News on the AI infrastructure buildout: data centers, networking, fiber, power, and the physical backbone of artificial intelligence.
The Broadcom Anthropic loan could cover a third of a $125.2 billion TPU lease, and Anthropic's own prospectus flags the supplier-lender overlap as a…
Nvidia's Open Agent Safety Platform moves agent enforcement into a BlueField-4 DPU. It is a strong design, and analysts say it covers less than…
AMD is buying a roughly 70-person world-model startup for $8.2 billion in stock, and the price looks like a payment for influence over how…
Eddie Wu announced Alibaba's Zhenwu V900 AI chip, claiming it triples performance over the previous model, alongside a Qwen model with 5 to 10…
Huawei's Ascend 960DT now lands in Q1 2027 and the 960PR in Q3, feeding a 4,096-NPU Atlas 960E SuperPoD built on near-packaged optics. The…
Apple, after fifteen years away from the enterprise server market, is reportedly planning a return by 2029, focusing on AI infrastructure with its M8…
Marc Camus announced BNP Paribas's five-year cloud and AI partnership with Google Cloud, emphasizing limitations in the deal. While utilizing Google's technology, BNP retains…
A Soviet machine called Setun ran on minus one, zero and plus one from 1958 and was killed off in 1965. Ternary LLM weights,…
Axelera Europa is now shipping in validated Dell and Supermicro systems and heading into the EU AI Factories in Italy and Luxembourg. The 629…
NVIDIA introduced open-source software, CUDA-Q Logical, at IEEE Quantum Week, significantly reducing Fermilab's fault-tolerant quantum design cycle from five months to three weeks. This…
The Salesforce Koa reasoning model is post-trained from Nvidia's Nemotron 3 Super and runs entirely inside Salesforce's trust boundary. A technical breakdown, how it…
The ASUS Ascent QN10 is the first Snapdragon X2 Elite mini PC you can actually buy, at $1,349. Eighteen Oryon cores and an 80…
This week, differing definitions of AI sovereignty emerged from New Delhi, Brussels, and Beijing. India proposed a stringent cloud sovereignty framework, bypassing diplomacy, while…
Meta's third-generation MTIA inference chip landed from TSMC on September 1 running within 2-3% of simulation, and a fourth generation is already a month…
TSMC plans to increase its 2nm capacity from 90,000 to 110,000 wafers monthly by mid-2027 and aims for significant growth in its 3nm and…
One maker's tech digest, weekly. Chips, AI, gaming, EVs — no spam.