MACOMX Intelligence
Market intelligence for AI compute, power and digital infrastructure.
What compute costs and where it sits
Pricing, availability and utilisation across GPU architectures and regions — the inputs a buyer needs before committing to a term.
- GPU pricing across regions and contract terms
- Capacity availability and provisioning lead time
- Regional supply and demand imbalance
- Utilisation of installed accelerators
What sits underneath the price
A GPU-hour is the output of a power contract, a building, a network and an operator. Compute price cannot be read without them.
- Data centre economics — capex, opex, PUE
- Power economics and tariff structure
- GPU infrastructure and cluster design
- Network and interconnect economics
Research coverage
Five standing programmes. Published research is added here as it is completed — nothing is listed before it exists.
MACOMX does not publish research it has not produced. Each programme will carry its own sources, reporting period and methodology when it is released.
Index products
Reserved. An index is only meaningful once it rests on executed transactions, so MACOMX will not estimate or model one before then.
Where the numbers come from
Intelligence is only as good as its attribution. Every figure MACOMX publishes states its source, its reporting period and whether it is reference data or an executable quote.
Official statistics establish market size. Public market information provides comparison. Direct provider submissions describe real capacity. Only MACOMX market data describes what actually transacted — and at V0.1 that category is still empty. Read the full methodology.