Version 1

Leopold Aschenbrenner — a former OpenAI researcher fired in 2024 who claimed it was retaliation for raising safety concerns [1][2] — disclosed a 5.6% stake in Nebius Group (NBIS) through his Situational Awareness LP hedge fund [3][16]. The disclosure, valued at roughly $2.86 billion [17], sent Nebius shares up more than 10% and drew coverage from Reuters and CNBC [8][9]. The fund is widely cited as having grown from $225 million to $13.7 billion in assets [3][6], making the Nebius position one of its largest disclosed holdings [18]. A parallel social media campaign amplified Aschenbrenner's AI safety views — especially his argument that AI honesty is the hardest unsolved problem in alignment [15].

Aschenbrenner sits at an unusual intersection: a safety-focused AI intellectual whose financial convictions are now publicly legible via regulatory disclosures, and whose choices move public markets. His Nebius stake signals that serious AGI believers are concentrating capital in compute infrastructure, while his ongoing critique of OpenAI keeps the question of whether frontier labs take existential risk seriously at the center of the public debate.

Leopold Aschenbrenner first drew broad attention when he was dismissed from OpenAI in mid-2024 under circumstances he described as retaliation for raising safety and security concerns internally [1][2]. Shortly after, he published a 165-page essay titled 'Situational Awareness: The Decade Ahead,' laying out a detailed argument that superhuman AI — AGI — would arrive within the current decade [3][4]. The essay circulated widely enough to prompt a substantive response from computer scientist Scott Aaronson [5] and established Aschenbrenner as a prominent voice on AI timelines from outside the major labs. He then founded Situational Awareness LP, a San Francisco-based fund, which social media reporting — citing 13F disclosures — claims grew from $225 million in initial capital to approximately $13.7 billion in assets [3][6], though the exact figure and its sourcing remain unclear.

In late May 2026, Situational Awareness LP disclosed a 5.6% stake in Nebius Group, a Nasdaq-listed AI infrastructure and cloud company backed by Nvidia [3][7]. The stake, reported at roughly $2.86 billion in value, became one of the fund's largest disclosed holdings and was covered by Reuters, CNBC, and Yahoo Finance as a market-moving event [8][9][10]. Nebius shares rose more than 10% on the news [8], and Nebius separately announced plans to raise its AI infrastructure spending to $25 billion [11]. The investment fits a broader thesis — articulated by commentary amplifying the disclosure — that AI conviction should be expressed through compute infrastructure rather than frontier model companies [12].

Running in parallel to the investment story, the Milk Road AI social media account ran a sustained five-post series using Aschenbrenner's track record to frame his intellectual arguments. The posts highlighted his claim that 'everyone is talking about AI safety but almost nobody is actually working on it' [13], pointed to OpenAI's 2019 decision to withhold GPT-2 as retrospective vindication of his early caution [14], and foregrounded his view that achieving AI honesty — making models reliably truthful — is 'surprisingly hard' and 'probably gets you pretty far' toward solving alignment [15]. Milk Road AI consistently framed these arguments through the lens of his investment returns, blending intellectual and financial credibility.

The overall picture is a story with three interlocking dimensions: a regulatory disclosure that moved markets; a persistent public argument that OpenAI failed on safety and that the broader industry's safety rhetoric is hollow; and a substantive but contested position on where the hardest technical problems in alignment actually lie. Aschenbrenner's concentration in Nebius rather than in frontier model developers suggests he believes durable value in the AI transition will accrue at the infrastructure layer — a bet that may also reflect his view that model development itself carries risks he is unwilling to finance.