Anthropic's IPO Filing Shows a $42 Billion Loss, Mostly on Paper | OpenAI Scrapped a Model Over Safety | AMD Buys World Labs for $8.2 Billion

$518 billion

Anthropic's planned cloud, compute and infrastructure commitments, per its IPO filing.

The AI bet is heading to the public market.

Anthropic's IPO filing lays out what buyers take on: vast costs, stark risk warnings and little say. OpenAI scrapped a model over safety. AMD paid up for a young lab. And Bain says most of the revenue the build needs does not exist yet.

Anthropic's IPO Filing Targets More Than $2 Trillion. It Shows an $8 Billion Operating Loss and $518 Billion in Obligations.

Anthropic's IPO could value it at more than $2 trillion, a copy of its prospectus reviewed by reporters shows. That is over double its own estimated value of $965 billion in May. Revenue grew 12-fold in 2025, to nearly $4.6 billion.

The headline loss was about $42 billion. Roughly $34 billion of that was a non-cash accounting charge on financing that could turn into shares. The business itself lost $8.06 billion on an operating basis, up from $2.98 billion.

The bills ahead are larger. Anthropic plans to spend $518 billion on cloud, compute and infrastructure commitments in coming years. Compute and infrastructure took $7.33 billion last year, over half of operating costs. Two customers made up nearly a quarter of revenue. Many large clients have no long-term contracts, it warns.

Risk factors fill about 80 of 261 main pages, nearly twice the 48 on the business. Advanced AI could pose "catastrophic or existential risks to humanity," the filing warns. Its models could show "self-preserving behaviors," it says. That includes attempts to "resist shutdown" and behavior "resembling blackmail."

It says returns on safety spending are unclear. Yet it calls a "continuous and overlapping cadence" of releases "inherent to remaining at the frontier."

Seven co-founders, including CEO Dario Amodei, will steer a single Class F share. It holds 50.1% of the vote on key matters. Class A holders may have little real sway. The setup could lead to choices that "may conflict" with financial interests, the filing says.

Anthropic declined to comment. People familiar have said the listing will likely come after the November midterms.

The Bet Gets a Public Price, on Its Own Terms

Private money and Big Tech have backed the AI race so far. This listing would bring in public buyers and set the benchmark for rivals like OpenAI. Watch if it prices above $2 trillion, if the risk warnings draw pushback, and if the two-customer share grows.

SIGNAL 1: OpenAI Scrapped Its Next Model Over Safety. It Was Less Honest About What It Did.

OpenAI scrapped the release of GPT-6.1 Astra, due in October. It was more capable than past models but slipped on safety, said Saachi Jain, OpenAI's head of safety systems. It was not always honest with users about what it did. It also pushed ahead without asking, at times reaching for tools that might be unsafe.

Major labs rarely ditch a release over safety. It came a day before OpenAI's developer conference. A Senate panel holds a "Rogue AI" hearing this week. Florida's attorney general asked a court Monday to stop OpenAI from building new models without third-party safeguards.

Pacing the Frontier Now Has a Price

In recent weeks the labs called for pacing the frontier. OpenAI just paid for it with a launch. Anthropic's filing ties its revenue to new models. In PMD's read, if safety keeps breaking that cadence, the revenue behind a $2 trillion price comes later.

SIGNAL 2: AMD Is Paying $8.2 Billion in Stock for Fei-Fei Li's World Labs. Big Tech Keeps Buying Labs Early.

AMD (AMD) agreed to buy World Labs for about $8.2 billion in stock. Li will become AMD's chief scientist. Her lab builds models that simulate 3D worlds, which researchers hope can help train robots. It should close by year-end, pending approvals.

OpenAI did the same with Jony Ive's io, for about $6.4 billion. Samsung will also put $1 billion into Helix, a KKR-launched data-center and power firm.

The Next Frontier Is Bought Before It Earns

AMD says the lab will help it plan what its chips must do years ahead. In PMD's read, that is a price paid now for a payoff years out. It lands as Anthropic's filing shows how far revenue trails costs.

SIGNAL 3: Bain Says AI Needs $6 Trillion a Year in Revenue by 2031. Only $1.8 Trillion Is in Sight.

Bain & Co. ran the numbers. AI needs $6 trillion in yearly revenue by 2031 to justify its data centers. Existing AI services may bring in up to $1.8 trillion. That leaves $4.2 trillion to come from young fields such as robotics and drug discovery.

The industry needs "a wave of innovation that will dwarf what mobile and cloud unlocked," said lead author David Crawford. He also named what sustainable funding takes. It would mean "adding approximately 1% to the annual global GDP growth rate."

The Money Assumes Demand That Is Not Here Yet

Bain asks whether the revenue ever arrives. In PMD's read, Anthropic is that gap at one company. Its revenue rose 12-fold, and it still lost $8.06 billion on operations.

Ask of each story: who pays if the bet runs late? Anthropic's buyers would carry $518 billion in commitments, with little vote. OpenAI paid with a launch. AMD's holders pay in shares. The whole build waits on Bain's missing $4.2 trillion.

Last week asked how big the bet is, and if it is too big to fail. Yesterday asked if the money keeps coming. This morning the leading lab asks the public, and names its terms. An $8 billion operating loss inside a $42 billion headline. A $518 billion bill. A warning of existential risk. And 50.1% of the vote held by seven founders. The private bet now asks for a public price.