The artificial intelligence investment cycle has reached a scale that would have seemed improbable only a few years ago. A proposed data center campus in Virginia became one of the clearest examples of that shift, with plans for thousands of acres of development, dozens of buildings and an estimated investment approaching $100 billion before the project was ultimately abandoned following years of opposition.

The project itself is no longer the important part of the story. What matters is what it revealed about the current phase of the AI cycle. The world’s largest technology companies are not treating artificial intelligence as a short-term product opportunity. They are investing enormous amounts of capital to build the computing infrastructure they believe will define the next decade of competition.

That conclusion is not particularly controversial anymore. Investors have spent the past several years watching the first phase of the AI boom unfold. The companies providing the scarce resources required to build these systems, especially semiconductors, networking equipment and data center infrastructure, became the clear beneficiaries because the immediate challenge was simple: there was not enough computing capacity to meet demand.

The next phase is more complicated. The question is no longer whether companies will spend money on artificial intelligence. They already are. The question is whether the returns generated by that spending will justify the enormous amount of capital being committed today.

That distinction is what separates a powerful technology trend from a successful investment cycle. Nvidia remains the clearest example of how quickly an investment cycle can reward the companies solving the initial bottleneck. The company has transformed itself from a leading semiconductor company into one of the most important suppliers of the AI economy because its products became essential infrastructure for companies building large-scale AI systems.

But investors are now looking beyond the extraordinary growth of the past several years. The next question for Nvidia is not whether customers want more computing power. Demand is clearly strong. The question is whether customers can continue increasing their AI spending at a pace that supports another period of exceptional growth, and whether the economic benefits created by those investments are large enough to sustain the cycle.

The same question applies to the companies making the largest investments. Microsoft, Amazon and other technology leaders are spending billions on AI infrastructure because they believe the opportunity is significant. The market will eventually judge those investments not by the size of the spending, but by whether they produce measurable improvements in revenue growth, margins and productivity.

That is usually how technology cycles mature. The early winners are often the companies that provide the essential tools. The later winners are often the companies that use those tools more effectively than competitors. The internet created enormous opportunities for infrastructure companies, but the lasting value was ultimately created by businesses that used the internet to transform existing industries.

Artificial intelligence is likely to follow a similar path. The companies building the technology will remain important, but the next stage of the cycle will depend increasingly on adoption. Businesses will need to demonstrate that AI can improve decision-making, automate processes, increase efficiency and create advantages that competitors cannot easily replicate.

This is also why we believe investors should be careful about viewing AI as a single investment theme with a single set of winners. Large economic transitions rarely benefit only one group of companies. The AI buildout requires not just chips and software, but also electricity, infrastructure, networking, cooling systems and the physical capacity required to support a more computational economy.

At the same time, investors should remember that a transformative technology does not guarantee attractive returns for every company associated with it. Railroads changed transportation. The internet changed commerce. Both created enormous economic value, but many investors still lost money by paying too much for companies whose future earnings could not justify their valuations.

That is the tension in artificial intelligence today. The technology appears to be as important as many investors believe. The challenge is determining which businesses are creating durable economic value and which are simply benefiting from a wave of spending.

Our view is that the AI investment cycle is not ending. It is becoming more selective. The first phase rewarded companies that enabled the technology. The next phase will reward companies that can prove the technology is creating real earnings power.

That is a healthier environment for long-term investors because successful bull markets do not require the same leaders forever. They broaden as more companies demonstrate the ability to convert powerful trends into sustainable financial results.

The largest data center projects in history are not evidence that artificial intelligence has gone too far. They are evidence that the world’s largest companies are making one of the biggest capital commitments in modern business history. The investment opportunity now depends on what comes next: not whether artificial intelligence will be built, but whether the businesses using it can generate returns worthy of the investment being made today.