Some Anthropic investors reportedly believe the company could go public at a valuation of $2 trillion or more. This is not an official target from Anthropic, but it would make the five-year-old Claude developer one of the world’s most valuable companies.
At first, the figure looks excessive. Anthropic may currently lead some important model rankings, but it is not the undisputed winner. OpenAI, Google and xAI remain close, regularly beating it on particular benchmarks and workloads.
But then there is SpaceX.
The SpaceX comparison
As of August 14, SpaceX had a market capitalization of approximately $1.84 trillion. More importantly, it is increasingly positioning itself as an AI and computing company rather than simply a rocket and satellite business.
Following its acquisition of xAI, SpaceX now owns Grok, X and the Colossus computing infrastructure. It also sells computing capacity to outside customers, including Anthropic.
Elon Musk recently told employees that AI revenue could exceed all SpaceX’s other business lines combined as early as September. He believes AI could eventually account for almost all the company’s value.
These are ambitious projections rather than audited guidance. SpaceX still has substantial launch and Starlink operations. But the strategic direction is clear: AI is becoming its central growth thesis.
If SpaceX can command a valuation approaching $2 trillion, a similar valuation for Anthropic no longer looks inherently absurd.
Anthropic is arguably the stronger pure AI business. Its annualized revenue crossed $47 billion in May, while SpaceX reported approximately $23 billion in trailing revenue. At $2 trillion, Anthropic would trade at about 43 times its current annualized revenue. SpaceX currently trades at roughly 80 times trailing revenue.
These are not directly equivalent measures, but Anthropic does not look obviously more expensive. Of course, comparing one aggressively valued company with another can explain a price without justifying it.
The cost of staying at the frontier
In May, Anthropic projected that second-quarter revenue would reach $10.9 billion, producing approximately $559 million in adjusted operating profit—potentially its first profitable quarter.
The numbers underneath that projection are more revealing.
In the first quarter, Anthropic reportedly spent 71 cents on computing for every dollar of revenue. It expected that ratio to decline to 56 cents in the second quarter.
That would be a significant improvement, but it would still leave only 44 cents to cover research, engineering, sales, administration and the continuing race to develop the next frontier model. The projected adjusted operating margin was approximately 5%, and the calculation reportedly excluded stock-based compensation—an important omission in perhaps the most expensive technical labor market in history.
The larger problem is the shortening economic life of each frontier model.
Anthropic launched Fable 5 as its leading public model in June. Approximately six weeks later, it introduced Opus 5, which reportedly surpassed Fable on several benchmarks while costing customers half as much.
That is extraordinary innovation. It is also an economic warning.
A model may remain useful for years, but its ability to command premium prices can disappear within months. Anthropic must recover enormous development costs before either a competitor—or Anthropic itself—releases something better and cheaper.
When temporary disruption becomes permanent migration
Enterprises are also becoming more cost-conscious.
CIOs, CTOs and chief AI officers are introducing spending limits, model-routing systems and stricter return-on-investment requirements. This does not mean companies will stop buying AI. It means they will become more selective about which model receives each token.
A difficult coding problem may justify Claude’s most advanced model. Summarizing a document or classifying a customer-support request may not.
That opening became more visible in June, when Anthropic was forced to suspend Fable 5 and Mythos 5 following a U.S. government export-control directive. Fable remained unavailable for nearly three weeks, while Mythos returned only to selected U.S. organizations.
The interruption showed customers that access to an advanced proprietary model could disappear overnight—not because of technical failure, but because of a government decision.
Chinese-model adoption on OpenRouter had already been rising before the suspension. Among U.S. organizations using the platform, Chinese models had captured more than 30% of weekly token volume since February, reaching as high as 46%.
During the Anthropic interruption, Chinese models moved into OpenRouter’s leading positions. Their usage remained elevated after Fable returned, suggesting that at least some customers who tested cheaper alternatives did not immediately move their workloads back.
OpenRouter is not representative of the entire enterprise market. Its users skew toward developers, startups and technically sophisticated customers. Nor does the available data prove that every former Fable user permanently migrated.
But it offers an important leading indicator: once customers are forced to test a cheaper substitute and discover that it is good enough, temporary disruption can become permanent migration.
Open-weight models offer another advantage: control. A model that can be downloaded, hosted and adapted independently cannot be withdrawn overnight by its developer. These models do not need to replace Anthropic completely. They only need to capture the high volume of routine work below the frontier.
Anthropic could remain the benchmark leader while losing token volume and pricing power.
The better layer of the AI economy
This reveals an important difference between SpaceX and Anthropic.
SpaceX increasingly sells the scarce infrastructure required to build and operate AI. It can generate revenue from the AI race regardless of which model wins. Anthropic must buy enormous amounts of that infrastructure, develop a leading model and sell access at a sufficient premium to recover its costs.
SpaceX is betting that demand for compute will continue expanding. Anthropic is betting that its intelligence will remain sufficiently differentiated to avoid becoming a commodity.
Anthropic may become one of the world’s most important companies, and its valuation does not look irrational relative to SpaceX. But $2 trillion assumes that Anthropic can maintain frontier leadership, improve margins and preserve premium pricing despite relentless competition and rapidly improving open-weight alternatives.
That is possible. It is not yet proven.
Anthropic may be the better AI company, but SpaceX may occupy the better layer of the AI economy.
Leave a Comment