The Verge reports that OpenAI is using non-disclosure agreements around its data centre work, a sign that the physical build-out behind AI services is becoming harder for outsiders to assess. For professionals buying or governing AI services, the issue is visibility into the infrastructure on which those services depend.
Why the NDA matters
The AI discussion often starts with model behaviour, application design and user controls. Data centres sit underneath that layer. Training and running large AI systems require compute, power, cooling and network capacity, and those dependencies affect cost, resilience and risk.
A non-disclosure agreement does not itself show wrongdoing. It does change the information environment around a project. It can limit what contractors, partners or other parties can say publicly, while not removing whatever planning, environmental or procurement obligations apply in a given jurisdiction.
What enterprise teams can ask
IT leaders should avoid treating AI capacity as an abstract cloud resource. If a service depends on large new facilities, procurement teams need questions about regions, continuity, energy constraints, support routes and contractual remedies. Some of those answers sit outside model documentation.
Where a provider will not disclose facility-level detail, customers can still ask for service-level commitments, data residency terms, incident notification language and exit options. Those are familiar controls from cloud contracting. AI demand gives them more operational weight because capacity constraints can affect rollout plans, latency assumptions and continuity planning.
The practical point is not whether every facility detail should be public. It is whether customers and public bodies have enough information to assess dependency, exposure and cost. If those answers are covered by confidentiality terms, the governance work moves into contracts, architecture decisions and risk reviews.
When AI services depend on data centre capacity that customers cannot fully inspect, the practical questions move to Azure architecture, hybrid cloud, cost and platform engineering in ECS's Infrastructure topic. Sessions and workshops give attendees room to compare architecture choices, contracting assumptions and practitioner experience from live platform work. Bring the capacity, region and supplier questions that are difficult to settle from public statements alone.
This article was written by the ECS editorial team with the assistance of Amal Agent, run.events' AI assistant, which supported the research, drafting and proofreading.