I don’t think the stock market is actually crazy right now.

AI investment is concentrated in a small number of companies. Those companies have enormous room to grow. We have barely started turning what they’re building into products people actually use.

The large labs can grow AND the companies building on top of them can grow.

AI Is Democratizing

AI is democratizing. Switching costs are low. Barriers to entry are surprisingly low. Building an AI monopoly will be much harder than building a search monopoly. There will be no single monolithic AI model dominating the planet.

There will always be opportunities to build specialized AI on top of frontier models that is better, faster, and cheaper for a specific industry vertical, job, or workflow. The AI arbitrage opportunities here mean that you can do something 10x better, 10x faster, and 10x cheaper than monolithic AI.

I see this personally in the legal industry. State of the art monolithic AI providers (OpenAI, Google, Anthropic) are unusable for rigorous legal analysis, legal drafting, legal research, etc. Their accuracy only gets you 70% of the way there.

The AI value layer comes from SaaS companies that optimize AI models, AI workflows, AI agents, and AI user experiences around targeted use cases — that take their usability from 70% to 95%.

A frontier provider won’t own every $1B/yr niche. It can supply the models for thousands of them. That’s still trillions in value. They fight and compete to be the AI provider selected by the niche industry value layer. As that layer grows, they grow.

Frontier models will get better. That gives you more to build with. It does not eliminate the work of turning a model into a product people can actually use.

Who Gets Left Behind

• Companies with cultures that refuse to adopt AI or leverage it to maintain competitiveness or productivity.

Opportunities: Cultural Market Turnover

The opportunities we have today are similar to what we have always had. Companies that refuse to adopt modern innovative practices (AI) will lose business to those that do. The reason this happens is mainly cultural. A lot of companies are very slow to change, and it’s hard to steer a Titanic. It requires firing friends, breaking up political factions, slaughtering sacred cows, and layoffs of people slowing down the change.

We all know people who refuse to use AI as a matter of principle:

• The developer who simply refuses to use it from ignorance or arrogance.
• A tech lead who discourages AI use on the team because of sloppy PRs — creating a culture of fear around it.
• A nontechnical C-Suite who saw quality metrics drop when GPT 3.5 was adopted and made it a policy to not use it.
• The VP who sees AI cannibalizing their business and ruining their KPIs and selfishly crushes innovation.

Companies that fall behind will quickly find that their dwindling revenues no longer support their business structure, and they won’t be able to change fast enough to continue as a going concern.

This will happen at every layer of the market, from IBM-sized professional services companies, to manufacturing, to your local 5-person web shop.

The S&P 500 Will Keep Rotating

The overall S&P 500 should continue to see ~10% annualized gains. But non-AI companies will rotate out for new AI-enabled players.

Even in 2025 — the year we see massive increases in AI model competence and companies like NVDA reaching $5T values — the S&P 500 has remained stable with a modest ~12% return. They quietly rotate underperforming legacy companies for those that are leveraging technology and AI to gain market share in the new economy.

The dying legacy companies incapable of change shrink and rotate out of the mix.

The S&P 500 has rotated 11 companies already this year. There are more opportunities being created outside the Magnificent 7. Other S&P 500 companies showing massive gains in profitability by out-competing legacy companies are primed for their own 5x, 10x return opportunities. The companies at the top do not have to shrink for this to happen.

Doomers: The Market Isn’t Broken

There is a lot more to build.

• The overall market is in good shape.
• The value layer is wide open. Thousands of industry-specific, workflow-specific, job-specific AI applications that turn 70% use-case-specific accuracy into 95%. You can build these on top of the large labs. You do not need to beat them at building a frontier model.
• The size of the AI investment pie will continue to increase. The infrastructure grows. The labs grow. The companies building useful products on top of them grow.
• It’s unlikely we’ll see a Google-Search style monopoly because provider switching costs are so low, and the gap between frontier labs is so small.
• Frontier AI will continue to get better, faster, and cheaper.
• The value layer is primed with opportunity for companies willing to move fast.

This isn’t an impending market crash. It’s a market catching up to our new paradigm.