On October 14, 2026, at an investor briefing ahead of its initial public offering (IPO), Anthropic said its July–September revenue reached $16.8 billion, up 45% from the previous quarter. According to Reuters, which based its report on two people who attended the briefing, the company also turned an adjusted profit for the quarter. But on October 15, Reuters reported that some investors were valuing the company at a lower level for the IPO than they had previously expected. Despite the surge in revenue, caution is spreading among investors, driven by slowing growth and the question of how much profit Anthropic can secure from ever-increasing sales.
$16.8 billion in quarterly revenue, but the dollar increase is shrinking
According to information Reuters obtained from briefing attendees, Anthropic's quarterly revenue in 2026 grew from $4.7 billion in January–March to $11.6 billion in April–June, and then to $16.8 billion in July–September. However, comparing the dollar increase and growth rate from the previous quarter shows a change in the pace of expansion.
| 2026 quarter | Revenue | Increase from prior quarter | Quarter-on-quarter growth |
|---|---|---|---|
| January–March | $4.7 billion | — | — |
| April–June | $11.6 billion | $6.9 billion | About 147% |
| July–September | $16.8 billion | $5.2 billion | About 45% |
The revenue figures are those Reuters reported on October 14; the dollar increases and growth rates were calculated from them. The increase is the difference between the current and prior quarter's revenue, and the growth rate is calculated as current-quarter revenue ÷ prior-quarter revenue − 1. These figures are based on numbers conveyed by attendees of the investor briefing and have not been verified against audited financial statements.
The quarterly dollar increase shrank from $6.9 billion in April–June to $5.2 billion in July–September. As revenue grows larger, the same dollar increase produces a lower growth rate, but in this case it is not only the growth rate that has fallen; the dollar amount added is also smaller. Still, revenue itself continues to rise, so the smaller increase does not immediately signal a decline in demand. Whether this trend is temporary or reflects a change in the pace of growth will need to be judged from upcoming quarterly results.
Note also that $16.8 billion is revenue for a three-month period. The "annualized revenue of more than $47 billion" that Anthropic disclosed when it raised funds in May converts the revenue pace at that time into a one-year figure, which is a different metric from quarterly revenue. Multiplying July–September revenue of $16.8 billion by four gives $67.2 billion, but that is merely a calculation assuming the quarter's average revenue continues for a full year. It is not a figure indicating the revenue pace at year-end or a forecast for full-year 2026 revenue.
Enterprise demand spreads into everyday banking work
On April 6, Anthropic announced that the number of business customers spending more than $1 million on an annualized basis had surpassed 1,000. Because the figure was more than 500 in February, the number of large customers rose sharply in a short period. This is not a simple count of registered companies but reflects a growing number of customers whose usage exceeds a certain amount. However, it is based on annualized spending and does not represent the cumulative amount each company actually paid over a year.
A concrete picture of how enterprise AI is being used also emerges from the expanded partnership with UK financial giant Barclays that Anthropic announced on October 1. According to Anthropic, Barclays has operated an internal knowledge-search assistance system since 2025, with more than 16,000 users and more than 1 million cumulative searches. In markets-related operations, it uses Claude to classify incoming customer emails, fill in missing information and route them to the appropriate person, handling about 120,000 items a day.
Searching internal documents and routing inquiries are tasks that recur in daily work. Usage in such areas is a concrete example showing that corporate AI adoption is moving beyond trials and into the stage of being built into existing business processes. Looking at Anthropic's enterprise business only through the software development tool Claude Code would mean overlooking this spread into other operational areas.
Barclays also plans to expand its adoption of Claude Code. It has set a goal of rolling it out to 50% of its developers by the end of 2026, and expects to extend it to a majority of its software engineers in 2027. However, these are future plans and should be distinguished from confirmed usage. In addition, the expanded partnership was announced after the July–September quarter ended, so it is not clear how much such initiatives contributed to revenue growth in that quarter.
The Barclays case confirms that continued use of Claude is advancing in specific operations. But the reported figures do not show how much of Anthropic's total revenue comes from Claude Code, general API use and enterprise contracts, respectively. To judge whether growth is sustainable, the revenue mix of each business and whether usage continues will matter, in addition to the increase in large customers.
How will model price cuts affect revenue and profit?
Anthropic said that Claude Haiku 5.5, announced on October 7, can cut average running costs by about 75% compared with the earlier Haiku 4.5. Haiku 5.5 is a small model intended for uses that repeatedly process large volumes of requests, such as summarization and classification. For Claude Sonnet 5.5, the company also halved the fee for cache reads, which reuse stored input, from $0.20 to $0.10 per million tokens.
For companies using the models, this could allow them to hand more work to AI on the same budget. In tasks that repeat the same kind of processing at scale, such as classifying inquiries or searching internal documents, lower unit prices help widen the range where AI can be applied. For Anthropic, however, a rise in usage does not necessarily translate into revenue growth at the same rate. If the amount billed for each task falls, the company must offset that decline through higher usage volume or new use cases.
However, these price changes took effect after the July–September quarter ended, so they are not the cause of the growth slowdown seen in that quarter. Rather, they are a factor to consider when assessing revenue growth from October–December onward. Haiku 5.5's pricing structure also varies with input length, so actual payments will not fall uniformly by 75% for every task. How much usage rises as a result of the price cuts will shape future revenue.
At the same time, Anthropic is also securing and improving the efficiency of the computing resources needed for inference. The company has said it will use AWS's Trainium, Google's TPUs and NVIDIA GPUs together, assigning work to the hardware best suited to each task. Under the agreement with Google and Broadcom announced in April, multi-gigawatt computing infrastructure using next-generation TPUs is scheduled to begin operating from 2027. Contracts to secure future computing resources should be distinguished from facilities that are already running.
If Anthropic can lower costs through hardware allocation and more efficient inference, it becomes easier to reconcile price cuts for customers with securing profit. But published API prices are what customers pay, not the processing costs Anthropic itself bears. The size of the price cuts alone therefore does not show whether profit margins will improve or worsen. Going forward, what matters is how far demand expands as prices fall and how far computing efficiency reduces costs.
The IPO will test the sustainability of revenue growth and profitability
According to Reuters on October 15, some existing investors and investors considering participating in the IPO are discussing a valuation of around $1.5 trillion for Anthropic. Three people familiar with the matter said that a valuation of around $2 trillion had been expected before that. Even if revenue is rising, a valuation premised on the same pace of growth as before could be revisited as the growth rate falls. Indeed, Reuters reported that some investors were disappointed by the slowdown in quarter-on-quarter growth.
However, about $1.5 trillion is not a finalized IPO valuation but merely a level being discussed among investors. Nor can it be simply compared with the $965 billion post-money valuation set in the May fundraising, because the timing and nature of the transactions differ. Anthropic announced that it confidentially submitted a draft registration statement (Form S-1) to the US Securities and Exchange Commission (SEC) on June 1, but at that point the number of shares and the price had not been determined. It also said the IPO itself depends on conditions such as market environment.
The reported adjusted profit is likewise only one piece of information for judging corporate value. From the figures said to have been presented at the briefing, it is unclear which costs or items were adjusted or how large the profit was. It therefore cannot be treated as equivalent to a net profit under generally accepted accounting standards or to an improvement in cash flow. Revenue and adjusted profit alone also do not show how much of the future spending on computing resources and facilities the existing business can fund with the cash it generates.
What will matter for Anthropic's IPO is not only revenue growth but whether enterprise customers keep increasing usage even after price cuts, and whether the company can improve profit after processing costs and cash flow. If the computing expansion planned from 2027 proceeds as scheduled, it will create room to broaden AI use in everyday work such as banks' internal search and inquiry handling. At the same time, the accompanying spending on computing resources must be taken into account. Whether Anthropic can turn rapid growth, with quarterly revenue of $16.8 billion, into sustained profit and cash flow will greatly influence its valuation going forward.