Anthropic: May 2026 News

Over the past month, Anthropic has made headlines for several reasons — from corporate partnerships and infrastructure deals to its first operating profit and the Mythos controversy. In this article, we've compiled the month's key events.

Anthropic is a core holding in the Raison portfolio, which is why we follow the company especially closely. For our investors, May's news is further validation of the entry point:

• In March 2023, Raison made its first investment in Anthropic at $13.7 per share.
• A second batch of shares was acquired at $36 per share.
• As of today, Anthropic shares trade at approximately $290 on the secondary market, representing a 21x increase from our initial entry.

This is a compelling case for getting in early, before valuations reflect the full picture.

Mythos: one month in, no consensus

Mythos is Anthropic's new model, so effective at finding vulnerabilities that the company restricted access to roughly forty organizations. We covered it in detail in the last issue.

The NYT piece took stock: a month after the release, the industry still can't agree on whether Anthropic made the right call. Some praise the caution; others argue that cutting off the broader research community makes it impossible to independently assess what the model can actually do. Logan Graham, head of Anthropic's Frontier Red Team, put it plainly:

We don’t truly know what is the best way to roll out models like this

The most politically significant consequence is that the White House is considering an executive order requiring government review of new AI models before release. Mythos, it seems, is what nudged the Trump administration away from a hands-off position.

PwC trains 30,000 staff on Claude

One of Anthropic's largest enterprise clients just got larger. Anthropic and PwC (PricewaterhouseCoopers, a global consulting and auditing network with over 364,000 people across 136 countries) announced a major expansion of their strategic alliance. 30,000 US professionals will be certified on Claude, with Claude Code and Claude Cowork rolling out across the firm's global workforce.

The partners are launching a joint center of excellence and a dedicated practice for banking, insurance, and healthcare, with Claude handling the operational grind. Early results are already in: insurance underwriting cycles have been cut from 10 weeks to 10 days; cybersecurity incident response has dropped from hours to minutes.

PwC has been leading AI’s expansion into the parts of the economy where accuracy and reliability are nonnegotiable — financial services, healthcare, life sciences, cybersecurity — and the results are clear.

Dario Amodei

co-founder and CEO of Anthropic

First operating profit and the SpaceX deal

Several sources reported two related developments.

• Anthropic turned an operating profit for the first time in a single quarter. According to Reuters, annualized revenue already exceeds $30 billion and, by some accounts, is approaching $40 billion. In an industry where infrastructure costs routinely consume any revenue generated, this is a landmark: Anthropic is arguably the first frontier AI company at this scale to reach operating breakeven.

• Meanwhile, SpaceX's IPO prospectus revealed that Anthropic pays Elon Musk's company $1.25 billion per month for computing power through May 2029. The figure speaks for itself — it's more than SpaceX earns annually from several of its business lines, and makes plain just how acute Anthropic's compute hunger has become as demand for Claude surges.

Anthropic goes to Washington

In April, Anthropic opened its first Washington office with large event spaces to demonstrate its technology to regulators and discuss AI's implications for national security and the economy. A NYT piece captured the broader picture: a quarter of all federal lobbyists in the capital — people whose job is to advance corporate interests before lawmakers — are now working on AI issues in some capacity. In 2023, that figure was 11%.

Over the past year, Anthropic increased its government affairs spending tenfold to $3 million, hired six specialist firms, tripled its policy team, and plans to triple it again. January saw the company appoint its first dedicated head of lobbying.

Anthropic's position in the regulatory debate stands apart from its rivals: while OpenAI, Meta, and Google push for minimal government intervention, Anthropic openly supports new legislation. As Dario Amodei said at the company's first public event in Washington in September:

Many of the risks we’re most worried about are approaching us. They’re on the horizon."

Microsoft Maia 200 chip talks

CNBC confirmed that Microsoft is in talks to supply Anthropic with its custom Maia 200 AI chips. No deal has been signed yet.

For Microsoft, this would be a meaningful win: among its cloud rivals, Amazon and Google, the company lags in supplying clients with purpose-built AI silicon. The Maia 200 was announced in January, but the chip isn't yet available through Azure. At Microsoft's April earnings call, CEO Satya Nadella said the Maia 200 delivers over 30% better cost-per-token performance than the company's previous hardware. The chips are already running in Microsoft data centers in Arizona and Iowa.

The context for these talks is clear: Dario Amodei acknowledged in early May that Anthropic has "difficulties with compute" as demand for Claude and Claude Code has surged. The company has already contracted chips from Amazon and Google, and has historically relied on Nvidia GPUs. Microsoft would be the fourth source.

The collective bet on Anthropic

An informal but telling signal: activity on Polymarket. As of late May, the market puts Anthropic's private valuation at roughly $949 billion and gives 93% odds that the company will cross the $1 trillion mark by year's end. The probability of hitting $1.5 trillion stands at 76%.

This is no longer about a future IPO. It's a collective forecast from people putting real money on it.

If you pull all of May's events together, a clear picture emerges: a company simultaneously tackling three challenges — scaling infrastructure by any means available, monetizing through major enterprise contracts, and turning a profit for the first time. A good month.