Ex‑OpenAI Researcher Aschenbrenner Shorts Nvidia, Oracle in $13.6 B AI‑Crypto Bet

Why It Matters

Aschenbrenner’s portfolio shift highlights a growing debate over what will ultimately limit AI scaling: chips or the physical infrastructure needed to power them. If power constraints prove decisive, investors may redirect capital from high‑valuation semiconductor firms to energy‑heavy data‑center operators and crypto miners that already possess the necessary grid connections. This could reshape the risk‑reward landscape for AI‑related equities and introduce new volatility drivers tied to energy policy, grid capacity and cryptocurrency regulation. The short positions also serve as a market signal that at least some sophisticated investors view the current AI‑chip rally as over‑inflated. Should Nvidia or Oracle miss earnings expectations, the ripple effect could trigger broader sector corrections, prompting traders to reassess exposure to AI‑driven growth stories.

Key Takeaways

• Leopold Aschenbrenner disclosed a $13.67 B portfolio, up from $5.5 B, in a May 18, 2026 13F filing.
• Short positions include $1.57 B on Nvidia, $1.07 B each on Oracle and VanEck Semiconductor ETF, and $1.01 B on Broadcom.
• Long holdings focus on power‑intensive crypto miners and AI‑compute firms such as Bloom Energy, CoreWeave and SanDisk.
• Core Scientific plans a 1.5‑GW AI data‑center campus, repurposing 300 MW of mining capacity.
• TeraWulf’s AI‑hosting revenue ($21 M) surpassed its Bitcoin mining revenue ($13 M) in Q1 2026.

Pulse Analysis

Aschenbrenner’s bet is a textbook example of contrarian positioning in a market caught up in hype. The AI narrative has been dominated by chip makers, with Nvidia alone accounting for a sizable share of AI‑related market cap growth. By shorting those names while loading up on assets that already have the physical bandwidth to host AI workloads, Aschenbrenner is effectively wagering on a supply‑side bottleneck. Historically, similar infrastructure constraints have surfaced in other technology cycles—think of the data‑center power crunch during the early cloud era. If the grid cannot keep pace with AI demand, the sector’s growth will be throttled, and the premium on chip makers could evaporate.

From a trading perspective, the size of the puts is noteworthy. A $1.57 B short on Nvidia translates to roughly 1.2 % of the company’s float, enough to move sentiment if the positions are disclosed publicly. The timing is also critical: Nvidia’s earnings are due in two days, and any miss could trigger a cascade of stop‑loss orders and algorithmic sell‑offs. Conversely, the long side in crypto‑linked data centres could benefit from a surge in demand for AI‑compute that bypasses traditional chip supply chains, especially if miners repurpose existing high‑density sites.

Looking ahead, the market will watch two variables: AI‑chip earnings and energy‑policy developments. A tightening of power‑grid regulations in key mining jurisdictions (e.g., China, Texas) could accelerate the shift toward AI‑compute farms that already have grid access, validating Aschenbrenner’s thesis. On the flip side, a strong earnings beat from Nvidia could reaffirm the chip‑centric view and force a re‑evaluation of the short bets. Traders should therefore monitor not just earnings, but also real‑time power‑usage data, grid capacity announcements, and crypto‑miner earnings reports for early signals of where the AI infrastructure battle is heading.