Chipmaking equipment billings reached $40.53bn in the second quarter, 23 percent above a year earlier

The AI stack on 3 September 2026. 12 moves, 9 with a primary source.

Chipmaking equipment billings reached $40.53bn in the second quarter, 23 percent above a year earlier

· L2 Silicon · $40.53bn · Confirmed

Before anyone can make a chip, someone has to sell the machines that make chips. Bills for those machines came to 40.53 billion dollars worldwide in the second quarter, almost a quarter more than in the same quarter a year earlier.

Source

Hidden instructions succeeded against Astra in an estimated 8.5 percent of attempts on Gray Swan's IPI Arena

· L4 Models · 8.5% · Reported

A prompt injection is a hidden instruction planted in a web page or a document that an AI assistant reads and obeys by mistake. On one such test set, hidden instructions succeeded against OpenAI's newest model in an estimated 8.5 percent of attempts, against 27 percent for the model before it.

Source

Nvidia's filing puts the Hugging Face purchase price at approximately $11.9bn to stockholders

· Band Money · approx $11.9bn · Confirmed

Nvidia's filing with the US securities regulator puts the price for Hugging Face at approximately 11.9 billion dollars payable to its stockholders. Up to approximately 1 billion dollars more is set aside in shares for Hugging Face employees who join Nvidia.

Source

Canada sets five expectations for data centres, including no shifting of electricity costs to Canadians

· Band Rules and society · Five expectations · Confirmed

Canada has set out five things it expects from anyone building a large data centre there, including that electricity customers across the country should not end up paying more because of it. The government presents these as expectations meant to work alongside existing regulation.

Source

Nvidia agrees to buy Hugging Face, the home of open models

· L5 Tooling and infrastructure · $12.93B · Confirmed

Nvidia makes the chips nearly all AI runs on. Hugging Face is the site where researchers publish and share their AI models freely. The company that makes the engines has bought the garage everyone parks in.

Source

Flex agrees to buy EPC Power, adding data centre power conversion

· L1 Power and land · $4.4B · Confirmed

Electricity arrives from the grid in a form computers cannot use directly, so it has to be converted. Flex has paid $4.4 billion for a company that builds those converters. Even the plumbing between the grid and the machines is now worth billions.

Source

Nscale and Figure sign a compute deal, contracted rather than spent

· L3 Iron and cloud · $3.5B · Confirmed

A robot company has agreed to buy $3.5 billion of computing power from a data centre operator. Contracted is not the same as spent: it is a promise to pay over years, and promises can be broken. That distinction runs through most large AI numbers.

Source

Sanders and Casar announce a bill to ban superintelligent AI outright

· Band Rules and society · 20 years · Confirmed

Two members of Congress announced a bill to make building AI smarter than humans a crime, with prison sentences of up to twenty years. It defines superintelligence in words rather than by any measurable threshold, which is unusual and will be hard to enforce.

Source

A trading firm, not an AI lab, is reported to sign a five-year cloud contract

· L3 Iron and cloud · $13B · Reported

A firm that trades shares for a living, not an AI company, is reported to have signed a $13 billion deal for computing power. The buyers of very large amounts of AI computing are no longer only AI companies.

Source

OpenAI prices GPT-6 Astra at $10 per million tokens in and $50 out

· L4 Models · $10 in, $50 out per 1M tokens · Confirmed

OpenAI's GPT-6 Astra API is priced at $10 per million input tokens and $50 per million output tokens, its release says. A token is a small piece of text. The release gives no comparison with earlier prices. For developers, this is the price of building on OpenAI's top model through its API.

Source

HUMAIN puts a 428 billion parameter Arabic model into research preview, built by China's MiniMax

· L4 Models · 428B params · Reported

HUMAIN is owned by Saudi Arabia's sovereign wealth fund. Its new Arabic model is in research preview, so access is limited rather than general. The announcement says MiniMax, a Chinese laboratory, developed it on top of that lab's own model and trained it further on more than one trillion tokens of Arabic text.

Source

Abu Dhabi publishes a training mixture, including approximately 10 trillion synthetic tokens

· L4 Models · ~10T synthetic · Confirmed

Most laboratories keep secret what their models were taught on. A research institute in Abu Dhabi has published its recipe for six models, from 0.9 billion to 375 billion parameters. Each saw approximately 20 trillion tokens, and approximately 10 trillion synthetic tokens were used across the whole set, with no breakdown per model.

Source

Strata, the whole AI stack, explained simply. Every number carries a source and a confidence label.

Today · The Stack · Stories · Archive · Glossary · Break the Chain · Focus · About · RSS feed

Privacy · Terms · Corrections · Report an error

Also from Dheeco: Dheeco · AP Fact Check

Published by dheeco.com · © 2026 Dheeco