The three challenges left by Jensen Huang: who will take the helm, can CUDA still be maintained, and how to fulfill the $279 billion commitment

Article | Li Hailun

Editor | Su Yang

In the future NVIDIA GTC conference, we may no longer see the familiar scene of "Jensen Huang" taking the stage and waving to the audience.

On November 12, local time in the United States, Jensen Huang announced his departure from NVIDIA, which he had led for 33 years. NVIDIA's stock price plummeted about 14% during trading, then narrowed its losses, ultimately closing down about 4.3%.

Martin Casado, a general partner at Andreessen Horowitz, referred to it as "a dark day."

Since founding NVIDIA in 1993, Jensen Huang has almost dominated every major transformation of the company. From early graphics processing chips to the CUDA computing ecosystem, and now to the AI data center business, NVIDIA's ability to seize several rounds of technological change is closely related to Huang's judgment on technological direction and long-term investment.

Especially in this wave of generative AI, Huang not only pushed NVIDIA to transform from a chip supplier to an AI infrastructure platform but also deeply participated in product roadmap formulation, supply chain coordination, and collaboration with large customers.

NVIDIA's current market position is largely built on this set of business strategies.

For NVIDIA, Huang's departure does not mean that existing business will be immediately impacted. After years of development, the company has established a mature technical team, product roadmap, and supply chain system, and these advantages will not disappear with the departure of a single manager.

However, after losing Huang as a core decision-maker, can NVIDIA continue its rapid growth of the past few years? What new challenges will this company, positioned at the core of the global AI industry, face next?

01 Successor Does Not Equal Helmsman

Currently, NVIDIA's interim CEO, Colette Kress, has served as CFO for over 13 years and participated in the entire process of NVIDIA's transformation from a gaming graphics card company to an AI infrastructure giant. She is familiar with the company's financial situation, customer contracts, supplier arrangements, and capital expenditures.

She has also long attended earnings call meetings with Jensen Huang and is a familiar face to Wall Street.

KeyBanc Capital Markets analyst John Vinh believes that Kress's appointment as interim CEO is a positive signal because of her long tenure and deep understanding of NVIDIA's business and strategy. eToro chief analyst Josh Gilbert also believes that Kress is an obvious choice to stabilize the company. She is familiar with the business and can represent the company when the market needs management to provide certainty.

However, this does not mean Kress can immediately take over Huang's decision-making responsibilities. According to The Information, the interim CEO is not the main driver of significant company decisions and is not responsible for setting the agenda for major decisions. An interim CEO and a long-term helmsman are two different things.

Previously, The Information cited sources saying that Kress is not the main driver of NVIDIA's significant decision agenda and tends to control risks in some high-stakes deals, with the OpenAI credit support arrangement being an example.

Huang had pushed NVIDIA to provide large-scale credit support for OpenAI, with initial discussions reaching $250 billion to support its data center construction. Kress took a cautious stance on this. Subsequently, the related arrangement was split, with the first part supported by NVIDIA amounting to about $100 billion.

Bernstein analyst Stacy Rasgon believes that Kress's appointment helps stabilize short-term business, and he is not worried that this management change will immediately affect NVIDIA's operations. However, he is more cautious about long-term issues: whether the new CEO can take over Huang's control over the product roadmap and his network of relationships with customers, suppliers, employees, and the entire industry ecosystem remains highly uncertain. Rasgon has thus downgraded NVIDIA's stock rating and target price.

Goldman Sachs analyst James Schneider also pointed out that NVIDIA does not lack a mature functional management team, with experienced leaders in chip design, software, systems, and finance, but the technical vision, industry influence, and cross-enterprise coordination ability that Huang possessed are difficult for an ordinary professional manager to directly replace.

Currently, among internal candidates, the most watched is NVIDIA senior technical executive Ian Buck. He joined the company in 2004 and has long participated in GPU general computing and CUDA ecosystem development, currently responsible for AI hardware and software products aimed at hyperscale cloud providers.

Among external candidates, Arm CEO Rene Haas has been mentioned by several analysts. He previously served as the head of computing products at NVIDIA, joined Arm in 2013, and later became CEO. Additionally, Apple Executive Chairman Tim Cook, Apple hardware head Johny Srouji, and Palo Alto Networks CEO Nikesh Arora have also been named.

Investor Armina Rosenberg proposed a possibility: NVIDIA could consider having Kress handle operations and finance, while Buck could be responsible for products and technology, forming a management arrangement similar to co-CEOs.

After all, more than 60 executives directly managed by Huang have been reported, and NVIDIA's decision-making system has long operated around him.

02 Can CUDA Be Maintained?

Looking back at NVIDIA's development history, one of Huang's most important roles has been to drive the company to continuously invest in technological directions that have not yet formed clear commercial returns.

The most representative case is CUDA.

In 2006, NVIDIA launched CUDA, enabling GPUs to handle general computing tasks beyond graphics rendering. In 2012, AlexNet, trained using NVIDIA GPUs, achieved breakthroughs in the ImageNet competition, making GPUs an important computing platform for deep learning. Since then, NVIDIA has continuously improved the CUDA toolchain and computing libraries, and developers have gradually built model training, optimization, and deployment processes around this system.

This is also why competitors find it difficult to catch up with NVIDIA solely based on chip performance. Even if companies purchase GPUs or dedicated chips from other manufacturers, they need to re-adapt existing code, models, and software tools. For customers who have already invested significant resources in using CUDA, migration is not simply a matter of changing hardware.

However, the way AI computing is changing, and CUDA must expand accordingly. In recent years, NVIDIA's advantage has mainly been in large model training; as the industry shifts to large-scale inference, customers are beginning to focus more on the computational cost per token, GPU utilization, and response speed. NVIDIA plans to launch Dynamo 1.0 in 2026, specifically for resource scheduling and request management for large-scale AI inference, attempting to extend competition from GPU performance to the entire inference system. It is also integrating with software like TensorRT-LLM and supporting mainstream open-source inference frameworks.

AI agents and robots have opened up new application spaces. Agents need to continuously call models, data, and external tools, while robots need to handle vision, spatial understanding, and real-time control. NVIDIA is extending its underlying computing capabilities to these scenarios through CUDA-X, the Isaac robotics platform, the Omniverse simulation environment, and the Cosmos model. In the future, the use of CUDA may no longer be limited to data centers but will also enter factories, vehicles, and robotic devices.

However, this software barrier is not unshakeable. Currently, AMD is improving the ROCm ecosystem, and Google TPU and Amazon Trainium are also developing their own software systems. Open-source tools like PyTorch and Triton are reducing developers' dependence on specific hardware platforms. If more and more models can adapt to different chips through a unified framework, the migration costs brought by CUDA may gradually decrease.

This is precisely the long-term issue that NVIDIA needs to face after Huang's departure. The success of CUDA comes from a long payback period of technological investment, while inference, agents, and physical AI may be the entry points for the next round of ecosystem expansion. Who decides the priority of investment in these directions and who drives the software platform to continuously adapt to new computing methods will affect whether NVIDIA can continue to leverage its accumulated ecological advantages.

03 How to Fulfill the $279 Billion Commitment?

After Huang's departure, another thing the market needs to pay attention to is NVIDIA's massive supply and capacity commitments.

NVIDIA's quarterly report as of July 26, 2026, disclosed that the total amount of supply and capacity commitments reached $279 billion, up from about $119 billion in the previous quarter, an increase of about $160 billion in a single quarter.

According to the distribution of fulfillment timelines, approximately $92 billion remains for the remainder of fiscal year 2027, about $87 billion for fiscal year 2028, about $88 billion for fiscal year 2029, and the remaining approximately $12 billion arranged for later periods.

This $279 billion is not customer orders that have already been received, nor is it cash that NVIDIA has already paid. It mainly corresponds to the procurement and capacity arrangements made by the company to meet future demand with suppliers and capacity partners. NVIDIA needs to lock in advanced processes, packaging, memory, and other key resources in advance to ensure that Blackwell and subsequent products are delivered as planned.

The problem here is that commitments on the supply side may not necessarily align with customer demand.

Large data centers require a long cycle from financing, site selection, power supply, to construction and deployment. Even if customers have planned to purchase GPUs, they may delay their purchases due to power access delays, obstacles in data center construction, insufficient financing, or changes in the deployment rhythm of new generation products. NVIDIA also warned in its 10-Q filing that customers might delay or cancel purchases due to infrastructure, funding, and technology adoption progress.

If the speed of customer demand fulfillment is lower than expected, while NVIDIA has already locked in upstream capacity, it may face inventory buildup, cancellation costs, and supply-demand mismatches. Even if it ultimately finds other buyers, generational changes in products may affect pricing and gross margins.

Another easily confused figure is the approximately $36 billion AI cloud service agreement commitments disclosed by NVIDIA.

These arrangements typically involve multi-year cloud service capacity collaborations and are not on the same basis as the $279 billion supply and capacity commitments. NVIDIA obtains cloud computing capacity through partners and may also resell part of that capacity to third parties; as capacity is used, related commitments will decrease accordingly. However, if external customer demand is insufficient, NVIDIA may still bear certain procurement responsibilities.

These arrangements reflect the changes in NVIDIA's business model: the company is no longer just procuring chips and delivering products based on orders, but is increasingly participating in the resource allocation of the entire AI infrastructure in advance. Huang excelled at driving project implementation through customer relationships and industry influence, while his successor will need to balance growth and financial risk control.

04 The Next Growth Engine?

In the past few years, NVIDIA has seized the wave of generative AI, becoming a major beneficiary of global AI infrastructure development. However, for a company whose quarterly revenue is approaching $100 billion, maintaining high-speed growth has become increasingly important. NVIDIA's most significant source of growth currently remains the data center. In the second quarter of fiscal year 2027, the company's revenue reached $96.221 billion, a year-on-year increase of 106%; data center revenue was approximately $89 billion, a year-on-year increase of 117%, accounting for the vast majority of total revenue. The company has provided third-quarter revenue guidance of $108 billion, with growth momentum still strong.

Additionally, as global large tech companies continue to expand their investments in AI infrastructure, the Blackwell platform is being deployed on a large scale, and the next-generation Vera Rubin platform has also begun to enter the production and shipping phase.

Moreover, NVIDIA is also looking for other new growth opportunities.

Among them, AI inference is an important direction. As large models transition from training to large-scale application, enterprises need to continuously call models to complete various tasks, and the proliferation of AI agents may further increase demand for computing power.

Compared to model training, inference business focuses more on computing costs, energy consumption, and operational efficiency. NVIDIA is enhancing inference performance through next-generation chips and complete systems, hoping to continue expanding the business scale in this market.

Fields such as robotics, autonomous driving, and industrial AI are also seen by NVIDIA as future growth directions. For example, as more countries and regions seek to build autonomous and controllable AI infrastructure, sovereign AI is bringing new chip procurement demands. Robotics and autonomous driving are expected to further expand the application range of GPUs and related computing platforms.

However, whether these new markets can reach the scale of the current data center business still requires time to verify.

At the same time, the competition NVIDIA faces is also changing.

In the past, NVIDIA primarily competed with chip manufacturers like AMD. Now, large customers such as Google, Amazon, and Microsoft are also developing their own AI chips, hoping to reduce computing costs and dependence on a single supplier. Broadcom is participating in the chip development of large tech companies through its custom ASIC business. These products may not fully replace NVIDIA GPUs in the short term, but they have already provided customers with more options for certain computing tasks.

Jensen Huang has led NVIDIA to seize the growth opportunities brought by generative AI. Moving forward, in the context of the constantly changing competition in the AI industry, who will determine NVIDIA's next steps and whether they can lead the company to seize the next round of growth opportunities will become a new test.