GoKawiilAnthropic has announced it's building an in-house chip development team to co-design its own custom ASIC processors for handling AI inferencing workloads. As described to Business Insider, Anthropic is starting to hire engineers to design the chips with an unspecified partner, and it looks set to do it at pace, with the job listing saying that any potential hiree would need to be ready to work to a schedule and get the chip design over the line.
This is just the latest major AI company to announce it's developing its own custom hardware. As the global economic shortages squeeze chip supply and models increasingly lean on optimizations to make workloads more efficient and potentially profitable, making custom silicon for your own data centers makes a lot of sense. Anthropic now joins the likes of Google, Meta, Microsoft, Amazon, and OpenAI in building their own chips for the job.
If you want something done right...
It's no secret that if you want to train an advanced AI model, you need Nvidia GPUs. Even Chinese AI developers, who have the ruling party leaning on them and limited access to Nvidia hardware, still use Nvidia GPUs — even if they have to smuggle them first.
Latest Videos From Tom's Hardware Watch full video here:
But if you're looking to run AI to perform inferencing workloads for agentic and generative AI models, you can use a much wider array of hardware. Nvidia GPUs are good, but they're expensive — custom silicon can have a lower total cost of ownership of up to 65% — and power-hungry, and there are much more efficient options available. Chinese labs are using domestic Chinese hardware, and many Western AI developers have their own solutions; those that don't are making them.
Google has been building its Tensor Processing Unit (TPU) chips for 12 years, working with Broadcom to develop each generation. Amazon has its Trainium and Inferentia chips, and Meta recently announced several new MTIA designs for deployment through 2027. Microsoft has its Maia line, and Tesla recently pivoted to its AI5 and AI6 chip designs after years of developing Dojo.
And now Anthropic is getting in on the act, and for much the same reasons. Anthropic told Business Insider that it was co-designing the chips so that they would allow Claude to run faster and more efficiently at the scale its customers need. Indeed, Anthropic has seen explosive growth in the past year, seeing huge expansion in the consumer space and taking on significant government contracts — not to mention agentic AI requiring far more tokens than traditional single-prompt interactions.
Anthropic hasn't revealed which firm it's working with on the design and development. Although Broadcom and Marvell are the two largest companies in the ASIC co-design market, representing some 95% of it, The Information reported last month that Anthropic was in talks with Samsung for manufacturing.
The shovel sellers always benefit
... continue reading