Nvidia’s $100 Billion Revenue Still Accelerates: Jensen Huang Personally Steps In to Defend, Crushing New Product Delay and ASIC Encroachment Rumors

Prior to the roadshow, market rumors circulated that Nvidia's next-generation flagship Rubin Ultra architecture might be delayed for delivery until 2028, triggering industry concerns over a gap in product iteration.

Jensen Huang directly denied these rumors on-site, explicitly confirming the timeline for the normal shipping of Rubin Ultra next year. Moore added that while Rubin's supporting system is indeed adjusting its rack design—replacing the original Kyber rack with an optimized solution aimed at adapting to larger-scale supercomputing clusters—this is merely a system-architecture-level optimization. Key technologies, such as the 800V high-voltage power supply and inter-rack optical interconnects, are progressing completely on track, and product delivery milestones remain unaffected.

This statement directly dispelled market doubts regarding the pace of Nvidia's product iterations. Over the past few years, Nvidia has solidified its leading position in the AI computing field through the on-time delivery of products like Hopper and Blackwell. As a core platform designed for the era of AI agents, the on-time shipment of Rubin Ultra is crucial for maintaining market confidence.

Regarding market concerns over cloud service providers' self-developed ASICs eroding Nvidia's market share, the customer data disclosed during the roadshow provides a clear answer.

In the report, Moore mentioned that for a frontier model customer who previously relied primarily on ASIC development with minimal involvement from Nvidia, Nvidia's share of its computing power has now risen to nearly 50%. Although the report did not name the client, given the characteristics of "frontier large models + heavy reliance on self-developed ASICs," the market generally speculates that this refers to AI startup giant Anthropic—whose backer, Amazon, is the primary driver behind the self-developed Trainium chip.

During the roadshow, management emphasized that cloud providers' self-developed ASICs and Nvidia's business growth are not a zero-sum game, and that both can grow in tandem.

The core criterion for customers purchasing computing power is the comprehensive cost per token across the entire training and inference process, rather than the price of a single chip.

Leveraging the advantages of its complete hardware and software stack, Nvidia's solutions offer a lower cost per token in the vast majority of commercial scenarios, enabling it to remain competitive in both training and inference. Data show that from 2024 to 2026, Nvidia's overall share in the global AI computing market actually increased rather than decreased.

The market had previously been concerned about Nvidia's over-reliance on top-tier cloud providers for its performance. During the roadshow, management broke down three differentiated growth tracks, demonstrating a trend of diversification in its customer structure.

The first major track is AI laboratories, which currently contribute approximately 20% of total demand. In addition to top-tier large model research and development organizations remaining deeply integrated with Nvidia, clients like Anthropic—which previously prioritized in-house chip development—are also ramping up their GPU procurement.

The second major track comprises traditional hyperscale cloud providers—Microsoft, Meta, Amazon, and Google—which collectively contribute half of Nvidia's revenue. Nvidia is transitioning from a pure-play GPU supplier to a provider of full-stack solutions, including CPUs and networking equipment, thereby broadening its revenue frontiers.

The third major track is the fastest-growing segment: sovereign AI, emerging AI clouds, and industrial enterprises. Driven by data security and industrial autonomy needs, various countries are aggressively building out localized computing infrastructure. Such projects show a stronger preference for Nvidia's integrated solutions and are less vulnerable to the impact of custom ASICs.

Notably, Nvidia reiterated during its roadshow a $20 billion revenue target for its CPU business this fiscal year, with nearly half of that revenue coming from standalone CPU racks. This indicates that the Vera CPU is no longer merely a companion management chip for GPU servers; rather, with its proprietary architecture optimized for single-threaded workloads, it has officially penetrated the general-purpose server market.

Coupled with the connectivity demands driven by the expansion of AI clusters, Nvidia has transitioned from a pure-play GPU vendor into a full-stack AI infrastructure provider spanning GPUs, CPUs, networking, and complete systems.

As Nvidia's market capitalization approaches $5 trillion, existing growth funds are nearing their holding limits, leaving limited room to increase positions. This roadshow also signals a major upgrade in Nvidia's capital strategy—proactively shifting its communication focus to attract value investors and broaden its shareholder base.

Management disclosed that the company will allocate over 50% of its future cash flow to share buybacks and dividends. This move allows Nvidia to possess both high-growth attributes and the stable cash flow characteristics of value stocks, catering to the needs of different types of investors.

Morgan Stanley estimates that Nvidia's revenue will grow by 82% in fiscal year 2026 and 52.4% in fiscal year 2027, and is expected to reach $598.8 billion in fiscal year 2028.

Nonetheless, the report also flagged potential risks, including an excessively rapid release of hardware supply, a sharp decline in AI R&D costs, competitors launching disruptive products, and cloud service providers accelerating development of their own chips.

However, Moore stressed that Nvidia's main challenge at present is not a lack of AI demand, but how to overcome constraints such as memory, networking, power, and data center space to convert its massive order book into actual revenue.

This roadshow not only dispelled various market doubts about Nvidia but also showcased the company's full-stack capabilities in the AI infrastructure sector and its diversified growth potential. The "accelerated growth" signal personally delivered by Jensen Huang will undoubtedly reshape market expectations for Nvidia's medium- to long-term performance and consolidate its core position in the semiconductor sector.

This content was translated using AI and reviewed for clarity. It is for informational purposes only.