AMD CEO Lisa Su says the company plans to "substantially increase" supply in 2027 and needs "more advanced wafer capacity", as she meets Foxconn and TSMC
THE SO WHAT
AMD planning a substantial supply ramp in 2027 and courting more advanced wafer capacity signals that today’s GPU scarcity is a timing problem, not a permanent condition. If you’re locking in long-term AI infra contracts, model scenarios where accelerator prices and availability look very different three years out.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechThe best 3D printer deals on October Prime Day: Creality, Bambu Lab, and more on offer right now
Discounted high-speed and multi-color 3D printers on Prime Day point to desktop additive moving from hobbyist to accessible shop-floor and lab tooling. If you run hardware prototyping or small-batch production, this is a low-friction window to stand up or expand in-house print capacity instead of waiting on external queues.
Deep & Emerging TechNvidia Is on the Verge of a $6 Trillion Market Value
A near-$6T valuation with another ~30% run-up this year tells you the market still prices Nvidia as the core toll operator for AI compute. For operators, that means continued pricing power and potential supply tightness — budget and architecture plans should assume GPU cost volatility, not relief.
Deep & Emerging TechAMD CEO Sees Very High Chip Demand, Urges AI Firms to Cooperate
“Very high” multi‑year chip demand from AMD is another data point that AI compute scarcity is structural, not a one‑quarter anomaly. If you depend on accelerators, lock in multi‑year supply where you can and design workloads to be portable across vendors to benefit from any cooperation push.
Deep & Emerging Tech10+ Prime Day power stations deals: We found Jackery, Anker, and Bluetti deals to shop
Discounted portable power stations from Jackery, Anker, and Bluetti push backup power deeper into the consumer base—raising expectations for always‑on connectivity during outages and travel. If your product is usage‑sensitive to power loss, assume more users will have local storage and design offline and recovery flows accordingly.