Short Thesis on CoreWeave ( CRWV 0.00%↑ )
Kerrisdale Capital describes CoreWeave as the perfect symbol of the AI infrastructure mania. Just five months after its IPO, shares are up ~200%, valuing the company at $75B. The business model is straightforward: lease GPUs from Nvidia, finance them with asset-backed loans, and then rent the compute capacity out to customers under long-term contracts.
The problem, Kerrisdale argues, is that there’s little substance beyond that. CRWV lacks proprietary software, differentiated infrastructure, or a vertically integrated platform. Instead, it’s essentially a leveraged lessor of GPUs — dependent on a small number of anchor customers and highly exposed if those relationships weaken.
Recent headlines illustrate this fragility. Microsoft struck a deal with Nebius, while Oracle disclosed that much of its cloud backlog comes from OpenAI, not CRWV. In other words, the most strategic workloads are not flowing to CoreWeave but to better-capitalized hyperscalers or their chosen partners.
Meanwhile, the capital structure is stretched. Billions in debt finance the GPU purchases, and those loans are tied to contracts that may look less favorable as hyperscalers scale their own AI infrastructure and squeeze pricing. The real bottleneck in AI infrastructure is power, not GPUs alone, and CoreWeave has no unique advantage there.
In Kerrisdale’s view, CRWV has been bid up by investors chasing anything AI-related, but it sits in a precarious spot: overvalued, overleveraged, and under-defensible. If demand growth slows or customers shift workloads in-house, the downside could be severe.
Short Thesis on Pure Storage ( PSTG 0.00%↑ )
Pure Storage built its reputation on all-flash arrays (AFAs) that outperformed traditional spinning hard drives in enterprise storage. For years, it was seen as a disruptive growth story, riding the secular shift from HDDs to flash. Today, Kerrisdale argues, that story has lost momentum.
Growth has slowed, competition from giants like Dell, HPE, and NetApp is fierce, and Pure’s technology lead has narrowed. The company’s latest narrative centers on QLC flash replacing HDDs at hyperscale — a claim Kerrisdale dismisses as unrealistic. Hyperscalers design their storage architectures around cheap, high-capacity HDDs for cold storage and increasingly customized, in-house solutions for flash. Pure’s premium arrays don’t fit into those economics, and the hyperscalers are unlikely to become meaningful hardware customers.
The much-hyped Meta software license deal is another example. PSTG bulls point to it as validation of their relevance to hyperscale, but Kerrisdale notes the deal is narrow in scope, replicable, and doesn’t require Meta to buy Pure hardware. Meta can source NAND directly and use white-box ODMs — bypassing Pure altogether. Importantly, the agreement hasn’t expanded to other Meta data centers, undermining the bullish narrative.
Within enterprise storage, PSTG is stuck in a mature, low-growth, cutthroat market. Market share shifts are incremental, margins are pressured, and competitors bundle flash arrays into broader IT offerings. The company touts its “software” revenue, but Kerrisdale highlights that most of it is support contracts and prepaid upgrades, not recurring SaaS with sticky economics.
At its current valuation, Kerrisdale argues, PSTG is being priced like a growth company on the cusp of hyperscaler adoption. In reality, it’s a hardware vendor in a saturated market with limited differentiation and slowing momentum.