NVIDIA: ‘The Party Is Going to End Soon’ According to CNBC

Nvidia posted Q4 FY2026 revenue of $68.13B, up 73% YoY, and guided Q1 FY2027 revenue to about $78B, while shares trade near $215 and 24x forward P/E. The article is constructive on AI infrastructure spending but highlights a bear case that hyperscaler CapEx growth of roughly 10% may not support Nvidia’s 40% guidance trajectory. Corning and AMD were also cited as beneficiaries of the AI buildout, with Corning optical communications revenue up 36% to $1.85B and AMD up 94% YTD on strong AI demand.

Analysis

The key rotation is away from the obvious compute bottleneck and toward the enabling layers where pricing power may actually be more durable. That is constructive for GLW and other optics/power-cooling vendors because capacity additions in AI are becoming constrained by utility interconnects, rack density, and interconnect bandwidth rather than by raw GPU availability; those constraints typically support longer order backlogs and less cyclical revenue than semis.

NVDA still looks fundamentally strong, but the bear case is no longer about whether demand exists — it is about whether hyperscaler spending can stay ahead of internal silicon substitution and budget normalization. The second-order risk is that customers do not need to “beat” NVDA to slow the stock; they only need to marginally diversify deployments, which can compress multiple before fundamentals roll over. That makes the next 1-2 quarters the critical window: if capital intensity shifts from GPU count to infrastructure mix, NVDA can grow but still underperform.

AMD is the clearest beneficiary of the “good enough” alternative thesis, especially if large platforms keep splitting workloads across vendors to preserve negotiation leverage. The market may be underappreciating that Meta-style deployments do not need to fully displace NVDA to matter; even modest share gains in frontier inference can force a broader re-rating of AMD’s AI credibility over the next 6-12 months. The risk is execution: if software maturity or yield ramps lag, the stock’s YTD move can outrun shipment reality quickly.

The consensus may be too anchored on GPU unit growth and not enough on the multiyear capex stack behind it. If AI spend rotates into power, networking, and optical layers, the winners will be those with quasi-infrastructure economics, while pure-play accelerators face faster multiple compression from customer concentration. In other words, the trade is less “AI is slowing” and more “AI is maturing into an ecosystem, and the margin pool is moving away from the chip SKU.”

Key Decisions for Investors

• Add GLW on pullbacks for a 3-6 month trade: the setup is favorable if AI spend keeps shifting into optics and interconnect. Use a 15-20% trailing stop; upside is multiple expansion if backlog visibility improves.
• Trim or hedge NVDA into strength with a 1-3 month horizon: buy downside protection or sell call spreads against core longs ahead of the next hyperscaler capex guide cycle. Risk/reward favors reducing beta because the stock already discounts a lot of near-term growth.
• Long AMD / short NVDA pair for 6-9 months: express the view that incremental AI budgets increasingly favor second-source accelerators and negotiation leverage rather than single-vendor dominance. Stop if NVDA reaccelerates on broad-based cloud spend or AMD software adoption disappoints.
• Watch META as the signaling stock for custom-silicon and AI capex mix; if it continues to deploy aggressively while diversifying suppliers, that supports long GLW and AMD but is a warning for NVDA multiple stability over the next 2 quarters.
• For tactical traders, consider NVDA put spreads into the next earnings/guidance window: defined-risk way to play the possibility that growth remains strong but forward expectations become harder to raise, which is the setup most likely to compress the multiple.