AI Hedge Fund Manager Aschenbrenner Bets 20% of SharonAI, Stock Doubles YTD

Leopold Aschenbrenner, whose Situational Awareness hedge fund returned more than 2,000% in 2025 and has more than doubled so far this year, disclosed a 20% ownership stake in SharonAI Holdings. The revelation drew attention to a small but fast-growing company betting on a critical choke point in AI infrastructure: megawatt capacity.

SharonAI operates as a neocloud provider—a model pioneered by competitors like CoreWeave—in which the company leases out data center space powered by its own proprietary AI chips. This approach allows SharonAI to scale its footprint without building data centers from the ground up. The company has landed notable customers in the first quarter, including Canva and GMI Computing, plus an unnamed U.S. company. More importantly, it has demonstrated execution speed: starting with a goal to grow from 50 megawatts to 100 megawatts by early 2027, SharonAI achieved 116 megawatts of contracted capacity by July—a substantial leap in a short time frame. As AI companies create more products and attract larger customer bases, the need for megawatts to sustain those workloads will only intensify, positioning SharonAI to capture a growing market.

Yet SharonAI trades at a steep discount to its main peer, CoreWeave. CoreWeave commands a $45 billion（約7.2兆円） market cap—18 times larger than SharonAI's $2.5 billion（約4000億円）—and boasts roughly nine times more active power (more than one gigawatt). If SharonAI were valued at the same rate per contracted megawatt, the stock would need to double from its recent price of about $72. CoreWeave's aggressive forecast of more than eight gigawatts of active power by 2030 justifies its premium; Aschenbrenner's large position in SharonAI suggests he believes the smaller player can close that gap through rapid megawatt accumulation.

However, SharonAI's model carries a structural vulnerability. Unlike CoreWeave and others that own their land and infrastructure, SharonAI leases its data center sites, exposing it to lease-renewal pressure. Landlords can demand higher rates when agreements expire, potentially crushing margins. The company is currently unprofitable, making it especially vulnerable to margin compression. Aschenbrenner is not betting solely on SharonAI; Nebius makes up 40% of his hedge fund's portfolio, while SharonAI represents just over 6%—a portfolio tilt that suggests he is hedging against the lease-rate risk that could undermine SharonAI's long-term viability.

SharonAI's appeal rests on addressing a structural bottleneck in AI infrastructure. As generative AI products multiply and customer bases grow, the computational power required scales exponentially—and that power must come from somewhere. Megawatt capacity has become a scarce resource, especially for providers that control both the data center footprint and the chips running inside. SharonAI's model—owning the chips but leasing the land—sits between pure chip makers and pure landlords, offering a middle ground that lets it expand quickly without the capital burden of real-estate ownership.

Aschenbrenner's 20% stake signals conviction that SharonAI can accumulate megawatts faster than its initial projections suggested. The company grew from a 50–100 megawatt target (for early 2027) to 116 megawatts of contracted capacity in just months, demonstrating execution speed. However, the body also flags that this model carries long-term risk: landlords can raise rates at renewal, squeezing margins. That SharonAI remains unprofitable amplifies the concern. Aschenbrenner's broader portfolio—with Nebius comprising 40% of his fund and SharonAI just over 6%—suggests he is diversifying across different infrastructure bets, hedging against the lease-rate risk that could undermine SharonAI's competitive position.

No comments yet. Be the first to share your thoughts!