The Rule of One or Two?
Who will be left standing in the AI race?
This week’s news largely centered around concentration of power. Two labs dominate the LLM space. In chips, Nvidia is moving up the stack. Is this dominance sustainable? What does it mean for smaller competitors? What does it mean for businesses like ours that increasingly depend on these providers?
When I led strategy at Sprint, the “Rule of Three” was a major focus. The wireless industry had been highly fragmented with many names that have now faded into history (Cingular, Sprint PCS, Alltel, Nextel, US Cellular, Western Wireless, Dobson, Rural Cellular, Leap, Qwest, Centennial, Triton). When I joined the company in 2003, there were 13 carriers with at least a million subscribers and #6 (Nextel) was about one-third the size of #1 (Verizon).
But in my 11 years there the industry consolidated dramatically. In 2004, AT&T and Cingular combined to jump into #1 (26% share), leaving Sprint a very distant #3 (10%). That year we announced a combination with Nextel to close the gap (19%), but a very poorly managed merger led to years of losing market share. By the time I left Sprint in 2014 T-Mobile was nipping at our heels.
Why does it matter, and what does that have to do with today’s AI industry?
In 1976, BCG’s Bruce Henderson published an insightful and influential paper titled “The Rule of Three and Four”. In it he posited “A stable competitive market never has more than three significant competitors, the largest of which has no more than four times the market share of the smallest.”
In other words, theoretically, if you aren’t in the top 3, or you have less than one quarter the market share of #1, then you don’t stand a chance.
This week’s news, and the general buzz around the AI model market would seem to imply that Anthropic/Claude and OpenAI/ChatGPT are running away from the field and no one else stands a chance. The real story is a bit more complicated.
Get Russell McGuire’s stories in your inbox
Join Medium for free to get updates from this writer.
Recent news indicated that Anthropic’s annualized revenue is around $65B. OpenAI’s annualized revenue is around $40B. This week’s news is that Anthropic is targeting a $2T valuation in its upcoming IPO, making it one of the 5 most valuable companies in the world. However, market share reports that Claude only has 9% share in the AI chatBot market, while Google Gemini has 28% share, and chatGPT has 54% share.
Meanwhile in the AI chip space, Nvidia has around 80+% market share, Google has around 7% and AMD has around 6% but growing fast.
Why would this “rule” be true?
It all comes down to money.
It takes a lot of money to train a new model (hundreds of millions of dollars). It takes even more money to respond to AI prompts against a model. OpenAI is expected to spend $14 billion this year on inference (responding to prompts). Anthropic and OpenAI are buying up all the hardware they can to build out data centers to respond to those prompts. Companies that don’t have paying users can’t afford to keep up in the capacity race and might not be able to afford to keep training models.
Similarly, it takes a lot of money to develop and produce a new AI chip. It is estimated that Nvidia spent $10 billion to develop their latest chipset. Companies that don’t have enough sales can’t afford to design and build the next generation chipsets.
What does that mean for those of us becoming increasingly dependent on AI?
For starters, it’s probably too early to call the AI race. Positions are changing rapidly. AMD and xAI are making bold moves in their respective spaces. Google hasn’t yet figured out how they want to play, but they have an extremely strong foundation on which to build. But in time, these markets likely will settle out to 2–3 dominant players who likely will have significant negotiating leverage. Some players will likely disappear. Others may emerge.
Pay attention to the changes that are happening and don’t become too reliant on a single provider.
If you want help thinking through how to manage AI vendors, reply to this. It’s part of what I do.
