Amazon is reportedly making a last-minute play to buy TikTok as the ban deadline looms

After that incident, Waymo instituted an “interim remedy” to make the vehicles “exclude additional operating conditions that present an elevated risk of encountering a flooded, higherspeed roadway,” but added that it was still “developing the final remedy for this recall.”

As we’ve noted, Waymo has mostly kept its rollout — now public in 11 cities — to more temperate climates, as severe weather poses more challenges to autonomous vehicles.

But as the role of inference has shifted to a top priority, with companies focusing on actually running models to make money, they’ve started shopping around, buying chips tailored for the task, and in some cases decided to make their own.

Additionally, Anthropic has become something of a victim of its own success at rolling out products that can be quickly adopted by enterprise clients. That rapid, wide-scale adoption has revealed significant compute constraints. Anthropic is now, effectively, looking for any and all compute capacity it can find, striking deals with CoreWeave, Amazon, Google and Broadcom, and even xAI.

Amazon and Google have both seen hot demand for their custom inference chips. But Microsoft is still trying to get its custom Maia chips into the mix, after encountering delays.

If Microsoft lands Anthropic as a customer for its Azure-based Maia computing services, it could open the door for other companies seeking another option for meeting the sky-high demand for AI inference, as agentic models gobble up trillions of tokens.