Anthropic’s fundraising deck set out an ambitious plan: raise as much as $5 billion over the next two years, enter more than a dozen major industries and build a far more capable AI model. The company said that developing the model would require a billion dollars over 18 months.
A new model with a large computing bill
The proposed system was tentatively named “Claude-Next.” Anthropic described it as a “frontier model” that could be 10 times more capable than today’s most powerful AI. The company framed the project as a successor to Claude, its chatbot for tasks such as writing, coding, summarizing documents and answering questions.
Anthropic estimated the model would need on the order of 10^25 FLOPs, or floating point operations. The deck also referred to computing clusters with “tens of thousands of GPUs.” Those figures point to the scale of the resources the company believed the work would demand; the time required would depend on the speed and scale of the computing system.
The company connected its plans to a training approach it calls “constitutional AI,” which it described as a “next-gen algorithm for AI self-teaching.” At a high level, the approach aims to align AI systems with human intentions by giving them guiding principles for answering questions and carrying out tasks.
From chatbot to broader applications
Anthropic presented the future model as a potential basis for virtual assistants that could answer emails, conduct research and generate art, books and other material. The deck suggested that such systems could automate large portions of the economy, while arguing that companies leading in 2025/26 models could establish a lasting advantage.
Claude was already positioned as a general-purpose chatbot, with features for searching across documents, summarizing, writing, coding and answering questions about specific topics. Anthropic compared it with ChatGPT and argued that constitutional AI made Claude “much less likely to produce harmful outputs,” “easier to converse with” and “more steerable.” These are claims made by the company in the deck.
Anthropic released Claude commercially in March, following a closed beta late last year. Around 15 partners had initial access. The deck’s wider expansion ambition was to move from early access toward adoption across more industries, building products around areas where Anthropic saw demand and where it believed its expertise and brand fit.
Fundraising and a shift toward commercialization
The deck was created for Anthropic’s Series C fundraising. It confirmed a target of $300 million at a $4.1 billion valuation, though at the time the document was created, only half of that amount had been raised from a “confidential investor.” The Information had reported the fundraising target in early March, along with a total raised of $1.3 billion.
The company’s plans came as it sharpened its commercial focus. The deck said Anthropic had concentrated on research during its first year and a half, then committed to commercialization in September [2022]. It described a go-to-market strategy and initial product specializations intended to match the company’s expertise and expectations for adoption over the next 12 months.
Anthropic was founded in 2020 by former OpenAI researchers. Dario Amodei, formerly OpenAI’s VP of research, launched the company in 2021 as a public benefit corporation, joined by several former OpenAI employees, including former policy lead Jack Clark. The source says Amodei left after a disagreement about OpenAI’s direction and its growing commercial focus.
Competition and investors
Anthropic’s field included OpenAI and AI startups Cohere and AI21 Labs, which were developing and commercializing text-generating systems, and in some cases image-generating ones. The article said OpenAI had raised substantially more capital, including a reported $10 billion from Microsoft at a $29 billion valuation.
The deck also described Alameda Research Ventures as a “silent investor” with “non-voting” shares, and said it spearheaded Anthropic’s $580 million Series B round. Anthropic expected Alameda’s shares to be disposed of in bankruptcy proceedings within the next few years.
Google was another investor. It had pledged $300 million for a 10% stake. Under the deal, Anthropic agreed to make Google Cloud its “preferred cloud provider,” and the companies would “co-develop AI computing systems.” Other backers named in the article included James McClave, Facebook and Asana co-founder Dustin Moskovitz, former Google CEO Eric Schmidt and founding Skype engineer Jaan Tallinn.
Together, the deck’s fundraising target, model ambitions and commercial plans describe a company trying to turn AI research into a competitive business. Its proposed model depended on significant funding and computing capacity, while the company sought customers and partnerships that could support wider deployment.