Last May, Anthropic, one of the world’s hottest AI startups, raised $450 million from investors including Google and Salesforce. It was the start of an amazing funding spree.
By August, Anthropic had raised $100 million from two Asian telcos. Amazon then poured $4 billion into it, followed by another $2 billion from Google.
This month, venture capital firm Menlo Ventures closed a deal to invest $750 million in Anthropic.
In total, the AI ​​startup raised $7.3 billion in one year. The five financing deals stood out not only for their speed and size, but also for their unusual structures.
In one of those deals, Anthropic agreed to use technology such as chips and cloud computing services from the companies that invested in it. This meant, in effect, that some of the money it raised would be funneled back to its investors. And to pool smaller investors interested in Anthropic, Menlo created a legal entity known as a “special purpose vehicle.”
“These deals are so complicated,” said Dave Brown, vice president of Amazon Web Services who was involved in Amazon’s deal with Anthropic.
Despite AI’s promise to transform every aspect of society, it began by disrupting Silicon Valley’s startup dealmaking. Start-ups typically raise money every 15 months or so after they show their businesses have grown. But once genetic AI — which can generate text, images, sounds and video — burst onto the scene in late 2022, the rulebook was toppled as investors jockeyed for a piece of the hottest developers.
Few companies better illustrate this shift than Anthropic, which makes a chatbot called Claude and sells various forms of its AI technology. In the last year, the start-up’s valuation has tripled to $15 billion, three people with knowledge of its finances said. It reached about $8 million in monthly revenue last year and expects to grow about eightfold this year, two of the people said.
Other AI startups, including OpenAI, Character.AI and Cohere, have struck similar kinds of investment deals as they scramble to raise the most money, form the most lucrative partnerships, hire the best talent and gain access to the most computer chips . OpenAI recently completed a deal that values ​​it at $80 billion or more.
Investors can’t afford to miss the action because “if you miss the winner in the space, you’re kind of out of the game,” said Ilya Strebulaev, a Stanford economics professor.
A number of investments in AI startups by the tech giants have recently attracted the attention of regulators. Last month, the Federal Trade Commission said it had opened an investigation into Amazon and Google’s investments in Anthropic for possible antitrust violations.
An Anthropic spokesman said it plans to cooperate with the FTC. The company declined to comment further. Anthropic’s funding from Menlo Ventures was previously reported by The Information.
Since Anthropic was founded in 2021, Dario Amodei, the CEO, and his sister, Daniela Amodei, the president, have positioned it as a start-up that will build AI with guardrails. In a podcast interview last year, Dario Amodei said there was a 10 to 25 percent chance that AI technology could destroy humanity.
But if that doesn’t happen, he said, “it’s not going to go well, it’s going to go very, very well.”
From the beginning, Anthropic’s funding was unconventional. In 2021, it raised $124 million from investors including Jaan Tallinn, an entrepreneur known for his focus on the existential risks of technology, as well as the Center for Emerging Risk Research, a Swiss nonprofit that aims to “build a future guided by wisdom and compassion for all sentient beings.” (The group changed its name to Polaris Ventures.)
In 2022, Anthropic raised $580 million for research into building powerful AI technologies and working to ensure they don’t cause harm. Most of that amount — which dwarfed what venture capitalists had invested in other AI startups — came from Sam Bankman-Fried, the founder of cryptocurrency exchange FTX, and his colleagues. They belonged to a community known as effective altruists, which has long viewed artificial intelligence as an existential threat.
When FTX filed for bankruptcy in November 2022 and control of its assets was handed over to new management, Anthropic was left with an uncertain future. Its prospects were reversed days later when OpenAI launched its AI-powered chatbot ChatGPT. The technology that underpinned ChatGPT was largely developed by Dario Amodei and others who had worked at OpenAI before leaving to found Anthropic.
This drew attention to Anthropic and Google made its first investment. Anthropic also agreed to buy computing power through Google’s cloud computing service, which it uses to train and service its technologies.
In September, Amazon signed a similar deal with Anthropic, investing up to $4 billion. Anthropic’s Claude chatbot was the most popular AI service offered on Amazon’s cloud computing system, Amazon Web Services, a person with knowledge of the matter said.
As part of the pact, Anthropic agreed to build its AI using specialized computer chips designed by Amazon. If Anthropic is successful, Amazon’s shares in the start-up could perform handsomely. Meanwhile, the cloud computing deal will boost Amazon’s results.
The deal was structured as convertible bonds, or debt that becomes equity when Anthropic reaches certain milestones, two people familiar with the structure said.
Amazon’s funding of Anthropic mirrors how OpenAI raised money. In 2019, OpenAI raised $1 billion from Microsoft and spent most of the money buying computing power through Microsoft’s Azure cloud service. Microsoft has since pumped an additional $12 billion into the company, and OpenAI has spent the most money on Microsoft’s cloud services.
(The Times has sued OpenAI and Microsoft, alleging copyright infringement.)
Some investors have respondent such deals because companies like Google and Amazon invest money that ends up boosting their own revenue. The companies said the arrangements were kosher.
Google’s investment in Anthropic is separate from the start-up’s agreement to use its cloud services, said Daniel Gabis, a Google spokesman. They were “always separate,” he said.
Amazon properly accounts for all revenues and expenses, said Casey McGee, an Amazon spokesman. “To suggest otherwise, or that AWS’s deal with Anthropic is anything but a normal business deal, is completely false,” he said.
Even after raising billions from Amazon and Google, Anthropic knew it would eventually need more money. AI startups are constantly updating, improving, and expanding their technology to make their product accurate, up-to-date, and more powerful, and that requires massive amounts of precise computing power.
Finding new investors was easy for Anthropic. But many of those interested wanted to invest $10 million to $25 million, while the company was aiming for a much larger amount.
In November, Neerav Kingsland, Anthropic’s head of business development, spoke at a conference organized by previous investor Menlo Ventures. Menlo offered to lead Anthropic’s next round of funding, with a twist: What if the company pulled all the small investors into a special purpose vehicle?
The setup will save Anthropic time and simplify the process. Mr. Kingsland and Anthropic’s founders agreed, said a person with knowledge of the talks.
Anthropic told investors that $15 billion was the lowest valuation it would accept, two people familiar with the situation said.
After raising $750 million this month, Anthropic is no longer running a formal process to raise money, a person familiar with the situation said. But investors may soon have another chance.
As part of FTX’s bankruptcy proceedings this month, the crypto firm asked the US Bankruptcy Court in Delaware for permission to sell its 8% stake in Anthropic. FTX lawyers said they tried to move quickly to sell the shares alongside Anthropic’s upcoming funding rounds.
FTX understood “that Anthropic will continue to seek additional rounds of equity financing,” the lawyers wrote.
