Google has announced its eighth-generation TPUs (Tensor Processing Units), the TPU 8t and TPU 8i, which are specialised ASICs for AI. They are specifically designed for AI training (TPU 8t) and ...

Google's eighth-generation TPUs split training and inference into two specialised chips. Here's how TPU 8t and TPU 8i work, ...

Google used Cloud Next '26 to announce agent platforms, new TPUs, cross-cloud infrastructure, networking and data services aimed at AI workloads.

Here is how you know that GenAI training and GenAI inference are very different computing and networking beasts, and ...

Google's Ironwood TPU is live with 4.6 petaFLOPS per chip. Its eighth-gen splits into two: Broadcom for training, MediaTek for inference, both at 2nm in late 2027 ...

Alphabet introduced its 8th-generation tensor processing units: TPU 8t optimized for training and TPU 8i for inference ...

Alphabet Inc's GOOGLGOOG Google is sharpening its AI hardware strategy by redesigning its chips to improve efficiency and ...

Google is sniffing around Marvell for yet more silicon. Talks between Google and Marvell have started on two new chips aimed at making AI inference less of a slog. One chip is a memory processing unit ...

Google is discussing two new chips with Marvell Technology for AI inference, adding a third design partner to its TPU supply chain as custom ASIC sales are set to grow 45% in 2026.

Modern AI is no longer powered by a single type of processor—it runs on a diverse ecosystem of specialized compute architectures, each making deliberate tradeoffs between flexibility, parallelism, and ...

An analyst worries about a line from a recent Broadcom filing that suggests the company may be somewhat involved in the funding of an Anthropic data center Broadcom's stock has more than doubled over ...

Discover how an NPU neural processing unit and AI chip enable fast, private on-device AI, and why your next AI PC laptop should include this powerful technology. Pixabay, jarmoluk Neural processing ...