Sections
00:00 What we're covering today
00:28 1. Anthropic commits $11.6 billion to Akamai's cloud
01:59 2. Microsoft plans more than $10 billion in Middle East spending
02:41 3. Claude guessed book preferences correctly in 61% of pairs
03:23 4. Trump and Xi meet without a disclosed AI breakthrough
04:00 5. Teams in a $380 million automated-lab program meet in Washington
04:48 Visit Hot Tea

Disclosure
Narration uses an AI-generated voice.

Transcript
Welcome to Hot Tea for Friday, September 25, 2026. Today, we're looking at Anthropic's large computing commitment and what Akamai must spend to deliver it.

We'll also cover Microsoft's Middle East plans and an experiment that let agents trade books. Then we'll turn to the unanswered A I questions after a Washington summit and efforts to connect automated laboratories.

This briefing uses an A I generated voice.

Anthropic commits $11.6 billion to Akamai's cloud. Akamai announced an $11.6 billion computing commitment from Anthropic on September 24. Over seven years, Akamai will provide CPU capacity from general-purpose processors in its cloud. The companies could add another $9 billion to the deal. That possible expansion is separate from the signed commitment. Akamai is describing its own deal and has a stake in how investors see it. Its capital spending estimate is a company forecast, and the contract is not revenue collected today. The possible $9 billion expansion still depends on further business. Akamai still has to make the capacity available and start recording revenue from the contract. The company says the deal does not change its 2026 revenue guidance. Akamai also gave Anthropic a warrant for up to roughly 5% of its common stock. More of the warrant vests if the relationship expands. Future filings should show how much has vested and whether spending stays near the initial estimate. Akamai expects to spend about $5.5 billion to provide the promised capacity. It raised its 2026 capital spending budget by $1.7 billion to secure components, including memory. SiliconANGLE reports that Akamai shares jumped more than 20% after the announcement. That jump does not show how much profit Akamai will make from the deal. Akamai must spend billions before the contract can produce its full return. The test is whether the future cash exceeds the cost of building the capacity.

Microsoft plans more than $10 billion in Middle East spending. Microsoft announced plans to spend more than $10 billion across the Middle East through 2030. The September 24 framework covers capital and operating expenses in Kuwait, Qatar, Saudi Arabia and the UAE. The company also described more than $400 million in planned subsea and terrestrial connectivity investment. Microsoft is describing its spending plans. The company announcement does not establish that all the capacity is operating, or that customers have earned returns from it. Watch when the projects open, how much capacity customers can use and what results those customers report.

Claude guessed book preferences correctly in 61% of pairs. Anthropic published Project Swap on September 24 after 201 employees took part in a summer book-trading experiment. Each employee told an agent about their reading preferences. Anthropic reports that the agents' rankings matched the employees' rankings for 61% of book pairs. Random guessing would score 50%. Those are company claims, and the experiment used employees in a controlled book market. It doesn't prove that agents can reliably represent customers in employment, housing or other costly decisions. Tests with people outside Anthropic should show whether agents can learn what people want before negotiating for them.

Trump and Xi meet without a disclosed A I breakthrough. Donald Trump and Xi Jinping met in Washington on September 24. Reuters reported that evening that there was no sign of an A I breakthrough. Xi publicly called for A I to remain under human control. His statement continues a safety debate covered earlier this week, but it isn't a signed agreement. Reuters didn't report shared testing rules or an operating emergency channel. The next concrete evidence would be a joint text that names the responsible officials and sets an implementation date.

Teams in a $380 million automated-lab program meet in Washington. Teams in the NSF programmable cloud laboratory test bed were due to meet in Washington on September 24 and 25. The Engineering Biology Research Consortium says NSF has invested $380 million in 20 teams working across seven scientific and engineering areas. They planned to discuss shared data standards, infrastructure and safety. That's the start of the coordination work. It isn't proof that the laboratories can already work together, or that automated experiments can produce reliable discoveries. The next evidence should be published standards and experiments that separate laboratories can reproduce. Repeated results would show whether the network works beyond a demonstration in one laboratory.

That is the signal before the noise. This briefing was produced from Hot Tea's verified daily edition. For the complete briefing and every source link, visit Hot Tea dot A I.