Why I’m Buying the AI Selloff: My Top Picks for 2027

Listen to the audio version of this article (generated by AI).

On the evening of May 25, 1943, millions of Americans were glued to their radio sets for a special CBS program.

It begins with a “new study in terror,” as a woman attempts to call her husband only for the telephone to glitch and patch in two strangers arranging a murder.

She decides to call for help, but the line is busy no matter who she dials out to. Each failed attempt leaves her with more questions, until she begins to understand that the men were discussing her.

That was the premise of Sorry, Wrong Number, the radio drama starring Agnes Moorehead.

It may come from a different age, but its suspense is familiar: hearing only part of a conversation and trying to work out what happens next.

That’s the problem investors face every earnings season…

We hear what management says. We examine the numbers. Then we make judgments about a future we can’t possibly know completely.

Right now, those judgments have pulled AI stock prices and business results in very different directions.

Could the stocks you’re holding be time bombs?

In our 100th episode of Being Exponential, recorded on Election Day before the results were known, I examine that gap.

Technology stocks were trading at roughly 20 times expected earnings, while analysts projected about 35% annual earnings growth over the longer term.

Those estimates depend heavily on continued spending by the largest technology companies. If those budgets shrink, suppliers will have less growth than Wall Street expects.

The opportunity depends on whether that fear has pushed prices below what the businesses can realistically deliver.

Since then, the election results have added information we need to incorporate. I now expect slower data-center construction and lower earnings estimates across parts of the supply chain. My preference is for chips and related components that can follow investment to other locations, with greater caution toward businesses tied to individual building projects.

The demand story remains compelling. The delivery schedule has become more complicated.

Watch our 100th episode below for the evidence behind my AI outlook, the risks that could change it, and the stocks I’m watching for a recovery:

Continued Spending Can Still Arrive More Slowly

Before the results, my focus was on whether the largest technology companies would voluntarily retreat from their AI investments.

We now have another issue to consider: They may want to build faster than local rules, permits, and power availability allow.

That can delay spending and supplier revenue even when customers remain committed to their projects. It also means some earnings estimates will need to come down.

I expect a slower data-center buildout over the next 12 to 24 months. The companies closest to construction face the greatest pressure: contractors, power providers, and suppliers of electrical and cooling equipment.

Chip and component suppliers have more options. If one location cannot accept equipment, another facility may be ready. That flexibility supports my preference for chips, networking, memory, and optics.

But these businesses can still feel the slowdown. They are simply better positioned to follow spending as its destination changes.

This is why I’m recommending reducing exposure to the most construction-dependent businesses while concentrating on the essential components that growing computing demand still requires.

Look Beyond the Processors

More computing power creates additional demands throughout an AI system.

Processors need to exchange information quickly. They need memory to handle their work and storage to retain data. A bottleneck in any of those areas can limit what the whole system accomplishes.

That is why my attention extends to Marvell Technology (MRVL), Lumentum (LITE), and Astera Labs (ALAB), along with memory and storage companies such as Micron Technology (MU), SanDisk (SNDK), Western Digital (WDC), and Seagate Technology (STX).

In the episode, I identified these as potential beneficiaries if continued AI investment helps restore confidence in future earnings.

Their appeal after the election also reflects where they sit in the buildout. Their products can serve computing demand across different locations, giving them more flexibility than a business dependent on a particular construction site.

But falling prices do not automatically make the whole group a bargain. Customer concentration, competition, financing needs, and the timing of orders differ from company to company.

And a stock that looks inexpensive against an old earnings forecast can become less attractive when that forecast falls.

Nebius Offers More Upside (and More Ways to Disappoint)

My most aggressive rebound idea in the Election Day discussion was Nebius Group (NBIS).

Nebius rents computing capacity to customers building and running AI applications. That makes its growth highly sensitive to continued investment in AI.

If demand remains strong, the company delivers its planned capacity, and investors regain confidence in its earnings outlook, the shares could respond strongly.

That is why I described Nebius as my highest-upside idea for 2027. I also made clear that it was not my preferred choice for balancing potential gains against risk.

The post-election construction outlook makes that distinction even more important.

Nebius needs facilities, electricity, financing, and equipment working together before it can deliver computing services. Customer demand is necessary, but it does not resolve every obstacle between a signed contract and revenue.

A chip supplier may redirect a shipment. A cloud operator cannot relocate an entire contracted deployment as easily.

For Nebius, I want evidence that funding remains available on workable terms and that capacity arrives when promised. Yet, it belongs in the discussion as a higher-risk opportunity whose potential depends on successful execution. It should not be treated as interchangeable with the larger chip suppliers.

The Bottom Line on the AI Trade

The next phase of the AI trade will require more than confidence in the technology.

We need to know which businesses can keep serving demand, which depend on projects staying on schedule, and which have already fallen enough to reflect a more difficult outlook.

That is also why opportunities in application delivery and cybersecurity deserve separate consideration. Businesses still need reliable, secure digital services, but those investment cases should stand on their own customer demand and financial results.

My conviction in AI remains strong. My approach to owning it is becoming more selective.

To help you take that next step, we’ve put together a new free report: 5 Stocks for the Next Wave of Growth.

Watch our 100th episode of Being Exponential With Luke Lango here to download that report.