My take on the current wave of "AI" companies
The breakthrough was real. Large language models genuinely changed what’s possible in computing. With enough data, compute, and tuning, we crossed a threshold. Machines could now generate coherent, fluid language, summarize complex information, and assist with tasks in ways that felt useful, even magical. It was a leap, the kind that doesn’t happen often.
OpenAI, Google, and Meta pulled it off. This wasn’t magic. It was brute force: scale, compute, and a pile of engineering duct tape layered on top of decades of research. Credit where it’s due. The tech is real, and it works.
Then the rest of the industry showed up.
Once the models started to work, the noise got loud fast. In 2024, over 70,000 AI-focused startups sprang up around the world, with the U.S. accounting for about a quarter of them. A lot of them weren’t building anything new. They were repackaging what already existed. And venture capital poured in. AI companies raised over $100 billion that year alone, nearly a third of global VC funding. In Q1 2025, they pulled in 58% of all funding. The signal got buried.
The result was predictable. Demo culture took over. Everyone raced to show something, anything, that looked like AI. Interfaces were polished, wrappers got dressed up as products, and the narrative shifted from "What problem are we solving?" to "How fast can we raise a round?"
I started seeing the same thing again and again: founders selling futures, not functionality. Demos were cinematic, but there wasn’t much under the hood. In fact, 90% of AI startups today are just thin layers on top of public APIs. That’s not inherently bad, but it’s not innovation either. The hard problems, he ones in healthcare, education and logistics, got left behind. In 2024, just $2.8 billion of VC healthcare investment went to AI companies, despite the sector's massive potential. That was only 21% of total healthcare VC spend.
Meanwhile, trust eroded. A Pew Research study showed 52% of Americans were more concerned than excited about AI, up from 38% just two years earlier. Among AI researchers at U.S. universities, 60% said they don’t trust companies to develop this stuff responsibly. And frankly, I don’t blame them.
We ended up with a generation of aesthetic startups with products optimized for screenshots and sizzle, not sustainability. It’s not that they’re malicious. It’s that the incentives are broken. Everyone’s building to be acquired, not adopted.
So where do we go from here?
We stop treating AI like a religion and start treating it like what it is: a tool. A powerful one, sure, but one that only matters when it's used well.
If you’re starting something new, start with the problem. Not the pitch deck, not the brand voice, not a ChatGPT wrapper. Figure out what’s broken and whether AI is even the right answer. Maybe it is. Maybe it isn’t.
If you’re inside a company trying to figure out how to "leverage AI," take a breath. You’re not late. You’re not missing out. You’re only behind if you’re chasing noise instead of solving problems.
Hype doesn’t build real companies. Thoughtful execution does. We’ve seen this movie before. Big tech waves, big promises, and eventually, the market flushes out what actually delivers.
You don’t win by wrapping APIs. You win by solving something that matters, and getting it into the right hands at the right time.
When the pitch decks dry up and the demo videos stop autoplaying, what’s left is the same question we’ve always had to answer: does this actually help someone?
That’s where I’m putting my focus.
Sources & References:
.png)
.png)