Token Economics Reshape AI Infrastructure
SiliconANGLE, Wednesday, September 30th, 2026
theCUBE analysts say inference growth makes tokens per watt the key metric, positioning CoreWeave beyond GPU compute.
In theCUBE's keynote analysis at CoreWeave's Fully Connected event, covered by SiliconANGLE, Dave Vellante and John Furrier argue token economics are becoming the defining measure of AI infrastructure efficiency.
Vellante says CoreWeave's customers are at a 50/50 training-to-inference split and expect 10/90 next year, a shift that favors system-level optimization across silicon, networking, storage and software, which CoreWeave is pursuing as it expands beyond GPU compute.
Customers are also pushing for shorter contracts, spot pricing and on-demand access, which carry higher prices than long-term take-or-pay deals.