AI's $600 Billion Question

Two pieces have been circulating among investors and founders over the past few weeks.

The first, from David Cahn at Sequoia, is titled “AI’s $600B Question”. The argument, roughly: companies are spending enormous amounts on AI infrastructure - GPUs, data centers, energy - and to justify that spending, AI needs to eventually generate something like $600 billion in revenue per year. Today it’s nowhere close.

The second is a Goldman Sachs report with an equally pointed title: “Gen AI: too much spend, too little benefit?” One of their analysts argues the technology isn’t designed to solve the complex problems that would justify the cost.

Meanwhile, Nvidia briefly became the most valuable company in the world last month.

So: is AI a bubble?

Yes. And it doesn’t matter as much as you think.

I think there’s clearly a bubble in AI investment. Too much money is chasing too many similar companies. Many of the AI startups funded this year will not exist in three years. Some of the infrastructure being built will sit underused for a while. Valuations are pricing in a future that may take much longer to arrive than investors hope.

And AI is still going to change everything.

These two statements are not in conflict. They’re how almost every major technology arrives.

We’ve seen this movie

In the late ’90s, the Internet was a bubble. Pets.com, Webvan, absurd valuations, billions poured into fiber optic cables that nobody used for years. When it popped, a lot of people lost a lot of money, and plenty of smart observers concluded the Internet had been overhyped.

It hadn’t been. The hype had been early, and pointed at the wrong companies. Amazon survived the crash and became Amazon. The “useless” fiber became the backbone of the broadband era. The Internet ended up being bigger than even the optimists predicted - it just took ten more years than the stock market wanted.

The same thing happened with railways in the 1800s. Massive overinvestment, spectacular busts, and a railway network that transformed the world.

Bubbles are how society funds infrastructure for the future, by accident.

What’s different about AI

One thing makes me even more optimistic about AI than I would have been about the early Internet: people are already getting real value from it, right now, daily.

I see it in my own work, and in every developer I talk to. Coding with AI is dramatically faster. Writing, research, analysis, customer support, translation - real, measurable productivity gains, today, with tools that cost tens of dollars a month. The Internet in 1998 was mostly a promise. AI in 2024 is a promise and a working tool.

The question isn’t whether AI creates value. It’s whether the value gets captured by the companies spending all that money, on the timeline they need. That’s an investor question, not a technology question.

What it means for builders

If you’re building, rather than investing:

Don’t build on the hype cycle. If your business only works if AI funding keeps flowing at today’s pace, that’s fragile. Build something people would pay for anyway.
Use the subsidies while they last. Model prices are dropping fast, in part because companies are competing for market share with investor money. That’s a gift to builders. Take it - but don’t marry one provider.
Focus on real workflows. The AI companies that survive the shakeout will be the ones solving specific, painful, expensive problems for real customers, not the ones with the best demo.
Invest in people who can use it. The value of AI ends up wherever skilled people apply it. That’s why experienced developers who’ve learned these tools are the best bet in the market right now.

My prediction

Somewhere in the next few years, there’ll be a correction. Some big names will fail. Headlines will declare AI “overhyped”. And quietly, underneath, the technology will keep getting better and keep spreading into everything, until in ten years we can’t remember how we worked without it.

That’s not a contradiction. That’s just what a revolution looks like from the inside.

The bubble is in the valuations. The revolution is in the work.