Weekly Wrap
Weekly Wrap: Anthropic's Record IPO, Nvidia, Shenzhen and Brussels
Anthropic's record IPO, Nvidia's $115.8B quarter, a Shenzhen summit with an AI line, a Brussels hack and Japan's defence AI plans: the week in AI by region.
AI finance, security and diplomacy all moved at once this week. Here is what happened in each region, how well it is sourced, and what to watch next.
Key takeaways
- Anthropic priced its IPO at $83 a share on 18 November, raising a reported $81.3 billion, the largest on record, and closed its first trading day, 19 November, up 17% [1][2].
- Nvidia reported $115.8 billion of third-quarter revenue and guided to $130 billion for the next quarter, assuming no China data-centre compute revenue [3].
- A Trump-Xi-Putin declaration in Shenzhen was symbolic and included no arms-control talks; China's readout also named AI and drug-control cooperation [4][5].
- The European Commission confirmed that attackers used AI tools against its internal systems; Japan's Defence Ministry will publish AI guidelines at the end of November [6][7].
United States: the money
Anthropic priced its offering at $83 a share on 18 November, the company said, for 980 million shares, about 24 million of them sold by existing stockholders [2][12]. CNBC puts the raise at $81.3 billion [1]. On its first trading day, 19 November, the stock opened at $111.50 and closed at $97.17, up 17%, a fully diluted market value near $1.8 trillion, per CNBC [1]. The prospectus, as reported by CNBC, shows $33.24 billion of revenue for the nine months to 30 September, with Amazon holding 14.7% and Google 13.7% [1]. CNBC also reported that Nvidia agreed to buy about $10 billion of stock in a private placement, and noted protests outside the company's San Francisco headquarters and a bill from Senator Warner on disclosure of AI financing [1].
Nvidia's own release gives the other half of the picture. Third-quarter fiscal 2027 revenue was $115.8 billion, up 103% from a year earlier, with Data Center at $107.6 billion; the company guided to $130 billion, plus or minus 3%, and said the outlook assumes no China Data Center compute revenue [3]. These are company-reported figures. The same release mentions a partnership with Firebird and an Australian buildout of up to 2 gigawatts by 2027 [3].
China and Russia: a summit with an AI line
At the APEC meetings in Shenzhen, Reuters reported a symbolic peace declaration signed by Donald Trump, Xi Jinping and Vladimir Putin, with no arms-control talks [4]. Bloomberg, via The Star, reported that China's readout names an action plan for a "constructive relationship of strategic stability" and cooperation on AI and drug control; no US account of the plan had appeared in the sources we reviewed, so treat the Chinese version as one side's description [5]. The same report says the US-China trade truce runs to 10 January [5].
A technical sign of the same direction: the Xingchen AGI Lab's Xing 4 report describes pretraining a 285-billion-parameter model on 8,192 Huawei Ascend 910C cards [8]. It is a systems paper, not a benchmark result.
Europe: AI on the attacker's side
The European Commission confirmed, through spokesperson Thomas Regnier, that AI tools were used in an attack detected on 22 October, according to the Brussels Times citing Politico [6]. An EU official told Politico that CERT-EU identified the use of ARTEX, a publicly available AI-powered tool for finding vulnerabilities. Several internal applications, mainly HR tools, were disrupted for about ten days; the Commission says it found no evidence of a data breach affecting its staff as a whole [6]. We have not seen the Commission's own statement, so this is labelled reported by Brussels Times, citing Politico.
Japan and Korea: rules and exposure
Defence Minister Shinjiro Koizumi said on 15 November that the ministry will release AI guidelines at the end of November, Jiji Press reports [7]. He named three pillars: humans make decisions and bear responsibility for AI use, data becomes a defence asset, and fast adoption is balanced with safety [7]. In Korea, the Korea Times reports that SK Telecom holds about 38.6 million Anthropic shares, worth about $3.75 billion at the closing price, and may consider a sale; it also lists LG CNS and Samsung SDS as Claude deployers [9].
Builder's view: what the research said
K2-Horizon, from the K2 Team, releases six models with data, checkpoints and training code; the authors say it is competitive with open models of its size [10]. SWE-CoWork found that the two leading models pass about 63% of fail-to-pass tests but complete fewer than 8% of full assignments [11]. Both are the authors' own results.
Strengths, weaknesses and open questions
| Region | What happened | Status | Evidence |
|---|---|---|---|
| US | Anthropic IPO; Nvidia results | Priced; company-reported | Anthropic, SEC filing, CNBC; Nvidia release |
| China, Russia | Shenzhen declaration | Symbolic | Reuters, Bloomberg via The Star |
| Europe | Commission AI-assisted attack | Confirmed (reported) | Brussels Times citing Politico |
| Japan | Defence AI guidelines | Due end-November | Jiji Press via nippon.com |
| Korea | SK Telecom stake | Possible sale | Korea Times |
| Australia | Nvidia buildout | Announced | Nvidia release |
Strengths: capital and compute are scaling quickly, and governments are writing rules. Weaknesses: the IPO's trading and prospectus figures come from press coverage, the Shenzhen plan has one side's readout, and the Commission incident rests on secondary reports. Open questions: how Nvidia's guidance holds without China, and what the "action plan" contains. This wrap has no new India or Gulf item; Thursday's Legal AI Pulse covered both.
Our take
In our analysis, the week's pattern is that markets and attackers adapt faster than rule-makers. The IPO prices AI as infrastructure, the Commission incident shows the same tools on the offence, and Shenzhen shows AI entering diplomacy as a line item. None of this is investment advice.
What to watch next
- Anthropic's trading and any index inclusion decisions [1].
- Whether Nvidia's $130 billion outlook holds without China [3].
- A US description of the Shenzhen action plan [5].
- Japan's guidelines at the end of November [7].
FAQ
How big was the Anthropic IPO? 980 million shares at $83 each, the company said; CNBC reports $81.3 billion raised, the largest IPO on record [1][2].
Did Nvidia count China in its outlook? No; its $130 billion guidance assumes no China Data Center compute revenue [3].
Was the Commission hack AI-assisted? The Commission confirmed AI tools were used, per Brussels Times citing Politico [6].
This article is analysis, not financial or investment advice. AI Worldwire uses AI assistance in research and drafting; an editor reviews every article. The AI system that assisted this article, Claude, is made by Anthropic, whose IPO is covered above.
Sources
- Anthropic IPO live updates, CNBC, 19 November 2026.
- Anthropic prices initial public offering, Anthropic, 18 November 2026. Primary; company statement.
- NVIDIA Announces Financial Results for Third Quarter Fiscal 2027, NVIDIA, 17 November 2026. Primary; company-reported.
- Trump's historic meeting with Xi and Putin, Reuters via Yahoo Finance, 19 November 2026.
- Xi embraces Trump as China's best option to contain chip row, Bloomberg via The Star, 20 November 2026.
- European Commission confirms AI-assisted cyberattack on internal systems, Brussels Times, 19 November 2026, citing Politico.
- Japan Defense Ministry to Release AI Guidelines at End-Nov., Jiji Press via nippon.com, 15 November 2026.
- TeleChat to Xing4, arXiv, 17 November 2026. Primary; authors' own.
- Anthropic's record IPO: which Korean companies stand to gain, Korea Times, 20 November 2026.
- K2-Horizon, arXiv, 16 November 2026. Primary; authors' own.
- SWE-CoWork, arXiv, 15 November 2026. Primary; authors' own.
- Anthropic, PBC prospectus, Form 424(b)(4), US Securities and Exchange Commission, November 2026. Primary; filing.