AI Analysis · Q1 2026

Harvard Management Company's $1.82 billion U.S.-listed public portfolio for March 2026 reflects an endowment manager executing a deliberate thematic rotation from crypto-asset exposure into the semiconductor and AI-infrastructure complex — a move that is both analytically conspicuous and consistent with the endowment's technology-investment philosophy. The headline trade in Q1 was a near-doubling of the Taiwan Semiconductor Manufacturing position, the portfolio's largest holding at $232 million (12.8% of assets), from 307,000 shares to 687,000 shares through the addition of 380,000 shares. The $139 million in new capital deployed into TSM at prices between $65 and $75 per share — a level that reflected the stock's appreciation year-to-date entering the quarter — represents the portfolio's highest-conviction deployment. Nvidia was also dramatically increased from 410,000 shares to 744,000 shares — an 81.3% share-count increase worth $53 million in new capital — confirming the endowment's commitment to the AI-capital-expenditure cycle across both the equipment layer (TSM makes the chips for Nvidia's customers) and the chip-design layer (Nvidia). The Goldman Crypto narrative is visible in what the endowment sold: the iShares Bitcoin Trust (IBIT) was reduced by 43.1% from 5.35 million shares to 3.04 million shares, removing $149 million in value, as Bitcoin's own price declined from the highs of late 2024 — taking profits off a position that had been carried at substantial weight. The iShares Ethereum Trust (ETHA) was liquidated completely in Q1 for a $87 million exit, eliminating Harvard's sole pure-play Ethereum exposure. Alphabet was reduced by 23.6% — from 808,000 to 617,000 shares — at the same time the position increased in value from $189 million to $177 million on price, a pure rotation away from a name that had been the portfolio's Information Alphabet-sector anchor. Meanwhile, the endowment added to Microsoft (by 4.8%), Broadcom (by 47.3%), Booking Holdings (by 3.8%), and Union Pacific (by 3.8%) in the face of mixed price action, deploying new capital into quality names rather than reducing to cash. The $263 million AUM contraction — from $2.08 billion to $1.82 billion — was substantially a function of crypto-markdowns (IBIT, ETHA, GLD) and Alphabet's price decline, not of realized losses on the tech core. This is a computer-and-storage thesis playing out inside the walls of one of the world's most recognizable institutional investors.

Taiwan Semiconductor Manufacturing at $232 million and 12.8% of the public sleeve — TSM is the world's largest dedicated semiconductor foundry, manufacturing chips for customers including Apple, Nvidia, AMD, Qualcomm, Broadcom, and dozens of other fabless semiconductor designers. TSM's technological leadership in advanced process nodes (currently shipping 3-nanometer and developing 2-nanometer process technology) gives it a structural position in the global semiconductor supply chain that is difficult to replicate. TSM's revenue is highly correlated with global semiconductor demand, and its earnings have benefited enormously from the AI-driven semiconductor capex cycle that has seen hyperscalers (Amazon, Microsoft, Google, Meta, Oracle) invest unprecedented sums in GPU and AI-accelerator manufacturing capacity. Harvard's decision to more than double the position in a single quarter — adding 380,000 shares representing $139 million of new capital at an average price between $65 and $75 (based on prevailing Q1 market prices) — reflects a view that the AI-semiconductor cycle has further to run and that TSM's technology leadership will allow it to capture a disproportionate share of that growth. The position was established through several prior quarters but was materially scaled up in Q1 as AI-capex guidance from hyperscalers remained constructive. At a 12.8% portfolio weight, TSM is Harvard's collective single-largest position across all four legs (together with the Zillow positions, Broadcom, and Nvidia), and the investment decision that drove the increase was likely made at the portfolio-manager level and approved by Harvard's investment committee.

Nvidia at $130 million, up 81.3% — Nvidia's addition is the portfolio's second most aggressive trade by percentage and is thematically linked to the TSM increase: Harvard is not simply betting on AI; it is making a chip-to-foundry-to-AI-infrastructure pair trade across two companies whose fates are tightly linked. Nvidia designs the GPU chips, TSM manufactures them; as demand for AI training and inference capacity grows, both companies benefit. The 334,000 additional shares deployed during Q1 represent $53 million in net new capital at an average cost somewhere in the $100-$135 range (pre-split-adjusted), a level that Harvard's investment team clearly believed was justified by the earnings-growth trajectory even after Nvidia's stock had already substantially appreciated in 2024 and 2025. The position at 744,000 shares and $130 million makes Nvidia the portfolio's seventh-largest holding by value and the portfolio's most direct way to express a view on AI hardware demand.

Quarter at a glance — Q1 2026

Top 10 holdings

Filing history