Leopold Aschenbrenner's Fund Buys AMD, CoreWeave Calls After $35B Wipeout

Leopold Aschenbrenner, the former OpenAI researcher who runs the hedge fund Situational Awareness, has bought call options on a group of AI-linked companies, CNBC reported, citing sources. The names are AMD, Bloom Energy, CoreWeave, SK Hynix and SanDisk. The trades ran from late last week into this week. The purchases came roughly six weeks after AI bets nearly wrecked the hedge fund in a $35 billion wipeout. Now the same manager is putting money back into the same corner of the market, this time through options rather than outright share purchases.

What exactly did the fund buy?

The fund bought call options, not common stock. A call option gives the buyer the right, though not the obligation, to buy the underlying shares at a set price before a set date. That structure caps the money at risk to the premium paid, while keeping exposure to upside moves. For a fund rebuilding after a large drawdown, options can offer a way to express a view without committing the same capital as a direct equity position.

The five names span different layers of the AI trade. AMD designs the GPUs and accelerators that compete with Nvidia in data centers. SK Hynix is a major supplier of high-bandwidth memory, the specialized DRAM stacked next to AI chips. SanDisk sits in the NAND flash and storage market. CoreWeave rents out GPU compute capacity to AI developers. Bloom Energy makes fuel cell systems used for on-site power, a piece of the electricity problem that large data centers keep running into.

CNBC reported the trades citing sources, which means the positions are not confirmed by the fund itself. That matters for anyone trying to read the move as a formal signal. Reported options flow can reflect hedging, a short-term tactical trade or part of a broader strategy, not only a directional bet. Still, the choice of names tells a clear story about where this manager sees the next leg of AI spending.

Why does this matter now?

The timing is the story. Six weeks before these trades, AI bets nearly wrecked the hedge fund. A $35 billion wipeout is the kind of loss that forces most managers to cut risk, face redemptions and spend months rebuilding trust. Aschenbrenner is instead re-entering the same theme, and he is doing it in public view. That makes the trades a test case for how fast the market will forgive a high-profile blowup when the underlying theme is still intact.

For traders, the reported positions matter because they put fresh attention on AMD, CoreWeave, SK Hynix, SanDisk and Bloom Energy. Options buying can influence the market in ways share purchases do not. Dealers who sell calls often hedge by buying the underlying stock, which can amplify moves when prices rise. That dynamic can make a cluster of call buying more visible in price action than the notional size alone would suggest.

The move also lands in a market that has spent months debating whether AI infrastructure spending is peaking or still accelerating. Hyperscaler capital expenditure plans, data center power constraints and memory pricing have all been part of that debate. A fund buying calls on memory, storage, compute and power names is effectively a bet that the buildout keeps pulling demand through several layers of the supply chain at once.

Who is Leopold Aschenbrenner?

Aschenbrenner is a former researcher at OpenAI. He left the company and later published a widely read essay on artificial general intelligence and national security. That background gives him a specific kind of credibility in the AI trade. He is not a generalist investor who added AI exposure to a diversified book. He built his public profile around the argument that AI capability is advancing fast and that the compute and power needed to sustain it will be enormous.

His fund, Situational Awareness, is named after that thesis. The fund's concentrated approach cuts both ways. When the theme works, a focused book can outperform broad indices by a wide margin. When positioning goes wrong, the losses compound quickly, which is what happened in the episode that produced the $35 billion figure. The reported return to AI names suggests the manager is treating that drawdown as a positioning problem rather than a broken thesis.

There is also a signaling angle. A former OpenAI insider buying AI infrastructure names draws more attention than the same trade from an unknown fund. It reinforces the narrative that people closest to the technology expect the hardware and power buildout to continue. That narrative has been a powerful driver of crypto and equity markets alike, since AI demand has become a major factor in semiconductor and energy valuations.

What does this mean for AI-linked stocks?

It means one prominent fund is willing to take fresh risk on the group after a painful episode. That is not the same as broad institutional confirmation. Reported options positions can be closed quickly and are not disclosed in the same way as large shareholdings. Traders watching the names should treat the report as one input among many, alongside earnings, guidance and capital expenditure data from the largest technology buyers.

The five companies sit at different points in the AI supply chain, so a single trade does not mean identical exposure. AMD is a direct play on accelerator demand and competes with Nvidia. SK Hynix and SanDisk depend on memory and storage pricing cycles, which can turn faster than GPU demand. CoreWeave is tied to the market for rented compute. Bloom Energy depends on data center power needs and on the economics of fuel cells versus grid power and gas turbines.

Call options also change the payoff profile. If the shares rise above the strike price before expiry, the fund can profit without having put up the full cost of the shares. If the shares stall or fall, the premium paid is the loss on those positions, unless the fund offsets them elsewhere. That asymmetry is why options are a common tool for re-entering a volatile theme after a drawdown.

How does this fit the bigger AI trade?

The AI trade has widened well beyond chip designers. Power generation, electrical equipment, memory, storage and data center operators have all been pulled into the story as the constraint has shifted from chips to electricity and infrastructure. The reported basket reflects that widening. It is not only a bet on faster processors. It is a bet on the physical buildout needed to run them.

That buildout has become one of the strongest cross-market themes, linking equities, energy and digital assets. Bitcoin miners have pivoted some capacity to AI hosting. Data center operators have signed large power agreements. Memory makers have seen demand shift toward high-bandwidth products used in AI servers. Each of those threads shows up in the reported positions, from SK Hynix and SanDisk to CoreWeave and Bloom Energy.

For crypto traders, the relevant channel is less direct than it was in earlier cycles. AI demand competes for the same chips, power and capital that miners and data center operators use. When AI infrastructure valuations rise, financing conditions for power-intensive operations tend to improve. When the theme cools, the same names can sell off together. That correlation has made AI headlines a regular input for anyone trading mining and infrastructure tokens.

The memory angle deserves particular attention. High-bandwidth memory has been one of the tightest parts of the AI supply chain, and pricing has reflected that. SK Hynix is a key supplier in that market. SanDisk, meanwhile, is exposed to NAND flash, where pricing cycles have been more volatile. Buying calls on both is a bet that the storage and memory layers keep benefiting from AI server demand, not just the GPU layer that gets most of the headlines.

What should traders watch next?

The first thing to watch is disclosure. The positions were reported by CNBC citing sources, so the market has no regulatory filing confirming the size, strike prices or expiry dates. Follow-up reporting or any fund comment would clarify whether these are short-dated tactical trades or longer-dated positions. The distinction matters for how much weight to give the signal.

The second is earnings and guidance from the companies involved. AMD's data center revenue and its progress against Nvidia set the tone for the accelerator trade. SK Hynix and SanDisk commentary on memory and storage pricing will show whether AI demand is still tightening those markets. CoreWeave's capacity and contract updates speak to the rented compute market. Bloom Energy's orders and power agreements tie the trade to the electricity constraint.

The third is the macro backdrop. AI-linked names have been sensitive to interest rate expectations because so much of their value sits in future cash flows. A shift in rate expectations can move the group regardless of company-specific news. With call options, that sensitivity is magnified, since time decay works against the buyer as expiry approaches.

Finally, watch whether other funds follow. A single manager re-entering a theme after a large loss is notable. A broader return of institutional money to AI infrastructure would be a stronger signal. The opposite is also possible: if the reported trades are followed by a pullback, the episode could reinforce the view that the AI trade has become overcrowded and prone to violent reversals.

For now, the facts are narrow and clear. A fund run by a former OpenAI researcher, six weeks after a $35 billion wipeout tied to AI bets, bought call options on AMD, Bloom Energy, CoreWeave, SK Hynix and SanDisk. The trades ran from late last week into this week. Everything beyond that, including the size and the intent, remains unconfirmed.