Listen to the audio version of this article (generated by AI).
There’s a reason professional athletes don’t train at full speed every single day.
If you push too hard for too long, you’ll break like a KitKat bar.
The body needs time to rebuild. So, sometimes, the smartest thing an athlete can do to get stronger is slow down.
Sound familiar? That is exactly what’s happening to the AI trade.
For the past few years, Big Tech has been running at full speed.
Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), and Meta Platforms (META) have been pouring hundreds of billions of dollars into AI infrastructure, building data centers as fast as humanly possible, rewarding investors handsomely.
But now we’re seeing the consequences of moving too fast.
Wall Street is shaking in its boots over AI infrastructure spending. Meanwhile, Washington is making it harder to build new data centers, and projects are being delayed due to critical equipment shortages.
In a nutshell, the AI boom is running into some very real limits.
That is going to create more pain for AI stocks before things get better…
In fact, we studied six similar selloffs in semiconductor stocks dating back to 1995. In every single instance, stocks fell further after breaking a critical long-term price level. The median additional decline? 29%.
Of course, history doesn’t repeat itself perfectly. But it’s a warning I’m taking seriously.
Here’s the interesting part, though. I’m actually more bullish on AI’s long-term future today than I’ve ever been.
Because slowing down now could prevent something much worse later. It could give companies time to get their finances in order, ease supply shortages, and build a healthier foundation for the next leg of this boom.
That’s why I’m not abandoning AI. Far from it.
I’m preparing for what could become one of the best buying opportunities of this entire cycle.
In today’s Being Exponential With Luke Lango, my co-host Brook and I discuss why I think the bottom may still be months away, the specific market signals I’m watching, and my three-stage plan for knowing when it’s time to start buying aggressively again.
Remember, the goal isn’t to buy every dip. It’s to have enough cash left to buy the one that matters. Watch the full podcast below:
Could Stocks Fall Another 30%?
A big rally after a selloff feels good. But it doesn’t necessarily mean the trouble is over.
That’s why we went back through semiconductor history to answer a simple question: What happens when chip stocks go from a euphoric rally to breaking below a major long-term trend line?
We focused on the PHLX Semiconductor Index (SOX) and its 50-week moving average, which is essentially its average closing price over roughly the past year.
In a healthy market, a pullback toward that line can attract buyers. Investors see an opportunity, step in, and help prices stabilize.
Last week, that didn’t happen. Semiconductor stocks fell through it.
So, we looked for previous instances of the same broad setup: a huge rally that stretched chip stocks far above their long-term averages, followed by a sharp break below the 50-week average.
Our study identified six precedents. Every one brought additional losses:
| Historical break | Additional decline |
|---|---|
| December 1995 | 24% |
| October 1997 | 34% |
| September 2000 | 59% |
| October 2018 | 16% |
| January 2022 | 35% |
| September 2024 | 21% |
That’s after the break below the 50-week average — on top of the losses that had already happened.
The median time to reach a bottom was 34 weeks, or roughly eight months. Applied to this selloff, that would put a potential bottom around July 2027.
Now, folks, six historical examples don’t tell us exactly what happens next. I’m not treating a 30% decline as inevitable or circling a date when we should buy everything.
But this is evidence we need to respect.
The recent rebound hasn’t yet repaired the damage described in our analysis. Major AI benchmarks remain below important trend lines and previous lows. Prices have bounced, but buyers still have to prove they can reclaim those levels and stay above them.
A market can rally sharply and still be in a downtrend.
So, what exactly could convince us to start buying again?
The Three Stages of an AI Market Recovery
We’ve gone defensive, but we’re not planning to stay in that posture forever.
I want to own great AI businesses for the next leg of this boom. I also want evidence that the market is finding its footing before we get aggressive.
Here’s how we’re breaking that process into three stages.
Stage One: The Initial Bounce
When stocks fall hard, they often bounce. Some investors buy what looks cheap. Others close bets against the market. Prices recover.
That alone doesn’t tell us whether a lasting recovery has begun.
It’s too early to dismiss every rebound as a temporary rally that will fail. It’s also too early to assume the bottom is behind us.
At this stage, our posture is to hold.
We don’t want to panic-sell after a plunge. We don’t want to chase the first rally, either. We want to watch what happens when stocks reach the levels they recently fell through.
Stage Two: The First Real Test
This is where the rebound starts to tell us something.
We’re watching whether the major AI benchmarks can reclaim their 200-day moving averages and important previous lows. Two reference points from our analysis are roughly $504 for the VanEck Semiconductor ETF (SMH) and $681 to $682 for the Invesco QQQ Trust (QQQ).
Getting above those levels is the first step. Holding above them is the next.
Think of it this way: A level where buyers previously stepped in can become a level where investors sell when they get another chance. We need to see buyers regain control.
If the rally stalls there and stocks turn lower, the case for another leg down strengthens. That would be a reason to use the rebound to reduce exposure.
If prices reclaim those levels and hold them on subsequent pullbacks, we can start getting more constructive and selectively buying again.
What matters is how the market behaves at those levels.
Stage Three: A Broader Recovery Takes Hold
Before we get aggressive, we want confirmation across the AI trade.
That means the major benchmarks trading above their 200-day moving averages and holding those lines when prices pull back.
For the semiconductor index, we’re also watching 11,684, its Oct. 28 high. A move above that level would break the pattern of progressively lower rally peaks and provide stronger evidence that the downtrend has ended.
We don’t need to buy at the exact bottom to participate in a meaningful recovery. We need enough evidence to justify putting more money to work.
For now, we’re watching the rebound approach its first important tests.
Let the market show us it can hold its ground. Then we can start moving back on offense.
Why a Slower AI Boom Could Be a Stronger One
Folks, I remain very bullish on the next two to five years for AI. But getting to that brighter future may require a difficult adjustment in how much companies spend, how quickly they build, and what investors expect them to earn.
To understand that adjustment, we’re watching five factors.
First, demand. This remains the strongest part of the story. AI coding tools, software that can complete tasks for users, and AI-powered machines all create reasons to use more computing power. I don’t see a demand collapse driving this selloff.
Second, returns on spending. Selling more AI services is one thing. Earning enough profit to justify the investment is another. Cloud businesses are growing quickly, but companies are still figuring out how to turn AI into better profits across search, advertising, logistics, and other operations. That makes this part of the picture less convincing.
Third, Washington and state governments. My expectation is that tighter rules and tougher approval processes will slow some data center projects. That doesn’t mean construction stops. It means spending that investors expected in 2027 could stretch into 2028, 2029, and beyond.
Fourth, Wall Street. When companies fund expansion with their own cash, their boards largely control the spending decisions. When they need to sell stock or borrow money, investors get a say, too. Falling share prices and more expensive borrowing make those decisions harder.
Fifth, supply shortages. A company can have customers waiting and money available, but it still can’t finish a data center without the necessary power, chips, and networking equipment. Missing components push back projects — and the spending attached to them.
That’s why strong demand alone isn’t enough to give us the all-clear.
Think back to the athlete pushing through every workout at full speed.
Taking time to recover may slow the immediate gains. But it can also make it possible to keep training for years.
I think AI could benefit from a similar reset.
A more measured construction pace would give power generation and equipment suppliers time to catch up. Spending that stays closer to what companies can afford from their own cash flow would reduce their dependence on willing lenders and rising stock prices. Addressing communities’ concerns could also reduce the risk of a harsher political backlash later.
I believe this shakeout could create an excellent opportunity. I also believe repeatedly buying on the way down can leave investors without enough cash when the evidence finally improves.
Our job is to preserve the ability to act.
I’m optimistic about where AI is headed, yet I’m staying patient about when to put more money to work.
