Compute availability is the primary bottleneck complicating most organizations’ AI strategies. Plenty of ink has been spilled regarding GPU/CPU availability, but many other core components are constrained too: Memory costs are up, with lead times as long as 30–40 weeks. The impact from memory shortages compounds as related aspects of AI infrastructure beyond just compute are also affected. One example is the availability of network…

The post Overcoming AI Supply-Chain Bottlenecks Isn’t Just About the Compute appeared first on Interconnections - The Equinix Blog.