Building the market infrastructure for global AI compute
The market is fragmented on both sides
AI compute infrastructure is expanding rapidly across the world, but the market remains fragmented. Capacity is distributed across regions, data centres and infrastructure providers, while global AI demand varies by GPU architecture, network configuration, workload, duration, jurisdiction and service level.
The result is that a buyer cannot easily discover what is available, and an operator cannot easily discover who wants it. Price is negotiated case by case, with no shared reference.
A standardised market interface
MACOMX is being developed to create a standardised market interface between global AI demand and compute infrastructure: a common way to describe an instrument, state a quote, verify capacity and record a delivery.
That interface has to be built before it can be traded on. The first step is collecting the two things it needs — real capacity and real demand.
What we are trying to improve
Five outcomes, in the order they become possible.
Transparency
Every number states where it came from, when it was reported and whether it can be transacted.
Utilisation
Idle accelerators are a stranded asset. A market for short terms is how they get used.
Price discovery
A price means something once the same instrument is quoted by more than one provider.
Delivery reliability
A quote is only worth what the delivery behind it is worth. Verification is part of the product.
International access
Demand and supply frequently sit in different jurisdictions. Routing between them is the hard part.
Who is behind it
MACOMX is initiated by H&N Labs Pte. Ltd., Singapore.
Contact
Direct lines by subject. Every message reaches a person, not a queue.
MACOMX becomes a market the moment someone quotes on it.
Until then it is an interface waiting for its first real number.