Nvidia Stock Has Underperformed the Semiconductor Sector in 2026. Will It Skyrocket in the Second Half?

Nvidia is expected to begin shipments of its next-generation Vera Rubin processors this fall, with management citing a $1 trillion order pipeline for Blackwell and Vera Rubin across 2026-2027. The article says Rubin could lift system pricing by 25%, helping Nvidia's fiscal 2027 EPS rise 88% to $8.96 and revenue increase 82% to $392 billion. The stock trades at 29x earnings versus 75x for the iShares Semiconductor ETF, suggesting room for re-rating if execution stays strong.

Analysis

The market is likely mispricing the shape of the next leg of NVDA’s cycle: this is less about unit growth alone and more about operating leverage from a higher-value mix, tighter supply allocation, and ecosystem lock-in. If Rubin really lands with materially better perf/watt, the second-order effect is that hyperscalers will tolerate higher upfront capex because the compute cost per training run falls, which can actually expand addressable demand rather than merely shift spend between vendors. That makes the bull case more durable than a simple product-refresh trade.

The underappreciated winner is the broader AI supply chain, not just NVDA. Faster adoption of Rubin should pull forward demand for advanced packaging, HBM, photolithography, and networking; any bottleneck in those layers becomes the gating factor, which means upside can leak into firms with pricing power there before NVDA even re-rates fully. Conversely, if capacity constraints or yields slip by a quarter, the market will punish the entire AI basket because expectations are front-loaded into 2H, so the catalyst is powerful but timing-sensitive.

The contrarian read is that the consensus is anchoring on earnings acceleration while ignoring cyclicality in expectations. NVDA at a low headline multiple can stay cheap if the market believes peak growth is already in the price of the AI capex cycle; the key test is whether Rubin converts from product narrative into visible enterprise orders and margin expansion by the next two reporting windows. A failure to see evidence in backlog, mix, or gross margin would likely compress the multiple even if revenue stays strong.

For META, MSFT, AMZN, and GOOGL, Rubin is a net positive only if it improves model economics enough to justify another round of AI spend; otherwise, these firms risk being trapped in a capex arms race with slower near-term ROI. That creates a subtle relative-value angle: winners are the vendors selling picks-and-shovels, while the largest buyers are exposed to spending pressure unless monetization improves faster than compute costs.

Key Decisions for Investors

• Long NVDA into the next earnings/guidance window with a 3-6 month horizon; best risk/reward is on a pullback or post-consolidation break, targeting a re-rating toward 35-40x forward earnings if Rubin demand translates into visible margin expansion.
• Pair trade: long NVDA / short SMH or an equal-weight semiconductor basket for 1-3 months to isolate product-specific upside from broader AI sentiment; this works if NVDA’s execution beats the sector but the group remains choppy.
• Long the AI infrastructure enablers (e.g., AMAT, ASML, AVGO, MRVL, or NBIS/NET if available in your universe) versus short one of the large AI spenders on a relative basis over 2-4 quarters, because the bottleneck value migrates to constrained supply layers.
• Buy NVDA call spreads for the next 2 earnings cycles instead of outright common if you want convexity with defined downside; structure around a 20-25% upside band to capture a re-rate without paying for a full melt-up.
• If NVDA fails to show margin upside or backlog conversion by the next report, fade the move with a tactical short or hedge via put spreads; the setup is vulnerable to a 10-15% derating if launch timing or yield assumptions slip.