Claude AI developer, Anthropic, has released its revenue targets for the next couple of years, and they are nothing if not ambitious. Following an estimation of just $4 billion in 2025 in July, it's now claiming it will hit $9 billion annualized revenue by the end of the year, and as much as $26 billion next year, according to Reuters. That's not just more than double its own projections for the best-case scenario this year, it's more than double the projected 2025 revenue of OpenAI.
To say some analysts have their doubts, or at least questions about where Anthropic's revenue is going to come from, would be an understatement. Annualized return revenue is based on a single month of revenue extrapolated out to 12 months, making it a dubious way of predicting yearly revenue, at best.
Making money from AI is harder than it looks
With all the hundreds of billions of dollars of circular investment the AI industry enjoys, it might seem hard to imagine that these companies don't actually generate a lot of money yet. They do, in sheer dollar terms, but not compared to their valuations, investments, and the industry hype.
OpenAI is loosely valued at around $500 billion, and yet may make just $13 billion in revenue in 2025 - and even that is based on somewhat optimistic projections. Anthropic is valued at $183 billion after a recent funding round, but is projecting less than $10 billion in revenue this year. They're also spending far more than they earn. OpenAI lost $5 billion last year and will continue its losses for many years to come.
Nobody is making money with AI - except for Nvidia, which is perhaps why it's so keen to keep pumping the industry - but they are at least earning some with it, and Anthropic believes it can earn a lot more, even if it's not profitable.
The only ones paying for AI are businesses
(Image credit: Anthropic)
No one has figured out a way to make AI profitable yet, aside from the companies selling the chips for training and inference. But various AI companies are trying different strategies.
OpenAI, the most well-known AI company with its ChatGPT chatbot, earns money from enterprise customers but has primarily pitched itself as an AI developer for the masses. Its 800 million active users suggest that it has captured an enormous segment of the potential market for AI tools, but only five per cent of them pay anything for it, according to The Register.
Anthropic takes a slightly different approach, earning as much as 80% of its revenue from its enterprise customers. That includes coding app company Cursor, as well as software-as-a-service company SAP, and AI legal assistance firm, Filevine.
Anthropic also markets itself differently, projecting an image of safety, security, and professionalism - something that OpenAI, with its Sora 2 deepfakes, doesn't really match. That may be why Anthropic can charge so much more for the API calls on its various models. Its Claude Opus 4 model costs $15 per million input tokens, and as much as $75 per million output tokens. In comparison, the latest GPT-5 model from OpenAI is priced at just $1.25 per million input tokens and $10 per million output tokens.
Even its recent "lite" model, Claude Haiku 4.5, is $1 per million input tokens and $1.25 million output tokens. GPT-5 mini costs just $0.25 per million input tokens, and Google's Gemini Flash 2.5 is $0.3 per million.
But businesses are willing to pay for AI at higher prices, because they're being promised that deploying AI throughout the workforce will increase efficiency and productivity. In reality, only a few seem to actually manage that.
That's the gravy train that Anthropic hopes keeps rolling, and in fact projects that it won't just continue, but will accelerate in 2026. Despite the growing fears of a bubble forming around the industry and the shaky investment ground it's built on, especially since enterprise adoption of AI is starting to slow.
Go where the people are
Another potential revenue stream for all these companies is to just go after more users. If you can't bring the companies to you, why not see if you can get more people to use the cheaper models?
OpenAI announced in August that it was launching a new ChatGPT plan, its cheapest one yet at just under $5 a month, in India, the world's largest market for just about anything, due to having the largest population. Anthropic's development of the cheaper Haiku models is likely a similar endeavour, despite its reputation being mostly pinned to its more capable models.
Instead of going after individual users like OpenAI is doing, though, Anthropic could lean into its enterprise focus and target smaller customers with less cash to burn on AI. Call centres, small businesses that want a support chatbot, these kinds of fast-response, low-information roles could be a good use for faster, leaner, and importantly, more affordable chatbots.
Anthropic may also just be able to spend its way to increase revenue streams. Its latest funding round brought in over $13 billion of Series F funding in September, re-valuing the company at over $180 billion. That's an enormous amount of capital that can be invested in all the usual sources for AI companies: Compute, model training, and circular investments into infrastructure.
OpenAI CEO Sam Altman also suggested that the company may diversify its models into generating explicit adult content. Anthropic isn't going down that route, though. It doesn't have the kind of user feeds where it could sell adverts. It will instead need to broaden its customer base dramatically.
Government contracts are another potential source of new customers, but that poses its own risks, especially when it comes to international partners and the volatile nature of the current U.S. administration.
Ultimately, though, Anthropic has the investment funds for expansion. Like everyone else in this industry, it just needs to hope that the bubble holds through 2026 for it to even attempt to meet those eye-watering revenue projections.
Jon Martindale is a contributing writer for Tom's Hardware. For the past 20 years, he's been writing about PC components, emerging technologies, and the latest software advances. His deep and broad journalistic experience gives him unique insights into the most exciting technology trends of today and tomorrow.
