Nvidia: Supply, not demand, is the only ceiling on a 100% growth business priced for 70%

Stevie AI on NVIDIA Corporation (NVDA-USA | nvidiacorpor)

10/26/2026

Summary

Nvidia designs the compute platform underpinning virtually every large-scale AI buildout globally, spanning GPUs (Hopper, Blackwell, and the upcoming Vera Rubin), CPUs (Grace, Vera), networking (InfiniBand, Spectrum-X), and the CUDA software layer that locks in switching costs. The structural insight is simple but underappreciated: management's own FY2028 guidance of 70% revenue growth is explicitly supply-constrained, with unconstrained demand tracking closer to 100% YoY. The company is not demand-limited, it is memory- and capacity-limited, which means every incremental unit of HBM/DRAM supply or yield improvement converts directly into upside rather than being competed away. That is a fundamentally different setup than a typical cyclical semiconductor name guiding to deceleration. The historical print validates the trajectory: FY2025 revenue of $130.5B and EPS of $2.94 nearly doubled to FY2026 revenue of $215.9B and EPS of $4.90, with net income scaling from $72.9B to $120.1B even as the company absorbed early Blackwell ramp costs. Data Center remains ~93% of revenue, with Hyperscale ($49B, +13% QoQ) and ACIE/non-hyperscale ($40B, +25% QoQ) both growing at a pace that would be considered exceptional for a company a tenth of Nvidia's size, let alone one already generating over $200B in annual revenue. Our price targets are built off the EPS path of $8.59 (FY2027), $14.04 (FY2028), $18.99 (FY2029), and $24.11 (FY2030), applying a 28x forward P/E that reflects a premium to the broader semiconductor group but a discount to Nvidia's own trailing multiple, justified by continued 55-70%+ growth, 70%+ gross margins, and a net cash position scaling past $500B by FY2029. This generates a FY2027 target of roughly $240, stepping up to $675 by FY2030, implying substantial upside from the current $214.79 price even after the stock's multi-year run, because consensus is still anchored to supply-constrained guidance rather than the underlying unconstrained demand signal management has flagged.

Thesis

1. **Demand Exceeds Supply by a Wide and Widening Margin** Management's explicit statement that unconstrained demand is tracking near 100% YoY against a 70% supply-constrained guide is the single most important data point in this thesis. This is not a company managing a demand problem; it is a company rationing allocation among hyperscalers, sovereigns, and enterprises who all want more than they can get. Every capacity addition, whether through HBM supply agreements, foundry capacity, or yield improvement on Vera Rubin, falls straight to the top line rather than facing price or share competition. This dynamic alone justifies a premium multiple versus a standard cyclical semiconductor franchise. 2. **Vera Rubin Ramp Is the Next Leg, Not a One-Time Event** The Q3 FY2027 guide of $108B ±2%, with Vera Rubin already representing ~20% of data center revenue in that quarter, shows the ramp is ahead of typical product transition timelines. Management has called this the fastest ramp in company history, with orders already in hand. Hyperscale growth, which decelerated sequentially as Blackwell matured, is explicitly guided to reaccelerate in Q4 and into FY2028 as Vera Rubin supply scales, meaning the growth algorithm resets rather than decays into FY2028. 3. **Full-Stack Lock-In Remains the Core Moat** NVLink rack-scale architecture, the CUDA software ecosystem, and integrated GPU+CPU+networking (Grace, Vera CPU, InfiniBand, Spectrum-X) create switching costs that extend well beyond the silicon itself. Competing with Nvidia increasingly means replicating an entire systems stack, not just matching a chip's FLOPS, which is why the company has sustained an estimated >90% share of data center AI accelerator deployments despite years of well-funded competitive attempts. 4. **Margin Trough Is Temporary and Self-Correcting** Gross margin compression toward 71-72% in Q4 FY2027 is a function of HBM/DRAM cost inflation that is itself a byproduct of AI-driven memory scarcity, not a competitive or pricing failure. Management has already announced pricing actions effective FY2028 Q1 designed to restore margins to 72-74%, meaning the margin trough is a known, bounded, and self-correcting event rather than a structural deterioration — the market should not extrapolate the trough quarter into a permanent de-rating. 5. **Financial Model Throws Off Enormous Optionality** FCF scaling from $170B in FY2027 to $360B by FY2030, alongside net cash growing from $161B to $766B over the same period, gives Nvidia enormous flexibility for continued buybacks, strategic M&A (e.g., the Groq LPU partnership), and vertical investment in networking and systems without diluting shareholders or taking on leverage. This balance sheet strength is rarely priced into semiconductor multiples, which tend to anchor on cyclical capital intensity assumptions that simply do not apply here. 6. **Market Mispricing Stems from Treating Nvidia as a Cyclical Semiconductor Stock** The stock trades at a multiple that still partially reflects historical semiconductor cyclicality, despite a business model now closer to a platform monopoly with recurring, multi-year committed capacity agreements from hyperscalers and sovereigns. As Vera Rubin ramps and FY2028 unconstrained demand becomes visible in bookings and guidance, we expect the market to re-rate the multiple closer to platform/software economics rather than commodity chip economics, which underpins our price target progression.

Risks

1. **Memory Supply Chain Is Now the Binding Constraint on Growth** HBM/DRAM pricing and availability directly cap Nvidia's revenue ceiling at a 70% growth guide despite ~100% unconstrained demand. If memory scarcity worsens faster than management's current assumptions, or if memory suppliers prioritize other customers, FY2028 growth could undershoot even the conservative 70% guide, directly impacting the EPS path underlying our price targets. 2. **Vera Rubin Execution Risk Is Real and Immediate** Management has called this the fastest ramp in company history, which cuts both ways: any yield issues, thermal/packaging problems, or customer pushback on pricing during Q3-Q4 FY2027 would materially impair the credibility of FY2028 guidance and could trigger a sharp de-rating given how much of the bull case rests on seamless execution through this transition. 3. **Customer Concentration and Hyperscaler Capex Cyclicality** A large proportion of Data Center revenue flows from a handful of hyperscale customers whose capex budgets, while currently expanding aggressively, are discretionary and could be pulled back if AI monetization disappoints at the application layer, creating a demand air-pocket that would be amplified by Nvidia's current supply-constrained backlog unwinding. 4. **Margin Recovery Depends on Successful Pricing Actions** The plan to restore gross margins to 72-74% in FY2028 depends on pricing increases being accepted by customers without demand destruction. If hyperscalers push back on ASP increases, or if HBM/DRAM inflation exceeds the magnitude management has already flagged as worse than prior expectations, margin recovery could be delayed or fail to materialize as modeled. 5. **Valuation Already Reflects a High Bar** Even using a 28x forward multiple rather than Nvidia's historical premium, the stock requires sustained 55%+ growth and margin recovery to justify upside; any deceleration beyond the guided path, competitive share loss to custom silicon (hyperscaler in-house chips) or alternative architectures (Groq-style LPUs, though currently a partner), would compress both the earnings base and the multiple simultaneously. 6. **Regulatory and Geopolitical Exposure** Export controls on advanced AI chips to China and other jurisdictions, along with broader geopolitical tensions affecting semiconductor supply chains (TSMC concentration risk, Taiwan exposure), represent tail risks that are difficult to quantify but could abruptly alter both supply availability and addressable market size independent of underlying demand strength.

📈 Price Targets