AI Startups Get Help from Big Tech

The day of reckoning for unsustainable LLM companies is fast approaching. This article summarizes a few of the many deals being made to acquire LLM startups while trying to avoid antitrust enforcement. It’s getting very difficult. The entire LLM sub-sector, valued at hundreds of billions of dollars, was built largely on data theft and anti-competitive quicksand. Because of LLM’s flawed model of being based on scale rather than quality and integrity of systems, they required massive amounts of computing power with an infinite appetite. The number of players in the world that had the infrastructure, capital, and strategic fit was very small and fairly obvious. With ChatGPT, OpenAI was starting an arms race full of risk, but it was a race that big tech companies felt they had to win, so they went all in. Within months, some LLM firms were running out of steam as they realized they didn’t have a sustainable business model, could incur significant personal liability for inherent security flaws, and the market was quickly drying up in favor of instant multi-billion dollar valuations. Even Musk had to make a significant effort to close his initial fundraising round, and he’s one of the richest people in the world with a very solid track record of return on investment. When the LLM craze began, it was assumed they would simply turn to their Big Tech strategic partners, but ironically, the deals themselves have prompted an acceleration in antitrust scrutiny, making it nearly impossible for Big Tech to acquire LLM startups. Even talent acquisitions are under scrutiny, as they should be — poaching the most top talent is as anti-competitive as any other tactic. The fundamental problem is that while language models offer great value, consumer LLM firms were built on quicksand from the get-go. Now, increasingly, they have nowhere to go. Either they have to pull off a small miracle and become financially stable while also finding a legitimate way to train on data, or they will fail. We had the opportunity to go down a similar path with Big Tech years before the OpenAI/Microsoft partnership. We didn’t go down that path because of what seemed obvious to me—the probabilities of the possible outcomes.