Now Playing
CXMT's HBM3 Breakthrough, Memory Bottlenecks & Broadcom's Punished Growth
Audio is available on Spreaker — see link below.
China's ChangXin Memory achieves HBM3 manufacturing capability, threatening SK Hynix's dominance as memory bandwidth becomes AI's new rate-limiting factor. Plus Broadcom posts $10.8B in AI revenue — and gets punished for it.
Audio is available on Spreaker — see link below.
China's ChangXin Memory Technologies has achieved HBM3 manufacturing capability. That's not a roadmap claim or a government target.
The numbers behind CXMT's plans are worth taking seriously. The company is targeting three hundred thousand total wafers per month by end of twenty twenty-six, with around twenty percent allocated to HBM3.
Jensen Huang's visit to Seoul last week wasn't incidental. Nvidia's Vera Rubin platform, now entering full production in Q3 twenty twenty-six, requires HBM4 from all three qualified suppliers: Samsung, SK Hynix, and Micron.
The broader shift worth tracking is where scarcity is moving. GPU availability drove the first phase of the AI hardware race.
One more datapoint that fits this frame. Broadcom posted ten point eight billion dollars in AI revenue this quarter, with a hundred and eighty percent year-on-year growth rate.
The near-term tests are clear. Watch CXMT's yield data as it scales toward that sixty thousand wafer target.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.