87 percent of companies are trying to hire AI engineers. But here’s the crazy part. 40 percent fewer AI-ready graduates are being produced than the market actually needs.
Universities can’t keep up. It’s not close.
Meanwhile, 70 percent of top AI talent is being absorbed directly by FAANG companies (Amazon, Apple, Facebook, Google, Netflix). The remaining 30 percent is split across thousands of companies.
So if you’re a mid-size company trying to hire an AI engineer? You’re competing with Facebook for the same person.
This creates a brutal dynamic. Mid-size companies can’t match FAANG salaries. So they lose. They end up hiring people who are almost qualified. Or they outsource. Or they use AI-as-a-service and hire people who know APIs instead of AI.
FAANG companies, meanwhile, have more AI engineers than they know what to do with. They’re shipping new models constantly. They’re pushing the frontier.
Everyone else is trying to apply last year’s AI to their specific problems.
This is actually creating a weird divergence. The frontier is advancing at one speed (very fast, driven by FAANG). But practical AI adoption is advancing at a different speed (slower, limited by talent availability).
It reminds me of how aerospace works. There are 10 companies that push the frontier. Everyone else uses their 10-year-old technology.
I think AI is heading the same direction. OpenAI, Google, Anthropic, Meta will keep pushing the frontier. Everyone else will be implementing and integrating.
Which means if you’re hiring AI people in 2026, you’re not competing for frontier researchers. You’re competing for implementers. People who can take existing models and apply them to your specific problem.
That’s a different skill set. And there’s actually more supply of those people. Because you don’t need a PhD. You need someone who understands your business and can use APIs well.
The shortage feels massive because everyone is trying to hire researchers. But if companies start hiring implementers instead, the shortage goes away.