Three stories set the terms for Wednesday. Nvidia reported the largest quarterly revenue in its history, confirming that AI infrastructure demand compounded through the removal of its founder. Anthropic closes its bookbuilding today, with Class A shares set to trade on Nasdaq tomorrow morning. And a new OpenAI science post shows GPT-6 Astra helping a particle theorist sketch a candidate answer to a 50-year open problem in fundamental physics, quietly enlarging the frontier of what these models are being asked to do.
Nvidia Posts $115.8 Billion in Q3, First Results Without Jensen Huang
Nvidia reported Q3 fiscal 2027 revenue of $115.8 billion, up 103 percent year-on-year and 20 percent quarter-on-quarter, well above an analyst consensus of roughly $109 billion. Data Centre generated $107.6 billion, 93 percent of total revenue, up 110 percent year-on-year and 21 percent quarter-on-quarter. Free cash flow reached $43.8 billion, up 98 percent year-on-year. GAAP diluted earnings per share was $2.70.
The Vera Rubin NVL72 began shipping in Q3 and contributed nearly $20 billion in its first production quarter. For Q4, Nvidia guided to revenue of $130 billion, plus or minus 3 percent. The company made clear it has assumed no data centre compute revenue from China in its Q4 outlook, a standing constraint since export controls tightened.
These are the first results delivered by Colette Kress, who has served as interim chief executive since Jensen Huang's resignation on 12 November. The numbers confirm what the product cycle suggested: the AI infrastructure buildout is not dependent on a single individual, and the Blackwell-to-Rubin transition is proceeding without disruption. For operators, the near-term implication is that GPU allocation, pricing, and cloud availability are unlikely to shift materially in the next two quarters.
Anthropic Finalises Its IPO Price Today; Amazon and Google to Hold Nearly 30 Percent
Anthropic finalises the price of its initial public offering tonight, with Class A shares set to begin trading on Nasdaq under the ticker ANT tomorrow, 19 November. The company set a range of $70 to $78 per share on 825 million shares, implying a valuation of roughly $1.4 trillion at the top of the range.
Disclosed in the amended prospectus and reported by TechCrunch on 17 November: Amazon and Google will together hold nearly 30 percent of Anthropic after the offering. Google's stake stands at approximately 13.7 percent; Amazon's is larger, reflecting its cumulative investment across two rounds. This sits alongside a $26 billion private placement involving seven technology companies, announced when the roadshow launched on 10 November.
The governance implication is material. With two hyperscalers holding nearly 30 percent between them, Anthropic's declared commitment to a safety-first, long-term mission now competes with the interests of its two largest shareholders, both of whom operate competing AI platforms. Operators that have made Claude a primary model should track whether public-market earnings pressure and board composition alter the model roadmap, pricing, or acceptable-use policies over the next four quarters.
OpenAI Opens a Science Blog with a Particle Physics Contribution
On 15 November, OpenAI published the inaugural post on a new science blog. The author is Nathaniel Craig, a particle theorist at UC Santa Barbara currently serving as a visiting scientist at OpenAI. The post describes using GPT-6 Astra to work toward a candidate lattice definition of the Standard Model.
The Standard Model of particle physics has no known nonperturbative definition. Placing chiral gauge theories, which the Standard Model contains, on a discrete lattice representing spacetime is a problem that has resisted solution for roughly 50 years. Craig circulated an October 2026 preprint, prepared with GPT-6 Astra's assistance, proposing a candidate approach. The blog post explains both the technique and the model's specific contributions to the work.
- This is the first time a frontier AI lab has published, on its own blog, an AI-assisted contribution to fundamental physics rather than applied science.
- The pattern of visiting domain experts paired with a frontier model producing a novel result signals a new class of research partnership that sits between tool use and co-authorship.
- The result is a preprint, not a peer-reviewed proof; it should be treated as an early indicator, not a confirmed advance.
For operators in research-intensive industries, pharmaceuticals, materials science, financial modelling, climate simulation, this is the signal that GPT-6 Astra-class reasoning is now being pointed at the hardest open problems in those fields. The question is not whether AI will contribute to your research agenda; it is how quickly you establish the workflows to capture it.
The near-term priority is understanding what Anthropic's new public-company obligations mean for your Claude contracts. The longer bet is mapping where GPT-6 Astra-class reasoning can accelerate your most research-intensive functions. And the evidence from Nvidia's Q4 guidance is that the infrastructure to run those models at scale will be available; the constraint is now on the application side.