AWS Chief Says Space AI Compute Not Viable Yet
AWS CEO Matt Garman stated on a podcast that space data centers for AI computing are unlikely to be commercially viable for the next 10 to 20 years. He cited high launch costs, payload limitations, and heat dissipation challenges as major obstacles.
Summarized by AI from one outlet
Source
- IT之家Read original ↗
Latest
- Chips
Baidu Announces Kunlunxin M300 Next Year, N-Arch Chips in 2028
At Baidu World 2026, Baidu announced its AI chip roadmap, with the flagship Kunlunxin M300 launching next year and multiple N-architecture chips planned for 2028. The company stated that inference demand will be the main driver for long-term AI chip growth, highlighting its commitment to compute infrastructure.
- Chips
Microsoft patent reveals renting idle GPUs for AI training
A new Microsoft patent details a system to rent idle consumer GPUs for AI model training. The scheme may offer rewards to participants, aiming to distribute compute resources for AI development.
- Chips
Intel reveals 256-core Diamond Rapids Xeon for AI infrastructure
Intel unveiled physical samples of its next-gen Diamond Rapids Xeon 7 processor, featuring 256 performance cores on an 18A-P process. The chip is optimized for agentic AI infrastructure and is expected to launch in 2027.
- Chips
Google launches TPU 8tx and 8i Max chips
Google announced two new 8th-generation chips, TPU 8tx and TPU 8i Max. TPU 8tx integrates with AMD to enhance confidential computing, while TPU 8i Max targets ultra-long context inference.
- Chips
AI Data Center Projects May Never Materialize Due to Hurdles
AI data center expansion is facing local opposition, permitting delays, and power shortages. Analysis suggests many announced multi-billion dollar AI infrastructure projects may never actually be built, affecting compute supply expectations.
All news → · All rights belong to the original creators