You’ve heard the story: AI spend is unstoppable, hyperscalers will never slow down, and names like Micron, Lam Research, and Datadog sit at the center of a structural supercycle. Top analysts are pounding the table—UBS just tripled its Micron price target to $1,625, Mizuho lifted Lam to $380 on surging wafer fab equipment forecasts, and BofA pushed Datadog to $260 after strong AI-related wins. The consensus view is clear: load up, because this time it’s different.
Reality is the punchline. Micron crossed $1 trillion market cap on the AI memory scarcity narrative—the same industry that has destroyed capital in every prior upcycle once supply normalized. That deadpan fact bomb should make you pause. Robust near-term AI spend will still deliver beats, but these names trade at levels assuming no normalization ever arrives. Micron and Lam face re-rating risk as supply catches up and hyperscalers optimize. Datadog’s premium valuation collides with potential usage efficiency gains in its largest AI accounts. Markets are pricing one path. History usually serves three.
Start with Micron. UBS didn’t just raise its target to a street-high $1,625 on May 27—they cited long-term agreements and HBM sold out for all of 2026, with EPS projections hitting $155 in 2027 and $167 in 2028 before stepping down. The stock popped hard and briefly touched $1T valuation. Sounds bulletproof until you remember memory’s track record. Competitors like SK Hynix and Samsung are already ramping next-gen capacity. Past cycles show pricing power evaporating within 12-18 months of peak tightness. You’re not buying a permanent shift here—you’re buying temporary scarcity that the entire supply chain is racing to flood. When that happens, the multiple on those peak earnings won’t hold.
Lam Research tells a similar story with different numbers. Mizuho hiked its price target to $380, pointing to WFE spend climbing to $153 billion in 2026, up 23% from prior estimates, with memory alone at $112 billion. Lam sits in the sweet spot for etch and deposition tools powering both logic and memory ramps. Near-term execution looks strong. But cleanroom constraints, meaningful China exposure, and the industry’s habit of post-peak corrections create real volatility. Wafer fab equipment has never sustained elevated levels indefinitely after a buildout frenzy. If hyperscalers start trimming intensity in late 2026 or early 2027, Lam’s recent rerating gets tested fast. The equipment cycle has always been exactly that—a cycle.
Datadog brings the software layer to this mix. BofA raised its target to $260 after Q1 results and highlighted AI deals proving mission-critical positioning. The company continues posting high-teens growth with expanding platform adoption. Yet it trades at a high teens revenue multiple that bakes in endless complexity and usage growth. Large AI accounts are already experimenting with efficiency gains—better observability tooling that reduces overall data ingestion and monitoring spend over time. OpenAI concentration risk adds another layer. If hyperscalers optimize workloads and usage patterns normalize, Datadog’s growth rate won’t compound at the rates the valuation demands. This isn’t a broken business. It’s a high-quality one priced for flawless execution in a world that rarely delivers it.
Tie it together with the hyperscaler math that everyone quotes but few pressure-test. Estimates for combined AI-related capex from the big players now sit around $600-700 billion for 2026. That’s real spend hitting the ecosystem today. But the revenue ROI on that capital has lagged expectations in recent data points—efficiency debates and utilization questions are already surfacing in earnings calls and industry checks. When you deploy hundreds of billions, even modest optimization (10-15% better utilization or lower intensity per model) cascades straight to equipment and memory demand. The contrast is stark: memory and WFE equipment are priced for a structural supercycle shift, while the data increasingly looks like a classic cyclical recovery with temporary pricing power layered on top.
This setup creates a clear variant perception. The market is early and lazy on margin durability and multiple sustainability. Near-term beats will likely keep the stocks elevated through summer. But as supply responses materialize and capex ROI gets scrutinized, re-rating pressure builds. You don’t need a recession for these names to derate—you just need normalization. And normalization is what memory and equipment cycles have delivered every single time.
The strongest Pulse pieces connect the headline to business impact, valuation, positioning, risk, and what would actually change the view. Here, the business impact is front-loaded wins followed by mean reversion. Valuation embeds perfection. Risk sits in concentrated customer exposure and historical cycle patterns. The setup favors caution over blind allocation.