Hyperscalers
What Does Oracle Do? The Hyperscaler Capacity Model
Oracle books AI cloud contracts, delivers halls and GPUs, and gets paid as that capacity is accepted — delivery, not the backlog print, is the stuck step.
The Stack
Hyperscalers — Capex Pull and the Delivery Bind
Status Tightening as of 24 Sep 2026: after record hyperscaler capacity adds, fleets stay near-full — cash sticks at delivery and attach versus contracted backlog, not the booking desk alone.
Home / The Stack / Hyperscalers
Hyperscalers are the large cloud and platform buyers whose capital spending pulls the whole AI stack. This layer defines them and maps how their cheques set pace.
Status: Stable · Updated: 2026-09-12
Investor read: They are demand aggregators, not the scarce finish-line. Capex guidance still runs high; the binding constraints usually sit in chips, power, or cooling beneath them. See Hyperscaler explained.
On this page
- What this layer is
- Where the money sticks
- Splits
- Who does what
- What would change the view
- Related reading
- FAQ
What this layer is
A hyperscaler operates cloud regions at massive scale and buys GPUs, networking, power, and buildings in multi-year waves. Their budgets set pull for every other layer.
Where the money sticks
Stable: they set volume. Money and scarcity often appear one step down the map when their orders hit packaging, HBM, or interconnects.
Splits
| Split | What to watch | Typical signal |
|---|---|---|
| Capex vs returns clock | Spend now vs prove ROI later | Guidance vs product monetization |
| Merchant GPU vs custom silicon | Bought accelerators vs ASICs | Supply diversification |
| Lease vs own | REIT campuses vs self-build | MW control and speed |
Who does what
Examples of exposure only — not recommendations.
| Role | Examples |
|---|---|
| Large public hyperscalers / platforms | Microsoft, Amazon, Alphabet, Meta, Oracle |
Always verify listing, ADR, and tax treatment for your jurisdiction.
What would change the view
Falsifiable flips:
- Sustained capex cuts that free upstream supply across chips and power.
- Clear proof that AI spend earns returns fast enough to extend the cycle.
- Custom silicon displaces a large share of merchant GPU demand.
Related reading
Explained children: Neocloud explained · Neocloud vs hyperscaler.
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 hyperscaler?
A large-scale cloud or platform operator whose infrastructure spending pulls suppliers across the AI stack.
Why Stable?
Capex remains large, but they are usually not the scarce physical bottleneck themselves.
How do they link to neoclouds?
Neoclouds often serve overflow or specialist demand when hyperscaler capacity is full or specialised.
Not investment advice. Do your own research.
Last updated: 2026-09-12 · Second Order by TKN