BlackRock Goes All-In on AI Compute Chain in Q2, Total AUM Rises to $6.73 Trillion
Claire Weston
BlackRock's Q2 U.S. equity holdings rose to $6.73 trillion, up 18% quarter-on-quarter, with new buying spread across every link in the AI compute chain — signaling the world's largest asset manager is betting the AI capex cycle is far from over.
What did a $6.73 trillion portfolio do in Q2?
BlackRock's total U.S. equity market value jumped from $5.72T to $6.73T, roughly +18% QoQ.
Activity breakdown: 274 new positions, 3,606 adds, 1,322 trims, 263 exits — the net direction is heavy accumulation.
This means → while hedge funds net-sold tech for four straight weeks, BlackRock was buying into the sell-off. The bull-bear divide is stark.
What do the top ten holdings look like?
Nvidia remains #1: ~1.94 billion shares, $388.6B market value, 5.77% of the portfolio, up 0.85% in share count.
Apple, Microsoft, Amazon, and Alphabet follow; top five alone account for ~19.24%. Top ten = 28.01% of total.
Key signal: every single top-ten position grew — Broadcom +3.18%, Alphabet A +3.13%, Tesla +2.90%. None were trimmed.
In plain terms = BlackRock's heaviest bets are all AI compute and cloud infrastructure names, and it added to all of them.
What was bought, what was sold — and how to read the signal?
Top five buys: Micron, AMD, Applied Materials, Marvell, KLA — mapping to HBM/DRAM (ultra-fast memory that feeds GPUs), AI GPUs, semiconductor equipment, optical interconnects (using light to move data between chips in a data center), and custom AI silicon.
Top five sells: Costco, Honeywell, Berkshire B, AT&T, Johnson & Johnson — spanning consumer staples, industrials, financials, telecom, and healthcare.
This means → the allocation call is unmistakable: "cut defense, add AI infrastructure." Capital is spreading along the entire compute chain, not concentrated in one name.
The market was selling — why was BlackRock buying?
Context: from late Q2 into early July, semis de-leveraged hard. The Philadelphia Semiconductor Index fell 4.2% in one week; hedge funds net-sold tech/semis for a fourth consecutive week.
BlackRock's counter-move implies its view is the opposite of near-term sentiment — it sees the pullback as a buying window, not a trend reversal.
This reflects a time-horizon gap between long-duration asset managers and short-cycle hedge funds: one is pricing the AI capex cycle's position; the other is trading quarterly momentum.
What is BlackRock's own stated logic?
Its BlackRock Investment Institute explicitly recommends overweighting U.S. equities to gain AI compute-infrastructure exposure.
Core thesis: the true scarcity in the AI boom is not any single model — it is the entire delivery chain: GPUs, CPUs, HBM, liquid cooling, optical interconnects, advanced packaging. Every link is a bottleneck.
In plain terms = BlackRock believes AI's constraint is hardware supply, not software. So the capital goes across the whole chain, not just into Nvidia.
An important footnote — what does this filing actually represent?
The 13F is a group-level consolidated filing covering iShares ETFs, index funds, mutual funds, and active mandates — all client assets under BlackRock's discretion.
This means → it reflects BlackRock's aggregate holdings as an asset manager, not a single portfolio manager's active bet.
Next verification point: whether the AI capex cycle is still in its early-to-mid phase will show up in cloud giants' capex disclosures and semiconductor-equipment order books — that is the next signal window.
Content is for reference only, not financial advice.