I've Covered Semiconductors for 5 Years. Nvidia Is My No. 1 Pick Heading Into 2027.

Nvidia expects fiscal 2028 revenue growth of ~70% YoY, following Q2 fiscal 2027 revenue up 106% YoY to $96.2B. Management sees supply remaining the bottleneck through fiscal 2028, and guides Q3 revenue to $108B (±2%), implying ~89% YoY growth at the midpoint. The article also highlights monetization beyond GPUs, including growing AI infrastructure networking and server CPU traction (Grace >$5B trailing-12-months; CPU demand ~doubles in fiscal 2028), though it flags execution risk from $119B to $279B in supply/capacity commitments that could be adjusted if demand slows.

Analysis

The incremental edge is not that one chip wins a socket; it is that the company is turning each AI buildout into a broader infrastructure toll booth. That shifts the debate from unit share to attach rate, which is structurally better for NVDA than for single-product substitutes like AMD and, to a lesser extent, merchant silicon/networking peers such as AVGO. TSM still looks like a quiet beneficiary because supply scarcity pushes more value into upstream capacity, but its upside is more cyclical and less self-directed than NVDA's.

The near-term setup is mostly about whether hyperscaler capex stays on its current glide path, not whether the street likes the long-term story. If order timing slips or customers stretch refresh cycles, the market will quickly de-rate the “supply-constrained” narrative because the company has deliberately pre-committed to a large amount of capacity. The first falsifier is not a headline, it is any sign that gross margin expansion stalls while commitments keep rising; that would imply the company is absorbing risk faster than demand is compounding.

Consensus is still underweighting the second-order effect: if NVDA captures more of networking and CPU content per rack, the competitive damage to AMD and parts of AVGO is larger than a GPU-only model suggests. That said, the move is not free money—if AI model monetization does not improve over the next 6-18 months, all of the infrastructure names can compress together, with the more cyclical names likely hit harder. No meaningful read-through to NFLX or GETY; this is an AI capex and semicap ecosystem trade, not a broad risk-on signal.

Key Decisions for Investors

• Long NVDA / short AMD for the next 1-3 months into earnings and hyperscaler capex updates; target relative outperformance of 10-15% with the hedge reducing single-name beta. Falsify if AMD shows clear accelerator share gains or NVDA guides margin compression.
• Use pullbacks in NVDA to add via 6-12 month call spreads rather than outright stock if implied vol stays reasonable; the best payoff is from continued monetization per data center, not multiple expansion. Risk/reward is attractive if revenue growth stays above 50% and supply remains tight.
• Overweight TSM as a lower-beta AI capex beneficiary only if you want upstream exposure without paying NVDA’s premium for execution. This works best on dips; trim if foundry lead times or AI-related wafer starts flatten over the next two quarters.
• Underweight AMD vs the semis basket until there is evidence that custom silicon displacement is slowing; this is a cleaner relative-value short than betting against the whole AI complex. Cover if hyperscaler custom accelerator commentary turns materially more supportive than expected.