Neocloud

Overview

Neocloud (also “neoclouds”) is the industry term for a new generation of cloud providers built
specifically around renting Nvidia GPU capacity for AI training and inference — “GPU-as-a-service” —
rather than the general-purpose compute, storage, and managed-services stacks of traditional
hyperscalers (AWS, Azure, Google Cloud). Named entities in this category include CoreWeave, Nebius,
IREN, Crusoe, Lambda, and Together AI.

Where hyperscalers offer breadth (databases, serverless, networking, dozens of regions), neoclouds
compete on depth: dense GPU fleets, software tuned for large distributed training/inference jobs, and
the newest Nvidia hardware generations (a commonly stated positioning, not verified here).

Business model

• Debt-financed buildouts — neoclouds are reported to borrow heavily (often via asset-backed loans collateralized
  by the GPUs themselves) to fund data-center construction and hardware purchases, betting that
  long-term customer contracts will cover debt service
• Long-term, creditworthy contracts — deals with major labs and hyperscalers (e.g. Microsoft, Meta,
  OpenAI) underwrite the debt; revenue is reportedly routed into structured “boxes” that prioritize opex and
  lender payments (as described in video commentary, not verified)
• Utilization is the core economic lever — profitability depends on keeping GPUs busy; utilization,
  availability, and scheduling efficiency drive unit economics far more than headline GPU count
• Nvidia as strategic partner — Nvidia has taken minority equity stakes in some neoclouds. For CoreWeave:
  a $6.3Bagreement(Sept2025)tobuyanyunusedcapacitythrough2032-04-13,andafurther$2B equity
  investment at $87.20/share (2026-01-26) to support >5 GW of AI factories by 2030. Some critics call this
  circular financing

Risks and bear case

• Depreciation vs. financing mismatch — commentators in the source videos argue GPUs are financed over a longer period than the useful life of a hardware generation (the specific 4–6 year and ~2 year figures are not verified), raising the risk that debt outlives the asset’s competitive value
• Capital intensity — data-center construction, power, and hardware costs are enormous and ongoing;
  commentary in the source videos warns that profitability may take longer than markets price in (opinion)
• Construction and power bottlenecks — per video commentary (not verified), near-term constraints are labor, supply chain and construction delays, with power expected to become binding later
• Case study: CoreWeave — reported to have grown from a crypto-mining origin. Early-2026 video figures (data-center, GPU, debt and cash-flow counts) are not verified and are omitted. Press-reported figures from Q2 2026 results (2026-08-11, secondary): revenue ~$2.58B(+112$104B (excluding >$25BofQ3commitments;anotherreportsays$129B), ~$35Btotaldebtatfacilityratesofroughly9-15datacenterswith1.5GWactiveand4.2GWcontractedpower,2026revenueguidance$12.4-13.2B.
  Insolvency risk is debated in the source videos; CoreWeave’s founders argue long-term contracts preserve GPU value
• Case study: Nebius — Amsterdam-based; Q2 2026 group revenue $582M(+454$3-3.4B
  revenue and $20-25Bcapex;>$40B contracted revenue from Microsoft and Meta, including a Meta deal of up to
  ~$27B (per secondary press summaries of the company shareholder letter; not read directly)
• Competitive squeeze — in the commentators’ view (opinion), neoclouds sit between hyperscalers and the risk that AI compute commoditizes

Why it matters

Neoclouds feature in the “AI bubble or infrastructure supercycle” debate: bulls frame them as the
AWS of the AI era — specialized infrastructure providers that will scale into indispensable picks-and-
shovels businesses — while skeptics point to their debt load, thin margins, and dependence on a small
number of AI-lab customers as classic markers of overbuilt capacity if demand growth slows.

See Also

• Nvidia — GPU supplier and, via equity stakes and compute backstops, strategic partner to several neoclouds
• CoreWeave, Nebius, Crusoe, Lambda, Together AI, IREN — named neocloud providers; none has a dedicated wiki note yet. As of 2026 reports: CoreWeave and Nebius are public; Lambda (reported Series E and IPO plans: not verified) and Crusoe are described as private in press (not verified); others named in 2026 rankings include Nscale and Fluidstack
• ai-api-platforms, inference-and-serving — related infrastructure notes

Open items

• Unverified: depreciation window (~4-6 years financed vs ~2 years obsolescence) and the early-2026 CoreWeave stats (33 DCs, 250K GPUs, $14Bdebt,-$110M/quarter) come only from video summaries; the Q2 2026 numbers come from press and aggregator reports (CNBC page blocked, 403), not the 10-Q. Backlog figure conflicts between sources ($104Bvs$129B).
• Together AI classed as neocloud in the videos; not confirmed in 2026 rankings.

Sources

• https://www.sec.gov/Archives/edgar/data/0001769628/000176962826000220/coreweave1q26earningspress.htm (CoreWeave 8-K, found via search, accessed 2026-10-02)
• https://www.investing.com/news/company-news/coreweave-q2-2026-slides-revenue-doubles-backlog-surges-246-93CH-4852949 and https://www.cnbc.com/2026/08/11/coreweave-crwv-q2-earnings-report-2026.html (search snippets, accessed 2026-10-02)
• https://techcrunch.com/2026/01/26/nvidia-invests-2b-to-help-debt-ridden-coreweave-add-5gw-of-ai-compute/ (accessed 2026-10-02)
• https://www.datacenters.com/news/coreweave-nvidia-s-6-3b-capacity-deal (accessed 2026-10-02)
• https://finance.yahoo.com/technology/ai/articles/nebius-q2-2026-earnings-beat-130010966.html (accessed 2026-10-02)
• https://www.marktechpost.com/2026/08/23/best-gpu-neoclouds-2026/ (accessed 2026-10-02)

Original sources:
yt-base video summaries (2025-10 through 2026-04): Caleb Writes Code (“The Economics of Neoclouds,” “How
CoreWeave is near insolvency”), CNBC (“Why Investors Are Bullish On Neoclouds”), Alex Kantrowitz
(interview with CoreWeave founders Michael Intrator and Brian Venturo), SiliconANGLE theCUBE (“Are
NeoClouds the Next AWS?”).