The Artificial Investor #58: Is Anthropic worth 180 billion dollars?
My name is Aris Xenofontos and I am an investor at Seaya Ventures. This is the weekly version of the Artificial Investor that covers the top AI developments of the last days.
The Most Interesting Story of Last Week: Anthropic gets valued more than 180 billion dollars
The most interesting story of the last few days was Anthropic’s 13-billion dollar funding round that brought the Enterprise AI market leader’s valuation to more than 180 billion dollars.
Is Anthropic overvalued? Are we in an AI bubble? Will investors get their desired returns? Let’s build it up by looking at the bull and the bear cases separately.
🐂 The bull case: Anthropic is undervalued
📈 What do valuation maths tell us?
Anthropic’s valuation looks aggressive at first glance: 183 billion dollars or 36 times its annualised recurring revenues (36x ARR). Nevertheless, the company’s growth trajectory is unprecedented, particularly for B2B businesses. Annualised recurring revenue (i.e. the last month’s recurring revenue multiplied by 12) has accelerated from 1 billion to 5 billion dollars in nine months, which is nearly 800% growth on an annual basis. The pace of scaling is only getting faster: it took about 21 months to add its first billion in ARR, then just three months for the next billion and only two months to move from 2 to 3 billion dollars.
By comparison, OpenAI took far longer to hit similar levels from founding and most enterprise software businesses never achieve this velocity at all. Against the backdrop of median software multiples at 5x forward revenues, Anthropic commands a premium.
Growth is very scarce in Tech public markets nowadays due to 1) high-growth companies staying private for a long time, and 2) the focus on profitability across the Tech sector post the 2022-23 market correction. As a result, we are back to a public market that assigns premium to growth. The few Tech companies that have more than 25% annual revenue growth have a median forward revenue multiple of 13x - more than double the revenue multiple of companies growing less than 25% annually.
Looking across the industry, the picture sharpens further. On a scatter of growth vs. valuation multiples, Anthropic sits alongside OpenAI in the “outlier” corner: extreme growth with extreme multiple. BigTech incumbents like Microsoft, Meta and Alphabet grow revenues at 10%–20% annually, with revenue multiples in the 7x–13x range. Nvidia, the AI hardware bellwether, trades at 24x revenues on 56% growth. Anthropic’s 300% growth rate and 36x multiple, in that context, doesn’t look stretched but rather aligned with its category-defining momentum. Snowflake and CoreWeave (both high-growth AI infrastructure peers) cluster in the high-teens multiples at 20%–200% growth, underscoring that public markets are already willing to pay richly for growth.
As a result, one could say that Anthropic’s valuation premium reflects not hype but a rational bet that it can consolidate leadership in the Enterprise AI market, where switching costs and safety-first branding make revenues both faster-growing and stickier than consumer AI models.
🦣 A mammoth market
The enterprise AI market is already valued at tens of billions of dollars and expected to grow into hundreds of billions of dollars within this decade. We found a number of different estimates from various consulting groups. Considering three reputable firms, Grand View, Mordor Intelligence and Coherent, we get a range of 25 to 45 billion dollars for the starting point in 2025 and 115 to 150 dollars for 2030.
It’s important to note that this compares to a 300-400 billion-dollar Consumer AI market, which means that Anthropic is addressing a smaller market than OpenAI. On the other hand, Anthropic’s narrower focus may come with strategic advantages: deeper integration, higher willingness to pay and recurring revenue from enterprise-grade AI deployments.
The fact that the market is large and addressable, doesn't mean that it will be captured by Anthropic. How well is the San Francisco rockstar positioned to take advantage of this market opportunity?
🥇 Riding a first-mover advantage
Anthropic has deliberately positioned itself as the Enterprise-first alternative in the GenAI AI race, differentiating from OpenAI’s consumer-heavy footprint, and it’s the market leader. Framing the company’s current traction against the market size estimates, a 5-billion-dollar revenue run rate today implies a market share of 10% to 20% (depending on which estimate you use) Even under the most conservative projections, this is an extraordinary footprint for a company less than a decade old.
Anthropic counts over 300,000 business customers with its flagship product, Claude Enterprise. Having become the backbone of many AI coding applications reaching 42% market share, it launched its own coding application, Claude Code, which has alone exceeded 500 million dollars of annualised revenue.
Anthropic clearly has a first-mover advantage in its market, which can prove decisive in a world where technological competitive advantages are slim.
🛡️ Holding the fort
Anthropic’s defensibility rests on a combination of capital, partnerships, product stickiness and regulatory positioning. With deep-pocketed backers like Amazon, Google and sovereign wealth funds, Anthropic has theoretically unlimited access to compute and funding, which is critical in a sector where model training costs keep increasing.
Unlike companies like ChatGPT that lead with consumer products, Anthropic is focusing on corporate clients where only 11% switched to a different provider in the last 12 months. Its integration into Enterprise workflows through Claude Enterprise and Claude Code creates real switching costs, especially as businesses build compliance-sensitive processes around its tools. Finally, the company has carved out a reputation as the “responsible AI” provider, a branding edge that could translate into durable differentiation as regulation tightens.
🧑🔬 It’s all about the people
At the heart of Anthropic’s rapid rise is its world-class team, led by co-founders Dario and Daniela Amodei, who left OpenAI to build a company explicitly focused on safe, reliable frontier models. Dario Amodei in particular brings a formidable track record: he led the development of GPT-2 and GPT-3 at OpenAI, and previously worked in deep learning at Google Brain. Daniela Amodei was formerly OpenAI’s VP of Safety and Policy and has been instrumental in building Anthropic’s governance and alignment-first culture.
Beyond the founders, Anthropic has attracted top researchers and engineers from OpenAI, DeepMind and Google Brain, such as Tom Brown (lead author of the GPT-3 paper), Sam McCandlish (pioneer of scaling law research) and Jared Kaplan (lead researcher in neural scaling laws).
🐻The bear case: Antropic is overvalued
🏰 Is there a real moat?
A critical structural disadvantage for Anthropic is that it does not own the underlying infrastructure. Microsoft owns a Cloud platform (Azure) and datacentres, and is well tied up to OpenAI giving it access to AI models. Google owns the entire stack: a Cloud platform (GCP) and datacentres, the TPU chips and the Gemini models. Meta designs its own AI accelerators (Artemis) while distributing LLaMA models broadly as open source. Amazon combines its custom chips (Trainium/Inferentia) with AWS dominance in the Cloud. Nvidia, the industry’s GPU king, sits even deeper in the stack by controlling the very hardware that all others depend on, and has recently launched its own AI models, Nemotron.
Anthropic, by contrast, must rent compute from partners (primarily Amazon and Google), both of whom are also direct competitors at the model layer. This creates a dependency that puts pressure on margins and limits strategic freedom. Without its own cloud or chips, Anthropic lacks the vertical integration that allows rivals to cut costs, experiment rapidly and tie model adoption into broader ecosystems (Microsoft Office, Google Search and Workspace, Whatsapp and Facebook/Instagram for Meta, Amazon Marketplace and Alexa). Competitors can subsidise model usage through their Cloud businesses or hardware sales, while Anthropic must charge enterprises directly, reinforcing the pricing pressure it already faces.
⚔️ Competition is fierce
Two structural forces are eroding Anthropic’s defensibility: performance convergence and price collapse. On performance, the Artificial Analysis Intelligence Index shows Anthropic’s top-tier Claude 4.1 Opus scoring 59, barely ahead of open-source contenders like GPT-OSS-120B (58) and Qwen3 235B (57). This convergence makes it hard to argue that Claude enjoys a durable technological edge - enterprises evaluating AI technologies increasingly see open-source delivering 95-98% of the performance.
On price, the moat looks equally fragile. A report from Epoch.ai and a16z shows that the cost per million tokens has fallen 10x each year, amounting to a 280x drop in less than two years for GPT-3.5-level performance. GPT-4 itself saw prices fall nearly 90% since launch, from 36 to 4 dollars per million tokens. Anthropic’s own pricing has trended downward across model families — but not as aggressively, leaving it exposed to being undercut. As open-source rivals improve and token prices race to the bottom, Anthropic faces the classic challenge: how do you defend premium positioning in a market that is rapidly commoditising?
Furthermore, as mentioned earlier, Anthropic has stood out with its “responsible AI” positioning, leveraging the background of one of its co-founders and targeting one of corporates’ most sensitive points and important purchase criterion. However, recently Anthropic departed from this position as it made a privacy policy u-turn: client data will now be used for training unless they opt out. This is, to say the least, confusing.
💲 Show me the money
Can the business ever make money?
Putting the moat and competitive challenges together, margins are thin because it must rent compute from AWS/GCP while the industry is seeing 90%+ price compression.
Another challenge of the business model is its capital intensiveness on the infrastructure side. The hardware and energy costs associated with training AI models are skyrocketing. xAI built Colossus, an AI super cluster of 200,000 Nvidia chips to train Grok 4. Dario Amodei, CEO of Anthropic, said last year that he expects training runs for AI models to reach 10 billion dollars by 2026.
In addition, training AI models has had some hidden costs that are gradually coming to the surface and will likely increase future training budgets. First, data used in training models has come “for free” so far. However, a series of lawsuits followed by partnerships with data publishers indicate that the free lunch is over.
Anthropic agreed to pay 1.5 billion dollars to settle a sweeping class-action lawsuit filed by authors whose works were allegedly used without permission to train Claude. A federal judge ruled that while AI training on copyrighted text could be protected under “fair use,” Anthropic violated the law by storing pirated works in a central repository. This settlement marks the largest known copyright recovery in U.S. history and the first of its kind in the AI era.
Overall, it is not surprising that Anthropic lost five billion dollars in 2024. Can it ever turn profitable with product prices declining and product development costs increasing?
🏦 Eventually investors need an exit
The best way for investors to materialise the value of their investments, particularly when their ownership stake in a company is worth billions of dollars, is a full exit of the business. However, Anthropic’s exit options are limited. One exit route would be an IPO, however, the IPO market remains volatile, having seen some green shots at the end of Q2’25 which are likely to continue with the traditional post-Labour Day window. Another route would be M&A. However, given the company’s size and the potential antitrust issues in a combination with any of the Magnificent 7, this may not be such a straightforward option.
⚖️ Risk/reward is everything
Going back to the investor maths, an investor in Anthropic’s current round would likely demand a 4x return of their money within five years, given that, despite its scale, the business is still loss-making, which bears risks. Taking into account dilution from future fundraising and option pools for the expensive talent, the target return would be at least 5x, i.e. about 920 billion dollars of valuation at exit.
Assuming that, at the point of exit the company would have a normalised growth rate and, thus, its valuation multiple would converge to the mean, say 10x ARR, then investors are betting that Anthropic would reach 90 billion dollars of ARR within five years. Even when taking the most ambitious market forecast of 150 billion dollars, then this would imply a 60% market share. Punchy.
🔮 Conclusions
In conclusion, whether Anthropic is overvalued or undervalued depends on your perspective. The bull case sees it as a market leader in a vast, untapped industry with defensible competitive advantages (albeit non technological), while the bear case highlights the competitive pressures and commoditisation of AI models, business model risks and risk/reward challenges for investors entering now at this valuation.
We recognise the potential large size of the Enterprise AI market and Anthropic’s market and innovation leadership. Nevertheless, Anthropic has a set of competitors that are vertically-integrated and have access to more cash, more data and equally-strong technical talent. Given the company’s size, we struggle to believe that it will continue to grow at an annualised rate of 900% or it will reach 60% market share in the Enterprise AI market. Albeit a great business, returns for new investors may not overweigh the risk taken.
See you next week for more AI insights.