Five items that matter to a small business, with sources. News moves fast and vendor announcements are vendor claims — check the linked source before acting on any of it.
Anthropic priced its initial public offering at $83.00 a share on Wednesday, November 18, above the $70–$78 range, and upsized the deal to 980 million Class A shares — 956,351,354 sold by the company and 23,648,646 by existing stockholders — for gross proceeds of $81.34 billion. The shares begin trading today, Thursday, November 19, on the Nasdaq Global Select Market under the symbol "ANT", and the offering is expected to close on November 20. Underwriters have a 30-day option to buy up to 147 million more shares. Alongside the offering, seven companies — Nvidia and Google among them — are buying $26 billion of non-voting stock in a private placement, bringing Anthropic's expected net proceeds to about $104.8 billion. News outlets put the fully diluted valuation at about $1.5 trillion, and Reuters called it the largest stock-market debut on record, ahead of SpaceX's June listing — Anthropic itself has not stated a valuation figure.
Why it matters: The first frontier AI lab is now a public company, and its first day of trading will set the market's price on the whole sector. For a small business, the signal is the funding cycle: a strong debut keeps the money flowing into AI tools; a weak one tightens it.
Anthropic: Anthropic prices initial public offering →
Nvidia has stopped shipping licensed Hopper-generation AI chips to China, interim chief executive Colette Kress said on the company's November 17 earnings call — despite selling $2.4 billion of the licensed data center products to Chinese customers in the quarter ended October 25. "Our guidance does not have any licensed Hopper data center products included, and we are no longer shipping licensed Hopper data center products to China," Kress said. She gave no reason and did not say whether the halt is permanent. The Hopper family includes the H200, which Washington had cleared for sale to approved Chinese buyers — including Alibaba, Tencent and ByteDance — under a licensing arrangement agreed earlier this year; Beijing has meanwhile restricted imports, with Reuters reporting about 600,000 H200 chips held in Hong Kong awaiting permission. Nvidia's $130 billion fourth-quarter revenue forecast assumes no further China data center sales.
Why it matters: The US–China chip channel that reopened this year is effectively closed again, and the world's biggest AI chipmaker is now guiding as if China contributes nothing. For a small business, the practical effect is on supply: chips that cannot go to China stay in the rest of the world's pipeline, but the politics around them keep getting sharper.
The Hindu BusinessLine: Nvidia stops selling AI chips to China despite US approval →
Google introduced Gemini 4 Neon on November 18, a mid-tier model aimed at complex software engineering, knowledge work and science workflows. It is priced at $3 per million input tokens ($0.15 for cached input) and $15 per million output tokens, with a 1-million-token context window, and is rolling out in the Gemini API, Google AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra subscribers. Google says Neon scores 61 on the Artificial Analysis Intelligence Index, third overall. One wrinkle worth knowing: until November 30 the flagship Gemini 4 Carbon is on a $2/$10 promotion, so Neon only becomes the cheaper option per token on December 1, when Carbon returns to $4/$20.
Why it matters: Google now has a model for nearly every price point, and the mid-tier is where most business work lands. For a small business, the useful comparison is per-task cost, not per-token — and the promo calendar matters: the "cheaper" model is not cheaper until December.
Google: Introducing Gemini 4 Neon →
At Ignite on November 17, Microsoft's in-house Microsoft AI group released seven models at once: MAI Code 2 and Code 2 Flash for coding, MAI Cyber 2 Flash for security, MAI Image 2.7, MAI Realtime 1 for live speech, Dragon Copilot Transcribe built on MAI Transcribe 2, and Decision 1.1, a routing model. The headline is the coding pair, live in GitHub Copilot the same day: Microsoft says MAI Code 2 reaches 96% of Claude Opus 5.5's coding performance at about 5% of the cost per task, and 98% of GPT-6.1 Sol's at about 40% — its own benchmarks, not independently verified. In Copilot, Code 2 costs $0.50 per million input tokens and $2.50 output on a promotion that ends May 17, 2027, when the prices double to $1 and $5; Flash costs $0.05 and $0.25, doubling on the same date.
Why it matters: Microsoft is now selling its own models against the ones it used to resell — and the pitch is price, not power. For a small business paying for Copilot, the coding model in the picker just got much cheaper; the catch is that the price you see today is a launch discount with a published end date.
OrcaRouter: Seven MAI Models at Ignite 2026: Where MAI Code 2 Fits In →
Citibank updated its Online User Agreement this week to say customers who give third-party AI agents access to their accounts are responsible for what those agents do — and if the tool leaks their credentials or is hacked, the customer is responsible for what the hackers do with the account too. PNC changed its terms on Wednesday, November 18, and Citizens emailed customers a similar warning; JPMorgan Chase and Bank of America moved in October. Citi says the update is meant to give customers "greater clarity about those risks". The banks' own AI tools are not covered by the same provisions. Whether the terms would hold up in a real dispute is untested: federal Regulation E caps liability for unauthorized transfers, and no court or regulator has yet decided how it applies when an agent is hijacked.
Why it matters: AI agents that can move money are arriving just as banks are disowning the risk. For a small business, the rule is simple for now: never hand a third-party agent your banking login — use the bank's own tools, or a connection the bank itself provides, and keep a human in the loop on every transfer.
TechCrunch: More banks say customers using AI agents are on their own — even if they get hacked →
This day has a permanent page: November 19, 2026. Every day is kept — see the News archive. The News page always shows the latest day.
← Latest day · News archive · Previous day: November 18, 2026
For what the AI companies and their leaders are saying on social media — claims and opinions, not verified news — see The Social Wire →