AI
The Suppliers Became the Shareholders as AI Went Public and Went Small at the Same Time
This week gave AI two debuts that point in opposite directions and fit together. Anthropic heads for the largest listing the industry has ever seen, with its own suppliers lining up to buy the shares, while Google shipped a faster model into products people already use and a 145 billion parameter model learned to run on a single laptop. The money is moving up, the models are moving down, and the rules are arriving from city halls and courtrooms at the same time.
Anthropic plans to price its offering on November 18 and begin trading on Nasdaq under the ticker ANT on November 19, selling 825 million shares at 70 to 78 dollars each for a raise of up to 64.4 billion dollars and a valuation near 1.4 trillion dollars fully diluted. The order book ran about six times oversubscribed as of November 12, up from about two times on November 10. The striking line sits beside the offering, a private placement of about 26 billion dollars at the IPO price, with Nvidia and Google each committing about 10 billion dollars, AMD committing 2.5 billion dollars, and Broadcom, Micron, and Microsoft each committing 1 billion dollars, plus 500 million dollars from TSMC Partners. Anthropic reported third quarter revenue of 16.956 billion dollars, roughly twelve times the 1.382 billion dollars it reported for the same quarter in 2025.
Google shipped Gemini 4 Hydrogen on November 13, replacing Gemini 3.8 Flash as its fast, lower cost model across Search, the Gemini app, Google Sheets, and developer tools. Google says it finishes tasks roughly 2.5 times faster than its predecessor, with a one million token context window and API pricing of 1.50 dollars per million input tokens and 7.50 dollars per million output tokens. Google also released Gemini 4 Carbon and Argon models that performed well on an accounting benchmark, a quiet signal that the frontier now competes on the work a model does per dollar as much as on raw capability.
The most surprising release of the week fits in a backpack. Edge0 released two models on November 12, Edge0 145B and Edge0 120B, with Edge0 145B billed as the first open model of its size built to run entirely on a personal computer, with the cloud left out entirely. The company reports a peak memory footprint of 29 gigabytes, decoding above 80 tokens per second, typical power draw near 120 watts, and context windows from 128,000 to 256,000 tokens. The fine print tells a fuller story, the model card's own measurement for the NVFP4 build puts total memory near 61.5 gigabytes, 23.25 gigabytes of video memory plus 38.27 gigabytes of host memory, and its metadata names Alibaba's Qwen3.8 Flash Next as the base model. Every benchmark so far comes from Edge0's own internal testing, the technical report that would let anyone reproduce them is still promised, and the weights ship under a custom Edge0 Community Model License 1.0, so teams should read the terms before building on it.
While the labs shipped, the rulebooks advanced from four directions at once. New York City voted 44 to 5 on November 12 to end personal data based price discrimination, the fifth such move in the country, with fines up to 1,000 dollars per violation and the measure taking effect one year after becoming law. President Ferdinand Marcos Jr. signed Executive Order 131 on November 12 to set the direction for artificial intelligence governance in the Philippines. India's market regulator SEBI said it will soon issue detailed guidelines on responsible AI use in capital markets. And four groups filed suit in a California court on November 11, asking for an emergency order pausing frontier training at five major labs, Anthropic, Google, Meta, OpenAI, and SpaceX's AI unit, on a public nuisance theory.
The money followed the builders, too. LangChain raised 300 million dollars in a Series C led by CapitalG at a 3 billion dollar valuation, a bet that managing agents matters as much as building them. For readers, the week reads as one story told three ways, the biggest lab in the world is about to belong to the public markets, the fastest models are arriving inside the tools already on your desk, and the rules for how all of it behaves are being written in the open, by cities, regulators, and courts, while the shipping continues.
Quick answers
What is this story about?
This week gave AI two debuts that point in opposite directions and fit together. Anthropic heads for the largest listing the industry has ever seen, with its own suppliers lining up to buy the shares, while Google shipped a faster model into products people already use and a 145 billion parameter model learned to run on a single laptop. The money is moving up, the models are moving down, and the rules are arriving from city halls and courtrooms at the same time.
Why does this story matter?
The money followed the builders, too. LangChain raised 300 million dollars in a Series C led by CapitalG at a 3 billion dollar valuation, a bet that managing agents matters as much as building them. For readers, the week reads as one story told three ways, the biggest lab in the world is about to belong to the public markets, the fastest models are arriving inside the tools already on your desk, and the rules for how all of it behaves are being written in the open, by cities, regulators, and courts, while the shipping continues.
Sources
- Ava Fund, Anthropic IPO order book
- iCameraPlus, weekly AI roundup
- Tokenstrom, Edge0 145B on a laptop
- The World of AI, daily AI news November 15
New to crypto? Read the crypto glossary, browse frequent questions, read our story, or explore the story archive.