Google Beat but AI Stocks Are Still Sinking. What’s Going On?

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

Luke got the earnings right, but the timing wrong … here’s what comes next… an AI event this Thursday… how to play the data center tsunami… readers sound off in the mailbag

Last week, in the run-up to Alphabet’s (GOOG) earnings, we highlighted analysis from technology expert Luke Lango, editor of Innovation Investor, who argued that if Alphabet held – or raised – Its AI capex guidance, it would help stabilize and support a rebound in the AI trade.

Sure enough, when Alphabet reported last Wednesday, it didn’t just hold its guidance. It raised it.

Cue the meteoric AI rebound rally, right?

Not exactly.

While part of the AI trade pushed higher last Thursday, most of that gain had slipped right back out the door by Friday. As I write on Monday, the AI trade continues to be under pressure.

So, what happened?

Luke: We got the earnings right – we got the timing wrong

That’s the quick summary Luke provided in last week’s Innovation Investor Daily Notes. But let’s go deeper.

After Luke walked through Alphabet’s blowout numbers, calling them “the strongest demand-side validation the AI infrastructure trade has produced in three years,” he pivoted to the ensuing AI selloff.

Here’s his take on why the rebound didn’t materialize:

While the earnings were landing, the Houthis opened a new front in the Iran War, attacking ships in the Red Sea on the other side of Saudi Arabia, sending oil above $90, the 10-year Treasury above 4.7%, and the 30-year toward 5.2%.

Frustrating? Yes. A reason to abandon the thesis? No.

In other words: the culprit was a macro-driven selloff, not a fundamentals-driven one. But Luke has seen this before – as well as what’s on the other side:

Every prior macro-driven correction in this AI bull market has resolved the same way: the geopolitics fades, the fundamentals retake center stage, and the prices correct toward the earnings rather than the earnings toward the prices.

Luke’s bottom line is patience. From a technical perspective, the AI trade sits in “no-man’s land” – meaning we’re mid-range between support and resistance. This leaves us in “wait and see” mode.

Luke’s take is that we’re looking at either a final flush toward the 100-day moving average or a consolidation-and-rebound setup. It’s frustrating – but it’s nothing that derails the AI trade.

Now, that’s Luke’s read on the next few weeks. But zoom out to the next few years, and there’s a different question entirely that has him fired up right now – one that has nothing to do with waiting out a selloff.

What has Luke especially excited today

Most investors assume the way to profit from AI is by owning the household names – Nvidia (NVDA), OpenAI, Anthropic, and the rest.

While Luke agrees, he says that it’s only half the picture…

As Big Tech’s cash pile has grown to hundreds of billions of dollars in recent years, the giants have increasingly chosen to buy breakthrough technology rather than build it themselves.

Look at Google’s DeepMind acquisition, Meta’s (META) $14.3 billion stake in Scale AI, or Microsoft’s (MSFT) $23 billion investment in OpenAI, before most investors had ever heard the name.

That pattern – enormous value created before the public ever gets a chance to buy a share – is what Luke’s calling the next phase of the AI boom.

This phase answers a new question: not “which companies will build AI,” but “which companies will AI’s biggest players eventually decide they can’t afford to compete with?”

For a deeper dive, put this Thursday, July 30, at 1 p.m. Eastern on your calendar. That’s when Luke is holding his 2026 AI MegaDeal Event.

During his presentation, Luke will walk through the framework he’s built for spotting these companies before Wall Street catches on – he calls it his “People, Product, Timing” filter. He’ll reveal one specific opportunity that fits this pattern today.

Plus, if you sign up and opt into Luke’s VIP text list, he’ll send over his bonus report, The AI Collectors’ Portfolio: 7 Stocks to Buy for the Biggest Tech Spending Boom of All-Time – totally free.

We’ll bring you more on this later this week, but to reserve your seat right now, just click here.

What’s powering the AI buildout (and how to profit from it)

Let’s now pivot to a question that’s easy to overlook amid all the AI headlines…

With Big Tech pouring more than a trillion dollars into data centers, chips, and infrastructure, where’s all the electricity to power this AI tsunami going to come from?

Senior Analyst, Brian Hunt, tackled exactly this question last week in his free e-letter Money & Megatrends.

After highlighting the unprecedented scale and velocity of the AI infrastructure capex boom, Brian pivots to the practical reality:

All that AI infrastructure is poised to consume huge amounts of electricity…

Unfortunately, we do not currently have adequate electric power generation and distribution capacity to meet the ravenous needs of AI and other power consumers.

In last Friday’s issue of Investing Insider, I touched upon the same challenge, adding some numbers in addition to Brian’s data.

Did you know that by the early 2030s, a single AI data center campus could consume more electricity each year than one million American homes?

To really grasp this, consider that there are only about 149 million homes in the entire United States.

Now, with this context, how many of these high-capacity facilities are planned or under construction?

267. (There are only 68 up and running today.)

So, if built, their combined electric demand would vastly exceed the electrical footprint of every single household in America.

That’s nearly 200 facilities still to come on top of the 68 already straining the grid today. So, Brian isn’t describing some vague future problem. He’s highlighting a very real math problem that gets worse every year it goes unsolved.

But it is getting resolved – which is where our opportunity lies…

How do we invest today?

One of Brian’s answers is IPPs – Independent Power Producers.

Unlike regulated utilities, which face strict price controls and limited capital flexibility, IPPs can build aggressively and charge market rates for their output. Higher risk, but higher upside.

Brian has flagged Vistra Corp. (VST) as a dominant IPP. But in last week’s issue, he named three smaller ones worth watching…

NRG Energy Inc. (NRG), held by hedge fund manager David Tepper; Talen Energy Corp. (TLN), held by top money manager Scott Ferguson; and Canada’s TransAlta Corp. (TAC).

Back to Brian:

IPP stocks enjoyed a big rally from mid-2023 to mid-2025. Since then, they have digested their gains and traded lower, with VST recently reaching a one-year low.

Given the electric power industry’s bullish long-term fundamentals, these stocks are likely to resume their long-term uptrends.

I agree with Brian. But don’t take it from either one of us. Connecting Brian’s point about “the electric power industry’s bullish long-term fundamentals” and last Friday’s Investing Insider issue, check out this chart I included from BloombergNEF.

It highlights U.S. data center electricity demand through 2035, forecasting it to reach 194 gigawatts – roughly four times today’s level – as hyperscalers race to build the infrastructure needed to power AI.

Bottom line: This is a huge multi-year growth story – invest accordingly.

For more of Brian’s analysis, he publishes Money & Megatrends every day the market is open, flagging opportunities like this before they become front-page news. His issues are loaded with trend analysis, actionable advice, and loads of specific tickers.

Best part: it’s 100% free. You can sign up right here.

Reader mailbag

Last week, I introduced an analytical framework I called “The Messy Middle” – the idea that AI won’t force us to choose between good and bad outcomes, but between two legitimate “goods” vying for priority. We won’t be able to have both in equal measure at the same time.

The conflict we flagged was “progress versus safety” when it comes to AI. Specifically, we looked at cheap, open-source AI models from China – tools that democratize access to frontier AI but also strip away every safety guardrail a closed, paid model comes with.

I asked what you thought, and we received some thoughtful responses. Here are two.

Paul M.:

This idea, that governments have some ability to control the progress of AI, fails completely unless all governments agree about how to do it.

And, what kind of draconian measures would be required to get organized crime to comply?

It seems that any such agreements would entail immense incentives to cheat.

Would you bet on the ability of the non-cheaters to win?

That’s a hard question to wrestle with, Paul – but such wrestling is needed.

The nuclear analogy might be instructive here. Non-proliferation treaties have had seven decades of the world’s most powerful governments behind them, and they’ve still only ever slowed proliferation, never stopped it.

Controlling AI is even harder. Obviously, no physical materials like uranium or centrifuges are required – just a laptop and a copy of the AI’s digital files leaked from a server.

So, maybe the real question isn’t whether we can control AI, but how we live in a world where we never fully can.

Wayne S. took the conversation somewhere different – less about government, more about a gap in the market itself:

There needs to be a strong, defensible, objective counterparty to the AI juggernaut. A company that will analyze the various releases on multiple dimensions and allow a more objective basis for discerning the differences between them. There is absolutely too much heat on this because of this lack.

Wayne, I think you’ve spotted a real gap. Something closer to an S&P rating for AI capability and safety doesn’t yet exist.

But let’s take it a step further…

After someone creates it, what then?

We could objectively measure what a model does – how capable it is, how easily it’s jailbroken, where it fails, and so on. But wouldn’t that data then need to be contextualized and translated into a ranking hierarchy to make it actionable?

If so, how do we objectively measure whether a certain level of risk is “acceptable” relative to the progress it offers?

After all, any attempt at that measurement points us back toward the core problem of the Messy Middle – conflicting values. Competing “goods” must be prioritized.

In this case, by whose standard, and whose authority, does anyone get to decide how much safety we’re willing to trade for some degree of progress?

I wonder – is the real issue whether someone builds such an analytical rating tool? Or whether our society can agree on how to interpret and act upon it once it exists?

If you want to chime in, email me at ipdigestfeedback@investorplace.com.

Have a good evening,

Jeff Remsburg

(Disclosure: I own GOOGL)