Anthropic’s proposed IPO puts a striking question about AI’s economics into view: can extraordinary growth produce profits when the infrastructure bill is extraordinary too? An Economic Times opinion column says the company reported revenue of $33.2bn for the first nine months of 2026, up from $2.5bn a year earlier, alongside about $900bn in commitments, largely for computing capacity.
Anthropic Watch analysis
What happened
The figures come from Anthropic’s IPO filing as described by Economic Times columnist Diva Jain. Her 17 November column says computing and infrastructure expenses amounted to 57% of revenue in the first nine months of 2026. It also notes that some commitments extend as far as 2045, with some conditional. These are figures reported in the column, not a verdict on what Anthropic will ultimately spend or earn.
Jain’s argument is that revenue growth alone cannot settle the question of future margins. AI services need computing power with each additional query, while the costs and benefits of building infrastructure are spread across model developers, cloud providers and hardware owners. The Economic Times column also raises the prospect that cheaper, more capable models could make providers interchangeable and weaken their pricing power.
Why it matters
The IPO is a test not only of investor appetite for Anthropic, but of how much value AI companies can keep when the technology depends on costly computing and infrastructure. The commitments also connect the fortunes of model makers to suppliers, data-centre owners and lenders. If demand disappoints, the exposure may not stay neatly inside one company’s accounts.
Jain’s column is analysis, not a forecast proved by the figures. But it points readers towards a more useful question than whether AI revenue is rising: whether companies can turn that revenue into durable returns after paying for the capacity required to serve it.
Our read
The growth figures are eye-catching. The commitments make them harder to read as a simple victory lap. A company can sell a lot of AI and still face a serious question about the cost of producing it. The IPO offers a timely chance to debate that tension, though one company’s filing cannot settle the economics of an entire industry.
What to watch
- How Anthropic’s final IPO terms compare with the proposed valuation.
- Whether its reported infrastructure costs and commitments translate into sustainable margins.
- Whether model providers retain pricing power as competing systems improve.
Discussion spark: If AI companies need enormous, long-term computing commitments to grow, should investors judge them mainly by revenue growth or by evidence those investments can earn durable returns?
Sources and evidence
- Source update (17 November 2026, 17:57 UTC)
Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.