Anthropic’s planned IPO is forcing public investors to confront the economics of frontier AI in unusually explicit terms.
A confidential prospectus reviewed by Reuters shows that the company generated nearly $4.6 billion in revenue in 2025, up roughly twelvefold from the prior year.
Operating losses exceeded $8 billion.
Net loss reached roughly $42 billion, although most of that figure reflected accounting charges tied to financing-related liabilities rather than normal operating cash expenses.
The company is seeking a valuation around $2 trillion.
At the same time, Anthropic expects to spend at least $518 billion on cloud, computing and infrastructure commitments over the coming decade.
Those numbers describe a company growing at extraordinary speed while taking on one of the largest infrastructure obligations in corporate history.
The IPO will therefore test whether public investors are willing to value frontier AI primarily on future growth or demand a much faster path to sustainable economics.
Revenue Growth Is Extraordinary
Revenue increased from roughly $400 million to nearly $4.6 billion in one year.
That is exceptional growth by any software standard.
Most of the revenue comes from usage-based consumption of Claude rather than traditional fixed subscriptions.
Anthropic reported roughly $3.8 billion from consumption-based usage and about $789 million from subscriptions.
That revenue mix matters.
Usage-based models can scale quickly when customers expand AI workloads.
They can also be volatile if customers optimize usage, shift models or reduce spending.
Enterprise AI increasingly behaves like cloud infrastructure: consumption can rise fast, but customers remain sensitive to price and performance.
The Operating Loss Is the More Important Number Than the Net Loss
The headline $42 billion net loss is dramatic.
It can also be misleading.
A large portion came from non-cash accounting adjustments related to the changing value of financing obligations.
The operating loss—more than $8 billion—is a better measure of the ongoing economics of the business.
Even that number is enormous.
Anthropic spent more than $7 billion on compute and infrastructure in 2025.
That demonstrates the central challenge of frontier AI.
Revenue can grow at software-like rates while the cost structure resembles infrastructure.
The business is not capital-light in the way traditional SaaS investors are accustomed to.
The $518 Billion Commitment Is the Real Balance-Sheet Story
Anthropic says it expects to spend at least $518 billion over the coming decade across six major infrastructure partners.
About 80% of those commitments are largely non-cancelable or require payment regardless of actual usage.
The company disclosed major commitments including roughly:
$111.1 billion with Google.
$110 billion with Amazon.
$31.4 billion with Microsoft.
It also has roughly $161.2 billion in Broadcom-related equipment lease obligations.
Additional arrangements include potential spending with xAI for Nvidia-based capacity and a major AMD relationship.
The scale means Anthropic is not simply renting cloud capacity as needed.
It is locking in infrastructure years in advance.
Why Anthropic Is Willing to Take That Risk
The company believes compute will be the main constraint on future AI growth.
If demand for advanced models grows faster than available capacity, companies that secure infrastructure early gain a strategic advantage.
That logic is similar to an airline locking in scarce aircraft or a semiconductor company reserving foundry capacity.
The risk is obvious.
If AI demand grows more slowly than expected, Anthropic could owe enormous amounts for capacity it does not fully use.
That is why the non-cancelable structure is so important.
Big Tech Is Partner, Supplier and Competitor at the Same Time
Anthropic depends heavily on Amazon and Google.
About 47% of its 2025 revenue flowed through cloud marketplaces operated by Amazon and Google.
Those companies also invest in Anthropic.
They provide compute.
They distribute Claude.
And they compete with Anthropic through their own AI products.
That creates unusually complex incentives.
Anthropic paid roughly $351 million in distribution fees to cloud platforms in 2025.
In other words, a meaningful share of revenue depends on companies that also compete for the same enterprise customers.
Customer Concentration Is Another Risk
Reuters reported that a small number of customers account for a meaningful share of Anthropic’s revenue.
Two customers each generated around 12% of sales.
That concentration can accelerate growth when major customers scale usage.
It also creates vulnerability.
A large customer that changes models, renegotiates pricing or reduces consumption can have an outsized impact.
Public investors will want to know whether Anthropic’s customer base becomes more diversified before the IPO.
The Safety Disclosures Are Unprecedented
Anthropic also plans to warn investors that advanced AI could create “catastrophic or existential risks to humanity.”
The filing discusses possible behaviors such as resisting shutdown, concealing information or acting in manipulative ways.
These are risk disclosures from Anthropic, not evidence that such outcomes are inevitable.
The important investment point is that safety is now part of the company’s legal and financial risk framework.
If regulators impose stronger requirements or if customers demand more monitoring, safety spending could rise further.
Anthropic is also a public benefit corporation, meaning management is legally required to consider its stated public mission alongside shareholder interests.
That structure may become an important governance issue for IPO investors.
What the $2 Trillion Valuation Implies
A $2 trillion valuation would put Anthropic among the most valuable technology companies in the world almost immediately.
That valuation cannot be justified by current earnings because the company is deeply loss-making.
It depends on expectations for future revenue, market share and margins.
The central debate will be whether Anthropic can grow into its infrastructure commitments faster than those commitments consume cash.
If it can, the huge compute buildout may look visionary.
If it cannot, the same contracts become a financial burden.
What It Means for Public AI Stocks
Anthropic’s spending plan is directly relevant to Nvidia, AMD, Broadcom, Amazon, Alphabet and Microsoft.
Those companies sit on the other side of Anthropic’s infrastructure commitments.
A successful IPO and continued growth would support demand for chips, cloud and networking.
A difficult IPO could make investors more skeptical of the circular financing relationships and enormous capital commitments across the AI ecosystem.
The filing therefore matters even before Anthropic becomes publicly traded.
What to Watch Next
Watch the public S-1 when it is filed.
Watch whether the $2 trillion valuation target changes.
Watch 2026 revenue growth.
Watch operating losses.
Watch customer concentration.
Watch cloud-marketplace dependence.
Watch infrastructure commitments.
Watch compute utilization.
And watch whether Anthropic can improve margins while still investing aggressively in safety and model development.
The company’s growth is real.
So are the obligations.
The IPO will force public markets to decide how much future AI dominance is worth when achieving it may require more than half a trillion dollars of infrastructure commitments.