AI Stocks Without The AI Price Tag

While we firmly believe AI stocks are in a bubble, it is undeniable that AI is powerful and likely a major driver of future earnings. Even with the dot-com bubble popping in devastating fashion, the internet upon which it was based is a clear source of value.

As fundamental-based value investors, AI poses an interesting puzzle: How do we invest in the technology and underlying growth without exposing ourselves to the risks of a potential bubble?

The headline AI names are trading at rather extreme valuations, essentially already building in tremendous success. Even those with seemingly reasonable multiples, such as the chip makers, are arguably bubble valuations if one adjusts for the cyclicality of earnings.

We believe there is a different category of stocks that simultaneously provides exposure to the upside of AI while remaining compliant with fundamental value principles.

We sought and continually purchased stocks of companies that were clear fundamental beneficiaries of the buildout of AI but had not yet experienced a bloom in valuation. Let us first walk through the phases of bubble formation as they played out and then discuss the opportunity set.

AI bubble formation resonating outward

As bubbles form, there is usually an epicenter where the hype is most concentrated and first appears. After the initial hype phase, it resonates outward to adjacent industries that participate somewhere along the supply chain.

The current AI bubble began when OpenAI released its LLM to the world, and individuals could experience for the first time how powerful the technology could be. Thus, the epicenter was the explicitly AI companies.

It was apparent that OpenAI could not do it alone. AI would need astronomical amounts of compute and infrastructure. So, the bubble resonated outward.

Hyperscalers like much of the Mag 7 already owned vast amounts of computing power.

Chip makers, led by NVIDIA but inclusive of the whole set, were the obvious picks and shovels of the AI gold rush.

All the incremental compute would need 2 factors to be possible:

1. Lots of power
2. Data centers in which to house and power the equipment

Independent power producers emerged as favorites because of their ability to sell power at market price rather than a regulated price. As auction prices spiked, their revenue multiplied immediately.

Many data centers were requesting green energy, but their 24/7 nature required on-demand power that was difficult to produce from wind or solar, so nuclear received the lion’s share of hype. Anything remotely related to nuclear traded up to bubble valuation, even speculative nuclear and SMR (small modular reactor) startups.

Data centers took a surprisingly long time to get bid up but eventually received bubble valuation.

Finally, electric utilities are being seen as the gatekeepers of the incremental electricity production necessary to fuel AI. Valuations across the sector crept up but remain reasonable.

Fundamentally responsible investing in AI

The 2 greatest pitfalls to investing in AI today are:

1. Bubble valuations
2. Temporary fundamental benefit

As the hype resonated outward, investors could have done very well by investing in each ring before the pricing went parabolic. Investing after the move seems a bit more dubious.

As value investors, we were only able to invest before the move because our valuation principles precluded investment once prices went haywire. GE Vernova (GEV) is simultaneously a point of pride and remorse. We saw it early but also exited way too early as the stock surpassed what we viewed as reasonable valuation.

It took a remarkably long time for the hype and extreme valuation to reach the outer rings in the diagram above. In 2025, data centers were still cheap. The market was so used to companies that experienced the demand boom in a more cyclical (high operating leverage) sort of way that Equinix (EQIX) got clobbered on its Investor Day presentation in 2025.

The market just didn’t seem to comprehend that the growth EQIX was talking about was secular, repeatable growth, while something like a chip maker was experiencing cyclical growth. All the market saw was that EQIX’s growth number was smaller. It sold off, affording a value entry point into a top performing company with clear long-term exposure to AI.

We think there is still substantial mispricing in AI-related stocks and a clear opportunity within that mispricing. The biggest remaining source of mispricing seems to be a lack of differentiation between temporary and permanent fundamental benefits.

Temporary fundamental benefit

Much of the temporary fundamental benefit from AI stocks is related to imbalances in supply chains that were created by a sudden surge in demand.

• Chip demand surges; production is insufficient, causing chip prices to soar.
• Power demand surges; production is insufficient, causing electricity prices to soar.
• Turbine demand surges; production is insufficient, so prices soar.

We consider this a temporary fundamental benefit because the margin expansion is directly related to the current imbalance. Over time, production will rise to meet demand, at which point prices will normalize.

Many of these stocks are priced as if the fundamental benefit is permanent. The earnings multiples are only appropriate if the margins stay high. However, there are already signs of supply chains normalizing.

• New chip production is being built.
• New power plants are in various stages of development.
• Increased turbine manufacturing is in progress.

While there may be 1-3 years before sufficient production comes online, we see eventual restoration of equilibrium as inevitable.

Thus, we believe the stocks in these categories that are trading at high multiples are at risk of the bubble popping.

In contrast, there are other companies that have either permanent fundamental benefits or locked-in enhanced earnings for a long time period.

Permanent beneficiaries

The contrast is most clearly seen in the difference between IPPs and regulated utilities.

• IPPs experienced extremely high growth, with many even reaching triple-digit growth. Almost all of that was based on the price at which they could sell.
• Regulated utilities had much more muted growth, around 8%. Their sale prices are regulated, so they didn’t get to participate in the price spike.

However, as sufficient power comes online, prices will come back down, and IPPs will lose earnings power. Regulated utilities will have grown permanently with their increased load. In 5 years, the regulated utilities will have earnings that are permanently ~40% higher because their loads will be substantially bigger, and they get a regulated return on their load.

The market seems to be dramatically overvaluing temporary beneficiaries, almost extrapolating the recent earnings surge. This could prove dangerous as earnings not only stop surging, but potentially come back down to where they were before the spike. In my opinion, GEV, chip makers, and IPPs are all susceptible to a bubble-style crash.

3 other sectors are closer to permanent beneficiaries:

1. Contracted power providers
2. Data centers
3. Regulated electric utilities

Contracted power providers like Clearway Energy (CWEN) and HA Sustainable Infrastructure (HASI) sign long contracts for their power production. During the surge, they have secured contractual earnings on incremental generation for terms north of 10 years. The pricing they secured was nowhere near as extreme as the IPPs, but it will last much longer.

Data centers are similarly being built in a build-to-suit fashion where they are constructed with contracts already in place at going-in cap rates north of 10%. Capital-intensive development at mid-teen cap rates will not create explosive earnings growth, but it is durable earnings growth. That said, data center multiples are getting a bit above our value range, so we only have a small stub position in EQIX left as well as ancillary exposure from Broadstone Net Lease (BNL) and American Tower (AMT).

Electric utilities are, in my opinion, the best remaining AI play. While the sector has performed well, earnings have kept up such that earnings multiples have remained in the normal range. In fact, regulated utilities are trading cheaper relative to the S&P 500 than they normally trade relative to the S&P 500.

It is a discounted sector with a PE multiple of 20.47X, yet the sector’s forward growth rate is higher than its normal. Almost all the major utilities are calling for growth in the 7%-10% range annually for the next 5+ years.

The math just works well for investors at this valuation. Dividend yields of 3%-4% with 7%-10% earnings growth imply well above market total return potential.

Avoid the bubble but participate in the technology

Investing in the way discussed above has 3 main benefits:

1. Reduced downside if/when the bubble pops. There could be some collateral damage to the whole market given the scale of the bubble, but companies with solid fundamentals and reasonable valuation should bounce back quickly.
2. Long-term upside as AI technology progresses.
3. Agnostic to which AI model wins

There are so many AI models, and the “best AI” keeps changing. We have no idea whether the ultimate winner will be Gemini, Anthropic, Grok.AI, or some other model that hasn’t even been announced yet. We also don’t know if it will be winner-take-all or split among dozens.

Investing in the underlying infrastructure at a reasonable valuation doesn’t care about the above unknowns. If AI succeeds in any form, data centers, utilities, and contractual power producers will win. The key is just buying at the right valuation.