The AI boom could run into trouble if companies build more data centers than they actually need.
But data centers aren’t the only part of the AI industry facing this kind of uncertainty.
Another surprising development could have enormous consequences for some of the industry’s biggest companies.
And it’s happening much faster than anyone expected.
This week’s chart shows just how quickly things are changing.
An Open-Weight Revolution
Our chart this week comes from Vercel, a technology company that helps developers build and run applications.
Vercel operates a service called AI Gateway that allows developers to access hundreds of different AI models through a single platform.
And its data reveals a surprising change in which models developers are choosing.
You see, there are two basic types of AI models.
Companies like OpenAI, Anthropic and Google offer powerful models that they generally keep under their control. You can pay to use them, but you typically can’t download the underlying technology and run it yourself.
These are known as closed-weight models.
But there’s another option. Open-weight models are available for anyone to download and run on their own computers or through other services.
That means businesses can run these models on their own systems without sending sensitive information to companies like OpenAI or Anthropic. They can also fine-tune the models using their own data, customizing them for specific jobs.
That’s especially valuable in fields like health care, where researchers often work with sensitive patient information.
And according to Vercel, developers are choosing open-weight models at an astonishing pace.
Take a look.
At the end of last year, open-weight models accounted for just 7% of the AI tokens processed through Vercel’s platform.
But as you can see, by September that number had climbed to more than 71%.
In less than a year, open-weight models went from handling a small fraction of the work on Vercel’s platform to handling most of it. But even though open-weight models accounted for nearly three-quarters of token usage in September, they represented only about 23% of the estimated spending.
Closed-weight models accounted for the remaining 77%.
That brings us back to something we’ve discussed before about the economics of AI.
Developers are turning to open-weight models to handle more of their work. But they’re still spending most of their money on proprietary models.
That’s partly because the most advanced closed-weight models can still outperform their open-weight competitors on difficult tasks like complex coding and reasoning.
And developers are often willing to pay more when they need that extra capability.
But as open-weight models improve, developers have more choices about where to spend their money. Being able to customize those models and keep sensitive data under their own control, makes them even more compelling.
And it could put pressure on companies like OpenAI, Anthropic and Google, which have invested hundreds of billions into developing proprietary AI models.
After all, these companies aren’t just competing to build the smartest AI. They also need to convince customers that their models are worth paying for.
And that’s becoming a tougher sell as open-weight alternatives continue to improve.
That said, I want to be careful about what this week’s chart tells us.
Vercel’s data represents activity on its own AI Gateway. It doesn’t measure every AI application or every model being used around the world, and a larger share of tokens doesn’t necessarily mean a larger share of customers or revenue.
But the speed of change it represents is hard to ignore.
Here’s My Take
I’m still convinced that AI will create some of the most valuable companies in history.
But that doesn’t mean today’s leaders will be tomorrow’s biggest winners.
OpenAI, Anthropic and Google are spending a stunning amount of money to develop their AI models. And they’ll have to keep spending that kind of money to stay ahead, even as cheaper alternatives become more capable.
That means AI could keep growing at an incredible pace while some of its biggest players struggle to make a profit.
Regards,
Ian King
Chief Strategist, Banyan Hill Publishing
Editor’s Note: We’d love to hear from you!
If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to dailydisruptor@banyanhill.com.
Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!