Good morning. The money behind super intelligence is moving faster than the models themselves today, with a funding race, a second giant compute deal and a direct challenge to Nvidia’s software hold.
OpenAI races to raise $30 billion before Anthropic lists. Reuters reports OpenAI is accelerating a roughly $30 billion round that could value the company above $1.4 trillion, aiming for commitments before Anthropic’s expected mid November stock market debut. SoftBank and Abu Dhabi based MGX are among the backers in talks, and OpenAI is telling investors annualized revenue has reached $50 billion after the launch of GPT-6 Astra, with $70 billion possible by year end. Why it matters: the two leading labs are now competing for the same sovereign and long term capital at the same moment, and whoever locks it in first shapes the comparison investors make at the listing.
Anthropic is in talks for a second $45 billion compute deal with Nscale. Bloomberg reports a second roughly $45 billion agreement covering another 460 megawatts at Nscale’s Monarch Compute Campus in West Virginia, running on Nvidia Vera Rubin accelerators, with a right of first refusal on a third building. It follows an August deal of the same size at the same site, and separate reporting puts Anthropic’s total compute commitments at about $517 billion over the past 11 months. Why it matters: compute is now contracted like real estate at national scale, and access to powered sites is becoming as decisive as model quality.
Amazon takes its biggest shot yet at Nvidia’s software advantage. Amazon has launched TorchNeuron, an open source tool that lets developers run existing PyTorch code on its Trainium AI chips with a configuration change rather than a rewrite. The public beta begins October 20 under the Apache 2.0 licence, supports Trainium2 and Trainium3, and more than 30 early customers were already using it. Why it matters: Nvidia’s lead rests as much on its CUDA software as on its chips, and making rival hardware easy to adopt is the clearest way to loosen that hold.
Chinese AI hardware makers seek $1.5 billion in Hong Kong. Two companies began marketing Hong Kong listings to raise a combined $1.5 billion, led by Suzhou Dongshan Precision Manufacturing, a maker of printed circuit boards for AI devices, seeking up to $1.4 billion with cornerstone backing from Alibaba’s investment arm and JPMorgan Asset Management. Changzhou Microintelligence, an industrial robot maker, is raising the remainder, and both are set to start trading on October 28. Why it matters: Hong Kong has raised a record amount across 116 listings in the first nine months, with more than half from AI and semiconductors, so this is a live test of investor appetite for the hardware layer.
MiniMax launches M3.1 Flash as China’s low cost model race intensifies. Shanghai based MiniMax has launched M3.1 Flash, claiming gains in agent tasks, coding and cybersecurity with faster inference and far lower usage costs, reported at about one yuan per million input tokens. A higher end Pro version may follow. Why it matters: when capable models get this cheap, agents become viable for everyday work, and pricing pressure reaches every provider, not just rivals in China.
Gairdner Science Week brings world leading health research to Toronto. From October 19 to 23, the Gairdner Foundation hosts free public lectures and symposia in Toronto and online with the 2026 Canada Gairdner Award laureates. This year’s honored work includes visualizing molecular structures inside cells, treatments for amyloid disease, oral cholera vaccines, new approaches to antibiotic resistant infections, and restoring blood pressure control after spinal cord injury. Why it matters: the biggest health breakthroughs can feel abstract until the people behind them explain them in public, and this week does exactly that.
That is the morning edit. We will be back tomorrow with the stories shaping tech and culture.
