Investing.com - Anthropic agreed to $517 billion in compute commitments covering 14.8 gigawatts of capacity in the 11 months through August 2026, according to The Information, even as its CEO publicly called on the industry to pump the brakes on AI development.

Nvidia, which is reportedly in talks to invest up to $10 billion as an anchor investor in Anthropic's expected Nasdaq IPO, is the most direct equity expression of that dynamic: every gigawatt Anthropic contracts translates into accelerated GPU and networking demand that flows straight to Nvidia's data-center revenue line.

The scale of Anthropic's compute commitments dwarfs a prior investor disclosure of roughly $180 billion through 2029, with four additional deals struck since July alone, The Information reported. That capital outlay is increasingly backed by real economics.

Anthropic told investors it expects a second consecutive quarter of adjusted operating profit, and inference gross margins exceeded 80% before distribution costs, the Financial Times said, citing investor briefings that Anthropic has not confirmed on the record. The company's revenue run rate climbed from roughly $9 billion to $65 billion in seven months, according to the same FT account.

The contrast with the rhetoric landing on the same week sharpened the tension considerably. On Saturday, Anthropic CEO Dario Amodei published an essay calling on AI labs to "slow the pace at which we improve the capabilities of AI models," proposing embedded third-party evaluators, industry-wide safety standards, and global coordination.

"Progress will still seem fast, and we must make wise use of the time we gain," Amodei wrote. OpenAI's Sam Altman and Elon Musk both publicly endorsed the framework.

"I don't know that investors are necessarily going to see it as a negative. Unless the companies are genuine and say, 'OK, we're not going to IPO, we're not going to use any more compute, we're not going to train any more models.' That's not what they're saying," Gil Luria, equity analyst at D.A. Davidson, told CNBC.

The market read the slowdown rhetoric with skepticism. AI-linked Asian stocks slumped following the Amodei' statements, and Wall Street's AI trade came under scrutiny. But the broader infrastructure thesis appeared intact.

Nvidia CEO Jensen Huang, speaking at Goldman Sachs's tech conference last week, forecast global AI infrastructure spending of $3 trillion to $4 trillion by 2030 and said Nvidia's roughly 70% fiscal 2027 revenue growth reflects a supply ceiling rather than a demand limit. His prior projections proved conservative: a February 2024 forecast that the world's installed data-center base would double within four to five years was reached by end of 2025.

On the IPO front, OpenAI will not go public in 2026. "Right now would be an ill-advised moment to go public. I would say not 2026, yeah. We got a lot of stuff to do," Altman said.

Anthropic is moving in the opposite direction, having confidentially filed its IPO prospectus with the SEC in June 2026. Sources cited by Reuters put its target valuation at around $2 trillion, a significant step up from its most recent private funding round valuation of $965 billion, with listing timing expected before the U.S. midterm elections in November 2026.

The offering would be among the largest in history and deliver the first GAAP-audited look at Anthropic's revenue and compute-cost structure. Nvidia's reported anchor bid of up to $10 billion would cement the chipmaker's strategic relationship with one of its largest customers while giving it direct upside in the lab's equity.

China's state-backed Global Times dismissed Amodei's slowdown call as a "Cold War playbook" targeting Beijing, underscoring that any coordinated Western pacing effort carries geopolitical weight well beyond the balance sheets of the labs involved.

"While the proposal uses fears of RSI and safety/alignment as its hammer, we wonder if China concerns are a large part of the driving force here," Bernstein analysts led by Stacy Rasgon wrote in a note.

"We suspect it is far too late to put the AI genie back in the bottle anyway, so we might as well find safer ways to let it out."