Anthropic’s operations have improved, but the timing mismatch between usage revenue and multiyear compute commitments remains

There is evidence of operating improvement; sustained profitability remains unproven

Anthropic has filed a preliminary prospectus to apply for a Nasdaq listing under the proposed ticker ANT. The offering price and share count remain blank. The operating changes shown in the filing deserve attention: unaudited revenue for the first half of 2026 was $16.286 billion, up 1,300% year over year; second-quarter revenue was $11.555 billion, with operating profit of $974 million. My assessment is that the company has provided evidence of revenue growth accompanied by operating improvement, but one quarter’s performance is not enough to establish sustained profitability. The first half as a whole still recorded an operating loss, and the listing application has yet to result in a completed offering.

Evidence: [1]

The enormous net loss needs to be broken down

The first-half net loss was $80.748 billion, including $72.388 billion in losses from changes in the fair value of convertible notes, derivative liabilities, and SAFEs. The company attributes these mainly to an increase in its valuation. These losses cannot be treated directly as cash burn. However, subtracting this item from the net loss does not yield the operating loss of $693 million either: the statements also include other expenses, net, interest income, income taxes, and other items. Operating profit measures income and expenses at the operating level, while net profit also includes the effects of these items. Positive second-quarter operating profit coexisted with a net loss of $66.309 billion. The two are compatible. Using the enormous net loss to summarize operating performance would therefore distort the picture, while declaring overall profitability on the strength of operating improvement would also be premature.

Evidence: [1]

The scale of paid demand is visible; future usage remains variable

About 80% of revenue in the first half of 2026 came from usage-based billing, with charges determined by token usage and the model selected. As of 2026-09-30, the company’s internal statistics showed more than 11,900 customers that had each contributed at least $100,000 in revenue over the preceding twelve months, and about 250 that had each contributed at least $10 million. These disclosures support the judgment that enterprise paid demand has reached considerable scale. But the customer measurement window differs from the first-half financial reporting period, and these figures cannot be treated as contracted backlog or assured renewals. The revenue mechanism supports the inference that continued customer usage helps sustain revenue, while reduced usage may directly affect it. Current disclosures are insufficient to determine how much these customers will use in the future.

Evidence: [1]

A lower expense ratio has not removed the pressure from commitments

Compute and infrastructure expenses fell from 160% of revenue for the full year of 2025 to 55% in the first half of 2026. This shows that revenue coverage of these expenses has improved. But the periods differ in length, so absolute expenses cannot be compared directly to claim that spending has declined, nor can long-term economies of scale be established from the ratio change alone. As of 2026-06-30, the company had approximately $265.6 billion in future fixed or minimum guaranteed purchase commitments over terms of 1 to 20 years. It also had approximately $321.4 billion in future undiscounted obligations for leases that had not yet commenced. The two categories of obligations use different measurement bases and should be considered separately. They cannot be added together and treated as cash spending or liabilities for the current period. The common issue they reveal is that some resource spending has been committed to in advance, while revenue still requires customers’ actual future usage to materialize.

Evidence: [1]

Cash flow still needs to withstand changes in payment timing

Net operating cash inflow in the first half was $3.929 billion. Management’s explanation included both improved performance after excluding noncash items and working-capital changes involving the timing of supplier payments and customer collections. This inflow therefore cannot be equated with free cash flow or sustainable cash earnings. Taken together, my greater concern is the timing mismatch between usage revenue and multiyear obligations: whether operating improvement can continue depends on customer demand, cash collections, and financing supporting future payments. The prospectus also identifies insufficient revenue growth or financing as a risk to meeting obligations. If subsequent disclosures show sustained customer usage and continued support for operating cash flow after changes in payment timing, a more positive assessment would be warranted. If revenue slows while committed payments continue to come due, the company’s capacity to sustain its expansion will need to be reassessed.

Evidence: [1]

What to watch next

• Whether operating profit can be sustained in subsequent quarters, and how compute and infrastructure expenses change as a share of revenue.
• Whether the company discloses customer usage, retention, or renewal information that helps assess whether the current scale of paid demand can be sustained.
• Whether operating cash flow remains positive after changes in the timing of supplier payments.
• Whether payment schedules for purchase commitments, conditions for lease commencement, and financing progress change the pressure on the company to meet its obligations.