Nvidia management explains: Where does the shocking 70% revenue guidance come from?

Nvidia Management Explains: Where Does the Shocking 70% Revenue Guidance Come From?

Nvidia has rarely provided the market with cross-year revenue forecasts, which has immediately attracted widespread attention.

J.P. Morgan stated in its latest report on September 2 that, according to recent discussions with Nvidia Vice President of Investor Relations and Strategic Finance, Toshiya Hari, this 70% year-on-year growth framework is not driven by a single factor, but is built upon the comprehensive acceleration in demand from hyperscale cloud providers, emerging cloud service providers, AI labs, sovereign AI initiatives, and enterprise customers. More importantly, management has made it clear that: Without supply constraints, the business growth rate could even double—the key variable currently limiting growth is supply, not demand.

The direct reason for Nvidia’s proactive disclosure of guidance through to fiscal 2028 is to bridge the significant gap between market consensus and internal company judgment. Management believes that if this gap is left unaddressed, it could present substantial challenges for planning among supply chain partners. This statement signals that the 70% growth target is seen by management as a well-supported "comfort zone" rather than an aggressive prediction, with significant upside far beyond what the market is currently pricing in.

Meanwhile, Nvidia has sent important signals across multiple dimensions such as customer composition, the proportion of inference business, supply bottlenecks, and financing arrangements, further sketching the chip giant’s medium-term growth outlook. J.P. Morgan maintains its Overweight rating on Nvidia with a target price of $320, implying approximately 43% upside from the current stock price of $224.41 (as of the September 2 close).

Supply, Not Demand, Is the Real Ceiling for Growth

According to the report, J.P. Morgan found through their discussion with Nvidia’s VP Toshiya Hari that management has expressed clear confidence in the 70% growth framework, but also emphasized its boundaries: this number is supply-constrained, not demand-constrained.

According to J.P. Morgan, Toshiya Hari pointed out during the conversation that, without supply limitations, Nvidia’s business growth rate could more than double year-on-year. This statement directly reveals the conservative nature of the current guidance—70% is the expected growth under realistic supply conditions, not the maximum demand the company could reach.

Management also specifically explained that the decision to provide a multi-year outlook partly stemmed from a "meaningful gap" between market consensus and the company’s internal view. Without proactive disclosure, this information asymmetry could lead to capacity planning mismatches among supply chain partners, which in turn would restrict Nvidia’s own delivery capabilities. In other words, this forward-looking guidance serves as both a signal to investors and proactive management of the supply chain ecosystem.

Advanced Wafers and Memory: The Two Core Bottlenecks in the Supply Chain

Regarding the specific composition of supply constraints, Toshiya Hari identified two of the most critical materials: advanced wafers and memory, ranking them as the two highest-weighted items in Nvidia’s bill of materials (BOM).

According to J.P. Morgan, Nvidia is currently maintaining deep communication with TSMC and the three major memory suppliers—Micron, SK Hynix, and Samsung— with the main focus on improving supply availability.

The bank believes that Toshiya Hari’s remarks indicate that Nvidia’s supply chain management has entered a phase of high-intensity proactive coordination, rather than passively waiting for capacity to ramp up. The pace at which these bottlenecks are alleviated will directly determine whether Nvidia can surpass the 70% growth benchmark in FY28. Improvements in the supply chain represent elasticity for financial performance.

Inference Business Share Continues to Expand, but Platform Fungibility Makes Quantification Difficult

The split between inference and training revenue has been a long-standing topic of market interest. In the discussion, Toshiya Hari provided the clearest directional judgment so far: about 18 months ago, training and inference each accounted for roughly half of revenue; now, inference has surpassed training, and this trend is expected to continue.

However, the report notes that management also pointed out the inherent difficulty in precisely quantifying this split. The reason lies in the high fungibility of Nvidia’s platform—for example, with Grace Blackwell products, customers can first use them for training workloads and subsequently switch the same hardware assets to inference tasks. This flexibility is a competitive strength for Nvidia’s platform but also limits the ability for external parties to break down the revenue structure with precision.

Customer Structure Remains Diversified, Emerging Cloud Providers Now Contribute Over 50%

Nvidia’s revenue sources are spreading from hyperscale cloud providers to a broader ecosystem. According to Toshiya Hari, OpenAI and Anthropic, the two leading frontier model builders, currently account for about 20% of Nvidia’s business from an end-consumer perspective, and this proportion is expected to rise to about 25% by FY28.

It is important to note that these figures reflect the share at the terminal consumption level, rather than among Nvidia’s direct customers—Nvidia typically sells computing power to hyperscale or emerging cloud providers, who then resell capacity to model builders.

At the direct customer level, the contribution from emerging cloud service providers (“neoclouds”) can no longer be ignored. According to J.P. Morgan’s report, emerging cloud service providers now account for more than 50% of Nvidia’s ACIE (Accelerated Computing & AI Infrastructure Ecosystem) business, indicating that Nvidia’s growth engine is no longer singularly dependent on a few major cloud providers, but is now jointly driven by a broader computing capacity construction and rental ecosystem.

Open Source vs. Closed Source: Nvidia’s Answer Is “Both Are Needed”

Regarding the market debate over the advantages of open source versus closed source large language models (LLM), Nvidia’s management has taken a clear stance: the two are not a zero-sum competition; the continued evolution of AI requires the coexistence and synergy of both open and closed source models.

Toshiya Hari stated that within Nvidia, closed-source models such as OpenAI and Claude are widely used, while for key tasks like chip design, a combination of closed and open source is adopted. The core logic of management is that: as long as model builders can commercialize and continuously improve their business models, demand will continue to flow through to chip suppliers such as Nvidia.

Additionally, management mentioned that model builders’ gross margin (GM) appears to be improving, in part thanks to the generation-over-generation per-token cost reductions enabled by Nvidia platforms. Improvements in model economics create a positive feedback loop for Nvidia chip demand.

Financing Arrangements Designed to Underpin Forward Demand—Management Refutes “Circular Financing” Concerns

Nvidia’s recently launched financing arrangements have caught market attention, and management systematically explained these in this discussion.

According to J.P. Morgan, these arrangements include: revenue-sharing agreements with some emerging cloud providers, the PORTS-Pike data center campus project, and a $500 billion private capital financing platform jointly set up with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, KKR, and other institutions.

For the revenue sharing agreements, Nvidia’s mechanism is: set a floor for computing power rental prices, and share upside revenue when the market rental rate exceeds the benchmark price, thereby creating recurring income opportunities in addition to selling core hardware.

In response to external concerns about “circular financing,” management emphasized that these financing arrangements are moderate in scale, have set caps, and are backed by strong underlying demand, ecosystem returns, and the credit of the ultimate capacity buyers. Nvidia positions these financing tools as a way to support forward demand for AI infrastructure rather than as a form of financial leverage.