Leopold Aschenbrenner's Situational Awareness hedge fund returned 270% in 2026, according to 13F filings that also disclosed Jane Street Capital as a new investor in the vehicle. The fund, launched in early 2025 by the former OpenAI safety researcher, concentrated positions in retrofitted data center operators, former bitcoin mining facilities converting to AI inference, and second-tier semiconductor suppliers serving hyperscale buildouts.
The filing shows Jane Street entered the fund in Q3 2026 alongside existing backers from family offices and sovereign wealth vehicles. Aschenbrenner's thesis—that AI compute demand would overwhelm hyperscaler capacity and force capital into improvised infrastructure—played out faster than consensus expected. The fund held stakes in twelve neocloud operators by year-end, including three former Compute North mining sites that secured long-term contracts with frontier model labs. Returns compressed in Q4 as competition for stranded power and retrofit sites intensified, but full-year performance still exceeded fifteen other technology-focused hedge funds tracked by Preqin.
Jane Street's participation signals quantitative validation of a thesis many dismissed as infrastructure tourism. The proprietary trading firm rarely discloses minority stakes in externally managed vehicles, suggesting the allocation reflects conviction rather than relationship capital. Aschenbrenner has not raised outside capital aggressively—the fund remains under $800 million AUM according to people familiar with the structure—but the investor composition now includes four quantitative shops and two multi-strategy platforms that entered in late 2026. The fund charges 2-and-30 with a two-year lockup, terms that have not deterred institutions seeking exposure to second-order AI infrastructure plays unavailable in public equity indices.
What matters is the position construction. Situational Awareness avoided hyperscaler equities and large-cap semiconductor names, instead building stakes in six private neocloud operators and eight public microcap data center REITs trading below 0.7x book value in early 2026. The fund also held convertible notes in two power infrastructure developers securing grid connections for AI campuses in Texas and North Dakota. These positions appreciated as model training runs extended beyond hyperscaler capacity and labs began signing eighteen-month contracts with regional compute providers. The strategy worked because Aschenbrenner correctly anticipated that inference demand would outpace training demand by mid-2026, creating pricing power for smaller operators with stranded GPU inventory and cheap electricity.
Allocators should monitor three follow-on developments. First, whether Situational Awareness raises a second vehicle or accepts outside capital into the existing fund—Aschenbrenner has historically resisted institutional money but Jane Street's entry suggests terms may soften. Second, how the fund repositions as neocloud valuations now reflect consensus demand; early entry into 2027 could favor different infrastructure chokepoints including custom silicon foundries or liquid cooling suppliers. Third, whether other quantitative shops follow Jane Street into thematic infrastructure funds, which would mark a reversal from the sector-neutral positioning that dominated 2024-2025.
The fund's AUM has not scaled proportionally to performance, which means Aschenbrenner is managing for alpha rather than fee revenue. Jane Street does not invest for signaling purposes.