Leopold Aschenbrenner's Situational Awareness hedge fund 13F reveals massive semiconductor puts and long Nvidia position

Leopold Aschenbrenner's 13F: Massive Semiconductor Puts and a Long Nvidia Bet

Leopold Aschenbrenner's Situational Awareness hedge fund disclosed holdings as of March 31, 2026, revealing a $2 billion put position on SMH, the Vanguard Semiconductor ETF, alongside a long position in Nvidia. The filing has drawn unusual attention across finance and tech circles, though interpreting it cleanly carries real pitfalls.

The puts signal a more targeted thesis than a blanket bearish bet on semiconductors. Rather than betting against the sector broadly, Aschenbrenner appears to be making a pointed call that much of the semiconductor industry is overpriced while specific players—notably Nvidia—remain undervalued. This fits his core investment thesis that frontier AI will drive unprecedented demand for compute and the infrastructure bottlenecks that follow.

What the filing actually shows and what it doesn't

A 13F is a snapshot frozen in time. These positions were in place during the early phase of the Iran war in March, before April and May saw significant portfolio rotation. The filing discloses notional value on options without reporting strike prices, expirations, premiums paid, or whether options are part of broader hedging structures. A reported $1 billion position in puts may represent far less actual capital deployed if those are low-delta convexity hedges rather than in-the-money positions.

The filing also omits outright shorts, private company stakes, and copy-trading positions—all of which could be material to Aschenbrenner's actual portfolio construction. His turnover and trade frequency are invisible. The positions disclosed may have been rotated out weeks or days after March 31.

The gap between snapshot and reality has spawned literal copy-trading trackers following Aschenbrenner's moves, though these come with massive tracking error and are "very accurate for three months ago," as one host notes. Making investment decisions based on stale options data sounds like a way to burn capital.

The solar subplot

One secondary signal worth considering: if Aschenbrenner holds a position in T1 Energy, a Chinese solar company that divested and rebranded, that suggests he may expect shorter timelines for solar deployment in the AI buildout than the AI-bull consensus allows. Most nuclear plays are still pricing in 2032 timelines for meaningful capacity. A solar position hints at more near-term feasibility than conventional wisdom has priced.

The broader AI backlash context

The 13F lands as data centers face coordinated political pushback from both left and right. The left targets job displacement and creative theft. The right frames them as surveillance infrastructure. Both sides are now using AI-generated graphics to oppose AI data centers, creating the kind of ironic recursion that has become routine in this space. This sustained opposition may be shifting the energy mix conversation toward faster-timeline renewables like solar, away from exclusive reliance on nuclear.