Investors question Nebius, CoreWeave amid Meta’s cloud compute plans
Meta Platforms is launching an internal division called “Meta Compute” aimed at monetizing its surplus AI infrastructure by offering raw compute power and hosted AI models to external developers. The news, first reported by Bloomberg, sent shares of both Nebius Group (NBIS) and CoreWeave (CRWV) tumbling roughly 15% in early trading on July 1.
The reason for the panic is straightforward. Meta isn’t just any customer for these two companies. It’s the customer. Nebius signed a long-term AI infrastructure deal with Meta worth $27 billion back in March 2026. CoreWeave expanded its own Meta partnership in April to approximately $21 billion, bringing its total Meta contract value to $35.2 billion.
When your largest revenue source starts building a business that looks a lot like yours, investors tend to get nervous. In this case, “nervous” translates to wiping out billions in market cap before lunch.
The neocloud model under pressure
Both Nebius and CoreWeave operate in what the industry calls the “neocloud” space, essentially GPU-as-a-service providers that specialize in renting out high-performance computing resources tailored for AI workloads.
Meta’s combined contracts with Nebius and CoreWeave total $48 billion. The fear now is that Meta Compute could gradually absorb workloads that would otherwise flow to these specialized providers, turning a revenue pipeline into a competitive moat pointed in the wrong direction.
Why some analysts aren’t panicking
Despite the dramatic stock moves, several market observers pointed out that Meta’s decision to commercialize its surplus compute actually validates the thesis that demand for AI infrastructure continues to outstrip supply. If Meta is building so much GPU capacity that it has meaningful surplus to sell, that signals the overall market is growing, not shrinking.
For Nebius and CoreWeave, diversification becomes the obvious strategic imperative. CoreWeave’s total Meta contracts of $35.2 billion represent a substantial portion of its business. Reducing that concentration by winning contracts from other hyperscalers, enterprises, and AI startups would meaningfully de-risk the business, even if Meta Compute becomes a formidable competitor in the medium term.
What this means for investors
Projects like Render, Akash, and io.net have positioned themselves as decentralized alternatives to exactly the kind of centralized GPU cloud that Meta is now entering.
For investors in Nebius and CoreWeave specifically, the 15% drop looks like a classic overreaction to headline risk. The contracts are signed. The demand is real. But the competitive landscape just got meaningfully more complicated, and the long-term margin profiles of neocloud providers may need to be revised downward if they’re competing with a company that treats compute as a byproduct rather than a primary business.
The key metric to watch going forward is customer concentration. If Nebius and CoreWeave can demonstrate meaningful revenue diversification away from Meta over the next few quarters, the sell-off will look like a buying opportunity. If Meta remains their dominant counterparty while simultaneously selling compute to Meta’s competitors, the structural tension in that relationship could become untenable.
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
Tonight, a "dovish rate hike"?
The Federal Reserve is almost certain to raise interest rates tonight, but the key issue is "what will be said after the hike." Citi characterizes this move as a "fine-tuning" adjustment, suggesting there is no inevitable future rate hike, yet warns that if Chairman Powell does not provide clear forward guidance, it will trigger significant market volatility. Goldman Sachs bluntly stated that there is insufficient economic foundation for this rate hike, with inflation being merely a one-off factor, and expects this to be a "signal-less rate hike."
JPMorgan: "Open source disruption" and "AI safety" are not issues, there is still room for capital expenditure in the next two years, semiconductor equipment will become the "new bottleneck"
JPMorgan believes that open-source models are not a threat, regulatory disruptions are only short-term, and cloud vendors’ leverage remains low—the fundamentals of computing power investment have not changed. It forecasts that the capital expenditure of the seven major tech giants will soar from $443 billion in 2025 to $1.577 trillion in 2027, with semiconductor equipment becoming the core bottleneck of the supply chain and a new round of price increases expected in wafer foundry and advanced packaging.
Micron executive: Storage determines AI limits, substantial new capacity will come after 2028
Micron executive Sumit Sadana stated that memory bandwidth and capacity have become the core factors determining the performance ceiling of AI systems. In the face of structural imbalances between supply and demand, Micron expects to increase capital expenditure to over $45 billion in fiscal year 2027, but due to process complexity, substantive new capacity will not be released until 2028. Additionally, long-term supply agreements are reshaping the industry's business model, and humanoid robots will trigger the next wave of massive demand.
Goldman Sachs warns top clients: AI momentum trading shows unprecedented cracks, recommends hedging
Goldman Sachs has issued a rare warning to top clients: deep structural cracks are emerging in AI momentum. The AI-themed basket has dropped nearly 45% from its peak, with the one-day performance gap between short-term and long-term momentum reaching a five-year high. Capital is accelerating its shift from semiconductors to software. Goldman Sachs explicitly recommends that investors with AI exposure start hedging.
