By / Daoge

From / Jiedian Finance

The record set by SpaceX’s IPO just a few months ago is about to be broken.

According to multiple media reports, Anthropic has confidentially submitted its S-1 prospectus to the U.S. SEC, planning to officially list on Nasdaq after the U.S. midterm elections in November. This time, it’s aiming for something staggering: raising up to $100 billion with a target valuation exceeding $2 trillion.

What does that mean? When SpaceX went public in June this year, it raised $85.7 billion with a valuation of $1.77 trillion—already the largest IPO in history. Anthropic’s move instantly raises the bar by an order of magnitude. If successful, this will be the largest public offering in human history—period.

But here’s the question: This company reported a net loss of $42 billion in 2025 on just $4.6 billion in revenue. Why is Wall Street valuing it at $2 trillion?

Let’s unpack the logic behind this "IPO of the century" and what it means for the entire AI industry—and for Chinese entrepreneurs.

The "Alchemy" of a $2 Trillion Valuation: Revenue Surges 12x, Losses Surges 12x

Take a look at these numbers—you might think you’re misreading the decimal point.

In 2025, Anthropic generated $4.6 billion in revenue, up 12x year-over-year. Its annualized revenue run rate (ARR) has skyrocketed to approximately $65 billion. For context, OpenAI’s ARR last year was just over 100 billion RMB—Anthropic’s growth rate ranks among the highest in tech history.

Even more staggering is its valuation trajectory. In May this year, Anthropic completed its Series H funding at a post-money valuation of $965 billion, already surpassing OpenAI. Just five months later, its IPO target valuation has doubled to $2 trillion—a 100% increase in five months. This isn’t fundraising; it’s printing money.

But on the flip side, losses are equally astronomical. Net loss in 2025: approximately $42 billion. Of course, about $34 billion of that is non-cash accounting expenses—basically paper gains from rising valuation, not actual cash burned. Even so, real operating losses exceed $8 billion.

Revenue up 12x, losses up even faster. This mirrors Amazon’s playbook from years past: burn cash relentlessly for growth, trading future stories for today’s valuation.

So what’s Anthropic’s story?

The answer is computing power. The prospectus reveals Anthropic has already committed approximately $518 billion to future computing power and infrastructure, with 80% non-cancellable. What does that mean? It’s betting the next decade’s ammunition on GPUs. If AI demand explodes, these commitments become money printers. If demand falters, they become an anchor dragging the company under.

Moreover, Anthropic’s revenue structure hides a risk: Nearly half (47%) of its revenue comes through distribution channels like Amazon Web Services and Google Cloud, while nearly a quarter comes from just two major customers. What does that imply? For all its talk of being an independent AI company, its lifeline is actually controlled by tech giants.

Surpassing SpaceX: A Coronation for AI’s New Era or a Signal of Peak Bubble?

Many compare Anthropic to SpaceX—both are "hard tech + mega-unicorn" stories. But dig deeper, and the parallels vanish.

SpaceX’s valuation logic is "real revenue + future options": It has Starlink’s cash flow, launch service orders, NASA contracts. Its $1.77 trillion valuation, while steep, is tangible.

Anthropic is different. At a $2 trillion valuation and $65 billion ARR, its price-to-sales (PS) ratio hits 30.8x. For comparison, SaaS star Snowflake trades at 8-10x forward PS, while NVIDIA trades at just over 20x. Anthropic is effectively priced higher than NVIDIA.

Wall Street’s rationale: Forget now, look to 2028. Anthropic itself predicts 2028 revenue will reach $190-200 billion. At that level, the $2 trillion valuation’s PS ratio drops to ~10x—"reasonable."

But how credible is this forecast?

Research firms like PitchBook pour cold water: Leaked financials show rapid growth but nothing justifying a $2 trillion valuation. To sustain it, Anthropic must simultaneously: boost gross margins, retain major clients, and exceed computing utilization targets. Fail on any front, and the valuation could halve.

Even more intriguing is the timing. Why after the U.S. midterm elections in November?

Subtle political calculus is at play. AI regulatory policies are being debated in Congress. An IPO before the elections carries too much regulatory uncertainty. Post-election, policy direction becomes clearer, allowing for more confident pricing. This suggests a significant "policy risk premium" is baked into Anthropic’s valuation, not just market pricing.

Another overlooked detail: Anthropic’s risk factors section actually lists "AI could threaten human survival." A company going public is flagging "human extinction" as a risk—you could call it honesty, or you could call it covering its back: "We warned you."

This dual role—player and referee—epitomizes this AI bubble: raise billions while shouting about risks.

Three Signals for China’s AI Industry: Amid the Euphoria, Where Are the Opportunities?

Anthropic’s IPO isn’t just a U.S. capital markets event. It sends three critical signals to China’s AI entrepreneurs, investors, and even retail investors.

Signal 1: The capital markets have finally "endorsed" the large language model (LLM) business model.

For years, China has debated: Can LLMs actually make money? BAT companies have poured hundreds of billions, but where’s the revenue? Anthropic’s financials—$42 billion in losses but 12x revenue growth—tell Wall Street: Forget profits for now, focus on growth. As long as growth is fast enough, losses are "strategic investments."

This is a boon for domestic LLM firms—now they have a valuation benchmark. Future fundraising can simply apply Anthropic’s PS multiples. But the flip side: If Anthropic’s stock tanks post-IPO, the global LLM valuation framework could collapse. Domestic firms would then struggle to sell "long-term stories."

Signal 2: The ceiling for the computing power arms race is being redefined.

$518 billion in computing commitments—what does that mean? Roughly 3.7 trillion RMB. This isn’t just a company’s investment; it’s on par with a nation’s annual fiscal budget. When an AI firm’s computing spend exceeds a country’s military budget, "computing power = national power" stops being a slogan.

For China, this is pressure—and motivation. Blocked from advanced GPUs, China must innovate in domestic alternatives, model efficiency, and applications. The crazier Anthropic gets, the clearer it becomes that computing power is this era’s oil—and China must control its own supply.

Signal 3: The "graduation season" for AI unicorns has arrived, but graduation ≠ safety.

Anthropic’s IPO marks the first batch of AI unicorns entering public market pricing. Next up: OpenAI, xAI, Mistral… all lining up for IPOs. What does this mean?

It means the AI industry is transitioning from "private fundraising storytelling" to "public market accountability." Stock prices fluctuate daily, quarterly reports expose all, and underperformance triggers sell-offs. PPT-driven fundraising is over; every dollar of revenue must now justify itself to shareholders.

For Chinese entrepreneurs, this is a crucial reference: Don’t envy Anthropic’s $2 trillion valuation—see the pressures behind it: $518 billion in computing debt, client concentration risks, policy uncertainties. When bubbles inflate, everyone wins. When they burst, the slowest to exit pay the price.

Final Thoughts

What does $2 trillion represent? It’s equivalent to Italy’s GDP, or a quarter of the entire A-share market’s value. A five-year-old, deeply unprofitable company seeking this valuation in public markets—call it a miracle of human innovation, or call it Wall Street’s collective madness.

But history teaches us: Bubbles and revolutions are two sides of the same coin. The railway bubble birthed modern logistics; the dot-com bubble birthed Google and Amazon. This AI bubble may well produce a company that changes the world.

The question remains—will it be Anthropic?

In November, we’ll find out.