Tiger Global Increases NVIDIA Stake in Q3, Significantly Cuts TSMC and Google
nashnova research
Tiger Global pushed Nvidia to its top holding in Q3 while slashing over a third of its TSMC stake and 42% of Alphabet — with total portfolio value down nearly 16%, the fund is concentrating its bets on AI chips themselves, not the equipment and platforms around them.
What changed in the overall portfolio?
As of September 30, 2026, Tiger Global's 13F filing shows total reported holdings of roughly $20.17 billion, down 15.88% from about $23.98 billion last quarter, across 45 positions.
This means → the portfolio is shrinking, but not evenly — some names were cut hard, others were added to. The real story is in the weight shifts.
Turnover hit 14.29%; the top ten holdings account for 66.75% of the portfolio. In plain terms = nearly seven dollars out of every ten sit in just ten stocks. Tiger Global still runs a high-conviction book.
Why did Nvidia rise to the top position?
Nvidia: roughly 12 million shares, worth about $2.74 billion, or 13.59% of the portfolio. Share count rose 7.15% quarter-on-quarter; weight climbed 4.24 percentage points — the single largest buy this quarter.
This means → Tiger Global didn't just hold Nvidia — it actively added shares while the total portfolio was shrinking, pushing it from number two to number one.
By contrast, Amazon ranks second (about $2.31 billion, 11.46% weight), but its share count actually fell 4.18%. The weight gain is passive — driven by price appreciation and a smaller denominator, not fresh buying.
Who got cut the hardest on the sell side?
Alphabet (Google's parent) was trimmed from roughly 5.81 million shares to about 3.35 million — a 42.31% cut, with weight dropping 2.94 percentage points. That makes it the quarter's single largest sell.
TSMC shares fell 35.79% to about 3.13 million; weight dropped 2.63 percentage points, the second-largest sell. This reflects a shift from "AI's upstream supply chain" toward "AI itself."
Spotify was virtually liquidated — from about 1.17 million shares down to just 2,561, a 99.78% drop. Semiconductor-equipment names Lam Research and Applied Materials were also cut by 30.67% and 26.40%, respectively.
Besides Nvidia, what else was bought?
Cerebras Systems — an AI chip company — saw its stake rise 48.26% to about 4.45 million shares, entering the top ten at 3.92% of the portfolio. This means → Tiger Global isn't just betting on Nvidia in AI chips; it's also backing a challenger.
Danaher — a life-sciences and diagnostics firm — surged from about 456,000 shares to roughly 1.27 million, up 178.25%, a major position build.
MercadoLibre, the Latin American e-commerce platform, was added by 59.75%, lifting its weight by 1.01 percentage points — one of the few non-AI additions this quarter.
What do the new positions and full exits signal?
Three new positions: Italian luxury carmaker Ferrari (about $237 million, 1.17% of the portfolio), Italian consumer-software company Bending Spoons (about $20 million), and Canadian quantum-computing firm Xanadu Quantum (about $4.5 million).
In plain terms = Ferrari is an off-beat luxury-consumption play; Xanadu Quantum's position is tiny (0.02% weight), more a toe-in-the-water than a conviction bet.
Four full exits — all clustered in crypto mining and AI-infrastructure operators: Chime Financial, Applied Digital, Cipher Mining, and Core Scientific. This reflects a systematic cleanup of "picks-and-shovels" periphery, pulling resources back to core AI.
What is the core logic behind this reshuffle?
Buys: Nvidia + Cerebras = AI chip makers. Sells: TSMC + Lam Research + Applied Materials = chip manufacturing and equipment. This means → Tiger Global shifted downstream along the same value chain — toward the designers, away from the builders.
Despite total portfolio value falling nearly 16%, top-ten concentration actually reached 66.75%. The trimming hit the margins; the core positions got bigger.
Whether this structural bet pays off in Q4 returns is the key thing to watch next. Put simply = Tiger Global is wagering that "the people who design AI chips" are worth more than "the people who build the machines to make them" — the next 13F will show if that call was right.
市场有风险,内容仅供研究参考,不构成投资建议。
