Anthropic's Mythos Model and Strategic Moves Attract $300 Billion in Funding

Article | LetterAI

Foreign media report that Anthropic is expected to complete a $30 billion funding round as early as next week, with a valuation exceeding $900 billion.

In February, Anthropic's valuation was $380 billion; just three months later, it nearly tripled.

But the most interesting aspect is that Google also participated in this funding round.

In April, Google pledged to invest up to $40 billion in Anthropic, initially committing $10 billion in cash at a valuation of approximately $350 billion, with up to an additional $30 billion in investment tied to Anthropic’s future performance.

Google has Gemini, and it recently released Gemini 3.5 at the I/O conference.

Wouldn’t it be better to keep that $4 billion as R&D funding for your own products—especially when investing in a competitor?

Therefore, I believe Google may not want to buy Anthropic’s model—they want to buy a location.

It’s not uncommon for competitors to invest in each other—Intel invested in AMD, Microsoft invested in Apple, and Sony invested in Epic Games.

Large companies often invest in promising startups to achieve financial returns and maintain exposure to emerging technologies.

But Google's scale and persistence of investment in Anthropic seem a bit excessive.

After all, that’s $40 billion—it could fund Google to build another AI team entirely!

So what exactly has Anthropic done to captivate Google so thoroughly? Aside from Opus 4.7 and a few minor product updates, the only other thing surrounding Anthropic is Mythos.

This is a model that Anthropic calls so powerful it should never be given to ordinary consumers, while the counterpart to Mythos is the cybersecurity collaboration initiative known as the Glasswing program.

It’s like I wrote a groundbreaking article but chose not to publish it, because I’m afraid that once you all read it, you’ll become completely absorbed and unable to let go.

This in itself is absurd. Yet Anthropic, through various explicit and implicit channels, has made Mythos’s power vividly tangible in everyone’s mind—even if you’ve never seen it, you can clearly outline the model’s shape through news reports and fragments of information.

That’s why I feel Anthropic’s model products aren’t really the company’s core strength—what’s truly impressive is its ability to tell stories.

It can turn something invisible and intangible into real gold and silver.

We must first acknowledge the fact that the market accepts hype.

On May 23, Anthropic published an article titled "First Update on the Glasswing Project," detailing the latest progress on Glasswing.

The article states that Mythos Preview has scanned over 1,000 open-source projects and identified 6,202 high-severity or critical vulnerabilities. Independent security research firms evaluated these findings, confirming that 90.6% were true positives, with 62.4% classified as high-severity or critical.

It wasn’t until I read the article twice that I realized it was actually filler content.

The most subtle aspect of these numbers is the lack of a reference point.

Anthropic did not tell you the scale of these 1,000+ projects, how much code was scanned overall, how long it took, or how much improvement was achieved compared to traditional security tools.

Of these 6,202 candidate vulnerabilities, only 1,752 were actually evaluated.

This itself is very un-Anthropic. Previously, Anthropic’s articles were reproducible—for example, when they used Qwen to simulate whether humans could control AI if it surpasses human intelligence in the future.

Anthropic chose Qwen over their own Claude so that you can replicate it yourself after reading the article.

But that article by Glasswing doesn't, so you can't reproduce it or verify it.

More importantly, Anthropic frames its decision not to publicly release the model as an expression of responsibility—it refrains from release because it fears misuse, and instead only releases it to specific users, using their feedback to indirectly validate the model.

Anthropic used this marketing narrative to turn its "unreleased" features into proof of its high technical capabilities.

This grants it a special exemption: it doesn’t need to prove it’s stronger than all competitors—only that it’s “strong enough not to be publicly disclosed.”

Anthropic also mentioned in the update that industry standard practice in software is to disclose vulnerabilities publicly after 90 days, or 45 days after a patch is released. This means that the details of the vulnerabilities discovered in Mythos Preview cannot yet be fully disclosed, as doing so would expose end users to risk.

This is a layer of protection.

Similar to the Windows vulnerability disclosure process, research teams that discover Windows vulnerabilities first submit them to a designated email address and only publicly disclose the details after Microsoft has released an update to patch the vulnerability.

Anthropic’s claim is logically sound, but it’s not Microsoft, and Claude isn’t Windows—it’s just a model. If you can’t use Claude, you can use ChatGPT; if your computer can’t run Windows, it’s just a brick (I’m being a bit hyperbolic—you could also use Linux, etc., as long as you know how to use it).

This is exactly where Mythos shines. It’s not just potentially powerful—it’s deliberately designed as a product narrative that grows stronger the less it’s revealed.

Anthropic seized on this characteristic. It doesn’t need to make Mythos accessible to everyone—it only needs to convince the market that Mythos represents a more advanced capability. And precisely because this capability cannot be widely verified, it appears even more mysterious and powerful.

This is what Anthropic is doing.

It cleverly transforms "unverifiable capabilities" into "imaginable value."

This value will ultimately be reflected in its valuation, its fundraising efforts, and the market’s and society’s overall assessment of Anthropic.

If only Anthropic claims that Mythos is strong, this is just corporate marketing.

But if the U.S. government is simultaneously labeling Anthropic as a supply chain risk while being revealed to be nearing agreements with intelligence agencies like the NSA to use its advanced models, the story becomes entirely different.

In February, Trump stated that the U.S. government would blacklist Anthropic, and the Pentagon referred to Anthropic as a supply chain risk.

The most interesting aspect of the situation at the time was that the White House didn’t say Claude was unusable. On the contrary, the conflict arose precisely because it was “too eager to use.”

Foreign media report that the Pentagon seeks greater freedom to use Claude in military and national security scenarios, but Anthropic insists on maintaining restrictions on two uses: mass surveillance within the United States and fully autonomous weapons.

At the time, it felt like the left brain was attacking the right brain: the White House viewed Anthropic’s restrictions as a supply chain risk, while simultaneously acknowledging that Anthropic’s technology was sufficiently important.

Just after Glasswing's first update article was published, foreign media suddenly reported that the White House had partnered with Anthropic to allow specific institutions to use Claude, and it is highly likely to be Mythos.

Even the White House is willing to contradict its own ban and reinstate Claude, which shows just how powerful this tool really is.

This contradiction itself serves as validation for Anthropic.

Although the article did not say so directly, it conveyed a signal that Anthropic’s technology holds unique value.

Just after the White House banned Claude, OpenAI partnered with the White House to become its AI provider. But now the White House’s actions are essentially telling you, “Anthropic is irreplaceable.”

Anthropic doesn’t need to prove to the public how powerful Mythos is—it only needs the market to know that Mythos has now been selected by the White House.

Government procurement, especially by agencies related to national security, means your technology has passed the highest standards of review.

Being on the procurement list means your company has met the U.S. government’s requirements for supply chain security, technical reliability, and long-term support capabilities.

More importantly, government procurement is often accompanied by long-term contracts and stable income.

This is important for investors, especially since Anthropic is about to go public.

With such a large order, it's clear that Anthropic not only has cutting-edge technology but also the ability to turn that technology into stable revenue and has the most reliable customers.

More interestingly, Anthropic had not previously adopted a conciliatory stance.

After being named by the Pentagon, it publicly emphasized that it is not refusing to serve U.S. defense, but cannot abandon two red lines: large-scale domestic surveillance and fully autonomous weapons.

Amody later added in the company's statement that while the tone of internal remarks could be apologized for, these principles would not change.

In other words, Anthropic first framed a procurement conflict as “I have principles,” then used subsequent messages from the White House and NSA to portray itself as “still essential to the U.S. government.” This narrative holds greater传播 value than simply securing a government contract.

However, Anthropic did not mention this partnership at all.

If we loudly advertised, “We’re collaborating with the NSA,” the market might suspect it’s just a marketing tactic. But this complete silence, paradoxically, makes the market even more convinced of Mythos’s strength—because it has compelled the U.S. government to use it, even at the cost of appearing to backtrack.

Even if the collaboration between the White House and the NSA ultimately does not fully materialize, and even if the specific details of the agreement differ from public expectations, the narrative itself has already had an impact.

It has convinced the market that Anthropic is a company capable of collaborating with national security agencies and possesses technical capabilities that meet White House standards.

Anthropic's greatest strength is not convincing consumers, but convincing the hardest-to-persuade buyers.

The decision-making process of the national security system is extremely complex, involving multiple stages such as technical evaluation, security review, policy consideration, and budget approval.

Being able to navigate these steps demonstrates the company’s overall strength.

Technical narratives can be hyped, security narratives can be discussed, and cooperation with the U.S. government may also involve various gray areas.

But financing is the most straightforward—capital ultimately turns stories into money.

Previously, investors evaluating a company would look at its revenue growth, profit margins, market share, and technological barriers.

However, in the AI industry, these traditional metrics are often insufficient, as no one has yet been able to fully assess the potential of an AI company.

Foreign media reported that Anthropic told investors that their annualized revenue will exceed $50 billion next month.

Anthropic's earliest commercial revenue growth came primarily from enterprise APIs, cloud platform distribution, and code generation demand, rather than from standard Claude subscriptions.

In May 2025, foreign media reported that Anthropic’s annualized revenue rose from nearly $1 billion in December 2024 to over $2 billion by the end of March 2025, and reached approximately $3 billion by the end of May, driven primarily by enterprise demand and code generation use cases.

In the second half of 2025, Claude Code became an independent growth engine. Foreign media reported that Claude Code's annualized revenue reached approximately $400 million in July 2025, later approaching $1 billion.

Anthropic later confirmed in its official announcement regarding the acquisition of Bun that Claude Code reached $1 billion in run-rate revenue within six months of its public release.

By February 2026, Anthropic officially disclosed that the company’s overall run-rate revenue reached $14 billion, with Claude Code alone exceeding $2.5 billion. This means Claude Code has transformed from a small tool for developers into a multi-billion-dollar product.

In April 2026, Anthropic stated in its announcement regarding its computing power partnership with Google and Broadcom that its overall run-rate revenue had surpassed $30 billion, a significant increase from approximately $9 billion at the end of 2025.

Meanwhile, the number of enterprise customers with annual spending exceeding $1 million increased from over 500 in February to over 1,000.

Investors are clearly willing to believe this number, as various facts support Anthropic’s narrative.

Google's continued investment further enriches Anthropic's story.

If a regular VC invests in Anthropic, the market might view it as a normal venture capital move. But if AI giants like Google are also investing in Anthropic, it signals that Anthropic truly has something special.

Google doesn’t lack money, technology, or talent. Its investment in Anthropic isn’t because Anthropic can fill any gaps in Google’s capabilities, but because Google believes Anthropic has the potential to become a major player in AI—and it doesn’t want to miss this opportunity to get on board.

This judgment itself has become an endorsement for Anthropic.

When making decisions, investors often look to the choices of other investors. This isn't herd behavior—it's a rational way of gathering information.

Each investor has their own information sources and judgment capabilities; when multiple leading investors make the same decision, the probability that this decision is correct increases.

Google's investment is a strong signal—it tells other investors that Anthropic is worth investing in, and worth investing in significantly.

As a result, other investors began to follow suit. Sovereign wealth funds, top-tier VCs, and major tech companies all expressed interest in Anthropic. This collective action further increased Anthropic’s valuation.

The increase in valuation, in turn, reinforces Anthropic's narrative.

When a company’s valuation reaches $900 billion, the market naturally assumes that the company must possess exceptional capabilities and tremendous growth potential. Otherwise, how could it be worth so much?

Money plays the most crucial role in the story.

You might not understand technology, and you might not understand the White House—but can you really not understand the dollar?

What Anthropic has now is this kind of real-money endorsement.

This is the pricing function of capital. It ultimately converts all narratives, all imaginings, and all expectations into a specific number.

Of course, this consensus isn't necessarily correct—everyone, including Altman and Huang, acknowledges that there is a bubble in AI.

A high valuation doesn't guarantee a company's success, and raising more funds doesn't mean the technology is necessarily superior.

Throughout history, many highly valued companies have ultimately failed.

But at this moment, Anthropic has indeed accomplished something remarkable: it has packaged an unverifiable capability for ordinary users, a collaboration not yet officially announced by the government, and a future revenue potential into a compelling narrative—and used that story to raise money from the market.

This is Anthropic's business model.

The power of this mechanism lies in the fact that Anthropic doesn’t need everyone to see Mythos firsthand—it only needs the wealthiest, most powerful, and most risk-savvy individuals to act as though they have.

The performance of these individuals has become the best proof of Mythos, more convincing than any performance metrics or benchmarks.