- AI model leaders may soon find themselves in shoes familiar to telcos on the wrong end of commoditization
- Freemium access might backfire
- Consolidation is likely on the horizon and that could have big implications for inflated AI company valuations
Telcos have resisted the label, but fixed and mobile broadband services have seemingly become utility-style services. They’re not “nice to have” but something consumers now need to have for basic work, school and entertainment functions. And while there is still plenty of differentiation between networks in terms of capabilities, broadband and mobile services are largely viewed as interchangeable. Could AI be next?
It depends on who you ask. But more on that in a second.
Fierce started pondering this question in light of Microsoft’s decision this week to sell its cloud and AI services to the U.S. government for a steep (we’re talking $3.1 billion here) discount. Importantly, Microsoft is throwing in a free year of its much-touted AI Copilot service for government users.
Microsoft could well be angling to become the only AI partner for government entities, offering the discount to edge out rivals like AWS, Google and other AI providers. They might be thinking that once users get hooked on AI, they’ll have no choice but to pay for it going forward.
Other AI players are similarly playing the “offer it for free to get users hooked” game. OpenAI, Google and Anthropic all offer so-called freemium tiers for AI tools including ChatGPT, Gemini and Claude.
The problem is that once people get something for free, they generally don’t ever want to start paying for it unless they are forced to.
This has notably been a problem for publishers, telecoms and streaming video services, all of which have coped by either offering consumers “perks” to justify upcharges or made users so miserable with ads and a poor experience that they’ve had to upgrade to a paid tier just to regain basic functionality.
Analyst perspectives on AI
Recon Analytics Founder Roger Entner told Fierce he’s not worried that AI will fall into the commoditization trap. Why? Because AI capabilities are very task specific. That is, you can’t really substitute one AI tool for another the way you can with, say, home broadband or mobile networks.
Entner used a fancy silverware analogy: In a proper set, there are different utensils for soup, salad, dessert and cutting. You’re not going to eat soup with a knife. Sure, you could try to, but you’ll probably have a bad time.
That specialization, he said, isn’t going anywhere anytime soon.
As for the freemium model? “It’s a gateway drug. You need to get people to try it,” he said. The big problem isn’t the technology being free, it’s beginners using it wrong and being offput by its performance, he said.
Gartner Distinguished VP Analyst Arun Chandrasekaran agreed, stating that there are still "significant [competitive] moats at the AI applications and infrastructure layers." On the latter front specifically, he noted raw performance, programmability and ecosystem integration remain "critical" differentiators.
But while AI may not be commoditized yet, he left the door open to the possibility in the future.
Entner’s Recon Analytics colleague Daryl Schoolar offered a different take.
Schoolar said a major challenge for AI companies today is differentiating their offerings in the eyes of consumers. “If consumers see all the AI solutions as being the same that will hasten commoditization,” he told Fierce.
He added that the freemium strategy AI companies are using is “risky, as it assumes consumers will find enough value in their services to pay for a premium experience. This is a big assumption, and it is doubtful there are enough people interested in paying to keep all these AI companies afloat.”
Sid Nag, an independent cloud and AI industry expert, similarly leaned toward AI traveling down the path toward becoming a utility-type technology. He told Fierce that instead of remaining the standalone technology it is today, AI will “evolve and mature as an embedded technology.”
Put another way, AI models will become the enablers of new applications and services rather than being the products themselves.
As AvidThink Founder Roy Chua pointed out, we’re already starting to see this shift with the rise of open weight models. These essentially amount to a kind of commoditization of the model layer.
“We’re not actually paying the full price for the value we’re getting. We’re not even paying anything near what it’s costing to give it to us,” Chua said. “The next level up is who tries to capture the value.”
Think about what the over-the-top streaming players did once connectivity became a given. With AI, the value will eventually go not to the model creators, but the applications that apply the models in a way consumers find valuable.
What it means for the market
So, what does this all mean for the marketplace and the oft-cited AI bubble?
Well, both Schoolar and Nag pointed to a forthcoming wave of consolidation. That means many vendors in the space will meet their demise and the market will be left with a “handful” of frontier model leaders, Nag said.
Such a shift, of course, means (arguably inflated) AI valuations will be adjusted accordingly.
“The valuation of AI companies will be recalibrated by the investor community as well as the financial analysts and it is likely that some investors will be impacted financially as a result of AI companies being wiped out or acquired at lower valuations than originally anticipated,” Nag predicted.
Schoolar similarly said some companies are “clearly overvalued” and likely won’t bring in the revenues needed to support those valuations. “I have more confidence in AI solutions that are part of a bigger digital solutions company like Gemini with Google than a standalone AI company that hopes to get all of its revenues from just AI,” he concluded.