Now Playing
Vera Rubin Ships, HBM4 Costs Surge 435% & China's CPU Supercomputer
Audio is available on Spreaker — see link below.
Nvidia's Vera Rubin NVL72 hits production with eight simultaneous cloud deployments as HBM4 memory costs soar 435% over Blackwell — plus Apple's 1.4nm defensive pivot, TSMC's 30% revenue surge, Intel Foundry's $5.4B loss-making quarter, and China's CPU-only TOP500 champion. The manufacturing layer of AI is the story right now.
Audio is available on Spreaker — see link below.
Apple is accelerating to a node that doesn't fully exist yet. The company is planning to move to TSMC's one-point-four nanometre process by twenty-twenty-eight, roughly two generations ahead of its historical cadence.
Nvidia's Vera Rubin is moving from announcement to production. The NVL seventy-two rack ships in the second half of twenty-twenty-six, and eight cloud providers are deploying simultaneously.
TSMC's May revenue came in at four hundred and sixteen-point-nine-eight billion New Taiwan dollars, roughly thirteen-point-two billion US dollars. That's a thirty percent year-on-year increase.
Intel Foundry posted five-point-four billion dollars in Q1 revenue, backed by partnerships with Alphabet, SpaceX and Tesla's Terafab operation, and other hyperscalers. The important distinction here is between revenue validation and profitability.
China reclaimed the top spot on the TOP500 supercomputer list with LineShine. The system runs two thousand, one hundred and ninety-eight exaflops using forty-five thousand domestic LX-two CPUs and Kylin OS.
Three signals worth tracking closely. Whether TSMC officially confirms Apple's one-point-four nanometre timeline.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.