The AI Winter That Isn't

Editor's note: In the AI market, costs have been falling. Recently, so have stock prices... But according to Jason Bodner of our corporate affiliate Brownstone Research, the technology itself is still advancing. In today's issue, adapted from the daily e-letter The Bleeding Edge, Jason explains how this disconnect emerged... and why it's unlikely to last.

I've never been a cryptocurrency guy, but I've watched markets long enough to recognize a familiar pattern...

Whenever cryptos collapse, headlines declare the experiment over. Investors lose fortunes, speculative projects disappear, and skeptics celebrate. They even have a name for these periods: crypto winters.

The crashes of 2018 and 2022 were brutal, with Bitcoin falling more than 75% each time. Yet after both collapses, Bitcoin recovered and set new all-time highs. Plenty of individual crypto projects failed along the way, but the broader market survived.

A similar disconnect is developing in AI. As I'll explain, the bears are once again mistaking falling prices for a lack of technological progress...

The AI-Earnings Reality Check

Wall Street is nervous about AI.

The world's largest technology companies are spending massive sums on data centers, processors, networks, and electricity. Investors understandably want to know when those investments will pay off.

Companies can't spend hundreds of billions indefinitely without generating attractive returns.

But AI-stock valuations aren't the same thing as AI demand. Right now, investors are confusing the two.

One of the biggest developments is agentic AI. Most people experience AI by asking ChatGPT a question and waiting for an answer. For them, it's a better search engine. That's it.

Agentic AI changes that equation. These autonomous or semi-autonomous systems can execute complex tasks and manage other AI agents independently, with little human oversight.

For instance, you could assign an AI agent to research an industry, analyze competitors, review financial statements, and prepare a presentation. It might deploy several specialized agents, each making dozens or hundreds of requests to models and databases.

Suddenly, your one request generates a massive amount of computing activity.

And that's where the economics get interesting.

When Cheaper Prices Mean More Activity

Skeptics point out that the cost of using AI is falling fast. They're right. Models are becoming more efficient, competition is increasing, and the cost of processing AI activity, measured in tokens, has roughly halved in recent months.

But cheaper prices don't necessarily mean a smaller market.

Think about cloud computing. As computing costs fell, startups could launch without expensive servers, businesses could analyze larger datasets, and software companies could offer services that were previously uneconomical.

Cheaper computing created more demand for computing.

AI will likely follow a similar path.

Suppose an AI task becomes 80% cheaper and businesses can afford to automate processes that previously weren't worth the expense. Multiply that across millions of companies and billions of tasks... and demand skyrockets.

And unlike chatbots, AI agents can work continuously, coordinating with other agents without someone initiating every request.

So even if the cost of each AI action falls dramatically, total AI spending would continue to rise.

Nothing is guaranteed. AI companies still need profitable business models. But investors focused only on falling processing costs may be missing an explosion in consumption.

Don't Confuse Price With Progress

When markets get volatile, I turn to the data. And recent AI earnings tell a different story from the stock prices.

Take Micron Technology (MU), a major supplier of AI memory, for example. Its latest quarterly revenue reached a record $54.2 billion, with operating cash flow approaching $44 billion. CEO Sanjay Mehrotra said that more than 75% of the company's 2027 production is already committed, with customer discussions extending into 2028 and long-term agreements reaching 2031.

That's an extraordinary look at future demand.

Corning (GLW) just reported optical-communications revenue of $2.38 billion, up 44% year over year. Generative-AI sales more than doubled, and adjusted earnings of $0.91 per share beat Wall Street's $0.88 estimate.

These results aren't about what AI might accomplish five years from now. They're being generated today.

And yet, both of these stocks are down over the past month:

Of course, not every AI stock is cheap, and some companies will struggle to generate acceptable returns. But when the businesses supplying this revolution keep delivering growth, a broad sell-off doesn't convince me the investment thesis is broken.

Stock prices reflect what investors will pay today. Earnings tell us what's happening inside the businesses.

Meanwhile, AI keeps getting cheaper, more capable, and more autonomous. AI agents could multiply computing demand far beyond today's chatbot usage, and the infrastructure supporting them still needs to be built.

I believe the AI revolution is alive and well. The technology is advancing, demand is growing, and the companies building the infrastructure are delivering real revenues and profits.

In a few months, investors may look back on this supposed "AI winter" as a missed opportunity. Because while stock prices have fallen, the businesses behind many of them are still growing.

And sooner or later, I believe the market will recognize that difference.

Regards,

Jason Bodner

Editor's note: AI chatbots and agents keep getting better... thanks to incredible advances in hardware. Right now, Jeff Brown – founder of our corporate affiliate Brownstone Research – is sharing how a new AI "super chip" can handle 120 trillion variables. That's 600 times more than market leader Nvidia's chip. With this powerful new tech, the AI bull market has plenty of room to run... and five little-known companies are poised to lead the rally.