What it is
“Neocloud” is industry shorthand for a cloud computing company built from the ground up to rent out GPU capacity for AI workloads, rather than a traditional cloud provider — Amazon Web Services, Microsoft Azure, Google Cloud — that added GPUs to an already-diversified catalog of services. Where hyperscalers sell hundreds of products (storage, databases, serverless functions, GPUs among them), neoclouds keep their offering narrow: bare-metal or thinly virtualized access to high-performance chips, fast interconnects, and little else. The business model is often called GPU-as-a-Service.
The term appears to have entered circulation in late 2024, and the analysis most frequently credited with giving it a structured definition is a SemiAnalysis piece by Dylan Patel and Daniel Nishball, “AI Neocloud Playbook and Anatomy” (October 3, 2024), which described neoclouds as “a new breed of cloud compute provider focused on offering GPU compute rental” and sorted the field into tiers — “neocloud giants” such as CoreWeave, Nebius, Lambda and Crusoe, which operate at large scale (SemiAnalysis’s benchmark was the equivalent of 100,000-plus Nvidia H100 GPUs), down through dozens of smaller “emerging” operators and capital-light brokers that resell access to GPUs they do not own.
CoreWeave is widely described as the prototype of the category — it began as a cryptocurrency-mining operation before pivoting to GPU rental for AI training. Notably, CoreWeave itself has recently pushed back on the label: as its cloud business has diversified toward enterprise customers and software tools, an executive told Bloomberg the “neocloud” tag is “temporary and without a purpose” for a company no longer selling raw compute alone.
Why it matters for AI governance and narratives
Neoclouds sit at the center of the capital-intensity story that increasingly structures AI coverage: who is financing the buildout of AI infrastructure, on what terms, and what happens if the revenue to justify it does not materialize. Because neoclouds typically carry heavier debt loads and narrower revenue diversification than hyperscalers — their business depends almost entirely on AI-optimized infrastructure margins — they function in public discourse as a kind of canary for the sustainability of the broader AI capital-spending cycle. Claims about neoclouds’ survival prospects are rarely neutral: they circulate from research firms, short-sellers, and critics with different incentives, and figures about their mortality rate can move markets within hours, as happened on October 14, 2026, when CoreWeave, Nebius and IREN shares fell on the Gartner forecast described in the editorial passage. Tracking how such figures are sourced, amplified, and contested is itself a narrative-formation story, separate from whether the underlying forecast proves accurate.
Key facts and dates
Gartner has published two notable neocloud forecasts in 2026 that point in different directions. In July 2026, Gartner analyst Enrique Castera projected that neocloud providers would capture roughly 20% of a $267 billion AI cloud market by 2030, driven by their performance advantages, flexible deployment, and data-sovereignty positioning relative to hyperscalers. Then, in an October 14, 2026 press release, Gartner analyst Will Sommer forecast that “cost pressure and a lack of a differentiating business model will force over 80% of neocloud providers to cease independent operations” by 2030, and that more than $3.5 trillion of AI infrastructure spending — against a projected $8.5 trillion in revenue needed to cover capital spending, financing costs and a 12% return on invested capital — could be exposed to impairment by then. Gartner’s own figures show AI providers spending nearly $1 trillion on AI infrastructure in 2026 alone, with a further $6 trillion forecast between 2027 and 2030, against enterprise and consumer AI revenue Gartner caps at roughly $3.6 trillion over the same window. Both forecasts come from the same research firm within months of each other and are not straightforwardly reconcilable — one emphasizes neocloud market-share gains, the other mass consolidation — a tension the editorial’s “order book and mortality table” framing captures.
Where to learn more
- Gartner: “Gartner Says Over $3.5 Trillion of AI Infrastructure Spending Could Be Exposed to Impairment by the End of 2030” — the primary source for the 80%/$3.5 trillion forecast (note: direct fetch was blocked by a paywall/bot gate during this research; details corroborated via secondary coverage below)
- SemiAnalysis: “AI Neocloud Playbook and Anatomy” (Dylan Patel and Daniel Nishball, October 3, 2024) — the most-cited structured definition and taxonomy of the category
- CIO Dive: “Neocloud providers will gain greater slice of AI cloud market by 2030” — covers Gartner’s earlier, more bullish July 2026 neocloud forecast
- TechnologyForYou: “By 2030, 80% of Neocloud Providers Are Expected to Cease Independent Operations” — secondary coverage quoting Gartner’s October 14, 2026 forecast directly, including analyst attribution