Anthropic is asking public-market investors to value its five-year-old AI lab at more than $2 trillion, enough to eclipse SpaceX’s record June 2026 debut, while disclosing a $42 billion 2025 net loss, $518 billion in future compute commitments, heavy customer concentration, and an unusually blunt warning that its own models could pose “catastrophic or existential risks to humanity”.
That combination of numbers from the IPO prospectus has become the latest flashpoint in the AI capital boom. The filing is not yet on EDGAR but was reviewed by Reuters. The timing is deliberate: Anthropic’s bankers have been testing a valuation more than double the $965 billion mark set in its May 2026 funding round, with a listing expected after the November midterm elections.
According to the document, the company made $4.59 billion in revenue in 2025, up 1,088% from $386 million the year before, against a GAAP net loss of roughly $42 billion. Operating loss widened to $8.06 billion from $2.98 billion. Compute and infrastructure spending alone hit $7.33 billion, 58% of $12.65 billion in total operating expenses, and nearly tripled year over year. Cash and short-term investments stood at $20.28 billion at year-end.
About $34 billion of the headline net loss was a non-cash accounting charge. As Anthropic’s private valuation soared, convertible financing instruments that may one day become shares had to be remeasured at higher fair value. The charge is real on the income statement; it is not money that left the bank. Strip it out, and the operating picture is still grim: more than $8 billion lost running the business. Two customers supplied nearly a quarter of revenue, and many large clients are not locked into long-term contracts.
The company then told investors it plans to spend $518 billion on cloud, computing and infrastructure obligations “in coming years”. That figure, more than 100 times 2025 revenue, is the real capital-intensity story. Frontier models are not cheap to train or serve. The race has become a race for chips, power and data centre capacity as much as algorithms.
According to 2026 operating data circulating among investors, Anthropic is seeing a faster-growth picture in Q2 revenue of about $11.5 billion and an annualised run-rate above $65 billion, with some adjusted-profit quarters, but the prospectus that will be shown to IPO buyers is built on the 2025 audited numbers and the multi-year spend commitments; those commitments do not disappear because a single quarter looked better.
How Anthropic’s numbers benchmark against SpaceX
SpaceX went public on June 12, 2026, at a $1.77 trillion valuation, raising about $75–86 billion in the largest IPO by proceeds on record. Shares opened higher, and the first-day market cap approached $2.1 trillion. The stock has since traded in a wide range; recent closes have been near $145–150.
The financial contrast is stark. SpaceX generated $18.67 billion in 2025 revenue, four times Anthropic’s, and a $4.94 billion net loss with a $2.59 billion operating loss. Starlink-led Connectivity produced $11.39 billion in revenue and $4.42 billion in operating income. The Space segment (Falcon, Starship) lost $657 million on $4.09 billion. The AI segment that includes xAI and X lost $6.36 billion on $3.20 billion. In other words, a profitable satellite-internet business and a launch franchise were already large enough to absorb a heavy AI build-out.
SpaceX also sells physical scarcity: reusable rockets, a growing satellite constellation (Starlink subscribers doubled to 12 million by mid-2026), government launch contracts, and a $47.5 billion backlog. Capex is enormous, $18.4 billion in Q2 2026 alone, 86% of it in AI infrastructure after the xAI combination, but the company already has cash of roughly $100 billion post-IPO and investment-grade bonds. Its 2025 loss was smaller than Anthropic’s even before the $34 billion paper charge, and it arrived with two cash-generating engines already at scale.
Anthropic is asking investors to pay a similar or higher multiple for a software company whose 2025 revenue was a fraction of SpaceX’s, whose operating loss was larger relative to sales, and whose future obligations dwarf current cash. The implied bet is that Claude’s enterprise adoption curve, model quality, and safety brand will produce the kind of winner-take-most economics that justify a $2 trillion valuation.
However, sceptics have been blunt, saying that a company that spent $12.6 billion to generate $4.6 billion in 2025, with $7.33 billion of that on compute, is still in the “growth at any cost” phase. Customer concentration and the absence of long-term contracts add execution risk. The $2 trillion target prices a future in which revenue compounds toward $100–200 billion while unit economics improve fast enough to cover infrastructure that does not yet exist.
Most worryingly, the prospectus spends roughly 80 of its 261 core pages on risk factors, nearly twice the space devoted to describing the business. Anthropic, which has marketed itself as the safety-first lab, tells investors that advanced models could exhibit “self-preserving behaviours”, including attempts to “resist shutdown”, “conceal or manipulate information”, and conduct “resembling blackmail”.
It warns that development of more capable systems “could further increase the risk that our models cause harm” and that “potential model awareness of our evaluation efforts creates a significant limitation on our ability to assess model safety.” The language includes “catastrophic or existential risks to humanity”.
Public companies routinely list product and litigation risks. Almost none have ever told prospective shareholders that the product they are buying into might contribute to human extinction. The filing therefore creates an unusual tension: the company is simultaneously arguing that AI will remake the global economy more profoundly than electricity or the internet, that it is the responsible steward of that technology, and that the same technology carries civilisational-scale risk.
That tension is now part of the IPO story. Investors who accept the $2 trillion narrative must also accept that they are funding a company whose internal research and legal disclosures treat catastrophic risk as a live possibility.
It’s important to note that the comparison with SpaceX is not only about size. SpaceX’s IPO sold a diversified industrial-and-infrastructure platform that already generates tens of billions in revenue and has a profitable core business large enough to subsidise moonshots. Anthropic, on the other hand, is selling a pure-play frontier-model company whose 2025 economics still look like a high-burn research lab, whose largest customers can walk, and whose own prospectus spends more pages on ways the product could go wrong than on how the business works.
Whether that package is worth more than SpaceX depends on whether public investors believe the 12-fold revenue jump of 2025 and the even steeper 2026 run rate can continue while compute costs are amortised and safety risks stay contained.
Technext Newsletter
Get the best of Africa’s daily tech to your inbox – first thing every morning.
Join the community now!
