If you looked at OpenAI prior to Friday afternoon, it had everything an enterprise buyer (and, for that matter, an investor) could possibly want in a startup: an absolutely killer product in ChatGPT, a rock star CEO and huge potential future revenue.
It looked as stable as any startup could possibly be — until it wasn’t. While the situation remains fluid, it appears that CEO Sam Altman is prepared to drive his tour bus to Microsoft, taking co-founder Greg Brockman and most of his employees with him.
When the shit hit the fan Friday, suddenly all the dangers that vendor lock-in represents were on display, and when the soap opera spilled into the weekend, startups that had invested heavily in OpenAI tech had to wonder how this was going to play out.
Vendor lock-in has been a problem for as long as there have been enterprise buyers. While there are plenty of large language models in the market from a variety of vendors, OpenAI’s GPT 3.5 and GPT 4 seemed to stand above the rest.
Many companies I’ve spoken with have stressed that even though OpenAI appeared to be the market leader (with help from Microsoft’s significant investment in the company), there was a general feeling that the current AI wave is still very early and that it pays to be flexible when choosing a model.
The companies that chose a flexible approach over depending on a single AI model vendor must be feeling pretty good today. If there is any object lesson to be learned from all this, even as the drama continues to play out in real time, it’s that it’s never, ever a good idea to go with a single vendor.
Founders proceed cautiously
Founders who put all of their eggs in the OpenAI basket now find themselves suddenly in a very uncomfortable situation, as the uncertainty around OpenAI continues to swirl. One startup founder, who chose to speak on background so that they could be candid, says their company was just about to sign a big contract with OpenAI, and this situation has left them in limbo.
“It’s still unclear how things will change, given that Sam and several other folks are joining Microsoft,” the founder said. “I’m keen to see what special projects they announce there . . . I’m hopeful that OpenAI continues to build a great developer platform. However, there are chances things/rate of new updates will slow down, so we will be watching.”
Arthur co-founder and chief scientist John Dickerson, whose startup helps customers monitor and evaluate AI models, says that while his company is a big fan of OpenAI, recent events show the need for companies to constantly evaluate alternatives.
“We make heavy usage of OpenAI products, both internally and supported via our customers. Yet, these last three days show the brittleness that comes with relying on a single provider or a single model to back production systems,” he said.
Box, for example, built the AI layer on its platform to handle multiple models. Though the company didn’t anticipate something like we’ve seen from OpenAI in recent days, that flexibility will come in handy now, CEO Aaron Levie told TechCrunch+.
“Our whole strategy was to build an abstraction layer that can work across any AI model and any AI vendor,” he said. “While we never could have imagined that this particular set of events playing out as the reason why you need an abstraction layer, we’re built for any amount of flexibility and optionality for where the underlying AI models come from, so we’re well prepared.”
VCs preach flexibility
It makes sense that executives are looking for options when it comes to LLM vendors, and that’s the advice they’re hearing from investors as well. Lily Lyman, general partner at Underscore VC, says the firm encourages the use of open source models, but they understand that not all startups are willing to go that route, at least not right out the gate.
“We recognize that many companies, including some of our own portfolio companies, have built upon OpenAI,” Lyman told TechCrunch+. “While this tumultuous leadership transition is unsettling, the underlying technology is so valuable to so many, including Microsoft, that we hope there will be a resolution that preserves the value, but also takes into account some of the challenges of safety and aligned incentives.
The turmoil may not slow the pace of overall AI progress. Kobie Fuller, general partner at Upfront Ventures, says while this has been dramatic and unsettling, he doesn’t see it slowing innovation in AI and large language models in any way. “It will be interesting to see how this plays out in the next couple of weeks or month [from now],” he said. “But I’m not worried about any of my companies that are building with LLMs, or that suddenly this whole LLM layer is going to go away. That’s not the case at all. So I’m not concerned about that.”
It’s natural for early companies to look for the most performant model. And to this point, that’s been OpenAI, said Jon Lehr, co-founder and general partner at early-stage venture capital firm Work-Bench. But he says he sees more open source options in the future, and he’s encouraging startups to turn to those options more.
“The bottom line is that companies want to avoid vendor lock-in with a single vendor. The recent OpenAI news will lead to more diversity in the ecosystem and help propel open source as the go-to community for venture-scale founders building with LLMs in order to decrease vendor lock-in and increase control over the model weights as well as the ultimate inference produced,” he said.
One thing is certain — companies have always known not to bet all their chips on one vendor. Nobody could have foreseen a situation like this, but it shows the danger inherent in relying on any single vendor, even one that looks entirely stable.
If anything good comes from this crazy situation, perhaps it will be relearning that lesson.
TechCrunch reporter Jagmeet Singh also contributed to this report.