Nvidia Is Still the King: 1 Clear Reason the Rubin Upgrade Cycle Means the Stock Is Not in a Bubble

Nvidia is expected to benefit from a Rubin architecture upgrade cycle later this year, with chips offering a 10x reduction in AI inference costs and a 4x reduction in training costs, while costing about 25% more than Blackwell. Analysts estimate FY2027 revenue growth at 81% and FY2028 growth at 41%, alongside a projected $650 billion in hyperscaler data center capex for 2026 and more than $1 trillion in 2027. The article argues the stock is undervaluing this growth runway and remains cheap relative to its long-term potential.

Analysis

The market is still treating NVDA like a single-cycle capex beneficiary, but the more important setup is that the company is moving from hardware unit growth to architecture-driven monetization just as hyperscaler spend is shifting from buildout to equip-out. That matters because the incremental dollar now has a higher attach rate to accelerators, networking, and software stack lock-in, so the revenue mix should get more resilient even if headline data-center capex growth decelerates. The second-order winner is not just NVDA margin expansion; it is the entire compute supply chain that can keep up with a faster refresh cadence, especially advanced packaging, HBM, and optical interconnects.

The consensus risk is that investors are extrapolating the upgrade cycle without fully pricing in execution friction. Rubin’s launch window creates a classic “good news into scrutiny” phase where supply constraints, customer qualification timing, and yield ramps can temporarily compress sentiment even if the multi-quarter demand thesis remains intact. If hyperscaler budgets slip from announced spend into actual deployments, the near-term catalyst window shifts from a clean rerate to a more volatile staggered adoption curve.

The contrarian view is that NVDA may be cheap on outer-year earnings precisely because the market is underweighting how much of the next leg is already embedded in customer planning, not cash flow. If the architecture transition delivers a step-function in performance per dollar, customers may choose to redeploy budget into more workloads rather than simply reducing unit counts, which supports revenue durability and keeps pricing power intact. The real bear case is not demand exhaustion; it is a temporary digestion period if the broader AI trade has to reconcile valuation with slower conversion of spend into visible revenue across the ecosystem.

Key Decisions for Investors

• Stay long NVDA into the next 3-6 months, but prefer call spreads over outright stock to capture the Rubin rerate while limiting downside if the market turns execution-sensitive; best risk/reward is on a post-consolidation breakout rather than pre-launch chase.
• Overweight the NVDA supply chain on a 6-12 month basis, with emphasis on HBM, advanced packaging, and optical interconnect beneficiaries; these names should outperform if architecture cadence accelerates and deployment dollars shift toward higher-content systems.
• Pair long NVDA / short a basket of higher-beta AI infrastructure beneficiaries that depend on capex growth without clear product differentiation, to express the view that architecture leaders keep pricing power while commoditized AI trades lose relative momentum.
• Use any 5-8% pullback in NVDA over the next 1-2 months as an entry point rather than fading strength; the setup is more likely to be noisy than broken unless hyperscaler spending guidance rolls over materially.
• If evidence emerges that 2027 capex commitments are slipping, reduce exposure quickly and rotate from outright NVDA longs into call spreads or paired structures; the key risk is not fundamental collapse but multiple compression during a timing mismatch.