Neoclouds, Explained: The Rented-GPU Layer
Rented GPUs and specialist compute — the neocloud layer of the AI Stock Map.
Neoclouds rent GPU capacity to builders who need clusters without owning the full stack. This layer covers the rental model, utilization risk, and links to chips and power.
Status: Stable · Updated: 2026-09-12
Investor read: Neoclouds sit between scarce GPUs and hungry users. Unit economics hinge on utilization, power cost, and customer concentration — not only headline GPU scarcity.
On this page
What this layer is
A neocloud is a specialist compute landlord: owned or financed GPUs, often in leased or partnered data centres, sold by the hour or contract. See also Neocloud explained.
Where the rent sits
Stable: rental pricing and utilization are mixed. Physical scarcity often still sits upstream in semiconductors and power.
Splits
| Split | What to watch | Typical signal |
|---|---|---|
| Owned vs leased GPUs | Balance sheet vs variable cost | Capex intensity and financing |
| Training vs inference mix | Sticky contracts vs spot | Utilization and pricing power |
| Hyperscaler vs specialist | Who wins overflow demand | Capacity and network effects |
Short roster
Examples of exposure only — not recommendations.
| Role | Examples |
|---|---|
| Listed neocloud / GPU cloud examples | CoreWeave, Nebius, Applied Digital |
| Related pivots | Some former miners / hosting names (verify model) |
Always verify listing, ADR, and tax treatment for your jurisdiction.
What would change the view
Falsifiable flips:
- GPU rental prices fall faster than cost of capital for a sustained period.
- Hyperscalers absorb overflow so specialist utilization stays weak.
- Power or chip delivery blocks fleet growth even when demand is firm.
Related reading
Who writes this
Second Order by TKN maps constraints for international readers. Educational only.
FAQ
Is this a buy list?
No. Exposure map only. Not investment advice. Do your own research.
What is a neocloud in one sentence?
A specialist operator that rents GPU clusters to customers who need AI compute without owning the full stack.
Why Stable?
Demand exists, but pricing/utilization are not the single scarcest physical gate right now.
What kills the model?
Chronic under-utilization, expensive power, or customers leaving for hyperscaler capacity.
Not investment advice. Do your own research.
Last updated: 2026-09-12 · Second Order by TKN