The veil of Mythos has become the lever for Anthropic to move trillions.

null

Text | Alphabet AI

According to foreign media reports, Anthropic is expected to complete approximately $30 billion in funding as early as next week, with a valuation exceeding $900 billion.

As early as February, Anthropic's valuation was still $380 billion; in just three months, it nearly tripled.

But what's most interesting here is that Google also participated in this round of financing.

Google once promised in April that they would invest up to $40 billion in Anthropic. An initial $10 billion in cash was invested at that time, with a transaction valuation of about $350 billion, and subsequent investments of up to $30 billion would be tied to Anthropic's performance.

Google has Gemini, which was just released at the I/O conference as Gemini 3.5.

Wouldn't it be better to use these $40 billion for R&D on their own products? Especially since they are investing in a competitor.

Therefore, I think Google may not want to buy Anthropic's models; they want to buy a position.

In fact, it’s not uncommon for competitors to invest in you—Intel invested in AMD, Microsoft invested in Apple, Sony also invested in Epic Games.

Large companies often invest in promising startups, both to gain financial returns and to stay connected with new technologies.

But Google's scale and ongoing commitment to its investment in Anthropic seem a bit “over the top.”

After all, that’s $40 billion! Enough for Google to build another AI team themselves.

So what exactly has Anthropic done to make Google so captivated? Besides Opus 4.7 and some minor product updates, the only thing left around Anthropic is Mythos.

This is a model Anthropic claims is so powerful that it cannot be used by ordinary consumers, and it is associated with the network security collaboration project Glasswing.

It’s like writing a sensational article that I don’t publish because I fear that once you all read it, you’ll be deeply immersed and unable to extricate yourselves.

This is absurd in itself. But Anthropic, through various overt and covert means, makes Mythos’s power tangible in everyone’s mind—so even if you haven’t seen it, you can sketch out its outline through news and snippets.

That’s why I believe that Anthropic’s model products are not the core of this company; what truly impresses people is its storytelling ability.

It can turn something invisible and intangible into real money.

Latest progress on Glasswing

We must first acknowledge a fact: the market accepts pie-in-the-sky promises.

On May 23, Anthropic published an article titled “First Update on the Glasswing Plan,” discussing the latest progress of Glasswing.

The article states that Mythos Preview has scanned over 1,000 open-source projects, discovering 6,202 high-risk or severe vulnerabilities. An independent security research firm evaluated these, confirming 90.6% as true positives, and 62.4% as high-risk or severe.

But until I read that article twice, I realized it was actually a “fluff piece.”

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

Anthropic didn’t tell you what scale these 1,000+ projects are, nor how much code was scanned, 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 in itself is very “un-Anthropic.” Previously, Anthropic’s articles were reproducible—for example, their use of Qianwen to simulate whether humans could control AI once it surpasses human intelligence.

They used Qianwen instead of their own Claude so that readers could reproduce the results after reading.

But the Glasswing article lacked that; you can’t reproduce it, nor verify it.

More critically, Anthropic packaged the “non-disclosure” as a sense of responsibility. They fear misuse, so they don’t publish. They only share with specific users, whose feedback is then used to indirectly validate the model.

Anthropic uses this marketing story to turn “not publicly released” into proof of their high technical level.

This grants them a special exemption. They don’t need to prove they’re better than all competitors; they only need to prove they’re “so strong that they can’t be casually disclosed.”

Anthropic also mentioned in the update that the industry practice is to disclose vulnerabilities 90 days after discovery or 45 days after patches are released. This means the vulnerabilities found in Mythos Preview cannot be fully disclosed yet, or it would pose risks to end users.

This is a layer of protection.

Just like Windows vulnerabilities, when a team finds a Windows flaw, they first send it to a designated email. Only after Microsoft releases an update patch do these teams publicly disclose the specific vulnerabilities.

Anthropic’s reasoning is logically impeccable, but it’s not Microsoft, Claude isn’t Windows—it’s just a model. If I can’t use Claude, I can use ChatGPT; if your computer can’t run Windows, it’s just a brick (I’m a bit full of myself saying this—you can also use Linux, etc., as long as you know how).

That’s the most powerful aspect of Mythos. It’s not just potentially very strong; it’s designed as a product narrative that “the less public, the stronger.”

Anthropic has grasped this feature. It doesn’t need everyone to use Mythos; it only needs the market to believe that Mythos represents a higher level of capability. And because this capability can’t be widely verified, it appears even more mysterious and powerful.

That’s what Anthropic is doing.

It’s cleverly turning “unverifiable capability” into “imaginable value.”

And this value ultimately reflects in valuation, funding, and the overall market and societal perception of Anthropic.

White House reactivates Claude

If only Anthropic claims Mythos is powerful, that’s just marketing.

But if the U.S. government views Anthropic as a supply chain risk and is also reported to be close to deploying its advanced models for NSA and other intelligence agencies, the story changes completely.

In February, Trump said the U.S. government would blacklist Anthropic, and the Pentagon labeled it a supply chain risk.

The most interesting part then was that the White House didn’t say Claude was bad. On the contrary, the conflict was precisely about “wanting to use it too much.”

Foreign media reported that the Pentagon hoped to use Claude more freely in military and national security scenarios, but Anthropic insisted on restricting two types of use: domestic mass surveillance and fully autonomous weapons.

At that time, it was like a split brain—White House thought Anthropic’s restrictions made it a supply chain risk, yet also recognized that Anthropic’s technology was sufficiently important.

But after the first update on Glasswing, foreign media suddenly reported that the White House had reached a cooperation with Anthropic, allowing certain agencies to use Claude, most likely Mythos.

Even the White House is willing to contradict its own ban and re-enable Claude—that shows this thing is truly formidable.

This contradiction itself is backing Anthropic.

Although the news didn’t directly say so, it sent a signal: Anthropic’s technology has unique value.

Because after the White House banned Claude, OpenAI partnered with the White House as an AI supplier. But now, the White House’s actions seem to tell you, “Anthropic is irreplaceable.”

Anthropic doesn’t need to prove to the public how strong Mythos is; it only needs to let the market know that Mythos has now been included in the White House’s carefully selected list.

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

Once on the procurement list, it indicates your company meets U.S. government requirements for supply chain security, technical reliability, and long-term support.

More importantly, government contracts often come with long-term agreements and stable revenue.

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

Having such a large order shows that Anthropic not only has technology but also the ability to turn that technology into stable commercial income and has the most reliable customers.

What’s more interesting is that Anthropic was not previously in a submissive or apologetic stance.

After being named by the Pentagon, it publicly emphasized that it does not refuse service to U.S. defense but cannot give up two red lines: large-scale domestic surveillance and fully autonomous weapons.

Amodei later said in a company statement that the tone of internal remarks could be apologetic, but these principles would not change.

In other words, Anthropic framed a procurement conflict as “I have principles,” and later, through White House and NSA messages, turned it into “the U.S. government still needs me.” This storytelling has more dissemination value than simply winning a U.S. government contract.

However, Anthropic made no mention of this cooperation.

If they had loudly announced “We are working with the NSA,” the market might suspect marketing hype. But this complete silence instead makes the market more convinced of Mythos’s strength, because even if the White House and NSA’s cooperation doesn’t fully materialize, or the details differ from expectations, the narrative has already taken effect.

It has already made the market believe that Anthropic is a company capable of working with national security agencies, a company whose technology meets White House standards.

Anthropic’s strongest point isn’t convincing consumers; it’s convincing the most difficult procurement parties.

The decision-making process in national security systems is extremely complex, involving technical assessments, security reviews, policy considerations, and budget approvals.

Being able to pass through these stages already demonstrates the company’s comprehensive strength.

All for funding and going public

Tech storytelling can be exaggerated, security narratives can be spun, and government cooperation can have gray areas.

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

In the past, investors would evaluate a company based on revenue growth, profit margins, market share, and technological barriers.

But in the AI industry, these traditional metrics are often insufficient. Because to this day, no one can clearly see the potential of an AI company.

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

Anthropic’s earliest commercial revenue growth was not mainly from ordinary Claude subscriptions but from enterprise APIs, cloud platform distribution, and code generation demands.

Foreign media reported in May 2025 that Anthropic’s annualized revenue approached $1 billion in December 2024, grew to over $2 billion by March 2025, and by May reached about $3 billion, mainly driven by enterprise demand and code generation scenarios.

In the second half of 2025, Claude Code became an independent growth engine. Foreign media said that in July 2025, Claude Code’s annualized revenue was about $400 million, then approaching $1 billion.

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

By February 2026, Anthropic disclosed that the company’s overall run-rate revenue reached $14 billion, with Claude Code alone exceeding $2.5 billion. That is, Claude Code had transformed from a developer’s small tool into a product worth tens of billions.

In April 2026, in the announcement of Google and Broadcom’s computing power collaboration, Anthropic stated that the company’s overall run-rate revenue had exceeded $30 billion, continuing to surge from about $9 billion at the end of 2025.

Meanwhile, the number of enterprise clients spending over $1 million annually increased from over 500 in February to more than 1,000.

Investors clearly believe in these figures. Because there are various facts supporting Anthropic’s storytelling.

Google’s continued investment makes Anthropic’s story even more compelling.

If an ordinary VC invests in Anthropic, the market might see it as normal risk investment. But if giants like Google are investing, it indicates that Anthropic truly has extraordinary potential.

Google isn’t short of money, technology, or talent. Its investment in Anthropic isn’t because Anthropic can fill gaps; it’s because it judges that Anthropic might become a major player in AI, and it doesn’t want to miss this opportunity.

This judgment itself also becomes a form of endorsement for Anthropic.

When making investment decisions, investors often look at what other investors choose. This isn’t just following the crowd; it’s a “rational way of gathering information.”

Each investor has their own information channels and judgment capabilities. When multiple top-tier investors make the same choice, the probability that the choice is correct increases.

Google’s investment is such a strong signal. It tells other investors that Anthropic is worth investing in, and worth large-scale investment.

As a result, other investors also start to follow suit. Sovereign funds, top VCs, major tech companies—all show interest in Anthropic. This collective action further boosts Anthropic’s valuation.

The rising valuation, in turn, reinforces Anthropic’s narrative.

When a company’s valuation hits $900 billion, the market naturally assumes that this company must have very strong capabilities and broad prospects. Otherwise, how could it be worth so much?

Money plays the most critical role in storytelling.

You may not understand the technology, nor the White House, but can you not understand dollars?

What Anthropic has now is precisely this kind of solid backing.

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

Of course, this consensus isn’t necessarily correct. Including Ultraman and Huang Renxun, everyone admits there are bubbles in AI.

High valuation doesn’t necessarily mean the company will succeed; lots of funding doesn’t guarantee technological leadership.

Many companies with high valuations in history eventually failed.

But at this moment, Anthropic has indeed achieved something impressive. It packaged an unverified capability for ordinary users, a cooperation not yet officially announced by the government, and a future revenue expectation into a complete story, then used that story to raise funds from the market.

This is Anthropic’s business model.

The brilliance of this mechanism is that Anthropic doesn’t need everyone to see Mythos with their own eyes; it only needs the wealthiest, most powerful, and most risk-aware people to act as if they’ve already seen it.

And their behavior becomes the best proof of Mythos—more convincing than performance metrics or benchmark scores.