BTC 
ETH 
XRP
SOL 
DOGE - This topic is empty.
-
Posts
-
Nvidia is weighing a shift to glass substrates for its AI chips — a packaging upgrade that could sharply increase how much high-bandwidth memory (HBM) each chip can carry.
What happened. The Seoul Economic Daily reported on September 24 that Nvidia is working with German equipment maker SCHMID and other companies on glass core substrates, citing Investing.com’s coverage of SCHMID’s first-half earnings presentation. SCHMID disclosed it is collaborating with key suppliers in the supply chains of Intel, Nvidia, and AMD on glass substrate equipment. But SCHMID chief strategy officer Roland Rettenmeier cautioned that metallization of through-glass vias remains a technical hurdle and that final customer qualification has not yet been completed — so this is a serious R&D push, not a done deal.
Why it matters. Conventional plastic substrates are hitting their physical limits as AI chips run hotter and denser: warping and weak heat resistance cap how much HBM can be mounted. Glass substrates offer superior flatness and rigidity — raising chip speeds by roughly 40% while cutting power consumption roughly in half, with better thermal stability and lower signal loss. For Nvidia, more HBM per package means faster, more efficient AI accelerators, which is exactly the currency of the AI buildout — and a direct answer to the memory crunch already reshaping its roadmap. The signal matters for the whole supply chain: HBM makers and packaging-equipment names stand to benefit if glass substrates move toward mass adoption.
What to watch. Customer qualification timelines at Nvidia, AMD, and Intel; whether SCHMID or rivals announce glass-substrate equipment orders; and how the technology fits into Rubin-era packaging designs. If through-glass-via metallization gets solved faster than expected, the packaging trade gets crowded quickly.
Related reading: Nvidia Halts China H200 Output, Shifts TSMC Capacity to Vera Rubin; Musk Plans 3 Nvidia Chip Waves for Colossus 2; Google Tests AI Chips in Orbit
Sources: Seoul Economic Daily
Auto-posted by the Bull Lounge news desk.
Not financial advice.Which part of the AI supply chain do you think is the most underappreciated right now — memory, packaging, or networking?
Log in to vote
