A well-funded AI neocloud provider recently failed to close a colocation deal despite terms that looked extremely favorable on paper: a 15-year commitment, six months of charges prepaid upfront, millions in liquid cooling infrastructure cost coverage, and pricing at $155 to $160 per kilowatt. The colocation provider turned it down. The reason: the neocloud operator could not demonstrate investment-grade creditworthiness.
Robert West, head of revenue at TRG Datacenters, documented the pattern publicly: "Investment grade credit is the price of admission right now. It is not a negotiating point. It is the minimum requirement to get to the table."
Current large-scale colocation pricing: $140–$155/kW (datacenterHawk). The deal that failed was offering $155–160/kW — above market rate — with a 15-year term and 6 months prepaid. It still didn't close. Colocation providers are now evaluating: credit strength, customer demand visibility, confidence in sustained multi-year utilization, and balance sheet stability. Pricing has become secondary to creditworthiness.
For two years, supply constraints meant colocation providers could lease capacity to almost any willing tenant and fill it immediately. The market rewarded speed. Underwriting rigor was a secondary concern when waitlists were measured in months and facilities were leased before they were built.
That dynamic has shifted. Colocation capacity is coming online faster. The hyperscaler backlog has eased somewhat as their own construction programs come online. Colocation providers are now underwriting multi-year commitments the way project finance lenders do: focused on the counterparty's ability to service the obligation for the full term of the lease, not just for the first 12 months. A neocloud operator with $20 million in the bank and an 18-month runway is not a 15-year counterparty — regardless of what the term sheet says.
Neocloud providers — CoreWeave, Lambda Labs, Crusoe, and their peers — lease GPU infrastructure and resell it as cloud capacity. Their business model requires colocation capacity to operate. If they cannot close colocation deals, they cannot scale. ABI Research pegs the neocloud market at $250 billion by 2030, but that projection assumes companies in the category can continue accessing infrastructure. The credit tightening is a direct threat to that assumption.
Matt Brown, COO at Core Scientific, identified the core issue as "credit and underwriting risk" and its impact on loan-to-cost ratios for data center financing. Colocation providers are themselves financed against projected lease income. A tenant who defaults on a 15-year commitment does not just create an empty rack — it creates a covenant violation on the colocation provider's construction debt. The underwriting shift reflects that downstream exposure being priced into deals at the colocation level.
The cooling industry sits downstream of neocloud deal flow. When a neocloud operator cannot close a colocation deal, the CDU purchase order that would have been issued for that deployment does not get written. The cold plates, manifolds, and cooling infrastructure that would have served that facility do not get ordered.
This is a different constraint than tariff exposure or lead time. It is a demand-side credit event — a market segment that was driving meaningful share of cooling procurement growth is now facing access issues to the infrastructure it needs to operate. Hyperscaler cooling demand remains strong. Neocloud cooling demand is dependent on a credit market that just tightened significantly. Cooling vendors who built their 2026 pipelines around neocloud deal assumptions should be updating their models.