Listen to the audio version of this article (generated by AI).
In October 1974, a boxer past his prime walked into a ring in Kinshasa against a champion nobody thought he could beat.
George Foreman hit harder than anyone in the sport. Muhammad Ali knew that head-on, he loses. So he did something that looked, to the crowd ringside, like defeat in slow motion. He leaned back against the ropes and let Foreman swing.
Round after round, Foreman unloaded everything he had, and Ali just absorbed it, covered up, and waited. By the eighth round, Foreman had punched himself into exhaustion. Ali uncoiled, landed five clean shots, and Foreman hit the canvas.
The crowd that thought Ali was losing had it backward the entire time. He was not retreating. He was setting up the knockout.
I bring this up because the artificial intelligence infrastructure trade is doing its own version of the rope-a-dope right now, and a lot of investors are reading the ropes as a loss.
I do not think this AI infrastructure pullback is the beginning of the end. I think it is the eighth round, and the knockout punch is coming by early August.
Since mid-May, AI infrastructure stocks have gone nowhere. Two months of consolidation. Peak-spending fears everywhere you look. And this past week, we got three enormous earnings reports that should have reignited the trade, and the trade shrugged them off entirely.
Taiwan Semiconductor Manufacturing Co. Ltd. (TSM) posted blowout numbers. ASML Holding N.V. (ASML) posted blowout numbers. Advantest posted blowout numbers. Samsung Electronics did the same the week before. Four major chip-and-equipment names, four strong reports, and AI stocks kept sliding anyway.
Here is the part almost nobody is talking about correctly. Those four companies were never going to be the ones to end this selloff. Not because their businesses are weak. Because they are the wrong messengers entirely.
So let me walk you through why the real catalyst has not shown up yet, why I believe it arrives within weeks, and why the technical setup right now looks like one of the best buying windows of this entire AI cycle:
Capex Spenders Set the Tone
Think about who actually cashes the check in this buildout. TSM, ASML, Advantest, and Samsung sit on the receiving end of the spending. They are capex takers. Every dollar that flows through the AI infrastructure buildout eventually lands in their revenue line, but they do not control the spigot. They cannot tell you what 2027 and 2028 capital spending looks like, because that decision does not belong to them.
TSM said it is largely sold out into 2027 and 2028. ASML implied the same. Advantest gave bullish 2027 guidance that points in an identical direction. That is real, useful information about current demand. What it is not is guidance on whether the hyperscalers keep the spending accelerating into the later innings of this cycle. Those four companies structurally cannot answer that question, no matter how good their earnings look.
The companies that can answer it have not reported yet. Alphabet Inc. (GOOGL), Amazon.com, Inc. (AMZN), Microsoft Corp. (MSFT), Oracle, and Meta Platforms, Inc. (META) are the capex spenders. They write the checks. Until they stand up and reaffirm, or better yet hike, their 2026 AI capex guidance, this selloff does not end. It is that simple. So yes, four companies hit home runs this week, and the market did not care, because the market is waiting on the hyperscalers to step up to the plate.
The Case for Optimism Builds by the Day
Now, here is why I expect the hyperscalers to deliver exactly what the market wants. Look at the news flow. Alphabet just signed a multibillion-dollar cloud computing deal with SpaceX (SPCX). Meta is building out its own cloud infrastructure business and pouring billions into a new compute cluster. Anthropic just launched new models. ChatGPT 5.6 just launched. None of that reads like a spending slowdown. It reads like convergence, every major player accelerating at once.
I also want you to look at the technical picture, because it tells the same story the fundamentals tell. Since ChatGPT launched in late 2022, every single garden-variety pullback in AI infrastructure stocks has bottomed in the 10% to 15% max drawdown range. That is exactly where the sector sits right now. The SOX and SMH both hold their 50-day moving averages. They have not broken lower. That is the tell. Wall Street is not dumping this trade. Wall Street is waiting for the hyperscaler catalyst before it buys the dip.
Add in a genuinely favorable macro backdrop. Consumer Price Index and Producer Price Index inflation both came in soft this week, which confirms May marked the peak of this inflation hump. Softer prints in June, softer estimates for July, and I expect August continues the trend. Disinflation pulls Treasury yields lower, and lower yields matter enormously here, because the AI boom increasingly runs on debt financing.
Capex-to-operating-cash-flow ratios for these hyperscalers sit near 100%, so fresh spending increasingly comes from the debt markets. Cheaper debt means easier financing, and easier financing means the spending keeps running.
My Verdict
Putting it all together…
Strong current demand from the capex takers… Plus, a believable catalyst on deck from the capex spenders…. And a technical setup holding its historical bottoming zone with a disinflation trend that eases financing costs for the whole complex…
This is a market absorbing punches on the ropes, waiting for the eighth round.
I expect this trade to reawaken by the end of July, with AI infrastructure names back at all-time highs by early August.
Across the broader complex, I look for bounces in the range of 15% to 20%, and in specific high-beta names, moves of 50% to 100% over the following weeks.
These stocks run in exactly that rhythm: fifteen steps forward, five steps back. We just took the fifth step back. It is time for the fifteen forward.
I recommend staying aggressive here, because the buy window like this one does not stay open long.
