This Chip CEO Just Warned the Data Center Market Is in a 'Bubble.' Which Stocks Are Most Exposed?

Microchip CEO Steve Sanghi called data-center spending a “bubble,” noting data centers account for 16.7% of Microchip’s business and arguing the company could outperform if spending corrects. From Tuesday’s close through Thursday, Nvidia fell 12%, while Vertiv and Comfort Systems USA each dropped 25%; Microchip fell 17% over that span. Microchip reported September-quarter sales up 42.2% year over year to $1.62 billion, above the top of its guidance, and forecast December-quarter sales ranging from down 3% to up 1% sequentially; it also cited lengthening lead times as it adds capacity.

Analysis

The key distinction is not simply “AI exposure” versus diversification; it is where each company sits in the capex chain and how quickly orders can be deferred. Data-center builders and power/cooling contractors such as Applied Digital, Vertiv, and Comfort Systems USA are exposed to project timing and financing decisions before a slowdown necessarily reaches chip demand. NVIDIA faces concentrated customer-spend risk, but a pause in new construction would not automatically eliminate compute demand; utilization, refresh cycles, and customer inventory are the swing variables. Ciena’s networking demand may be less immediately tied to building starts, so treating all these names as one short basket risks missing relative resilience.

Microchip Technology’s lower data-center mix offers relative insulation, not immunity: its stated capacity constraints could limit its ability to convert demand into shipments, while industrial and aerospace recovery remains unverified. The CEO’s bubble framing is also an interested party’s view, not independent evidence of cancellations.

Near term, positioning and liquidity can extend the selloff even without estimate cuts. Over 1–3 months, watch hyperscaler capex commentary, project postponements, and orders/backlog at contractors and equipment vendors. Over 6–18 months, the structural question is whether AI workloads monetize enough to sustain infrastructure returns. A recovery in project awards or sustained utilization would challenge the bear case; cancellations, weaker backlog conversion, or downward guidance revisions would confirm it.

Key Decisions for Investors

• Consider a small, beta-aware relative-value position: long MCHP versus short a basket of VRT and FIX, but wait for post-selloff price stabilization and confirm MCHP orders convert into shipments. The thesis is relative exposure, not an outright claim that MCHP is cheap. Reassess if capacity constraints worsen or industrial demand weakens.
• Do not add to shorts in APLD, VRT, or FIX solely because of the one-day drawdown. For any tactical short, require evidence of project deferrals, weaker backlog/orders, or customer financing pressure; a rapid rebound in awards or guidance would invalidate it.
• Keep NVDA risk tied to evidence from hyperscaler spending and NVIDIA’s next results, not the CEO’s characterization of a bubble. Track data-center growth, customer concentration commentary, and inventory; continued spending and strong utilization would make a broad AI-capex short vulnerable.
• Monitor CIEN separately from construction-linked names. If networking orders and backlog hold while new-build activity slows, relative performance could improve; a deterioration in orders would remove that distinction.