Anthropic CEO Dario Amodei admits his company struggles to balance safety with profits

In an interview with podcast host Dwarkesh Patel, Dario Amodei emphasized that Anthropic faces the same pressure as its competitors to keep innovating and ultimately become profitable 

“We’re under an incredible amount of commercial pressure and make it even harder for ourselves because we have all this safety stuff we do that I think we do more than other companies,” he said on the Dwarkesh Podcast.

Amodei was previously a vice president of research focusing on safety at OpenAI before jumping ship to start Anthropic with other ex-employees, like his sister and cofounder Daniela Amodei, who thought OpenAI was not focusing enough on safety. At Anthropic, Amodei has led the effort to build its large language model, Claude, with safety in mind. The company has done this in part through its Constitutional AI approach, which, rather than promote rules, gives Claude values so it can learn to be “a good AI.” Anthropic has also said it would not t release any AI capable of catastrophic harm, as outlined in its Responsible Scaling Policy. 

Anthropic has come a long way in the five years since it was founded, last week announcing a $30 billion fundraise at a $380 billion post-money valuation, solidifying it as one of the most valuable private companies in the world.

And yet, the company is, at the same time, managing a high-stakes balancing act between maintaining its mission and simply surviving, Amodei said.

“The pressure to survive economically, while also keeping our values is just incredible. We’re trying to keep this 10x revenue curve going,” he said. 

Amodei has previously recognized the pressure the company faces to compete in an industry that has set a new standard for rapid innovation. With a new model being released by any of the Big 4 AI companies, including Anthropic, OpenAI, Google, and xAI, seemingly every couple of months, Amodei has said if Anthropic sits on the sidelines, “we’re just going to lose and stop existing as a company.”

At the same time, the investors who are pouring billions of dollars into AI are desperate to get a return on their investment, Brian Jackson, the principal research director at technology research and advisory company Info-Tech Research Group, told Fortune.

OpenAI has already started placing ads within ChatGPT for free users and subscribers of its lowest tier plan—a move that Anthropic has called out publicly, including with a commercial at the Super Bowl earlier this month.

Yet, while Amodei has previously said he predicts the company will generate as much as $70 billion in revenue by 2028, Anthropic’s backers may still be reasonably anxious, Jackson said.

While earlier tech companies like Google or Meta (formerly Facebook) took just a few years to achieve profitability, AI companies like Anthropic and OpenAI have been around for several years and predict it will still be some time before they turn a profit. The longer they need to turn a profit, the riskier it is for the investors who backed them.

To be sure, in the announcement of its most recent fundraise, Anthropic said its run-rate revenue has grown 10x annually in each of the past three years since it brought in its first dollar in revenue. Its run-rate revenue sits at $14 billion, according to the announcement.

Yet, part of the reason these AI companies may be delayed in turning a profit is the sheer cost of compute, including capital spending on data centers and GPUs or ongoing cloud bills, Jackson said. Costs like these weigh on the companies’ finances.

While a search may cost Google next to nothing while also bringing in revenue from ads, the price to prompt an LLM is higher, Jackson said.

“As AI scales and as more usage grows, they’re not necessarily going to get to that profitability as easily or as quickly, because the cost per prompt is so high,” he said.