Situational Awareness LP doubled its 13F notional to $13.67B in Q1 — and exited Hut 8

Leopold Aschenbrenner's Situational Awareness LP filed its Q1 2026 13F with the SEC on May 18. The fund's total disclosed notional exposure — long equity plus the underlying value of all options positions — rose from $5.52 billion at end-2025 to $13.67 billion as of March 31. That figure is 13F notional, not assets under management: the fund's Form ADV, filed April 27, shows $9.28 billion in discretionary AUM.

The restructuring has two sides. On the long side, Aschenbrenner expanded positions across the Bitcoin miner basket he has held since at least Q3 2025. On the short side, he layered roughly $7.7 billion in new put notional across the AI semiconductor complex — the biggest single line being a $2.04 billion put on the VanEck Semiconductor ETF (SMH), followed by $1.57 billion against Nvidia, $1.07 billion against Oracle, and $1.01 billion against Broadcom.

Hut 8 is gone. The fund fully exited its $39.5 million Hut 8 position during Q1. The Q4 2025 filing had listed HUT as a position; it does not appear in the Q1 2026 holdings. The remaining miner book is: IREN (increased), Core Scientific (CORZ, increased), Riot Platforms (up from 6.17M to 11.50M shares), CleanSpark (up from 1.64M to 12.28M shares), Bitfarms (up from 6.90M to 19.88M shares), Bitdeer (up from 1.79M to 3.44M shares), and HIVE Digital (new, 3.39M shares, $6.4M). Applied Digital (APLD) was also increased.

The infrastructure thesis in Aschenbrenner's own words

The logic behind the miner longs is not a crypto thesis. In his 2024 essay "Situational Awareness: The Decade Ahead" — the 165-page monograph that preceded the fund — Aschenbrenner argued that the real bottleneck to AGI is physical, not algorithmic: compute infrastructure, power, and land. Bitcoin miners enter the argument because they already hold high-density grid connections and data center capacity that AI hyperscalers cannot replicate on short notice.

Fortune, which profiled Aschenbrenner in March 2026, described the fund's view as holding that "the most valuable assets in the AI era may not be algorithms, but electricity and computing power." The essay itself argued that scaling laws governing large language models dictate that capabilities increase predictably with compute, making power access a hard constraint on AGI timelines.

The fund was seeded by Nat Friedman (now Meta AI product lead), Daniel Gross (Meta Compute co-lead), and Patrick and John Collison. Aschenbrenner has invested nearly all of his net worth in the fund, according to a spokesperson quoted by Fortune.

Core Scientific is executing the pivot; Hut 8 was not

Core Scientific is the clearest example of the miner-to-AI-infrastructure transition. The company has contracted to deliver approximately 500MW of critical IT load to CoreWeave, the AI hyperscaler, by H2 2026. That total accumulated across multiple option exercises on the original June 2024 hosting contract. A separate 8-K from February 2025 confirmed the scale: Core Scientific has secured $8.7 billion in HPC hosting contract value.

Hut 8's exit from the portfolio is consistent with relative execution pace. The company has been slower than Core Scientific, Applied Digital, or IREN in converting mining capacity to AI hosting at scale. Aschenbrenner's reduction from a $39.5M position to zero during Q1 suggests the fund distinguished between miners actively pivoting to AI infrastructure and those still primarily running bitcoin operations.

Applied Digital, also held and increased, has built out a dedicated AI data center campus separate from its mining operations — a structural distinction the fund appears to weight.

What the semiconductor short tells you

The put book is internally consistent with the long book, not a contradiction of it. If AI compute value accrues to power sites and managed infrastructure rather than chip manufacturers, then semiconductor valuations face pressure even as companies like Core Scientific and IREN gain. The $7.7 billion in put notional across SMH, Nvidia, Oracle, and Broadcom encodes that view.

One critical caveat, flagged by TrendSpider's analysis of the filing: 13F disclosures show the notional value of put and call positions but do not specify whether those options are long or short. A $1.57 billion notional Nvidia put could be a directional short, a hedge against other positions, or a premium-selling income structure. The filing alone cannot confirm intent.

The Q1 2026 13F was filed May 18, 2026, covering positions held as of March 31. The fund's current book may differ materially.

Sources: SEC EDGAR accession 0002045724-26-000008 (Q1 2026 13F-HR, filed 2026-05-18); SEC EDGAR Form ADV for Situational Awareness LP (filed 2026-04-27); Core Scientific press releases via investors.corescientific.com; Fortune profile of Aschenbrenner, March 5, 2026; TrendSpider analysis of Q1 2026 13F positions.