Anthropic launches reporting tools to track enterprise AI spending
The company is also creating shared team budgets as businesses seek a clearer view of AI coding costs and results.
Alphabet's chip business brings billions in sales but weighs on cash
Google Cloud expects TPU system sales to rise significantly in 2027, while Alphabet must pay for hardware before receiving customer payments.
Alphabet is generating billions of dollars from selling its proprietary AI processors, but stocking the hardware is consuming cash alongside heavy data center investment. The chips, previously used only internally, compete with Nvidia's products.
On Alphabet's earnings call, finance chief Anat Ashkenazi said inventory now includes unsold tensor processing unit systems, known as TPUs. Hardware costs come before customer payments, creating a different cash cycle from the company's search and cloud operations.
Alphabet stock dropped nearly 11% on October 29 after another quarter of cash burn. The company also repeated its expectation that AI infrastructure spending would increase sharply next year.
The company said TPU sales were a meaningful contributor to Google Cloud's September-quarter revenue, which more than doubled. Alphabet provides no separate revenue or margin figures for the processors.
Google Cloud chief Thomas Kurian expects sales of the systems to increase in the current quarter and accelerate significantly in 2027. The systems package chips together with storage, software and networking.
Alphabet said last year it would supply Anthropic with tens of billions of dollars' worth of TPUs. Discussions with OpenAI about a comparable arrangement have also been described in media reports.