NVIDIA has made a rare move by offering the market a multi-year revenue outlook, immediately capturing widespread attention.

In its latest report dated September 2, JPMorgan revealed that based on recent discussions with NVIDIA's Vice President of Investor Relations and Strategic Finance, Toshiya Hari, this 70% year-over-year growth framework does not stem from a single driver. Instead, it is built upon accelerating demand across hyperscale cloud providers, emerging cloud service providers, AI labs, sovereign AI initiatives, and enterprise demand. More critically, management made it clear that without supply constraints, the business could potentially grow twice as fast—the current limiting factor for growth is supply, not demand.

NVIDIA's decision to proactively disclose its forward-looking guidance for fiscal 2028 is directly driven by the need to bridge a significant gap between market consensus and the company's internal projections. Management believes that if this disparity persists, it could pose substantial challenges to supply chain partners' planning capabilities. This stance indicates that the 70% growth target represents a well-supported "comfort zone" in management's view rather than an aggressive projection, with implied upside potential far exceeding current market pricing.

Meanwhile, NVIDIA has released important signals across multiple dimensions including customer structure, inference business share, supply bottlenecks, and financing arrangements, further outlining the mid-term growth trajectory for this AI chip giant. JPMorgan maintains its Overweight rating on NVIDIA with a price target of $320, implying approximately 43% upside from the current share price of $224.41 (as of September 2's close).

Supply, Not Demand, Represents the True Ceiling for Growth

According to the report, JPMorgan's interactions with Toshiya Hari revealed that NVIDIA's management team demonstrates clear confidence in the 70% growth framework while also defining its boundaries: this is a supply-constrained figure rather than a demand-constrained one.

Per the JPMorgan report, Toshiya Hari indicated during discussions that if supply restrictions were removed, NVIDIA's business could potentially grow at a rate exceeding double year-over-year. This statement directly reveals the conservative nature of the current guidance—70% represents expectations under realistic supply conditions, not the demand ceiling the company could potentially reach.

Management also specifically explained that the decision to provide multi-year outlook partly stems from a "meaningful gap" between market consensus and internal observations. Without proactive disclosure, this information asymmetry could lead supply chain partners to misalign their capacity planning, which would in turn constrain NVIDIA's own delivery capabilities. In other words, this forward guidance serves both as a signal to investors and as proactive management of the supply chain ecosystem.

Advanced Wafers and Memory: The Two Critical Supply Chain Bottlenecks

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

According to the JPMorgan report, NVIDIA is currently maintaining in-depth communications with TSMC and three memory suppliers—Micron, SK Hynix, and Samsung—with core discussions all centered on improving supply availability.

The bank believes that Toshiya Hari's statements indicate NVIDIA's supply chain management has entered an intensive phase of proactive coordination rather than passively waiting for capacity release. The pace at which these bottlenecks ease will directly determine whether NVIDIA can achieve growth exceeding the 70% baseline in FY28. The room for supply chain improvement equates to the elasticity potential for performance.

Inference Business Share Continues to Expand, Yet Platform Fungibility Complicates Quantification

The revenue structure split between inference and training has long been a topic of market focus. Toshiya Hari provided the clearest directional assessment to date during the discussions: approximately 18 months ago, training and inference revenue contributions were roughly equal; currently, inference has surpassed training, and this trend is expected to continue.

However, the report states that management also pointed out the inherent difficulty in precisely quantifying the exact split between the two. The reason lies in the high fungibility of NVIDIA's platform—taking the Grace Blackwell product as an example, customers can first deploy it for training workloads and later repurpose the same hardware assets for inference tasks. This flexibility demonstrates the competitiveness of NVIDIA's platform but also limits external analysis of revenue structure breakdowns.

Customer Base Continues to Diversify, with Emerging Cloud Providers Contributing Over Half

NVIDIA's revenue sources are spreading from hyperscale cloud providers to a broader ecosystem. According to Toshiya Hari, OpenAI and Anthropic, two leading frontier model builders, currently account for approximately 20% of NVIDIA's business on an end-consumption basis, with this proportion expected to rise to around 25% by FY28.

Notably, these figures reflect the proportion at the end-consumption level rather than NVIDIA's direct customer structure—NVIDIA typically sells computing power to hyperscale cloud providers or emerging cloud service providers, who then offer computing capacity to model builders.

At the direct customer level, the contribution from emerging cloud service providers (neoclouds) has become impossible to ignore. According to the JPMorgan report, neoclouds currently account for over 50% of NVIDIA's ACIE (Accelerated Computing and AI Infrastructure Ecosystem) business, indicating that NVIDIA's growth engine no longer relies solely on a few major cloud providers but is collectively driven by a broader computing construction and leasing ecosystem.

The Open-Source vs. Closed-Source Debate: NVIDIA's Answer Is "Both Are Needed"

Regarding the market debate over the superiority of open-source versus closed-source large language models (LLMs), NVIDIA's management has taken a clear stance: the two are not mutually exclusive competitors, and the continued evolution of AI requires the coordinated development of both open-source and closed-source models.

Toshiya Hari stated that NVIDIA internally makes extensive use of closed-source models such as OpenAI and Claude, while employing a combination of closed-source and open-source approaches in critical tasks like chip design. Management's core logic is that as long as model builders can achieve commercialization and continuously improve their economic models, demand will continue to flow through to chip suppliers like NVIDIA.

Additionally, management mentioned that model builders' gross margins (GM) appear to be improving, partly attributable to NVIDIA's platform-driven reduction in per-token costs across generations. Improvements in model economics constitute a positive feedback loop driving demand for NVIDIA chips.

Financing Arrangements Aim to Support Forward Demand; Management Rebuts "Circular Financing" Concerns

NVIDIA's recently launched multiple financing arrangements have drawn market attention, and management provided a systematic explanation during this discussion.

According to the JPMorgan report, these arrangements include: revenue-sharing agreements with certain emerging cloud service providers, the PORTS-Pike data center campus plan, and a $500 billion private capital financing platform built together with institutions such as Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR.

Regarding revenue-sharing agreements, NVIDIA's mechanism works as follows: establishing a floor price for computing lease, and sharing upside gains when market rental prices exceed the benchmark, thereby creating recurring revenue options beyond core hardware sales.

In response to external concerns about "circular financing," management clarified that the relevant financing arrangements are moderate in scale, capped, and supported by strong underlying demand, ecosystem returns, and the creditworthiness of end computing purchasers. NVIDIA positions these financing tools as a means to support forward demand for AI infrastructure rather than financial leverage operations.