What is the story about?
The generative AI sector often feels like a bewildering pond of self-proclaimed futurists, each eager to prove their vision is the one that will change the world. Investors and founders alike bombard the market with flashy product demos and bold predictions.
Yet beneath this veneer of innovation lies confusion about what truly constitutes a lasting business in artificial intelligence. Everyone’s convinced they’re building the future, but everyone’s version of the future looks different.
According to Google’s Vice President Darren Mowry, the honeymoon phase for AI startups built on thin layers of innovation is ending fast.
In the early cloud computing era, many startups made a living by repackaging infrastructure from Amazon Web Services (AWS). They weren’t creating anything new, just reselling what already existed. The same pattern has now reappeared in AI.
Over the past two years, countless companies have launched “wrappers”, interfaces built on top of large language models (LLMs) like OpenAI’s GPT or Google’s Gemini. These startups offered sleek, user-friendly designs but relied entirely on someone else’s technology behind the scenes.
Mowry believes the market’s patience for such business models is running thin. “If you’re really just counting on the back-end model to do all the work… the industry doesn’t have a lot of patience for that anymore,” he said.
The message is clear that simply putting a pretty face on someone else’s model won’t cut it. The same goes for “aggregators,” which combine several AI models into one interface. Once hailed as innovative, these tools now risk being seen as redundant middlemen in a market hungry for original ideas and proprietary tech.
Mowry draws a striking comparison between today’s AI landscape and the early days of cloud computing. Back in the late 2000s, startups that merely resold AWS infrastructure were quickly wiped out once Amazon began offering its own enterprise features.
Only those that provided genuine services, like cybersecurity, DevOps, or performance optimisation, managed to stay afloat.
A similar squeeze is now unfolding in AI. Model providers such as Google, OpenAI, and Anthropic are expanding rapidly, adding enterprise-grade tools that leave little oxygen for startups that simply bundle existing APIs.
According to Mowry, if your business doesn’t own any intellectual property or solve a specific problem better than anyone else, the giants will do it for you, and probably cheaper.
According to Mowry, the future belongs to startups that build genuine moats. Those that focus deeply on a niche, whether it’s legal analysis tools like Harvey AI, or coding assistants such as Cursor, can survive by integrating domain expertise that generic chatbots can’t replicate.
Developer-focused platforms that have nurtured thriving communities, like Replit and Lovable, are also proving their resilience by offering tools developers actually need.
Yet Mowry’s comments also hint at something bigger — a looming AI bubble. With investors pouring billions into anything remotely tied to “AI,” many companies are propped up by hype rather than solid fundamentals.
The pattern looks uncomfortably familiar, an echo of the dot-com and crypto eras, where optimism outpaced reality. As the novelty wears off, the market will begin to filter out the noise. The AI industry may not collapse, but many startups will. The real survivors will be the ones who treat AI as a tool, not a buzzword, and focus on solving problems the world actually cares about.
Yet beneath this veneer of innovation lies confusion about what truly constitutes a lasting business in artificial intelligence. Everyone’s convinced they’re building the future, but everyone’s version of the future looks different.
According to Google’s Vice President Darren Mowry, the honeymoon phase for AI startups built on thin layers of innovation is ending fast.
The era of easy AI is ending
In the early cloud computing era, many startups made a living by repackaging infrastructure from Amazon Web Services (AWS). They weren’t creating anything new, just reselling what already existed. The same pattern has now reappeared in AI.
Over the past two years, countless companies have launched “wrappers”, interfaces built on top of large language models (LLMs) like OpenAI’s GPT or Google’s Gemini. These startups offered sleek, user-friendly designs but relied entirely on someone else’s technology behind the scenes.
Mowry believes the market’s patience for such business models is running thin. “If you’re really just counting on the back-end model to do all the work… the industry doesn’t have a lot of patience for that anymore,” he said.
The message is clear that simply putting a pretty face on someone else’s model won’t cut it. The same goes for “aggregators,” which combine several AI models into one interface. Once hailed as innovative, these tools now risk being seen as redundant middlemen in a market hungry for original ideas and proprietary tech.
Déjà vu from the cloud era
Mowry draws a striking comparison between today’s AI landscape and the early days of cloud computing. Back in the late 2000s, startups that merely resold AWS infrastructure were quickly wiped out once Amazon began offering its own enterprise features.
Only those that provided genuine services, like cybersecurity, DevOps, or performance optimisation, managed to stay afloat.
A similar squeeze is now unfolding in AI. Model providers such as Google, OpenAI, and Anthropic are expanding rapidly, adding enterprise-grade tools that leave little oxygen for startups that simply bundle existing APIs.
According to Mowry, if your business doesn’t own any intellectual property or solve a specific problem better than anyone else, the giants will do it for you, and probably cheaper.
Survive the shakeout or pop with the AI Bubble?
According to Mowry, the future belongs to startups that build genuine moats. Those that focus deeply on a niche, whether it’s legal analysis tools like Harvey AI, or coding assistants such as Cursor, can survive by integrating domain expertise that generic chatbots can’t replicate.
Developer-focused platforms that have nurtured thriving communities, like Replit and Lovable, are also proving their resilience by offering tools developers actually need.
Yet Mowry’s comments also hint at something bigger — a looming AI bubble. With investors pouring billions into anything remotely tied to “AI,” many companies are propped up by hype rather than solid fundamentals.
The pattern looks uncomfortably familiar, an echo of the dot-com and crypto eras, where optimism outpaced reality. As the novelty wears off, the market will begin to filter out the noise. The AI industry may not collapse, but many startups will. The real survivors will be the ones who treat AI as a tool, not a buzzword, and focus on solving problems the world actually cares about.
