Are We Living Through an AI Bubble? Here's Why Investors Don't Need to Know.

Michael Burry thinks we're living through an enormous AI bubble. Elon Musk thinks SpaceX could generate as much as $1 trillion in annualized revenue by the end of 2027. Who is right?

It's an interesting question, particularly as technology companies are spending hundreds of billions of dollars building the infrastructure needed to power artificial intelligence. My view? There are compelling arguments on both sides.

Are win in an AI bubble? I think we're witnessing a technological revolution that could fundamentally change how the world operates. But I also think it's entirely possible that we're experiencing a speculative investment bubble, at least in certain parts of the market. In fact, both could be true.

Here's what I think is far more important: As an investor, you don't need to predict whether we're in an AI bubble, or when that bubble might pop, to generate attractive long-term returns.

Let me explain.

The Bear Case: Are We Spending Too Much on AI?

Let's start with the skeptics, because I think their concerns deserve to be taken seriously. The amount of money being invested in AI infrastructure is staggering.

According to

Goldman Sachs Research, approximately $1 trillion is expected to be invested in AI-related infrastructure globally in 2026, including approximately $581 billion in the United States, equivalent to 1.8% of U.S. GDP.

And the spending projections for the years ahead are even more extraordinary. Goldman Sachs estimates that AI investment in the United States could increase to 2.5% of GDP in 2027 and 2.8% in 2028. For perspective, these levels are comparable to the investment booms that accompanied some of the most significant technological transformations in history.

We're talking about enormous amounts of capital devoted to building data centers, purchasing semiconductors, expanding electricity infrastructure and developing increasingly sophisticated AI models.

That raises a simple question. Will these investments ultimately generate attractive returns for the companies making them?

Because technological progress and attractive investment returns are two very different things.

Consider the internet bubble of the late 1990s. The internet really did transform the global economy. Enormous amounts of capital were invested in infrastructure that eventually became essential to modern life. But many of the companies financing that infrastructure generated terrible returns for shareholders.

The technology wasn't the problem. The prices investors paid, the assumptions about future profitability and the financing structures were.

Could something similar happen with AI? Absolutely.

Michael Burry's Warning

Michael Burry, the investor famous for correctly identifying the risks in subprime mortgages before the 2008 financial crisis, has been particularly outspoken about the dangers of the AI investment boom.

In his August 2026 commentary, Burry raised concerns about what he calls circular financing.

Here's the basic idea. Imagine Company A invests billions of dollars in Company B. Company B then uses that money to purchase computing infrastructure or services from Company A. Company A reports impressive revenue growth, which investors interpret as proof of incredible demand.

But where did that demand originate?

Ultimately, some of it originated with capital supplied by Company A itself. That doesn't mean the revenue is fictitious, or that the arrangement necessarily makes no economic sense. But it raises questions about the sustainability of demand and what happens when the financing stops.

Burry has also questioned the economics of expensive AI hardware. Companies are investing enormous amounts of money in GPUs and other infrastructure that could become technologically outdated relatively quickly. If the economic value of that equipment declines faster than accounting depreciation schedules suggest, reported earnings could overstate the underlying profitability of the business.

And if companies finance these assets with substantial debt, the consequences of disappointing demand or falling rental prices could be particularly severe.

Burry has suggested that 2028 could be a potential turning point for an unwinding of the AI investment boom. I don't know whether he's right. But I think the underlying questions are legitimate.

What if AI infrastructure spending is growing much faster than the ultimate economic demand? What if the equipment being purchased today becomes substantially less valuable tomorrow? And what happens to the companies that borrowed heavily to finance all this construction?

Those are real risks, and investors shouldn't dismiss them simply because AI is an exciting technology.

The Bull Case: What if AI Demand Is Dramatically Larger Than We Imagine?

Now let's consider the other side. What if all this spending isn't irrational exuberance? What if the extraordinary investment is actually necessary to support a market that could become much larger than most people currently appreciate?

There's a compelling argument here, too.

Artificial intelligence is different from many previous technological innovations because it has the potential to replace or augment enormous amounts of human labor. Think about the costs businesses incur every year for customer service, software development, sales, accounting, legal services, marketing, administration and countless other functions.

If AI agents can perform meaningful portions of those activities at a fraction of their current costs, the economic opportunity could be extraordinary.

Imagine a business that currently spends $100,000 annually employing someone to perform a particular set of tasks. What if an AI agent could perform much of that work for $10,000 annually?

The business could potentially save tens of thousands of dollars. Meanwhile, the provider of that AI agent could generate $10,000 in annual revenue from a single customer.

And that brings us to an important distinction.

The economic value created by AI could be significantly greater than the revenue captured by the companies providing it.

If an AI agent creates $100,000 in annual economic value, that doesn't mean the AI company earns $100,000 in revenue. The customer might pay only a fraction of that amount, keeping the majority of the financial benefit.

However, if AI agents eventually perform trillions of dollars of economically valuable work across the global economy, even capturing a relatively small percentage of that value could create an enormous market for AI providers.

The important distinction is that AI wouldn't necessarily need to create entirely new categories of spending. It could capture a portion of the enormous amount of money businesses already spend employing people and purchasing services.

And that brings me to some fascinating comments Elon Musk made during SpaceX's October 20, 2026, earnings call.

Elon Musk's Extraordinary $1 Trillion Revenue Prediction

During SpaceX's third-quarter earnings call, Musk suggested that SpaceX could achieve between $500 billion and $1 trillion in annualized recurring revenue by the end of 2027.

That's not a typo.

For context, SpaceX reported approximately $14.6 billion in third-quarter 2026 revenue, an increase of 161% year over year. Of that amount, approximately $7.6 billion came from its Super Intelligence segment, which includes AI infrastructure and software. Meanwhile, SpaceX spent approximately $19.6 billion on capital expenditures during the quarter, much of it directed toward expanding its AI computing infrastructure.

These figures illustrate both sides of the debate. The company is generating substantial and rapidly growing revenue, but it is also spending enormous amounts of money to support that growth.

And Musk believes the opportunity is just getting started.

One of the most interesting aspects of his comments involved Grok Bot, an AI agent designed to function as a virtual employee. Rather than simply answering questions, the goal is to have AI agents performing meaningful work within businesses.

Musk believes demand for these agents could become so enormous that they ultimately consume most of SpaceX's AI computing capacity. The company also reported approximately 800,000 paid weekly Grok Bot users earlier in October and discussed expanding its enterprise deployments.

In addition to its AI products, SpaceX is generating substantial revenue by renting computing infrastructure to other companies. Management described extraordinary demand for its available computing capacity and suggested that newly deployed infrastructure is currently generating payback periods of less than a year at prevailing rental prices.

That's impressive, although it also raises questions about how sustainable those economics will prove as additional computing capacity becomes available throughout the industry.

So how could SpaceX go from its current revenue base to $500 billion or even $1 trillion in annualized revenue?

To put Musk's prediction in perspective, reaching $500 billion in annualized revenue would require SpaceX to generate roughly $42 billion in monthly sales by the end of 2027. At $1 trillion, that figure rises to approximately $83 billion per month.

For comparison, SpaceX is targeting an annualized revenue run rate of $100 billion by the end of 2026. Even achieving that ambitious target would still leave the company needing to increase its revenue run rate fivefold to tenfold the following year.

That's an extraordinary assumption.

And it's worth remembering that generating economic value for customers doesn't necessarily translate into an equivalent amount of revenue for SpaceX.

An AI agent might perform work that saves a business $100,000 annually, but perhaps the business only pays SpaceX $5,000 or $10,000 for that capability. Alternatively, some customers might pay based on their actual computing usage through inference or API fees.

The economics depend on how customers are charged, how much computing capacity they consume, how much competition drives down prices and, ultimately, how much profit SpaceX can retain after paying for the infrastructure needed to provide those services.

Musk's projection also includes SpaceX's broader businesses, so it would be a mistake to attribute the entire revenue opportunity to AI agents alone.

This is precisely where the bull and bear cases begin to overlap.

The demand for AI could genuinely become enormous. AI agents might eventually perform trillions of dollars of economically valuable work every year. But that doesn't automatically mean the companies providing those services will capture all, or even most, of that economic value.

And even if AI becomes as transformative as Musk envisions, his timeline could still be wildly optimistic.

I don't know whether SpaceX will achieve those targets. But I think the underlying opportunity is worth taking seriously.

What if the total market opportunity for AI is ultimately measured not just in software spending, but in a meaningful portion of the world's labor costs?

That's a very different way of thinking about the industry's potential. And it's why I wouldn't dismiss the AI boom simply because current spending appears extraordinary.

AI Could Be a Revolution and a Bubble at the Same Time

This is probably my most important observation about the bull and bear arguments. They aren't necessarily mutually exclusive.

Michael Burry could be correct that certain parts of the AI infrastructure industry are overfinanced, that some equipment will generate disappointing returns and that investors have become overly optimistic. Elon Musk could simultaneously be correct that AI agents will eventually create hundreds of billions, or even trillions, of dollars in annual economic value.

The problem is timing, capital allocation and valuation.

An industry can have phenomenal long-term growth potential and still suffer a painful period of overinvestment. And a company can generate remarkable revenue growth without necessarily creating attractive returns for shareholders.

In fact, some of the biggest beneficiaries of AI could ultimately be the companies using cheaper intelligence rather than the companies spending billions to build it.

Consider what happens if AI becomes dramatically cheaper to provide. A business using AI could enjoy lower operating costs and greater productivity, while the companies providing computing infrastructure might experience declining prices as competition increases.

In that scenario, AI could be a tremendous economic success while certain AI infrastructure investments produce disappointing financial returns.

The question is whether the economic value generated by AI accrues to infrastructure providers, software developers, customers or some combination of the three. And that may change substantially over time.

That's why I think it's a mistake to treat the question of whether AI is a bubble as the same question as whether AI will change the world.

They're fundamentally different questions.

Which brings me to what I think investors should actually focus on.

Consider Micron: AI Growth Alone Isn't Enough

Let's look at a company like Micron Technology (MU). Micron manufactures semiconductor memory and storage products, including DRAM, high-bandwidth memory (HBM) and NAND flash. These products are critical to modern computing. In particular, HBM helps powerful AI processors access the enormous quantities of data needed to train and run increasingly sophisticated models.

As demand for AI infrastructure has exploded, so has demand for advanced memory. And Micron's results have been remarkable.

In fiscal 2026, Micron generated approximately $133 billion in revenue, compared with about $37 billion the previous year.

That's astonishing growth.

But here's the critical question. What will the supply and demand environment for memory look like three to five years from now?

Because memory has historically been a cyclical industry. When demand significantly exceeds supply, prices and profit margins can soar. Those profits encourage manufacturers to invest in additional production capacity. Eventually, if supply grows faster than demand, prices can collapse.

You could have an environment where AI demand continues growing rapidly, yet memory prices and profitability decline because manufacturers have expanded capacity even faster.

Conversely, if demand remains exceptionally strong and new manufacturing capacity proves difficult to build, elevated profitability could last much longer than investors expect.

Determining which scenario is more likely requires understanding Micron's manufacturing investments, the expansion plans of competitors like Samsung and SK Hynix, the economics of new memory technologies, and the potential growth in demand.

It's not enough to simply conclude that AI is the future. You have to figure out how much money Micron could reasonably earn through the cycle. And importantly, what price are you paying for those future earnings?

A company can look incredibly cheap based on temporarily elevated profits and turn out to be quite expensive once the industry normalizes. Or it could genuinely be undervalued because the market is underestimating a structural improvement in its long-term economics.

I don't need to know whether the broader AI investment boom qualifies as a bubble to analyze those questions. But I do need to understand the risks and whether the potential return justifies them.

And if I conclude that there are too many uncertainties surrounding future supply, demand or profitability, I don't have to invest at all.

That's the beauty of investing. You don't need to have an opinion on every stock or participate in every opportunity.

You can focus your attention on the situations where you believe the future is easier to understand and the potential returns are particularly compelling.

My Approach: Find Businesses Where the Future Looks Obvious

One of the most important lessons I've learned from investing is that you don't need to have an opinion on everything.

I don't need to know exactly where interest rates will be next year. I don't need to know whether the S&P 500 will fall 20%. And I certainly don't need to predict the exact timing of a potential AI bubble bursting.

What I want to find are businesses where I have a high degree of confidence that their products or services will be significantly more valuable to customers five or ten years from now.

I want to understand why that growth should occur. I want management teams who are aligned with shareholders and make sensible capital allocation decisions. I want business models that generate real cash flow, with strong enough balance sheets to survive challenging economic environments without relying on constantly raising additional capital.

And I want to pay a compelling price. Not a compelling price assuming that absolutely everything goes right. A compelling price based on what I believe are reasonable, and preferably conservative, assumptions about the future.

It's why I focus on obvious growth, aligned management, compelling valuation, avoiding blow-up risks and creating win-win outcomes for customers and shareholders.

For example, if a business can potentially double its earnings over five years, has minimal financial leverage, provides an essential service and trades at a reasonable valuation today, I might find that investment interesting. Even if the broader stock market is expensive. Even if AI infrastructure spending disappoints. And even if other investors are worried about an impending crash.

Of course, I also need to consider whether that business is directly dependent on the AI investment boom. A company whose revenues disappear if data center construction slows is very different from one serving an established and growing customer need.

A strong balance sheet and an attractive valuation don't eliminate the possibility of significant share-price declines. Even wonderful businesses can trade substantially lower during a market panic.

But they can improve the odds that a company survives the downturn and continues creating value over time.

And ultimately, that's what matters.

You Don't Need to Predict the Next Crash to Be a Successful Investor

Perhaps we're approaching the peak of an extraordinary AI investment bubble. Perhaps we're still in the earliest stages of an economic transformation that will ultimately dwarf today's expectations. I genuinely don't know. And I'm comfortable admitting that.

I suspect certain companies will spend far too much money chasing opportunities that never generate satisfactory returns. I also suspect other companies will create enormous amounts of shareholder value as artificial intelligence becomes increasingly integrated into everyday life.

The challenge is separating those two groups and understanding what you're paying for.

Being right about the future of AI doesn't automatically make you right about an AI stock. And being wrong about the next market correction doesn't mean you can't achieve excellent long-term investment results.

I'd much rather spend my time identifying exceptional businesses, understanding their competitive advantages, assessing management and estimating their long-term cash-generating potential than trying to predict the exact moment investors become too optimistic or pessimistic.

Because ultimately, investing isn't about having a definitive opinion on every major economic question. It's about identifying situations where the odds of an attractive outcome are meaningfully in your favor.

And when you find those opportunities, having the discipline to act.

Of course, even the most compelling investments won't always work out as expected. Businesses change. Competitive advantages weaken. Management teams make mistakes. New information emerges that can challenge even our highest-conviction investment theses.

That's why I view investing as an ongoing process, not a single decision.

You don't need to predict whether the AI bubble will pop. You need to make sure your investment strategy doesn't depend on getting that prediction right.

Want to Follow the Investing Journey, Not Just the Stock Picks?

At UNRIVALED INVESTING, I believe successful investing isn't about finding one great stock idea. It's about the ongoing journey of identifying opportunities, evaluating businesses, managing risk and adapting as new information becomes available.

That's why I share my real-money portfolio, timely updates when I buy or sell, and the reasoning behind my decisions. You'll see how my investment theses develop over time, what changes my thinking and what I learn when things don't go according to plan.

The goal isn't just to share what I'm buying today. It's to help you become a better investor over the years ahead.

If that approach resonates with you, I'd encourage you to start with my free eight-part course on finding stocks with 4X potential.

And if you'd like to follow my ongoing research, real-time portfolio decisions and investment journey, you can explore UNRIVALED INVESTING membership here.

Daniel

Founder, UNRIVALED INVESTING

This article is for educational purposes and does not constitute personalized investment advice.