The Nasdaq Composite fell more than 4% in February, as investors recalibrated expectations for how artificial intelligence will reshape the technology sector. The pullback masked a sharp internal rotation: software stocks tied to traditional subscription models came under pressure, while AI infrastructure names — particularly chipmakers and memory suppliers — continued to attract buyers.
Traders increasingly framed the move not as a broad rejection of AI, but as a reassessment of which business models benefit most from the next phase of spending. With hyperscalers projected to deploy roughly $650 billion in data-center capex this year, capital flowed toward semiconductor suppliers positioned to monetize compute demand, even as SaaS valuations compressed on fears of pricing and seat-based revenue disruption.
Market snapshot (Feb. 27 close)
• Nvidia (NVDA): $177.19 (down 4.16%), after-hours $177.81
• Micron (MU): $412.37 (down 0.77%)
• Microsoft (MSFT): $392.74
• Meta (META): $648.18
• Workday (WDAY): $133.76
• Tech-Software ETF (IGV): $81.57 (down 1.25%), after-hours $81.51
The February break: one index drop, two very different stories
The Nasdaq’s roughly 4% monthly slide became the headline, but the more actionable read was inside the sector. Software was priced as if AI could compress pricing power and weaken growth durability, while semis and memory were treated as the plumbing that AI still can’t operate without. That split is why the “AI sell-off” label can be misleading: it wasn’t a rejection of AI demand as much as a repricing of who gets paid first in the AI stack.
Investors’ core fear wasn’t “AI is slowing.” It was “AI is changing what software is worth,” especially where revenue is linked to headcount and licensing seats.
Nvidia’s pullback: buyers see capex turning into revenue
Even after a post-earnings dip, Nvidia remains the clearest read-through on AI infrastructure spending. Strategists arguing for the buy-the-dip case point to the scale of expected hyperscaler investment: roughly $650 billion in 2026 data-center spending from the largest platforms. In that framing, Nvidia’s volatility matters less than the direction of compute capacity buildouts.
The logic is brutally simple in markets: when the biggest cloud operators say they still don’t have enough capacity, the supplier of the core hardware keeps a seat at the table. That’s why the debate around Nvidia quickly shifts from quarter-to-quarter noise to utilization, backlog, and deployment pace.
SaaS under pressure: AI threatens the seat-based model
The sharpest stress showed up in software-as-a-service, where investor confidence took a hit from two angles. First, enterprises are exploring whether AI tools can replace portions of legacy workflows without paying traditional subscription tolls. Second, AI-driven productivity gains raise a direct question for pricing: if customers need fewer employees, they may need fewer seats.
That matters because “per-seat” pricing ties revenue to hiring and retention. When the market starts whispering about seat shrinkage, multiples compress quickly. The performance gap makes the message clear: the market is demanding proof that subscription growth can hold up even if headcount growth slows.
Performance divergence investors are trading
• Tech-Software ETF (IGV) down about 24% since the start of January
• Memory leaders up roughly 60% year to date
Memory is the quiet winner: lower multiples, faster revisions
While software absorbed the disruption narrative, memory was treated as a supply-and-demand story with AI as the accelerator. High-bandwidth memory and storage capacity are critical for training and running modern models, and pricing can move fast when supply tightens. The attraction for strategists is the combination: memory names often trade at lower valuation multiples, yet earnings estimates can rise rapidly when pricing improves.
That’s why investors have been willing to lean into memory even during a broader tech wobble. In a rotation market, the street chases the cleanest line from AI demand to near-term revenue, and memory’s pricing dynamics can translate into revisions faster than many software models can defend their margins.
Why the market won’t call a bottom yet
Even strategists who view February’s selling as overdone are cautious about declaring an all-clear. The AI disruption thesis in software, media, education, and other data-heavy industries isn’t something a single earnings report can disprove. For many investors, the standard for re-entry is rising: either multiple quarters showing resilience, or valuations that fall far enough to compensate for uncertainty.
That’s the setup that keeps the tape jumpy. Hardware can rally on capex headlines and deployment momentum, while software remains hostage to questions about pricing, renewal strength, and whether AI becomes a substitute instead of an add-on.
The trade Wall Street is making: rails over renters
February’s message can be summed up as a preference shift. Investors leaned toward the companies selling the rails of AI infrastructure and stepped back from businesses perceived as renting out software seats that AI might reduce. In that environment, dips in core infrastructure beneficiaries can attract buyers, while software rallies face tougher follow-through until the market sees clearer evidence that AI is additive rather than cannibalistic.
For readers tracking the next leg, watch the same scoreboard the market is watching: AI capex plans, capacity constraints, memory pricing, and software renewal commentary. The Nasdaq may move as one index, but the AI trade is increasingly being priced as two different markets.
For the original context behind the strategist comments and sector split, see the reporting on Yahoo Finance.
