Anthropic is now embarking on its Series H funding round to raise another US$50 billion at a whopping US$900 billion valuation. That’s a tad higher than OpenAI’s post-money valuation of US$852 billion following its US$122 billion funding round in March. Anthropic raised US$30 billion in its Series G round at a US$380 billion valuation in February. That’s not counting US$65 billion in funding commitments Claude’s developer has secured from its two top shareholders — search giant Google and e-commerce powerhouse Amazon.com recently. In late April, Google committed to investing up to US$40 billion in Anthropic — an initial US$10 billion in cash to immediately boost computing capacity and US$30 billion linked to future performance targets. Last week, Amazon committed an additional US$5 billion immediately, and up to US$25 billion in total in the future tied to commercial milestones. Amazon also uses Anthropic’s models to power AWS Bedrock, its generative AI cloud platform.
Google now owns 14% of Anthropic compared with Amazon’s 7.8% even though Amazon has injected much more cash. Because Google was one of Anthropic’s earliest backers, its cash earned it a far bigger stake. Another early investor was the now defunct crypto group FTX, which, under Sam Bankman-Fried, invested US$500 million for an 8% stake at a US$2.5 billion valuation, just before it collapsed with over US$11.2 billion in debt. FTX eventually recovered over US$16 billion and its creditors were paid in full. As part of its bankruptcy, FTX was forced to sell its Anthropic shares for US$1.4 billion. At the US$900 billion valuation that Anthropic is seeking in its current funding round, that 8% stake would be worth around US$72 billion.
The rise and fall of ChatGPT
OpenAI, xAI, Anthropic and Google create the four advanced, large-scale foundational AI models trained on massive datasets to excel in reasoning, coding and multimodal tasks — text, image, audio and video. Frontier AI models are evolving rapidly and pose both significant security risks and transformative opportunities. OpenAI was the first out of the gate with its ChatGPT chatbot in Nov 2022. ChatGPT became the fastest adopted consumer technology in history with 100 million monthly active users in just two months and driving one of the fastest revenue ramps in tech history. In comparison, Instagram took 2½ years to reach 100 million users, TikTok achieved the same in nine months while the internet took over a decade to do so.
The total addressable market for developing AI models is currently estimated to be around US$4.2 trillion. Citigroup expects large, fast growing firms like Anthropic, OpenAI and Gemini creator Google to dominate AI model development, though it notes that the near term will be defined by compute capacity constraints, capital requirements, pace of enterprise adoption and fierce competition among AI labs, hyperscalers and open-source providers.
See also: Trump says he discussed AI guardrails, Nvidia chips with Xi
Anthropic was founded in 2021 by seven OpenAI executives including Dario Amodei and his younger sister Daniela Amodei, who split from the Sam Altman-led firm mainly due to disagreements over its direction, specifically a shift towards prioritising commercialisation over AI safety. They felt that OpenAI was not sufficiently committed to safety and wanted to build a company where it would be the primary focus. Anthropic pioneered “constitutional AI”, a technique for training AI to follow principles and self-critique. The firm derives its name from the word “anthropic”, which relates to human existence — fitting for an entity whose entire mission revolves around making AI work safely alongside humanity.
At OpenAI, Amodei had relentlessly battled for a more cautious approach to developing and releasing AI models, fearing that the rapid, profit-driven business that his then boss Altman was building could lead to catastrophic risks. Amodei argued for holding back releases to ensure safety, which clashed with the faster, more aggressive commercial approach that Altman was pushing at the time.
OpenAI, which was initially a non-profit, was founded and funded by Musk. The CEO of the pioneering electric vehicle maker Tesla now has his own AI firm xAI, which purchased X, formerly Twitter, and earlier this year merged with his rocket firm SpaceX. Musk and Altman are now battling in court over allegations that OpenAI CEO “stole” the OpenAI charity.
Here’s how the big four frontier AI models different from each other: OpenAI’s GPT is best known for its general tasks, image generation (DALL-E), plug-ins and coding. Its strengths include its GPT store akin to smartphone app stores, and voice mode. Its weaknesses include enterprise privacy concerns and the fact that its responses can be very verbose. Gemini has deep integration with the Google ecosystem — Gmail, Docs, Google Drive, Android and search. It also has image/video understanding. Claude, meanwhile, has a strong focus on enterprise and is best known for its nuanced and careful tone, and strong reasoning. Claude prioritises application programming interface (API), access, long-context document processing, and strict safety/privacy controls that appeal to companies.
See also: Kioxia prepares US shares after riding AI boom to big profit
According to Similarweb data for April 2026, ChatGPT topped global daily average web visits last month at 183 million (up 7% y-o-y but down 1% m-o-m), Gemini was at 92 million (up 572% y-o-y and up 9% m-o-m), and Claude was at 27 million (up 745% y-o-y and up 3% m-o-m). Clearly, Claude has momentum. Its mobile app average daily active users surged 44% from 16 million in March to 23 million in April, while Gemini grew roughly 19% over the same period. According to Counterpoint Research’s 1Q2026 data, Anthropic captured 31.4% of the global large language model revenue share, overtaking OpenAI’s 29%. While ChatGPT continues to dominate raw web traffic by a wide margin, the gap is narrowing fast — particularly for Claude and Gemini on a y-o-y basis.
Yet the real difference is in revenues. Claude, due to its focus on businesses and enterprises, had an annual revenue run rate of US$30 billion in April compared to OpenAI’s annual run rate of US$24 billion. ChatGPT did not reach 1 billion weekly active users by the end of 2025 as forecast, and currently has around 900 million. Open AI, which is forecast to lose over US$14 billion this year, expects to spend over US$125 billion a year on compute by 2030. Although it has begun to take its enterprise business more seriously, revenue mix still leans fairly heavily towards consumer subscribers. Anthropic’s share of US enterprise AI spending climbed to 40% in the January–March quarter, while OpenAI’s fell from 50% to 27% over the same period.
Here’s how Anthropic’s US$30 billion annual run rate compares with software-as-a-service (SaaS) players: Microsoft took 20 years to reach the US$30 billion level of revenue while Salesforce took 15 years. For Anthropic, 80% of revenue comes from enterprises. That includes over 1,000 business customers spending at least US$1 million a year or more on Claude. Anthropic projects positive free cash flow by 2027. OpenAI is not expecting breakeven until after 2030. Alphabet, which owns Google, has not yet disclosed revenue numbers or profitability for Gemini.
Anthropic the disrupter
In late January, Anthropic released 11 plug-ins for the Claude Cowork platform, triggering a massive sell-off in software stocks dubbed the “SaaSpocalypse”. The release of the tools, designed to automate complex, multi-step tasks in the areas of legal, sales and marketing, caused up to US$830 billion in market value to evaporate from software firms over a week. In February, Anthropic released Claude Opus 4.6 and Claude Code, accelerating the shift towards “vibe coding” and agentic workflows that threatened traditional per-seat SaaS models.
Last month, Claude Mythos, an advanced, unreleased “frontier” AI model developed by Anthropic, disrupted the cybersecurity landscape by demonstrating the ability to autonomously identify, analyse and exploit software vulnerabilities at a speed and depth far exceeding human capabilities. Unlike other models that assist with coding tasks, Mythos can autonomously scan vast, complex codebases and create full working exploits for bugs without human guidance. Due to its huge offensive potential, Anthropic restricted access to the model, choosing to release it only through “Project Glasswing” — a controlled initiative with major tech firms designed to harden systems rather than expose them to attack.
In early May, at an event in New York attended among others by JPMorgan CEO Jamie Dimon, Anthropic launched 10 AI agents targeting Wall Street banks and insurers. The tools can build investment bankers’ pitchbooks, draft commercial banks’ credit memos, screen private bankers’ Know Your Client files and have a variety of auditing functions. “Anthropic stopped selling AI to Wall Street and started becoming Wall Street OS,” says tech blogger Linas Beliunas. AI agents assist human analysts in real time, or through Claude Managed Agents, a hosted model in which Anthropic provides the underlying production infrastructure for more autonomous operation. Each template is designed to be customised around a bank or insurance firm’s internal standards, including how it structures models, manages risk and routes decisions for approval. Claude agents offer research and client coverage — including a meeting preparer, earnings reviewer and financial model builder; credit, risk and compliance work through a market researcher; and finance and operations tasks through a valuation reviewer, general ledger reconciler and statement auditor.
On May 4, Anthropic signed a US$1.5 billion joint venture with Goldman Sachs and private equity firms Blackstone and Hellman & Friedman to help bring Claude into the day-to-day operations of mid-market companies, in direct competition with the global consulting firms. Last week, Anthropic reached a deal to tap the computing resources of SpaceX. It will use the full computing power of SpaceX’s Colossus 1 super computer in Memphis, Tennessee, which houses more than 220,000 Nvidia processors. The deal, which will give the Claude chatbot maker 300mw of new capacity, marks a détente with Musk who once called Anthropic “misanthropic” and “evil” and claimed the firm “hates western civilisation”. Anthropic’s deal will generate up to US$5 billion in annual revenue and over US$3 billion in cash profit for SpaceX, which is now a hyperscaler and a very profitable one. A growing close partnership between two trillion-dollar firms seeking imminent listing is music to the ears of investors.
Assif Shameen is a technology and business writer based in North America
