ANALYSIS: Anthropic, the $65 billion company that wants to buy the future

65 billion, and the dizziness sets in

A number that outpaces the calendar

Some numbers describe a company. This one is already trying to describe an era. By the end of July, Anthropic’s annualized revenue had reportedly surpassed $65 billion, according to a report by Reuters and separately confirmed by CNBC. In May, the company was still talking about $47 billion. At the end of 2025, it stood at around $9 billion. The curve isn’t just rising—it’s practically tearing the page off.

But we must correctly name what dazzles us. An annualized rate takes a recent sales figure and projects it over twelve months. It is neither $65 billion already collected over the course of a year, nor a profit, nor a public audit. It is velocity transformed into a promise. In the world of artificial intelligence finance, velocity has become a currency even before it has become a balance sheet item.

The first duty is not to confuse the two

This clarification does not diminish the significance of the event. It makes it more troubling—and therefore more interesting. A private company can now present a surge lasting just a few weeks as the foundation for a colossal future market capitalization. Investors no longer pay only for what exists. They pay for the right to be there if the projection refuses to slow down.

The figure of 65 billion is not a full safe. It is an arrow launched toward the market, and everyone is already financing the spot where it promises to land.

The vertigo does not come from the amount. It comes from compressed time.

The run rate: the engine driving tomorrow

A useful metric, an immense temptation

The run rate serves an honest purpose: to show the current pace of a company whose growth quickly renders its annual financial statements outdated. For Anthropic, it reflects a shift in demand toward Claude, its interfaces, its API, and above all, its programming tools. It helps capture a surge in business that the financial statements of a private company do not yet allow the public to fully see.

The temptation begins when this pace is treated as destiny. A customer who signs up today may scale back tomorrow. A surge in usage may level off. A price may drop. A competitor could shift the market. Computing costs could eat into a portion that gross revenue doesn’t reveal. Annualizing assumes that the present will repeat itself with discipline. Artificial intelligence, however, does not repeat anything with discipline.

The projection does not account for risk

Risk, however, remains in the shadows: infrastructure expenses, talent compensation, chip purchases, energy, cloud computing contracts, and pressure on margins. AP noted during the May funding round that the private giants in this race continued to lose more money than they made. Sales are accelerating. This does not yet prove that the underlying economics of each query have stabilized.

A revenue line can grow faster than a cost line in a single quarter. A sustainable company must win the race over the course of years, under the watchful eyes of shareholders who won’t forgive indefinitely.

You can annualize a growth rate. You cannot annualize trust.

From 14 to 47, then to 65

Three milestones, six months

On February 12, Anthropic announced a $30 billion Series G funding round, a post-funding valuation of $380 billion, and annualized revenue of $14 billion. On May 28, the Series H round reached 65 billion, the valuation hit 965 billion, and the run rate stood at 47 billion. Then came the estimate at the end of July: over 65 billion. Three snapshots. Six months. A financial transformation almost too rapid for ordinary corporate vocabulary.

These numbers do not all come from the same light. The February and May figures appear in Anthropic’s announcements. The July figure was shared with investors and reported by the media based on confidential sources; Anthropic declined to comment on it to CNBC. This difference matters. It distinguishes a statement officially made by the company from information that is solid but still indirect.

Growth creates its own pressure

The steeper the climb, the more each subsequent milestone must justify the previous one. Growth from 14 to 47 billion changes the scale of the narrative. A jump from 47 to 65 in a matter of weeks changes its pace. Soon, slowing down won’t mean falling behind. Yet the market may interpret it that way. This is the trap of financial miracles: they set a standard that even the next miracle struggles to meet.

Anthropic no longer just has to sell Claude. It must now prove that the extraordinary can become ordinary without costs, customers, or technology rebelling.

Growth fuels expectations; expectations always end up demanding accountability.

Claude Code, the Gateway

The software that makes its way in through work

In February, Anthropic reported that Claude Code had surpassed $2.5 billion in annualized revenue. Its weekly active users had doubled since the start of the year, enterprise subscriptions had quadrupled, and enterprise usage accounted for more than half of the tool’s revenue. This isn’t a gimmick looking for an audience. It’s a production tool that finds its way into teams through a painful task: writing, debugging, understanding, and maintaining code.

That’s where its strategic strength lies. A consumer-facing assistant can be tried out and then forgotten. A tool integrated into the development cycle learns habits, adapts to code commits, becomes part of code reviews, and shifts deadlines. It becomes less visible at the very moment it becomes harder to remove. Adoption is no longer just a matter of preference. It’s starting to resemble infrastructure.

When Habit Becomes Dependence

Anthropic claims that customers then expand its use to financial analysis, data, sales, cybersecurity, and research. The commercial promise is crystal clear: first a door, then the hallway, then the building. Each expansion increases revenue. It also increases the organizational cost of walking away, because processes, training, and expectations are rewritten around the model.

Claude’s greatest victory would not be to be admired. It would be to be awaited every morning by teams who no longer know how to measure their day without him.

The product becomes powerful when its absence begins to take a toll.

Capital buys computing power

Billions Turned into Gigawatts

The 65 billion raised in May isn’t just to pad a valuation. Anthropic has announced agreements for up to five gigawatts of new capacity with Amazon, five gigawatts of next-generation TPUs with Google and Broadcom, as well as access to SpaceX’s GPU capacity. The terms may sound technical. Their meaning is stark: the company wants to secure a massive share of the physical resources that will power its models.

This race begins in data centers, power lines, and memory, storage, and semiconductor supply chains. It ends in a chat window that seems so light and airy. In between, there’s tied-up capital, long-term contracts, and global competition for limited resources. The cloud has always had a floor. AI simply makes it impossible to ignore.

Power Before Profit

Raising more capital allows for greater control. Greater control can improve models, reduce wait times, and attract customers. These customers then bolster annualized revenue, which justifies a new valuation. That’s the cycle. It can be virtuous. It can also become a wheel that demands ever more capital just to keep from slowing down.

Revenue is no longer just the result of a sale. It becomes a weapon for securing electricity, chips, engineers, and time before rivals take them.

In this industry, money doesn’t follow power. It precedes it.

The IPO as an accelerator, not a crowning achievement

Going Public to Open Another Reservoir

Reuters reports that Anthropic has confidentially filed for an initial public offering and that a listing could occur later in 2026. No public date, price, or number of shares has been announced yet, so it’s too early to call the deal a done deal. The IPO remains a possibility. But the logic behind it is already clear: to tap into a larger pool of capital at a time when the cost of the race is itself becoming astronomical.

An IPO does not make a technology mature. It changes the nature of future creditors. Private funds accept uncertainty in exchange for an early stake. Public markets then demand quarterly reports, explanations, margins, and clearly identified risks. The narrative will have to leave behind the confines of private meetings and withstand public scrutiny.

The bell doesn’t settle any questions

The listing could give Anthropic shares it can use to acquire, recruit, and compensate employees. It could also impose a discipline that the private company resists: disclosure of expenses, dependencies on suppliers, customer concentration, the nature of revenue, and the path to profitability. Public capital doesn’t just provide oxygen; it brings a clock on the wall.

An IPO would not be a medal awarded to Anthropic for its growth. It would be the moment when millions of investors would begin to pay to ensure that this growth remains credible.

The stock market does not complete the story. It removes the door to the room where the story was being told.

Valuation: A Vote on the Invisible

965 billion before the annual proof

In May, Anthropic was valued at $965 billion following its Series H funding round. In February, it was worth $380 billion. This increase isn’t the result of a long history of dividends. It reflects a massive bet on future revenue, a technological position, and the possibility that Claude will capture a lasting share of the world’s intellectual work. The private market is putting a price tag on a territory that hasn’t yet been fully built.

The price already reflects several victories

At this level, investors aren’t just betting on a good company. They’re assuming that Anthropic will remain among the leaders, that demand for AI will continue, that customers will accept the prices, that suppliers will deliver the capacity, and that regulation won’t break the model. Each of these conditions may be reasonable. Together, they form a cathedral of assumptions.

A valuation of 965 billion does not reward a single success. It is already pricing in an almost perfect series of technological, commercial, political, and industrial victories that have yet to be achieved.

The higher the price rises, the less room there is for the future to falter.

Companies are buying time

Why they’re paying now

Customers aren’t buying Claude because artificial intelligence is trendy. They buy it when it shortens a task, expands a team, or makes possible work that had been on hold. Anthropic says it serves eight Fortune 10 companies and has seen the number of customers spending more than $100,000 per year increase sevenfold in twelve months. More than 500 customers were then exceeding $1 million in annualized spending.

These numbers reveal a shift. Budgets are moving away from experimentation and into production. The CFO is no longer paying for a Friday demo. He’s paying for a capability expected by Monday. When the tool touches code, analytics, legal, or cybersecurity, the promised benefit isn’t just a cost savings. It’s the opportunity to get ahead of a competitor.

Time saved has an owner

But efficiency raises another question: who controls the shortcut? A company can save hours while transferring part of its expertise, data, and operational continuity to an external provider. If the model changes, if the price rises, if access is restricted, or if a government regulation intervenes, the time saved yesterday may become tomorrow’s dependency.

Claude sells thousands of saved minutes. Behind those minutes, Anthropic accumulates something even more valuable: a place in its clients’ daily operations.

Whoever sells time often ends up controlling the schedule.

The human cost isn’t in the presentation

A work revolution without the full picture

Annualized revenue measures what customers pay. It does not measure what professions are becoming. Claude Code can help a developer navigate a codebase, reduce a tedious task, or speed up a fix. It can also shift the line between junior and senior roles, alter productivity expectations, and make certain learning processes less visible. The same tool that eases a team’s burden can raise the bar for everyone.

It would be too simplistic to predict a widespread disappearance of jobs. The sources for this report do not allow us to quantify this, much less assert it. But they do allow us to observe the scale of adoption and the diversity of the roles being targeted. When a vendor says it is entering the fields of finance, law, sales, data, and research, it is not just selling software. It is negotiating the future shape of work.

Productivity always has a destination

The gain may go to the employee in the form of time, to the company in the form of profit margin, to the customer in the form of price, or to no one at all if the new pace immediately becomes the new obligation. This distribution is not reflected in the 65 billion figure. Yet it will determine whether AI is experienced as a lever or as an accelerating treadmill.

The question isn’t just how much Claude produces. It’s about who gets to breathe thanks to the time saved—and who simply has to run faster because that time exists.

Productivity is just a number until it becomes part of a human day.

The government: another client and another risk

Commercial power meets sovereignty

Anthropic’s trajectory also intersects with the state. In 2026, CNBC and AP reported tensions with the Trump administration and the Pentagon over military uses, restrictions, and government access to certain models. The details of these conflicts do not alter the financial core of the issue; rather, they show that a laboratory that has become infrastructure inevitably encounters the boundaries of national security.

A company can refuse a particular use, advocate for a security rule, or negotiate an exception. A government can impose a directive, revoke access, or classify a supplier as a risk. As soon as the models enter government agencies, defense, and cybersecurity, the commercial relationship is no longer ordinary. The customer also has the power to regulate the seller.

The market no longer decides alone

Anthropic sells intelligence to institutions capable of changing the rules of the trade. That is why its commercial power remains tied to a permanent political vulnerability.

When the customer writes the law, no contract is merely a contract.

The gross figure hides the net bill

Revenue, Margins, and Recognition

The figure of 65 billion says a lot about the momentum and very little about the exact economic quality of that momentum. AI companies sell directly, via APIs, and through cloud platforms. They pay for computing power, sometimes share revenue, and choose accounting methods that can complicate comparisons. Axios also points out that major tech firms do not necessarily measure their revenue in the same way.

The public will need more than just a numbers game. It will be necessary to distinguish between gross and net revenue, committed contracts and actual usage, organic growth and the impact of pricing, and margins before and after infrastructure costs. We’ll need to understand the concentration of major clients. Spectacular growth can be real while still painting an incomplete picture of what’s driving it.

Transparency will be the true outcome of the IPO

Public filings, if they materialize, will not eliminate all ambiguities. They will, however, force Anthropic to present a coherent, repeatable, and comparable financial framework. The market will then discover whether growth is based on a broad base of businesses, on a few massive contracts, on Claude Code, on the API, or on an even more powerful combination.

The next key figure may not be any higher than 65 billion. It will be the one that shows how much is left after paying for the machinery needed to generate it.

Revenue impresses. Margin reveals the truth.

Consolidation is happening quietly

Three clouds, a few labs

Anthropic presents Claude as the first state-of-the-art model available on AWS, Google Cloud, and Microsoft Azure. This distribution multiplies the entry points. It also concentrates a growing share of digital work around a few platforms, a few chip suppliers, and a few labs capable of funding the most resource-intensive training runs. Competition exists. Yet the cost of entering the market is becoming astronomical.

Dependency can take many forms

Offering Claude across three cloud platforms provides customers with a practical choice. This choice does not eliminate concentration if the same chip manufacturers, the same energy infrastructures, and the same small group of models are behind them. Resilience is not measured by the number of options in a catalog, but by the actual number of options capable of surviving an outage, a price hike, or a political conflict.

We may have several visible suppliers and a single deep dependency: that on costly, centralized intelligence controlled by companies that almost no one can compete with.

The diversity of storefronts does not guarantee the diversity of foundations.

The public will inherit a private gamble

From patient investors to impatient savers

So far, Anthropic’s greatest risks have been borne by funds, tech conglomerates, and investors capable of absorbing uncertainty. An IPO would shift part of that bet to pension plans, index funds, and individual investors. Even those who would never voluntarily buy the stock could be exposed to it if it were ever included in major indices.

This shift is not unethical. Public markets have long funded innovation. It does, however, impose a special responsibility: not to present a run rate as if it were a full fiscal year, not to present a valuation as proof of profitability, and not to confuse strategic urgency with a lack of choice.

The prospectus will need to take the heat off

The most important public document will not be the initial offering announcement. It will be the one listing the risks: dependence on partners, computing costs, competition, intellectual property, regulation, revenue concentration, and losses. The language there will be less melodious than that of funding announcements. It may be more valuable.

Private capital paid to accelerate the story. If public capital comes in, it will have to demand the right to read the pages where the story could fall apart as well.

A bet becomes a collective one at the very moment when caution ceases to be optional.

What 65 billion doesn’t tell you

No profit, no fiscal year

The filing does not allow us to assert that Anthropic generated 65 billion in revenue over twelve months. It does not publicly confirm an IPO date, a market valuation, or the terms of the offering. Nor does it establish the company’s sustainable profitability. These omissions are not embarrassing details to be hidden. They are the exact state of what we know today.

We know that Anthropic itself announced annualized revenue of 14 billion in February and 47 billion in May. We know that several credible media outlets reported 65 billion at the end of July. We know that colossal sums have been raised and that considerable physical capacity has been reserved. The rest remains to be seen in future documents.

The uncertainty does not undermine the argument

On the contrary. If even a partial indicator is already enough to reshape expectations, it’s because AI financing is advancing faster than its public transparency. The problem isn’t that the figure is necessarily false. The problem would be to ask it to answer questions it was never designed to resolve.

Rejecting exaggeration does not diminish Anthropic. It better reveals its power: even stripped of unproven promises, the company is already mobilizing capital and infrastructure on a historic scale.

The truth takes nothing away from the sense of wonder. It gives it a foundation.

The real race begins after the bell

Turning Momentum into an Institution

Anthropic must now accomplish what all financial rockets fear: becoming an institution without losing its momentum. It will have to serve customers, maintain models, fund infrastructure, uphold safety standards, respond to governments, and explain the numbers. It will have to do all this while competitors, in turn, raise funds, build, and sell.

Success won’t be measured solely by the first day of trading. It will be measured when the novelty has worn off, when quarterly comparisons become routine, and when a slowdown sets in. That’s when we’ll know whether Claude has built a foundation or merely ridden a massive wave.

The number is back, but it’s changed

65 billion. At the opening, the number looked like a triumph. By the end, it looks more like an obligation. It obliges Anthropic to prove its utility, margins, governance, and longevity. It obliges investors to distinguish between speed and destination. It obliges us to look beyond the sleek interface and see the sheer amount of power that is gathering.

The race to the IPO won’t just determine how much Anthropic is worth. It will determine how much of our work, our infrastructure, and our future we’re willing to entrust to a promise that’s still private.

65 billion. This time, it’s no longer just a rising number. It’s a question that concerns us all.

By Maxime Marquette, columnist

Columnist’s Transparency Box

Editorial Position

I am not a journalist, but a columnist and analyst. My expertise lies in observing and analyzing the geopolitical, economic, and strategic dynamics that shape our world. My work consists of dissecting political strategies, understanding global economic trends, contextualizing the decisions of international actors, and offering analytical perspectives on the transformations that are redefining our societies.

I do not claim to possess the clinical objectivity of traditional journalism, which is limited to factual reporting. Instead, I strive for analytical clarity, rigorous interpretation, and a deep understanding of the complex issues that affect us all. My role is to make sense of the facts, situate them within their historical and strategic context, and offer a critical interpretation of events.

Methodology and Sources

This text respects the fundamental distinction between verified facts and interpretive analysis. The methodological rule is consistent: factual information is published only if it is supported by a verifiable source, and the sources actually used in this article are listed under “Sources,” never here.

Categories of primary sources used by the publication, when applicable: official press releases from governments and international institutions, public statements by political leaders, reports from intergovernmental organizations, and dispatches from recognized international news agencies.

Types of secondary sources: specialized publications, internationally recognized news media, analyses from established research institutions, and reports from sector-specific organizations.

When an article cites statistical, economic, or geopolitical data, it comes from data-producing institutions (intergovernmental organizations, central banks, national statistical institutes), and the specific institution is named under “Sources.”

Nature of the Analysis

The analyses, interpretations, and perspectives presented in the analytical sections of this article constitute a critical and contextual synthesis based on available information, observed trends, and expert commentary cited in the sources consulted.

My role is to interpret these facts, contextualize them within the framework of contemporary geopolitical and economic dynamics, and give them coherent meaning within the broader narrative of the transformations shaping our era. These analyses reflect expertise developed through continuous observation of international affairs and an understanding of the strategic mechanisms that drive global actors.

This article describes a situation documented as of its publication date, not a prediction: subsequent developments may alter these perspectives. No updates are promised in advance; when an article is corrected or supplemented, the change is dated within the text.

ANALYSIS: Anthropic, the $65 billion company that wants to buy the future