With OpenAI’s future in limbo, its startup customers are looking to jump ship to other AI models offered by rivals Anthropic and Meta

With Sam Altman’s potential return to OpenAI still in question, one group is also left with a lot of uncertainty: OpenAI’s startup customers.

For the millions of developers and startup founders that use OpenAI’s models and cloud provider for their own products or businesses, the board’s decision to fire Altman as CEO Friday caused many to anxiously explore other business partners in fear of service disruptions. Those apprehensions were further worsened Monday, when over 700 of OpenAI’s employees threatened to resign if he wasn’t reinstated.

Several startup founders told Business Insider that they were considering switching to an open source model like Meta’s Llama 2, or Anthropic’s Claude. And some said they were looking to switch cloud providers to Google or Amazon Web Services from Microsoft Azure.

“We will likely offer Llama 2 as an option for users to choose from in the future and look into fine tuning our own models if the OpenAI situation worsens,” said James Blackwell, the founder of the AI-generated quiz maker QuizGecko, whose company currently uses OpenAI’s model.

Still, several customers like Takeoff AI founder McKay Wrigley are switching to Microsoft’s Azure service, which sells copies of OpenAI’s models and other models. He said this option seemed even more appealing after Microsoft announced that it hired Sam Altman and former OpenAI president Greg Brockman to lead the company’s advanced AI research team. Even Morgan Stanley, one of OpenAI’s flagship customers, had said before the shakeup that it was shifting more of their AI software to rely on Microsoft’s Azure cloud, The Information reported.

Wrigley told Business Insider that he was sticking with OpenAI’s API for now, but spent the weekend switching all of his startup’s cloud services to the Azure OpenAI Service.

“I have full confidence in the team that is there,” Wrigley said about the OpenAI staff. “My concern, however, is that the team may soon not be there. If things aren’t resolved, they will all jump ship to the new Microsoft outfit, and at that point, who knows what will still be online. “

Other founders are now looking at Anthropic’s Claude to be an attractive alternative.

“We chose OpenAI because of its perceived stability,” said Charlie Dolan, the founder of a tech-enabled waste maintenance startup DSQ Solutions. Given the “turmoil” over the last 72 hours, he said he’s also considering Azure, but could just as easily transfer parts of his code that are reliant on OpenAI to Anthropic.

More than 100 OpenAI customers contacted Anthropic over the weekend, The Information reported.

However, founders said that this crisis has made them question their dependency on one vertically-integrated model, and have been thinking about diversifying across models or even developing their own.

“There’s definitely a lot of anxiety in the developer ecosystem right now,” said Guillermo Rauch, CEO of Vercel, a startup that helps developers build websites that integrate with many of the biggest AI models.

Matt McIlwain, Madrona’s managing director, told Business Insider that several AI founders in his portfolio reached out to him over the weekend asking for advice on how to avoid a crisis if OpenAI sees a mass exodus of engineers.

One trend he’s seen and recommends to some founders has been utilizing different models for different needs, “doing some training across OpenAI and then using Llama for another component or stable diffusion for another component,” he said.

He compared the issue to startups finding new banking providers after the Silicon Valley Bank crisis.

“We’ve had two big situations here in 2023 where you don’t want to be single-threaded — in one case it was our banking partners, and in one case it’s our model partners,” McIlwain said.

Alistair Barr contributed to this report.