Follow your favourite podcasts, listen offline and in the car with CarPlay and Android Auto, and always pick up where you left off. Free to try.
Two Million Dollars of Memory: Inside Nvidia's 15% AI Server Price Hike
Nvidia has told Microsoft, Alphabet and Oracle that AI servers built on Vera Rubin and Grace Blackwell will cost more than 15% extra in early 2027, and the reason is not the GPU. Morgan Stanley estimates the next generation rack costs about $7.8 million to build, with roughly $2 million of that in memory alone, a jump of about 435% in a single generation. Chris and Laura dig into the inversion now reshaping AI infrastructure economics: graphics processors have fallen from more than 80% of an AI server's cost to roughly half, HBM and flash are becoming the real chokepoint, and custom silicon does not let anyone escape it.
Hosted by Chris and Laura.
The DX Today Podcast brings you daily deep dives into the most consequential stories in the AI ecosystem.
