- Venture firms are increasingly investing in competing LLM startups like OpenAI and xAI.
- Traditionally, VCs avoided backing competitors like Uber and Lyft.
- Some VCs argue investing in multiple LLMs is strategic, while others see it as unethical.
When venture firms pull out their checkbooks, there has traditionally been an unspoken rule: Do not back a competitor. Choose either Lyft or Uber, but not both.
"I have never seen it before, and I assumed no one would do it," said Joe Aaron, founding partner at TRAC. "Would you want a company that invested with you investing in your competitor?"
However, that norm seems to be disappearing as investors pour billions into competing large language model (LLM) startups.
I’m seeing more VC’s investing in competing companies than ever before (all in AI)
— Sheel Mohnot (@pitdesi) October 18, 2024
Andreessen Horowitz has backed OpenAI, Elon Musk's XAI, and Safe Superintelligence (SSI), the AI startup co-founded by former OpenAI chief scientist Ilya Sutskever. Sequoia Capital invested in OpenAI in 2021 and then backed SSI in September. Fidelity and Ark Invest have stakes in both OpenAI and XAi. Sound Ventures and Wisdom Ventures backed both OpenAI and Anthropic.
"I think it's really flawed and very unethical," said Umesh Padval, managing director at Thomvest Ventures, which backed Cohere, a Canadian LLM developer. "I will never invest in Anthropic or OpenAI, because how can you say 'I'm going to be all in with you, but I'm hedging my bet."'
To Padval, investing in similar companies is an affront to what he sees as the central function of a VC, which is to identify a company and be strongly committed to it.
"The hedging of the bet happens when people are not convinced about their thesis," Padval said. "If the company doesn't do well, that's venture capital. But in order to cover yourself up, don't do another company."
Padval is also concerned about VCs who are privy to confidential information and potentially sharing it with a competitor.
"People make it sound like it's firewalls, but there's no firewalls," he said.
However, one person whose firm invested in both OpenAI and Anthropic denied they had access to private information.
"We are really not privy to the inner workings of either a company to a degree where it would be concerning that we would be carrying information into the enemy camp," said the investor, who asked not to be identified because they were discussing internal dealings.
OpenAI reportedly asked investors in its latest funding to refrain from investing in five competitors, something it is uniquely positioned to do, according to Gregg Hill, co-founder and general managing partner at Parkway Venture Capital.
"OpenAI wouldn't make such a request if they couldn't secure the billions needed for their funding rounds," said Hill. "Ultimately, the market will decide what succeeds."
Some VCs don't have a problem with investing in LLM competitors
It could make sense to back multiple LLMs at this relatively early stage because there will not be a single winner, argues S. Somasegar, managing partner at Madrona Ventures.
"Every company that is building an AI-driven application is using multiple models, and nobody thinks they are going to restrict themselves to using just one model," said Somasegar. "I think the rules for investments might have been more murky in these early days and I expect them to sort themselves out as we get more understanding of what these model companies are evolving into."
There is also the matter of money, with only a handful of firms capable of writing the colossal checks required to fund LLM companies. The top ones have raised so much money that investing in them is equivalent to buying shares of similar public companies on the NASDAQ, said a VC at a firm that has backed both OpenAI and xAI.
"You wouldn't do this for small private companies where the next round really matters, but once you break billions of dollars, honestly, what's the difference?" said the VC. "They might as well be public companies even how much money they've raised."