Sign up to save your podcasts
Or
Share 'The Use Cases are There, the Infrastructure is Not'
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
'The Use Cases are There, the Infrastructure is Not'
Hyperscalers increasingly will realize they don’t need mega data centers, which, by the way, are becoming targets in military conflicts, according to industry veteran Tony Grayson.
As the former President of Compass Datacenters and then Northstar Enterprise and defense, Grayson has long been at the forefront of modular data centers that serve the edge compute needs of the digital infrastructure landscape.
On the sidelines of the 2026 PTC event in Honolulu, Grayson spoke with Cool Vector about a range of topics. As a former commander of a US Navy nuclear submarine, Grayson has strongly informed views on technology in mission-critical operations, nuclear energy in digital infrastructure, and the growing awareness of data centers as being vulnerable to armed conflict.
Among the key takeaways from this Cool Vector interview:
• The data center industry’s declared capacity pipeline is largely fictional, built on LOIs, inflated announcements, and wishful timelines that serious capital allocators are only now beginning to challenge.“You can sell what you don’t have with software,” says Grayson. “It doesn’t work that way for infrastructure because there are real costs.”
• The era of the mega data center as the default build thesis is running headlong into a hardware reality where rapid chip-set obsolescence can strand billions in capital before a single rack is powered on. Says Grayson: “The data center that takes you 12 to 24 months to build that you’re halfway through for Grace Blackwell is now not built for the latest chip set — and these are assets you haven’t even got yet that you spent a lot of capital on.”
• The real money in digital infrastructure will be made not in training campuses but in distributed, latency-sensitive inference infrastructure — a build-out that demands an entirely different architecture, operating model, and geographic logic than what the market is currently chasing. “The use cases are there. The problem is the infrastructure’s not,” Grayson tells Cool Vector.
• The US military’s post-Ukraine pivot to distributed compute has made defense — historically a follower of enterprise infrastructure trends — the unexpected leading indicator for where the entire industry is headed: “You’re better off taking that one data center, put it in 10 different spots, and getting resiliency, backup, replication across those sites — or just make them harder to go after in the early stages of a conflict. That’s what Russia did with Ukraine.”
Access the transcript and a searchable archive of primary-source market intelligence on the Cool Vector Substack: https://coolvector.substack.com/p/the-use-cases-are-there-the-infrastructure
#digitalinfrastructure #datacenter #coolvector #navy #energy
View all episodes
By david95a
5
33 ratings
Hyperscalers increasingly will realize they don’t need mega data centers, which, by the way, are becoming targets in military conflicts, according to industry veteran Tony Grayson.
As the former President of Compass Datacenters and then Northstar Enterprise and defense, Grayson has long been at the forefront of modular data centers that serve the edge compute needs of the digital infrastructure landscape.
On the sidelines of the 2026 PTC event in Honolulu, Grayson spoke with Cool Vector about a range of topics. As a former commander of a US Navy nuclear submarine, Grayson has strongly informed views on technology in mission-critical operations, nuclear energy in digital infrastructure, and the growing awareness of data centers as being vulnerable to armed conflict.
Among the key takeaways from this Cool Vector interview:
• The data center industry’s declared capacity pipeline is largely fictional, built on LOIs, inflated announcements, and wishful timelines that serious capital allocators are only now beginning to challenge.“You can sell what you don’t have with software,” says Grayson. “It doesn’t work that way for infrastructure because there are real costs.”
• The era of the mega data center as the default build thesis is running headlong into a hardware reality where rapid chip-set obsolescence can strand billions in capital before a single rack is powered on. Says Grayson: “The data center that takes you 12 to 24 months to build that you’re halfway through for Grace Blackwell is now not built for the latest chip set — and these are assets you haven’t even got yet that you spent a lot of capital on.”
• The real money in digital infrastructure will be made not in training campuses but in distributed, latency-sensitive inference infrastructure — a build-out that demands an entirely different architecture, operating model, and geographic logic than what the market is currently chasing. “The use cases are there. The problem is the infrastructure’s not,” Grayson tells Cool Vector.
• The US military’s post-Ukraine pivot to distributed compute has made defense — historically a follower of enterprise infrastructure trends — the unexpected leading indicator for where the entire industry is headed: “You’re better off taking that one data center, put it in 10 different spots, and getting resiliency, backup, replication across those sites — or just make them harder to go after in the early stages of a conflict. That’s what Russia did with Ukraine.”
Access the transcript and a searchable archive of primary-source market intelligence on the Cool Vector Substack: https://coolvector.substack.com/p/the-use-cases-are-there-the-infrastructure
#digitalinfrastructure #datacenter #coolvector #navy #energy