AI Startup Anthropic Pursues $5B Investment for OpenAI Showdown — Report
OpenAI may have serious competition soon, and not just from Google. Documents obtained by TechCrunch revealed that AI startup Anthropic is looking to raise $5 billion over the next two years to take on the ChatGPT creator and gain entry to a dozen major industries. These documents, in the form of a pitch deck for a Series C funding round, show Anthropic is planning to build a “frontier model” over the next 18 months called “Claude-next” that would be 10 times more powerful than current AI models and would cost a billion dollars to develop. The frontier model is outlined in the documents as a “next-gen algorithm for AI self-teaching,” with references to the company’s proprietary AI training technique called constitutional AI which involves both a supervised learning and a reinforcement learning phase. The Claude-next model will be a successor to Anthropic’s Claude chatbot which was released in March. Claude is currently used for tasks such as legal document summary and analysis, customer service emails and chat, coding, and natural language search. Anthropic asserts its constitutional AI training approach makes it less harmful and more steerable than OpenAI’s ChatGPT. The proposed model looks to be larger than any current model. The company estimates developing the Claude-next model will require 10^25 FLOPs, or floating point operations, requiring serious computing power. To put it in perspective, the world’s fastest supercomputer, Frontier, has a peak performance of 1.1 exaflops (an exaflop is 10^18). Frontier is housed at the Oak Ridge Leadership Computing Facility and draws nearly 21 megawatts of power during a run. Training time for Anthropic’s planned model will depend on the scale and power of the system used, and the documents mention “tens of thousands of GPUs.” This ambitious model could explain the $5 billion fundraising goal. Anthropic has raised $704 million in Series A and Series B funding rounds from 2022 and was reportedly looking to raise another $300 million last month. The Financial Times reported that Google is a 10% stakeholder in the company with an investment between $300 and $400 million. Another Anthropic investor is Salesforce Ventures, the investment arm of Salesforce. The company recently launched a $250 million Generative AI Fund and listed Anthropic as one of four companies the fund will initially invest in. TechCrunch reports that the documents it obtained confirm the company is seeking to raise $300 million at a $4.1 billion valuation, and that at the time of the document’s creation, only half was raised from a “confidential investor.” According to the article, Anthropic is looking ahead with its fundraising and development goals: “These models could begin to automate large portions of the economy,” the pitch deck reads. “We believe that companies that train the best 2025/26 models will be too far ahead for anyone to catch up in subsequent cycles.” Anthropic was founded in 2021 by former senior staff from OpenAI, including Anthropic CEO Dario Amodei, OpenAI’s former VP of research. Amodei and his sibling, Anthropic President Daniela Amodei, were among a group who left OpenAI due to “directional differences.” We've reached out to Anthropic for comment will update this article as we learn more.
OpenAI may have serious competition soon, and not just from Google. Documents obtained by TechCrunch revealed that AI startup Anthropic is looking to raise $5 billion over the next two years to take on the ChatGPT creator and gain entry to a dozen major industries.
These documents, in the form of a pitch deck for a Series C funding round, show Anthropic is planning to build a “frontier model” over the next 18 months called “Claude-next” that would be 10 times more powerful than current AI models and would cost a billion dollars to develop. The frontier model is outlined in the documents as a “next-gen algorithm for AI self-teaching,” with references to the company’s proprietary AI training technique called constitutional AI which involves both a supervised learning and a reinforcement learning phase.
The Claude-next model will be a successor to Anthropic’s Claude chatbot which was released in March. Claude is currently used for tasks such as legal document summary and analysis, customer service emails and chat, coding, and natural language search. Anthropic asserts its constitutional AI training approach makes it less harmful and more steerable than OpenAI’s ChatGPT.
The proposed model looks to be larger than any current model. The company estimates developing the Claude-next model will require 10^25 FLOPs, or floating point operations, requiring serious computing power. To put it in perspective, the world’s fastest supercomputer, Frontier, has a peak performance of 1.1 exaflops (an exaflop is 10^18). Frontier is housed at the Oak Ridge Leadership Computing Facility and draws nearly 21 megawatts of power during a run. Training time for Anthropic’s planned model will depend on the scale and power of the system used, and the documents mention “tens of thousands of GPUs.”
This ambitious model could explain the $5 billion fundraising goal. Anthropic has raised $704 million in Series A and Series B funding rounds from 2022 and was reportedly looking to raise another $300 million last month. The Financial Times reported that Google is a 10% stakeholder in the company with an investment between $300 and $400 million. Another Anthropic investor is Salesforce Ventures, the investment arm of Salesforce. The company recently launched a $250 million Generative AI Fund and listed Anthropic as one of four companies the fund will initially invest in.
TechCrunch reports that the documents it obtained confirm the company is seeking to raise $300 million at a $4.1 billion valuation, and that at the time of the document’s creation, only half was raised from a “confidential investor.” According to the article, Anthropic is looking ahead with its fundraising and development goals: “These models could begin to automate large portions of the economy,” the pitch deck reads. “We believe that companies that train the best 2025/26 models will be too far ahead for anyone to catch up in subsequent cycles.”
Anthropic was founded in 2021 by former senior staff from OpenAI, including Anthropic CEO Dario Amodei, OpenAI’s former VP of research. Amodei and his sibling, Anthropic President Daniela Amodei, were among a group who left OpenAI due to “directional differences.”
We've reached out to Anthropic for comment will update this article as we learn more.
What's Your Reaction?
