What Just Happened
The first two weeks of June 2026 have been brutal for anyone holding AI stocks. NVIDIA dropped roughly 18% from its late-May peak. The broader AI and semiconductor sector shed somewhere around $800 billion in market value. Companies that had been riding the AI wave — server makers, AI software platforms, cloud providers — got hit hard. The Nasdaq had its worst week since 2022.
I was watching my own portfolio during the worst of it. I've held NVIDIA since early 2024, bought in around $680 pre-split. Even after the drop, I'm still up significantly. But watching six figures of unrealized gains evaporate in a week is not an experience I'd recommend.
The trigger was a combination of things. A disappointing earnings forecast from a major cloud provider. New export control rumors out of Washington. A widely-read analyst note suggesting enterprise AI adoption was "measurably slower than consensus expectations." And probably the most important factor: after eighteen months of nearly uninterrupted gains, AI stocks were priced for perfection. Any crack in the narrative was going to cause a selloff.
Let's Look at the Numbers
Here's what actually happened, stripped of the panic headlines.
NVIDIA went from about $145 at the end of May to around $119 by June 7. That's a meaningful drop. But here's the thing: NVIDIA was at $49 at the start of 2024. Even after an 18% correction, anyone who bought before March 2024 is still sitting on massive gains. The company is still projecting revenue growth of 60%+ year-over-year. Its next-generation Rubin architecture is on schedule for late 2026. The fundamental story — insatiable demand for AI compute — has not changed.
The rest of the AI supply chain tells a similar story. TSMC shares fell about 12%. The company is still booked solid through 2027 on advanced packaging orders. Super Micro Computer got hit harder, down about 25%, but Super Micro is notoriously volatile — it's dropped 25% or more four times in the last three years and recovered each time.
The AI software names took damage too. Palantir fell 20%. C3.ai dropped 15%. These companies were trading at valuations that assumed near-perfect execution, so a correction was arguably overdue. But again: their customers aren't canceling contracts. The underlying demand for AI capabilities in enterprise settings continues to grow. The selloff is about valuations, not business deterioration.
This Has Happened Before
Every major technology wave has had moments like this.
The internet buildout of the late 1990s saw Cisco drop 50% in 1996 — two full years before the actual dot-com peak and crash. People who sold during that dip missed a 4x run. Amazon fell 90% from its 1999 peak to its 2001 trough. The people who bought during that panic made a hundred times their money over the next two decades.
The smartphone boom had its corrections too. Apple dropped 40% in 2008, a year after the iPhone launched. If you'd bought during that panic, you'd have made about 50x by 2020.
Cloud computing stocks got hammered in 2016, with the emerging SaaS companies dropping 30-40% as a group. Five years later, most of those companies were worth 5-10x what they'd been at the 2016 lows.
Every single one of these corrections felt like "the end" at the time. The headlines were the same. The analyst downgrades were the same. The "I told you so" articles from people who'd been calling it a bubble all along were the same. And every single time, the underlying technology trend continued, and the companies executing well recovered and went on to new highs.
Now, does that mean this time is the same? No. Maybe this time really is different. But history suggests that corrections during genuine technology shifts are buying opportunities, not get-out-now signals.
The Bear Case (Which I Take Seriously)
I'm not going to pretend there are no reasons to worry. There are real concerns that the recent selloff might be more than just a technical correction.
Enterprise AI adoption is happening, but it's happening slower than the stock prices implied. A McKinsey survey from April 2026 found that while 72% of companies are experimenting with generative AI, only 21% have deployed it in production at scale. The gap between "trying it out" and "running critical business processes on it" is significant, and it might take years to close, not quarters.
The ROI question is still unresolved. Companies are spending enormous amounts on AI infrastructure — GPU clusters, model training, AI talent. The productivity gains are real in specific areas like software development and content creation. But whether those gains justify the aggregate spend is an open question. If enterprises start pulling back on AI investment because the ROI isn't there yet, the revenue growth that AI hardware companies are priced for won't materialize.
Valuations, especially in the AI software space, remain stretched even after the correction. A company growing revenue at 30% and trading at 25x sales needs everything to go right for several years to justify its price. In a higher interest rate environment, that math gets harder.
And there's the geopolitical wildcard. If export controls tighten further, if the US-China tech decoupling accelerates, if the chip supply chain gets disrupted — any of these could materially impact AI industry growth.
These are real risks. Anyone who dismisses them completely is not being honest with themselves.
The Bull Case (Which I Find More Compelling)
Here's my counterargument.
The AI infrastructure buildout is not speculative. Microsoft, Google, Amazon, and Meta have committed over $250 billion in combined capex for 2026, a significant portion of which is AI infrastructure. These aren't speculative bets. These are signed contracts with concrete delivery timelines. The hyperscalers aren't buying GPUs because they hope AI will be useful someday. They're buying them because their customers are demanding AI capacity and they're capacity-constrained right now.
The productivity data is starting to show up. A June 2026 NBER working paper found measurable productivity improvements in software development, legal document review, and customer service operations — three large sectors of the economy — attributable to AI adoption. The magnitude is modest so far (2-5% productivity gains), but it's accelerating. If AI delivers even half the productivity improvement that its advocates project, the ROI case is solid.
Most importantly, the fundamental driver of AI demand — the desire to automate cognitive work — is not going away. It's the same force that drove the adoption of computers, the internet, and smartphones. Each of those waves had corrections. Each of those corrections was a buying opportunity.
What I'm Actually Doing
I want to be transparent about my own actions, not just spout analysis.
I added to my NVIDIA position on June 5, when the stock was down about 15% from its peak. It's a small addition, maybe 10% of what I already held. The stock could easily drop another 10-20% from here. I'm not trying to call the bottom — nobody can do that consistently. I'm betting that five years from now, the level I bought at will look reasonable relative to where NVIDIA is as a business.
I also bought a small position in a semiconductor ETF during the selloff, giving me broader exposure to the AI hardware supply chain without betting everything on one company. And I sold nothing. Not because I'm confident the selling is over — I'm not — but because my investment thesis hasn't changed, and selling during a panic is how retail investors consistently underperform.
If you're considering buying AI stocks during this dip, the only advice I feel comfortable giving: only invest money you won't need for at least three to five years. Size your positions so that a further 30% drop won't force you to sell. And understand that picking individual stocks is risky — the AI winners of 2030 might not be the winners of 2026.
Is the AI Boom Over?
No. The boom in AI stock prices might be taking a breather. The boom in AI technology and adoption is not.
The distinction matters. Stock prices reflect sentiment, positioning, and expectations. Technology adoption reflects utility, productivity, and value creation. They're correlated over the long term but can diverge sharply in the short term. Right now, AI stock prices are correcting while AI adoption is still accelerating. That disconnect, historically, has been where the best long-term returns are generated.
Whether this correction is a buying opportunity or the start of something worse depends on whether you think AI will genuinely transform how work gets done over the next decade. I think it will. I might be wrong. But that's the bet I'm making.
