NVIDIA Management: 70% Revenue Growth Guidance Is Supply-Constrained, Actual Demand Is Much Higher
nashnova research
Nvidia confirmed to JPMorgan that its FY2028 70% YoY growth guidance reflects supply limits, not demand — unconstrained, growth could more than double. The bottleneck, not the market, is the ceiling.
Why isn't 70% growth the upper bound?
Nvidia's VP of investor relations Toshiya Hari stated plainly: 70% YoY growth is a supply-constrained expectation, not a demand ceiling.
This means → if supply-chain capacity fully materializes, actual growth could more than double YoY. The 70% figure is simply "how much we can ship."
Management disclosed this framework proactively to close a "meaningful gap" between market consensus and internal forecasts — without clarity, supply-chain partners could misplan capacity and throttle Nvidia's own deliveries.
What exactly is the bottleneck?
Hari named two critical constraints: advanced wafers and memory, the highest-weighted items on Nvidia's bill of materials.
In plain terms = the two things GPU production depends on most — TSMC's leading-edge node capacity and high-end memory from Micron, SK Hynix, and Samsung — are both undersupplied.
Nvidia is in deep discussions with all four suppliers, focused on improving supply availability. The pace at which these bottlenecks ease will directly determine whether the 70% baseline can be exceeded.
What does inference overtaking training revenue mean?
Hari offered his clearest revenue-mix signal yet: about eighteen months ago training and inference were roughly even; today inference has surpassed training, and the trend will continue.
This means → the AI industry is accelerating from the "build models" phase to the "use models" phase. The inference demand surge is Nvidia's new primary growth engine.
Management acknowledged, however, that precisely splitting the two is inherently difficult — Grace Blackwell, Nvidia's next-gen GPU platform, is highly versatile: a customer can train on it first, then redeploy the same hardware for inference, making external revenue attribution hard.
Is customer concentration still rising?
OpenAI and Anthropic together account for roughly 20% of Nvidia's business at the end-consumption level, expected to reach about 25% in FY2028.
The more telling figure sits at the direct-customer level: emerging cloud providers now account for over 50% of Nvidia's accelerated-computing and AI infrastructure ecosystem.
In plain terms = Nvidia's growth engine no longer depends on a handful of hyperscalers — a broad wave of newer players is buying AI infrastructure at scale. Customer concentration risk is declining.
Is the "circular financing" concern justified?
Management addressed market concerns about customer financing arrangements head-on: the facilities are modest in scale, capped, and backed by strong underlying demand and the creditworthiness of ultimate compute buyers.
This means → Nvidia positions these financing tools as instruments to front-load AI infrastructure demand, not leverage plays — essentially helping customers take delivery now and pay later.
Where does JPMorgan stand?
JPMorgan reiterated its Overweight rating with a $320 price target, implying roughly 43% upside from the September 2 close of $224.41.
This reflects JPMorgan's conviction that supply-chain improvements will gradually materialize, giving the 70% baseline room to be exceeded.
The pace of supply-chain relief — especially advanced wafer and memory capacity ramp — is the key variable for validating this thesis.
市场有风险，内容仅供研究参考，不构成投资建议。