For the past year, the neocloud trade was one of the cleanest stories in tech investing. Rent GPUs to AI companies, watch demand outpace supply, collect revenue. Nebius, CoreWeave, and Iris Energy all rode that thesis to triple, quadruple, even eight-times gains. Then Meta showed up with its own product.
On July 1, 2026, reports emerged that Meta Platforms is preparing to launch “Meta Compute,” a service designed to sell off the company’s excess AI infrastructure capacity to outside customers. The market’s response was immediate and lopsided. Nebius and CoreWeave each fell roughly 15% in morning trading. Iris Energy dropped around 6.5%. Meta’s own stock surged more than 10%.
The customer concentration problem, made visible
CoreWeave holds a $21 billion commitment from Meta for AI infrastructure services. Nebius is tied to a potential $27 billion infrastructure agreement with Meta, including a separate $12 billion tranche slated to begin in early 2027.
When your biggest customer announces it is building the same product you sell, the relationship gets complicated fast.
CoreWeave’s financials illustrate how exposed these companies actually are. The company posted $2.078 billion in Q1 2026 revenue, a 111.7% increase year-over-year. That growth number is genuinely impressive. The $740 million net loss sitting underneath it is less so.
IREN’s different story, same bad day
Iris Energy’s situation is worth separating from the other two. The company spent years as a Bitcoin miner before pivoting hard into AI and high-performance computing data centers. Its differentiator is infrastructure powered by renewable energy. IREN also holds a significant contract with Microsoft, adding some customer diversification relative to CoreWeave and Nebius.
IREN had secured approximately 5 GW of power capacity heading into this period, a significant asset in a market where power availability is genuinely scarce.
What this means for investors watching the neocloud space
Nebius, CoreWeave, and IREN had all posted gains of somewhere between 100% and 800% in the preceding year. Short interest across all three names was running between 18% and 24% leading into the Meta news.
The deeper structural question is whether Meta Compute represents a one-time shift or the beginning of a pattern. If the largest hyperscalers collectively decide to monetize excess capacity rather than route overflow demand to independent providers, the total addressable market for neocloud companies shrinks.
The companies most at risk are those whose growth models depend on hyperscalers as both customers and as the source of overflow demand. CoreWeave’s $21 billion Meta commitment is a strength in one reading and a single point of failure in another.