Jen-Hsun Huang personally stepped out to protect Nvidia: shattering triple negative rumors and embracing more value investors
[Huang Renxun personally stepped out to protect Nvidia: Defeating three negative rumors and embracing more value investors] Faced with endless negative rumors in the market, Nvidia CEO Huang Renxun and all senior executives went to the road show to fight back strongly, and produced solid data and product planning to prove: Nvidia's growth is far from the end.
Faced with endless negative rumors in the market, Nvidia CEO Huang Jenxun and all senior executives went to the road show to fight back strongly, and produced solid data and product planning to prove that Nvidia's growth is far from the end. Recently, NVIDIA CEO Jen-Hsun Huang attended Morgan Stanley's closed-door non-deal roadshow in California with his CFO and core executives, and faced four major questions that have emerged in the market recently: new product delays, self-developed ASICs eating into share, peaking growth, and a single customer structure. Morgan Stanley analyst Joseph Moore pointed out in the post-meeting report that the atmosphere of the entire meeting was positive, and management repeatedly emphasized the core conclusion: Even though quarterly revenue is close to the US$100 billion mark, the company's growth curve is still in an accelerating upward stage. Heavy negative rumors refuted: Rubin Ultra will ship as scheduled next year The first negative rumor that Huang Renxun refuted was a rumor that had been quietly circulated in the industry and was enough to impact the chip sector: NVIDIA's next-generation flagship Rubin Ultra architecture will be delayed until 2028. This once made the market worry that faulty product iterations would directly lose its leading advantage in computing power. In response to this rumor, Huang Jensen directly denied it on the spot and clearly finalized the timeline for normal shipment of Rubin Ultra next year, and there will be no postponement arrangements. Moore further disclosed the real details of the hardware adjustment: Rubin supporting system is indeed adjusting the rack supporting solution. The original Kyber rack is replaced by a new optimization solution. The original intention of the transformation is to adapt to the implementation of larger-scale supercomputing clusters. However, this adjustment only stays at the system architecture optimization level and will not change any product delivery nodes. He emphasized that the research and development progress of the two core underlying technologies, 800V high-voltage power supply and inter-rack optical interconnection, is fully in line with the early planning, and there is no substantial delay in the new generation product cycle. The second round of bad news was defeated: self-developed ASIC will not seize Nvidia’s share Previously, another anxiety in the capital market about Nvidia focused on the threat to market share: as cloud vendors such as Amazon and Google continue to launch self-developed ASIC chips, will Nvidia's GPU survival space continue to be squeezed? The customer data presented by Huang Renxun in this roadshow directly reversed the market's pessimistic expectations. Moore disclosed key cases in the investigation minutes: “A representative cutting-edge model that was previously mainly based on ASIC development and with extremely low participation from NVIDIA, now accounts for nearly 50% of NVIDIA’s computing power.” Although the report does not directly name the customer, combined with the two labels of "cutting-edge large model + heavily self-developed ASIC", the market unanimously speculates that this implies the AI start-up giant Anthropic. Its partner cloud vendor Amazon continues to develop self-developed Trainium chips and is Nvidia's core ASIC competitor. The management made a new industry judgment: the expansion of cloud vendors’ self-developed ASICs and Nvidia’s business growth are not a zero-sum game, and the two can develop simultaneously. This is because when customers purchase computing power, they do not simply compare the price of a single chip. The core evaluation criterion is the comprehensive cost per Token of the entire process of training and inference. Relying on the advantages of a complete software and hardware stack, NVIDIA's solution has a better single-token cost in most commercial scenarios, and firmly maintains competitiveness in the two core scenarios of training and inference. Past data also supports this logic. From 2024 to 2026, NVIDIA's overall share in the global AI computing market will not fall but rise. The third negative factor was cleared: three major growth curves broke out simultaneously The third negative thing that market institutions have for NVIDIA is that its performance is overly dependent on leading cloud vendors. In response to this concern, Nvidia dismantled three differentiated growth tracks on site, emphasizing the continued diversification of its customer structure and paying attention to effectively diversifying the risk of fluctuations in single customers. The first major track is the AI laboratory, which currently contributes about 20% of Nvidia's total demand. The world's leading large model R&D institutions continue to be deeply integrated into the NVIDIA platform, and customers like Anthropic, which once prioritized self-developed chips, are now continuing to increase NVIDIA GPU procurement, bringing stable new increases.
The second largest track is the traditional ultra-large-scale cloud vendors. Microsoft, Meta, Amazon, and Google together contribute half of the company's revenue and are the pillars of the basic market. However, the expansion of such enterprises is strictly constrained by the construction cycle of electricity, land, and computer rooms, and there is an obvious ceiling for growth. Nvidia no longer only sells GPUs to it, but also outputs Vera CPUs and high-speed network equipment simultaneously, broadening the revenue boundaries of a single customer. The third major track is the fastest-growing sovereign AI, emerging AI cloud and industrial enterprise markets. At present, countries around the world are vigorously building localized computing infrastructure due to data security and industrial independence needs. However, such projects are weakly affected by self-developed ASICs, and NVIDIA's integrated full-stack solutions are preferred. Huang Renxun emphasized at the roadshow that Nvidia has previously announced a CPU business target of US$20 billion for this fiscal year, nearly half of which comes from independent CPU racks. This means that Vera CPU is no longer just a management chip supporting GPU servers. With its exclusive architecture adapted to single-threaded workloads, Vera CPU has officially entered the huge general server market. Coupled with the interconnection needs brought about by the expansion of AI clusters, NVIDIA has transformed from a single GPU supplier to a full-stack AI infrastructure service provider covering GPU, CPU, network, and complete system. The importance of this roadshow: embracing value investors For a giant company like Nvidia with a market value of trillions of dollars, one of the important purposes behind the executive team’s choice to participate in the roadshow may be to promote a comprehensive switch in its capital strategy. As NVIDIA's market capitalization approaches US$5 trillion, its original investment group structure has reached a bottleneck: most growth fund holdings have reached the upper limit of individual stock holdings, and there is almost zero room for additional positions. To this end, the road show in which Huang Renxun and a group of senior executives participated may be aimed at proactively adjusting the focus of external communication, listing value investors as key communication targets, and broadening the circle of shareholders. The core support for this strategic switch to "embracing value investors" is Nvidia's strong operating cash flow. Institutional estimates indicate that the company will be able to use more than 50% of its cash flow for stock repurchases, dividends and other shareholder returns in the future. This characteristic is enough to allow Nvidia to have both high growth rate and stable cash flow defensive attributes of value stocks, and can perfectly balance the investment preferences of both types of funds. Growth concerns remain: the core problem lies in delivery Although management is confident about mid- to long-term growth, Morgan Stanley still fully warned four potential risks in its report: First, if the hardware supply is released much faster than expected, the growth of data center business will slow down significantly; second, AI R&D costs will plummet; third, competing products will launch disruptive computing power products to seize the market; fourth, cloud manufacturers will accelerate the implementation of self-developed customized chips and divert GPU orders. Despite this, Morgan Stanley's Moore still emphasized that this roadshow has released multiple positive surprises that exceeded expectations, which not only completely dispelled the market's pessimistic conjectures, but also introduced new growth logic and capital planning, which is expected to refresh the agency's judgment on Nvidia's mid- to long-term performance. Moore maintained the preferred target of Nvidia's semiconductor sector and gave a target price of US$288. Compared with the current stock price, there is 42% upside potential, and continued to give it an "overweight" rating. (
The second largest track is the traditional ultra-large-scale cloud vendors. Microsoft, Meta, Amazon, and Google together contribute half of the company's revenue and are the pillars of the basic market. However, the expansion of such enterprises is strictly constrained by the construction cycle of electricity, land, and computer rooms, and there is an obvious ceiling for growth. Nvidia no longer only sells GPUs to it, but also outputs Vera CPUs and high-speed network equipment simultaneously, broadening the revenue boundaries of a single customer. The third major track is the fastest-growing sovereign AI, emerging AI cloud and industrial enterprise markets. At present, countries around the world are vigorously building localized computing infrastructure due to data security and industrial independence needs. However, such projects are weakly affected by self-developed ASICs, and NVIDIA's integrated full-stack solutions are preferred. Huang Renxun emphasized at the roadshow that Nvidia has previously announced a CPU business target of US$20 billion for this fiscal year, nearly half of which comes from independent CPU racks. This means that Vera CPU is no longer just a management chip supporting GPU servers. With its exclusive architecture adapted to single-threaded workloads, Vera CPU has officially entered the huge general server market. Coupled with the interconnection needs brought about by the expansion of AI clusters, NVIDIA has transformed from a single GPU supplier to a full-stack AI infrastructure service provider covering GPU, CPU, network, and complete system. The importance of this roadshow: embracing value investors For a giant company like Nvidia with a market value of trillions of dollars, one of the important purposes behind the executive team’s choice to participate in the roadshow may be to promote a comprehensive switch in its capital strategy. As NVIDIA's market capitalization approaches US$5 trillion, its original investment group structure has reached a bottleneck: most growth fund holdings have reached the upper limit of individual stock holdings, and there is almost zero room for additional positions. To this end, the road show in which Huang Renxun and a group of senior executives participated may be aimed at proactively adjusting the focus of external communication, listing value investors as key communication targets, and broadening the circle of shareholders. The core support for this strategic switch to "embracing value investors" is Nvidia's strong operating cash flow. Institutional estimates indicate that the company will be able to use more than 50% of its cash flow for stock repurchases, dividends and other shareholder returns in the future. This characteristic is enough to allow Nvidia to have both high growth rate and stable cash flow defensive attributes of value stocks, and can perfectly balance the investment preferences of both types of funds. Growth concerns remain: the core problem lies in delivery Although management is confident about mid- to long-term growth, Morgan Stanley still fully warned four potential risks in its report: First, if the hardware supply is released much faster than expected, the growth of data center business will slow down significantly; second, AI R&D costs will plummet; third, competing products will launch disruptive computing power products to seize the market; fourth, cloud manufacturers will accelerate the implementation of self-developed customized chips and divert GPU orders. Despite this, Morgan Stanley's Moore still emphasized that this roadshow has released multiple positive surprises that exceeded expectations, which not only completely dispelled the market's pessimistic conjectures, but also introduced new growth logic and capital planning, which is expected to refresh the agency's judgment on Nvidia's mid- to long-term performance. Moore maintained the preferred target of Nvidia's semiconductor sector and gave a target price of US$288. Compared with the current stock price, there is 42% upside potential, and continued to give it an "overweight" rating. (