcom (AVGO) shares — a 96% reduction that took the position from $245.7 million (5.23%) to just $11.1 million (0.17%). Why keep 32,129 shares instead of exiting entirely? It could be a placeholder for re-entry, a tax lot consideration, or simply a statement that Broadcom isn't bad — it's just not NVIDIA. With NVDA at 22.68% and AVGO at 0.17%, Gerstner has made his semiconductor conviction unmistakably clear.

ARM Holdings, Instacart, Synopsys — Clean Sweep

Beyond the 13F's explicit exits of Gemini Therapeutics ($38.3M), Patron Technology ($9.9M), and NightHawk Biosciences ($7.4M), web research confirms Gerstner also exited ARM Holdings, Instacart (CART), and Synopsys. The ARM exit is particularly telling: at valuations stretched to perfection, NVIDIA offers a purer and more liquid AI silicon bet. Instacart and Synopsys were likely position-sizing decisions as capital was redeployed to higher-conviction names.

The Gerstner AI Playbook: Three Big Ideas

In a recent CNBC interview, Gerstner stated: "We think we're really early in the super cycle" for software and cloud companies benefiting from AI. His Q4 portfolio maps directly to three interconnected theses:

1. AI Infrastructure — The Picks and Shovels ($2.12B, ~32%)

NVIDIA designs the GPUs. TSMC fabricates them. CoreWeave rents them out. This vertical stack captures value at every layer of AI compute. Gerstner isn't just betting on AI — he's betting on the physical infrastructure that makes AI possible. The Broadcom slash reinforces this: he doesn't want diversified semiconductor exposure, he wants concentrated AI silicon exposure.

2. Software Platforms — Where AI Value Accrues ($2.59B, ~39%)

Meta, Microsoft, Amazon, Snowflake, Uber, Confluent — these are platform companies embedding AI into existing products with massive distribution. Meta's AI-powered ad targeting, Microsoft's Copilot suite, Amazon's Bedrock and Alexa+, Snowflake's data-AI bridge, Uber's logistics optimization, Confluent's real-time data streaming. Gerstner's thesis: the companies that already have the users and the data will capture most of the economic value from AI.

3. Global Commerce — Structural Growth Markets ($870M, ~13%)

Coupang (Korea), MercadoLibre (Latin America), Shopify (global SMB), and Grab (Southeast Asia). These are commerce platforms in markets where e-commerce penetration is still growing. The Coupang increase (+56%) is the standout — Gerstner nearly doubled his position, signaling deep conviction in Korean e-commerce growth.

What Analysts Might Misread

"AUM grew 42% — Gerstner is crushing it"

Not exactly. Altimeter's Q4 performance was actually -4.92% according to tracking data. The 42% AUM growth from $4.70B to $6.66B was driven primarily by new capital inflows, not investment returns. The gap between AUM growth and fund performance suggests Altimeter attracted roughly $2.2 billion in new capital during Q4 — a massive vote of confidence from LPs even as the portfolio dipped. This distinction matters: the new positions weren't funded by profits, they were funded by fresh capital.

"Exiting China means he's bearish on Chinese companies"

It means he's bearish on the ability to own Chinese companies from a U.S.-domiciled fund without unacceptable regulatory risk. Alibaba and PDD may be excellent businesses — Gerstner's exit is about regime risk, not business quality. An investor following this signal should distinguish between "this company is bad" and "this position is uninvestable for this fund."

"CoreWeave is a speculative AI play"

At 3.45% of a $6.7B portfolio, this is a measured bet, not speculation. CoreWeave has $55 billion in contracted revenue backlog and 33 operational data centers. The real risk isn't business viability — it's customer concentration, with Microsoft representing 62-71% of revenue. If Microsoft builds competing capacity or renegotiates terms, CoreWeave's economics shift dramatically.

"The Broadcom stub position means he might buy back"

Maybe, but the more likely read is tax optimization or administrative convenience. At 0.17% of the portfolio, Broadcom is effectively a rounding error. Don't anchor on the 32,129 remaining shares — anchor on the 712,647 shares he sold.

Frequently Asked Questions

Why did Gerstner slash Broadcom 96% instead of exiting completely?

The most likely explanations are tax lot management (spreading realized gains across quarters), maintaining a placeholder for potential re-entry, or simply the mechanics of unwinding a large position through block trades. At $11.1 million in a $6.66 billion portfolio, the remaining position is functionally irrelevant — representing just 0.17% of assets. The signal is in the 712,647 shares sold, not the 32,129 retained.

Is Altimeter's 42% AUM growth sustainable?

The growth was driven by capital inflows rather than investment performance (the fund returned -4.92% in Q4). Whether inflows continue depends on Gerstner's ability to deliver returns on the new capital. The eight new positions represent his investment thesis for the next cycle — if AI infrastructure spending accelerates as he expects, the deployed capital should compound. But LPs who allocated in Q4 are buying his thesis, not his track record for that quarter.

Should retail investors follow Gerstner into CoreWeave?

CoreWeave offers genuine exposure to AI infrastructure demand, but comes with concentration risk that hedge funds can manage better than retail investors. With Microsoft at 62-71% of revenue, a single client decision could move the stock materially. The $55 billion backlog provides visibility, but retail investors should size this as a speculative allocation — which is exactly what Gerstner did at 3.45% of his portfolio.

What does the Alibaba exit signal about Chinese tech stocks broadly?

Gerstner's exit — combined with PDD and all China-adjacent positions — signals that he views U.S.-China regulatory risk as unquantifiable. This doesn't mean Chinese tech stocks can't perform well. It means that for a U.S.-based institutional manager, the tail risks (delisting, sanctions, data sovereignty conflicts) make position sizing impossible. Retail investors with higher risk tolerance and no LP obligations face a different calculus.

How does Gerstner's portfolio compare to other prominent investors?

Interestingly, Nancy Pelosi owns five of the same stocks as Gerstner: NVIDIA, Microsoft, Amazon, Alphabet, and Broadcom. The overlap speaks to a broad institutional consensus around mega-cap AI beneficiaries. Where Gerstner diverges is in his willingness to take concentrated positions in mid-cap names like CoreWeave, Snowflake, and Coupang — bets that institutional consensus hasn't fully formed around yet.

What is Gerstner's background and investment philosophy?

Brad Gerstner founded Altimeter Capital in 2008 during the financial crisis. He also hosts the BG2Pod podcast with legendary venture capitalist Bill Gurley. His investment philosophy centers on identifying technology platforms with durable competitive advantages during periods of structural change — exactly the framework driving his current AI infrastructure thesis. His willingness to make dramatic portfolio rotations (like this Q4 overhaul) distinguishes him from more incremental institutional managers.