Skip to content

Journal / AI News.

Anthropic's Record IPO, Nvidia Denies China Chip, Google's $12B Marvell Deal — AI News Briefing

Anthropic could file IPO paperwork as soon as this month and expects to match or top SpaceX's record raise, per Bloomberg. Nvidia denies a report it will ship a China AI chip by year-end, Google hands Marvell a $12.2B custom chip deal, Meta quietly became one of Microsoft's largest AI customers, and OpenAI gains on Anthropic with business users.

CinaGroup Automation Desk AI News 6 min read

Top 7 Stories

1. Anthropic could file for its record-setting IPO as soon as this month

Bloomberg reports Anthropic may publicly file IPO paperwork as soon as this month, and expects the listing to match or top SpaceX’s record-breaking raise. The company is reportedly adding Citigroup to its top IPO banks, signaling the mega-listing is moving toward execution. CNBC notes experts are already debating whether public markets can absorb both Anthropic and OpenAI going public in the same window.

The filings would bring the AI industry’s most closely watched financials into public view — including whether enterprise AI revenue can justify the valuations private markets have assigned. For investors, the answer will hinge on data retention battles, model stickiness, and how fast agentic workloads convert into durable spend.

2. Nvidia denies report it is rolling out a China AI chip by year-end

The Information reported Nvidia is plotting a China comeback with a new AI chip designed around export restrictions, potentially shipping by year-end. Reuters says Nvidia denied the report. The story lands just ahead of Nvidia’s quarterly earnings, where analysts like Jefferies expect a Q2 revenue beat of roughly $3 billion — but where questions about China, export controls, and hyperscaler capex will dominate the call.

The denial may be more about timing than intent: every major chipmaker is weighing how to serve China’s demand within tightening rules. Whatever Nvidia says on its earnings call will move not just its own stock but the entire AI supply chain narrative.

3. OpenAI is gaining on Anthropic with business users, new data indicates

TechCrunch reports new data from Ramp, the corporate card and expense company, shows OpenAI reclaiming ground among US business users after losing its lead to Anthropic earlier. The data underscores how volatile enterprise AI spending is — businesses flop between labs with each new model release, raising questions about how “sticky” AI contracts really are ahead of both companies’ IPOs.

In a related move, Bloomberg and Reuters report Anthropic plans to change its enterprise data retention policy after customer concerns — a direct competitive response to OpenAI’s new Zero Data Retention offering for frontier models. Data-handling terms are becoming first-line weapons in the fight for enterprise trust.

4. Google strikes a $12.2B custom AI chip deal with Marvell

Reuters reports Google has signed a custom AI chip deal with Marvell that includes an option for Google to buy up to $12.2 billion in Marvell shares. Analysts estimate the partnership could unlock up to $120 billion in revenue over its life, and The Register frames it as Google pitting Marvell against Broadcom as it chases the AI crown. Marvell stock jumped about 10% on the news.

The deal signals two things: hyperscalers are aggressively diversifying away from single-supplier silicon, and equity-for-orders is becoming the AI chip industry’s new playbook. Expect Broadcom, AMD, and Nvidia to respond as the custom ASIC war intensifies.

5. Meta has quietly become one of Microsoft’s largest AI customers

Bloomberg reports Meta now spends hundreds of millions of dollars a year on Microsoft Azure capacity for AI training and inference, placing it among Microsoft’s biggest AI customers. The revelation is striking because Meta has spent aggressively on its own data center buildout — yet still finds it cheaper or faster to buy capacity from a competitor’s cloud.

The arrangement gives Microsoft a strategically important anchor customer and underscores that even the largest AI builders cannot fully self-supply compute. It also deepens the interdependence between two companies that compete on consumer AI, and raises the stakes for Meta’s own capex plans.

6. Google gives publishers a new way to fight AI-driven traffic losses

TechCrunch reports Google is letting publishers embed a “Preferred Sources” button on their websites, allowing readers to mark them as favorites that get highlighted more often across Search, Discover, and Google News. The launch follows earlier publisher tools as Google tries to manage the traffic collapse caused by AI-generated search answers, which send fewer clicks to the open web.

The button is a concession that AI search has fundamentally changed the publisher economy — and that Google needs publishers onside politically and commercially. Whether a reader-driven signal can meaningfully offset AI traffic losses is the open question.

7. It’s Greg Brockman’s OpenAI now

The Verge reports OpenAI president and cofounder Greg Brockman has quietly amassed more power and authority through months of change at the company. The consolidation comes as OpenAI signals it is “hitting the brakes” on frontier model development, embracing voluntary pacing after detecting worrying capabilities — and as it launches “AI Futures,” a new blog exploring how transformative AI could reshape power, governance, the economy, and individual freedom.

Together the moves paint a picture of an OpenAI reorienting from pure acceleration toward safety-first pacing and policy engagement, likely with an eye on its own IPO narrative. The question is whether self-pacing holds as competitive pressure from Anthropic, Google, and open-source models intensifies.

Trend Watch

StoryImpactWhy it Matters
Anthropic’s record IPO plansFiling could come this month; Citigroup added to top banksFirst full public look at frontier-lab financials — and a test of market appetite for AI listings
Nvidia denies China chip reportExport-rules question resurfaces ahead of earningsThe denial sets up a high-stakes China Q&A on Nvidia’s earnings call
OpenAI gains on Anthropic in businessRamp data shows enterprise loyalty is shallowData retention and model cadence, not brand, will decide the enterprise winner
Google–Marvell $12.2B chip dealGoogle gets equity option; up to $120B revenue potentialHyperscalers diversify silicon supply and cement equity-for-orders as the ASIC playbook
Meta as a top Azure customerHundreds of millions a year flowing to MicrosoftEven the biggest builders outsource compute — a read-through for cloud demand and capex
Google’s Preferred Sources buttonPublishers get a reader-driven traffic signalA marker of how badly AI search has disrupted the web economy
Brockman consolidates at OpenAILeadership and strategy pivot to pacing and policySignals a slower, safer OpenAI — just as rivals accelerate

What to Watch

  • IPO filings: Watch for Anthropic’s S-1 this month, and how markets price two frontier labs going public in the same window.
  • Nvidia earnings: The China question is unavoidable — listen for how Nvidia squares its denial with reported plans.
  • Enterprise stickiness: Ramp-style spending data will keep exposing how quickly businesses switch labs between model releases.
  • ASIC war: Broadcom’s response to the Marvell–Google deal will shape the custom silicon market for years.
  • Meta’s cloud strategy: Whether Meta keeps renting Azure or pulls more compute in-house as its own data centers come online.
  • Publisher economics: Early adoption of the Preferred Sources button will show if reader signals can move AI-era traffic.
  • Agent momentum: Slack’s new collaborative vibe-coding channels, Ramp’s AI model router, and AWS agent guardrails are worth tracking as agentic work enters the mainstream.
Back to blog