Bull has secured a €387.8 million contract to build LUMI-AI, a EuroHPC supercomputer set for 2027. Notably, it uses AMD chips to diversify from…
Seven. That’s how many generations of Google’s Tensor Processing Unit Amir Salek shipped before he left in 2022, and it matters more than his…
Cerebras unveiled the CS-4, a powerful server rack featuring three wafer-scale chips, promising significant throughput improvements. However, despite reported advancements and a $180.1 million…
Sony's PSSR 2.0 significantly reduced training time from four GPU-months to under one week by simplifying its neural network's functions. It switched from a…
TSMC's 3nm production is set to reach 180,000 wafers monthly by early Q4, ahead of schedule due to high demand from Nvidia, AMD, and…
AMD has agreed to acquire Taalas, a startup focusing on specialized inference silicon that embeds trained model weights directly into chips, enhancing efficiency but…
AMD and Anthropic announced a partnership involving 2 gigawatts of AMD GPUs in Helios systems, supported by a $5 billion equity investment. The collaboration…
TSMC is expanding its CoWoS outsourcing to OSAT partners due to overwhelming demand, particularly from Nvidia. Despite doubling capacity, TSMC will still fall short…
The European AI sector faces a significant challenge: while purchasing AI accelerators is quick, connecting them to the power grid can take seven to…
AMD Helios, unveiled at the Advancing AI 2026 event, is entering production, shipping with major clients like Microsoft. Corporate VP Andrew Dieckman downplayed CUDA's…
Zhaoxin, a joint venture with VIA Technologies, holds a unique position as China's lone x86 chipmaker without U.S. sanctions. Despite being behind in technology,…
Hygon, a Chinese server chip manufacturer, evolved from a 2017 AMD EPYC architecture, leveraging a 2016 licensing deal. Despite U.S. sanctions halting new IP…
Chiplets and UCIe are unbundling AI hardware the same way Modular unbundles CUDA. Why the moat doesn't vanish, it just relocates to TSMC's packaging…
MIT's 2024 reasoning model and AMD's Ryzen AI chips proved on-device AI code generation could work. Two years later, the shift is complete, and…
128 GB of unified memory, a 40-core RDNA GPU, and local LLM inference nobody can revoke. The GMKtec EVO-X2 is the hardware argument for…
One maker's tech digest, weekly. Chips, AI, gaming, EVs — no spam.