The AI infrastructure market just crossed a threshold. With GPU-backed debt now underwritten at scale, the constraint on AI compute growth is no longer capital — it is creditworthy enterprise demand. Analysts project AI debt financing will become a multi-trillion-dollar credit market by the end of the decade, second only to U.S. mortgages. Every dollar of that financing rests on one thing: enterprises willing to sign multi-year compute commitments.

Those enterprises are the ONUG Community.

For 15 years, ONUG has been where enterprise IT leaders work through infrastructure transitions together — 100,000 members across 600+ Global 2000 companies, representing $1.5 trillion in IT spend. Critically, 86% of our members hold direct purchase authority. These are the executives who will sign the offtake agreements that anchor the next wave of AI infrastructure financing.

There’s just one problem: no playbook exists. How should an enterprise structure a multi-year GPU commitment? What tenor, what pricing model, what escape clauses? How do you evaluate a neocloud provider against a hyperscaler? What does creditworthy offtake even look like from the buyer’s side of the table?

Today we’re announcing the program to write that playbook.**

WG4: From PoC to Production: AI FinOps, Governance & Compute Offtake

Under the ONUG Collaborative, Working Group 4 brings together Fortune 500 infrastructure, finance, and procurement leaders to put AI FinOps and governance in the context of multi-year compute deals. The charter: define how enterprises forecast utilization across a 3–5 year commitment, account for compute obligations, govern internal drawdown against them, and evaluate and negotiate the agreements themselves. The output — evaluation criteria, contract structures, governance frameworks — will shape how the Global 2000 buys AI compute for the next decade.

The Compute Offtake Track at the AI Networking Summit NYC

At the AI Networking Summit, October 28–29 in New York City, a dedicated track will put suppliers and buyers in the same room: enterprise operators presenting their requirements, providers presenting their models, and structured sessions where the two sides work the problem together in real time in private rooms.

An Invitation to the Supply Side

This program only works with both sides of the market at the table. We are inviting the companies building the AI infrastructure stack to participate:

Neoclouds — CoreWeave, Nebius, Crusoe, Lambda, Together AI, and their peers — for whom enterprise offtake diversifies the customer book and directly improves financing terms.

– **AI infrastructure suppliers** — Nvidia, Cisco, Arista, HPE, Dell and their peer plus the networking, compute, and power ecosystem — whose roadmaps should be informed by the requirements enterprises define here.

Providers who participate help shape the frameworks. Providers who don’t will respond to them later in RFPs.

Participation runs through sponsorship of the AI Networking Summit and membership in the ONUG Collaborative, which includes a seat in WG4, track participation, and direct engagement with the enterprise executives structuring these commitments.

To get involved: Companies A–L, contact Paul at paul@onug.net. Companies M–Z, contact Henry at henry@onug.net.

The frameworks are being written now. The seats are being set now. Be in the room.