Although it may seem like the best model will win the AI race, we're overlooking the real factors that will determine which companies survive — and which disappear.
-
As consumers, we're constantly focused on model capabilities and which one delivers the best results. But for an AI model to survive, it has to be profitable.
Right now, every active AI company is still burning cash at an alarming rate. Investors are already beginning to tighten the purse strings and demand results, and any company whose business model isn't stable enough will be forced to shrink dramatically or shut down entirely.
-
For a language model to be profitable, it needs four foundational conditions:
processors to train the model,
a global cloud infrastructure to host it,
large volumes of high-quality data to train it on,
and an existing ecosystem of services and customers that can be upgraded by the model's capabilities.
The more of these conditions a company meets, the higher its profit margins — and therefore the better its chances of surviving once the bubble bursts.
-
Surprisingly, Google is the only company that owns all four.
It has in-house processors that eliminate dependence on Nvidia, Google Cloud as a global cloud infrastructure, an endless reservoir of high-quality data from its search engine, YouTube, Google Photos, and more — plus one of the largest user bases and app ecosystems in the world.
Even if the next models from Claude and OpenAI outperform Gemini by a wide margin, these four foundations will likely position Google as the ultimate winner of the race for AI dominance, with everyone else playing a supporting role.
-
Ironically, Claude and OpenAI — widely seen as the market leaders — lack all four conditions.
That means they're forced to pay enormous sums for processors, cloud infrastructure, data, and customer acquisition, with their profit margins shrinking accordingly.
Claude is already hedging its position by signing long-term agreements with enterprise clients, which will allow it to rely on a predictable revenue stream for years ahead.
OpenAI, however, continues to depend on individual consumers — users who could switch to a better or cheaper model at the click of a button — which puts a serious question mark over its ability to sustain its current market valuation over time.
There are, of course, many other interesting players in the space: Apple, X, Microsoft, and Perplexity.
If you're interested in taking a deeper dive into what the future holds for each of them, you're welcome to read the full analysis I wrote on the subject, published on Channel 10.
--
👋 Hi, I'm Shlomo Strauss — follow me for more content on science and technology.
In the image:
Three of Google's seventh-generation TPU processors, connected to a board and liquid cooling system. The seventh generation — a true breakthrough — is called Ironwood, and it's shaping up to be the beginning of real competition for Nvidia.