Now Playing
Kyber Delay, Micron's Long-Term AI Contracts & Singapore GPU Diversion | Ep 1
Audio is available on Spreaker — see link below.
Nvidia's Kyber NVL144 slips to 2028 due to a PCB manufacturing limit, dragging NVL576 cluster timelines with it — and the downstream risk for hyperscalers is real. Plus: Micron locks 16 long-term HBM contracts through 2030, a Singapore executive faces money laundering charges over diverted Nvidia GPU servers, and Intel's foundry gap widens.
Audio is available on Spreaker — see link below.
Nvidia's most ambitious rack-scale system just got pushed back a year, and the reason matters more than the headline. The Kyber NVL144, the system that packs one hundred and forty-four GPUs into a single rack using a vertical-mount tray design, was widely expected to reach hyperscalers in twenty-twenty-seven.
The downstream risk is what makes this more than a product schedule story. Kyber racks are designed to link together into the NVL576 system, eight racks connected optically to form a single massive compute cluster.
While Nvidia's supply chain shows stress, Micron is moving in the opposite direction. The company has locked sixteen long-term AI memory contracts through twenty-thirty, with price floors and ceilings built in.
On the export control front, a case in Singapore is getting more serious. Executives at Aperia Group now face money laundering charges alongside the original fraud allegations.
One more number worth keeping in mind. Intel's market cap sits at five hundred and twenty-eight billion dollars.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.