The generative artificial intelligence (AI) industry is undergoing one of the most dramatic shifts in tech history, with successful third-party AI services becoming more important by the day. The past year has complicated that story as foundational models like ChatGPT continue to evolve and the competition heats up. As industry leaders grapple with these changes, the notion that any single company can dominate the market is becoming less tenable.
Since summer, we’ve seen hundreds—thousands—of third party AI services pop up, all using these foundational models in an interchangeable and agnostic fashion. These companies are more interested in tailoring existing AI solutions to specific tasks as opposed to creating their own large models. Startups are pivoting their business model. Furthermore, they understand that foundational models are commodities that are easy to swap in and out so that their needs can be frequently met.
As one example, a venture capitalist at a16z, and former founder of firm’s IP, Martin Casado, has been very public on this changing infrastructure. He asserts that despite the significant investments in AI technology, “As far as we can tell, there is no inherent moat in the technology stack for AI.” This perspective underscores the growing realization that no single organization possesses a lasting competitive advantage in an industry characterized by rapid technological advancements.
This past year, the excitement of increasingly developing larger and larger foundation models is waning. Big companies, especially Meta, have invested substantial resources into this reticulated despair. Now, they are under the harsh light of public scrutiny, wondering if their strategies are actually working. The billion-dollar spending spree on expansive model development is beginning to look risky, as startups demonstrate agility by tailoring AI models to specific use cases.
As we continue on in this age of all-consuming foundation models, there have been incredible breakthroughs. The initial buzz about these post-training approaches will likely fizzle out in six months. This volatility speaks to the fact that companies need to be flexible and agile to evolving market conditions in order to succeed. That’s what happens when rapid evolution of AI capabilities has put even industry frontrunners like OpenAI at the mercy of their competitors. OpenAI was the first lab to release coding models (Codex) and generative models for images and video. It is in these innovation sectors that it has since fallen behind to new challengers.
The story of AI thus far has highlighted a paradox: while foundational models serve as vital tools for innovation, their widespread availability means that successful differentiation will increasingly rely on how effectively companies tailor these technologies to meet specific user demands.
Startup teams are embracing this idea, focusing on creating specialized interfaces and solutions rather than solely competing with larger firms on model size. Because of their flexibility, they can shift course rapidly when forces change the equation. This underscores why it’s so critical for incumbents to wake up and rethink their approaches.
Among other things, the explosion of third-party services is redefining how we imagine AI. Companies are learning that sometimes teamwork and tailored solutions are more effective than going it alone to capture the entire market.
“It’s like selling coffee beans to Starbucks.” – A founder