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's IPO roadshow is winding up ahead of pricing expected on Wednesday, November 18, with shares expected to start trading on the Nasdaq under the ticker "ANT" on Thursday, November 19. Chief executive Dario Amodei has called buying compute "the deepest business dilemma" the company faces, according to reports from the roadshow, and the numbers behind the question are in Anthropic's own filing: the company has committed $895.4 billion to compute and infrastructure, with nearly $70 billion due within 12 months. Third-quarter revenue was $16.96 billion — up 47% from the second quarter — but compute and infrastructure took 58% of revenue, turning a $974 million operating profit in the second quarter into a $567 million operating loss in the third. The offering is 825 million shares at $70 to $78, aiming to raise up to $64.4 billion, plus a $26 billion private placement from suppliers including Google, Nvidia, AMD, Broadcom, Micron, Microsoft and TSMC.
Why it matters: Demand for the shares is not the question — the order book was reportedly about six times oversubscribed — the question is whether a company that owes $70 billion in a year can earn its way out of its own compute bill. For a small business, the filing is the useful document: it shows that even the fastest-growing AI company in history pays more than half its revenue to run its models, which is why the prices you pay for AI are unlikely to fall as fast as the models improve.
GO Markets: Anthropic's IPO reality check →
LangChain announced on Friday, November 13, a $300 million Series C round at a $3 billion valuation, led by existing investor CapitalG, Alphabet's growth fund, with new investors GIC and DFJ Growth and existing backers IVP and Sapphire Ventures also participating. The money backs LangSmith, LangChain's platform for running, improving and governing AI agents in production, whatever framework they were built with. LangChain says more than 7,000 organizations use LangSmith, including Nvidia, Toyota and Salesforce, and it plans to show new LangSmith capabilities at a release event on November 20. The valuation is more than double the $1.25 billion set when LangChain raised $125 million about a year ago, TechCrunch reported. The figures are the company's own; it disclosed no revenue.
Why it matters: Agent governance is where the money is going because agents are where the risk is: an agent that can act on your behalf can also spend, email and delete on your behalf. A neutral platform that watches agents whatever model they run on is an easier sell to a business than another model — and a $3 billion valuation says investors agree. For a small business, the practical point is the direction of travel: the cost of AI is shifting from the model to the controls around it, and the November 20 event is where LangChain has to show what the money buys.
Unite.AI: LangChain Secures $300M Series C at $3B Valuation for Agent Governance →
An online petition calling on Nvidia's board to reinstate Jensen Huang passed 100,000 signatures within two days of his resignation on Thursday, November 12, according to the petition's own count — including more than 5,000 people who identify as current Nvidia employees. Elon Musk said on X on Friday that he signed it. Separately, Altimeter Capital chief executive Brad Gerstner and Atreides Management founder Gavin Baker wrote to the board asking it to run "a slow, measured, and deliberate process" on leadership — and to keep the possibility of Huang's return open — rather than calling for his immediate reinstatement. The board has said nothing about bringing Huang back; interim chief executive Colette Kress remains in charge, with third-quarter earnings due on Tuesday, November 17. The petition's counts are self-reported and unverified.
Why it matters: A petition does not reinstate a chief executive, and the board has given no sign it will. But 100,000 signatures — including thousands of the company's own employees — plus two large shareholders asking for the door to stay open changes the politics of the search: the next permanent chief will be measured against the founder, in public. For a small business, the lesson is narrower: when a supplier's leadership is in question, the contracts and the roadmap matter more than the personalities, and Nvidia's board has said both continue unchanged.
WCCFTech: Over 100,000 People Now Call For Jensen Huang's Return To NVIDIA →
Malaysia's trade ministry said on Friday, November 13, that it is working with customs to determine whether AI chips imported by Speedmatrix, the Malaysian subsidiary of Singapore-based cloud firm Megaspeed, were re-exported — and whether any such shipments complied with the law. Megaspeed is under investigation in Singapore over allegations it diverted restricted Nvidia chips to China, and it is the firm at the centre of the affair that preceded Jensen Huang's resignation. Prime Minister Anwar Ibrahim said there is no indication so far of chips being shipped out of Malaysia, and that Malaysia has received no formal request for assistance from Singapore or the US. Megaspeed, which denies wrongdoing, said on November 12 that it had hired a US law firm to take a full inventory of its Nvidia GPUs and assess its export-control compliance. Le Monde reported on November 13 that Megaspeed and Speedmatrix bought at least $4.8 billion of Nvidia equipment between June 2024 and February 2026.
Why it matters: The chip story keeps moving from Washington to the countries in between: Malaysia is now checking the same question Singapore and the US are asking, while drafting a new data-centre law of its own. Nothing has been established — no re-export has been proven, and the prime minister says there is no sign of one — but the checks themselves are tightening the route, with every jurisdiction in the chain now on notice. For a small business, this is the export-control regime becoming real in the supply chain you buy from, not just in headlines.
Reuters: Malaysia probing whether chips linked to Megaspeed unit were re-exported →
A forensic report published on Thursday, November 12, by Will DePue of the nonprofit Unemployment Labs found that six open-weight models from four Chinese labs — DeepSeek's V4-Flash and V4-Pro, Moonshot's Kimi K3 and K3.1-Pro, MiniMax-M3 and Ant Group's Ling-3.1-flash — carry numerical patterns matching Nvidia's Blackwell chips, which need a US export licence for China. The method reads the random starting values left in the published weights and matches them to the chip's processor-core layout; earlier models from the same labs match the older Hopper generation. DePue reproduced two of the matches on physical Blackwell hardware, and the report is explicit about its limits: it shows access to Blackwell-class hardware at the moment the models were initialised, not where the chips were, how they were obtained, or whether export controls were broken.
Why it matters: This is the first hard evidence in the chip-diversion story that anyone can check on a laptop: the fingerprint sits in public files, and the method is public code. It still proves access, not smuggling — the report says so itself — but it moves the question from "did it happen?" to "how did the chips get there?", which is the question Malaysia and Singapore are now investigating. For a small business, the direct effect is nil; the indirect one is the export-control tightening that follows, which is already raising the cost and the paperwork of buying advanced hardware.
ExplainX: Blackwell GPU Fingerprints in DeepSeek & Kimi Weights →
This day has a permanent page: November 15, 2026. Every day is kept — see the News archive. The News page always shows the latest day.
← Latest day · News archive · Previous day: November 14, 2026
For what the AI companies and their leaders are saying on social media — claims and opinions, not verified news — see The Social Wire →