Ecosystem

Twenty portfolio companies, five domains, one network. Our partners are both consumers of compute and the people who land it in production — the platform turns cards into dependable tokens, they turn tokens into something that runs at the customer site.

20
Portfolio companies
12
Listed
5
Domains
40+
Customer countries
Portfolio companies

Who is on the network

Grouped by role: consumers burn compute on the platform, suppliers feed compute or scheduling capacity into it, and channel partners carry platform capability to their own customers. The twentieth portfolio company, Shengshu Technology, appears in the next section as a model partner.

—
Consumer Runs inference and training workloads on the platform Supplier Supplies compute, silicon or scheduling capacity Channel Carries platform capability to its own customer sites
Model partners

Five leading models,
all deployable on our own nodes

Four open-weight flagships plus one closed model from our portfolio, all deployed as weights on our own nodes rather than proxied through anyone’s public API. Model version, scheduling policy and data residency therefore stay under platform control.

Why “deployable” is the word that matters. Reselling someone else’s public API means latency, availability and data flow are all outside your control. Only when the weights sit on nodes you operate can you talk about heterogeneous scheduling, in-country residency and inference acceleration. View the compatibility matrix →

Become a partner

If you work on vertical deployment, model development or compute supply, we would like to keep extending this network. Let us talk.

Contact us → Browse solutions