Jensen Huang Leaves Three Major Challenges: Who Will Take the Helm, Whether CUDA Can Maintain Its Dominance, and How to Fulfill the $279 Billion Commitment

Author: Li Hailun, Tencent Technology

At future NVIDIA GTC conferences, it may never again be possible to see the familiar scene of "Jensen Huang" stepping onto the stage and waving to the audience.

On November 12 local time in the United States, Jensen Huang announced he was stepping down from NVIDIA, the company he had led for 33 years. NVIDIA's stock price plummeted nearly 14% intraday before narrowing its losses, ultimately closing down about 4.3%.

Andreessen Horowitz General Partner Martin Casado dubbed it "a dark day."

Since founding NVIDIA in 1993, Huang has virtually spearheaded every major transformation at the company. From early graphics processing chips to the CUDA computing ecosystem, and now to today's AI data center business, NVIDIA's ability to capitalize on successive waves of technological change is closely tied to Huang's judgment on technological directions and his long-term investments.

Especially during this wave of generative AI, Huang not only drove NVIDIA's transformation from a chip supplier into an AI infrastructure platform but also played a deep role in shaping product roadmaps, coordinating supply chains, and collaborating with major customers.

NVIDIA's current market position is built largely upon this operational strategy.

For NVIDIA, Huang's departure does not mean existing operations will be immediately impacted. After years of development, the company already possesses a mature technology team, product roadmap, and supply chain system; these advantages will not disappear simply because one executive leaves.

However, without a core decision-maker like Huang, can NVIDIA continue its high-growth trajectory of past years? What new tests will this company, positioned at the core of the global AI industry, face next?

The Successor Is Not the Helmsman

Colette Kress, who is currently serving as NVIDIA's interim CEO, has held the CFO role for over 13 years, participated in the entire transformation process of NVIDIA from a gaming GPU company to an AI infrastructure giant, and is well-versed in the company's financials, customer contracts, vendor arrangements, and capital expenditures.

She has also long co-hosted earnings calls with Huang, making her a familiar figure to Wall Street management.

KeyBanc Capital Markets analyst John Vinh believes Kress's appointment as interim CEO is a positive signal due to her lengthy tenure and deep understanding of NVIDIA's business and strategy. eToro chief analyst Josh Gilbert also considers Kress the obvious choice to stabilize the company's situation. She knows the business inside out and can speak on behalf of the company when the market needs leadership certainty.

However, this does not mean Kress can immediately take over Huang's original decision-making responsibilities. According to The Information, NVIDIA's interim CEO is not the driver of major corporate decisions, nor responsible for setting the agenda for those decisions. An interim CEO and a permanent leader are two different things.

Prior to this, The Information cited informed sources stating that Kress is not the primary driver of NVIDIA's major decision agenda. On some high-risk transactions, she prefers risk control, with the OpenAI credit support arrangement being a prime example.

Huang once pushed NVIDIA to provide large-scale credit support to OpenAI, with initial discussions reaching $250 billion intended to fund its data center construction. Kress took a cautious stance on this. Subsequently, the arrangement was split, with the first phase of NVIDIA's support coming in at approximately $100 billion.

Bernstein analyst Stacy Rasgon believes Kress's appointment will help stabilize short-term operations, noting no immediate concern that leadership changes will impact NVIDIA's business. However, he is more cautious about long-term issues: there remains significant uncertainty as to whether the new CEO can inherit Huang's control over product roadmaps and his relationship network with customers, suppliers, employees, and the broader industry ecosystem. Consequently, Rasgon 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 across chip design, software, systems, and finance. However, Huang's technological vision, industry influence, and cross-enterprise coordination capabilities are difficult for a standard professional manager to directly replace.

Among internal candidates, the most watched is NVIDIA senior technical executive Ian Buck. He joined the company in 2004, has long been involved in GPU general-purpose computing and CUDA ecosystem development, and currently oversees AI hardware and software products for hyperscale cloud providers.

Among external candidates, Arm CEO Rene Haas was mentioned by several analysts. He previously served as head of Computing Products Business at NVIDIA, joined Arm in 2013, and later became CEO. Additionally, Apple Executive Chairman Tim Cook, Apple Hardware Chief Johny Srouji, and Palo Alto Networks CEO Nikesh Arora were also named.

Investor Armina Rosenberg proposed a possibility: NVIDIA could consider having Kress oversee operations and finance, while Buck handles products and technology, forming a joint CEO-like management structure.

After all, reports indicate that Huang directly managed over 60 executives in the past, and NVIDIA's decision-making system has long revolved around him personally.

Can CUDA Be Defended?

Looking back at NVIDIA's history, one of Huang's most crucial roles was driving continuous investment in technological directions that had not yet yielded clear commercial returns.

The most representative case is CUDA.

In 2006, NVIDIA launched CUDA, enabling GPUs to handle general-purpose computing tasks beyond graphics rendering. In 2012, AlexNet, trained using NVIDIA GPUs, achieved a breakthrough in the ImageNet competition, establishing GPUs as a vital computational platform for deep learning. Since then, NVIDIA has continuously refined the CUDA toolchain and computing libraries, allowing developers to gradually build model training, optimization, and deployment workflows around this ecosystem.

This is also why competitors struggle to catch up with NVIDIA relying solely on chip performance. Even if enterprises procure GPUs or specialized chips from other vendors, they must readapt existing code, models, and software tools. For customers who have heavily invested resources into using CUDA, migration is far from a simple hardware swap.

However, AI computing methods are evolving, and CUDA must expand accordingly. Over the past few years, NVIDIA's advantage mainly lay in large model training; as the industry shifts toward large-scale inference, customers are increasingly focusing on per-token computation costs, GPU utilization rates, and response speeds. NVIDIA plans to launch Dynamo 1.0 in 2026, specifically designed for resource scheduling and request management in large-scale AI inference, attempting to extend competition from raw GPU performance to the entire inference system. It integrates with software like TensorRT-LLM and supports mainstream open-source inference frameworks.

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

However, this software moat is not unshakeable. Currently, AMD is refining the ROCm ecosystem, while Google TPUs and Amazon Trainiums are developing their own software stacks. Open-source tools like PyTorch and Triton are reducing developers' dependency on specific hardware platforms. As more models become adaptable to different chips through unified frameworks, the migration costs associated with CUDA may gradually decrease.

This is precisely the long-term challenge NVIDIA must face after Huang's departure. CUDA's success stemmed from a technical investment with a very long return cycle, whereas inference, agents, and physical AI could be the entry points for the next round of ecosystem expansion. Who will determine investment priorities in these areas, and who will drive the software platform to continuously adapt to new computing paradigms, will dictate whether NVIDIA can sustain the ecosystem advantages it has accumulated.

How Will the $279 Billion Commitment Be Fulfilled?

Another matter requiring market attention following Huang's departure is NVIDIA's massive supply and capacity commitments already made.

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

Distributed by fulfillment timeline, approximately $92 billion covers the remainder of FY2027, $87 billion covers FY2028, $88 billion covers FY2029, with the remaining ~$12 billion scheduled for a later period.

This $279 billion does not represent received customer orders, nor cash already paid by NVIDIA. It primarily corresponds to procurement and capacity arrangements made with suppliers and manufacturing partners to meet future demand. NVIDIA needs to lock in advanced processes, packaging, memory, and other critical resources in advance to ensure the scheduled delivery of Blackwell and subsequent products.

The issue here is that supply-side commitments may not always be fulfilled in sync with customer demand.

Large-scale data centers require a lengthy timeline from financing, site selection, and power provisioning to construction and deployment. Even if customers plan to purchase GPUs, procurement might be delayed due to power connection delays, datacenter construction setbacks, insufficient funding, or shifts in the deployment rhythm of next-generation products. NVIDIA also noted in its 10-Q filings that customers may delay or cancel purchases due to factors such as infrastructure readiness, capital availability, and technology adoption progress.

If customer demand materializes slower than expected while NVIDIA has already locked in upstream capacity, it could face inventory buildup, cancellation costs, and supply-demand mismatches. Even if alternative buyers are eventually found, shifts in product generations could affect selling prices and gross margins.

Another figure prone to confusion is NVIDIA's disclosed commitment of approximately $36 billion for AI cloud service agreements.

Such arrangements typically involve multi-year cloud service capacity partnerships and fall under a different accounting category than the $279 billion supply and capacity commitments. NVIDIA obtains cloud computing capacity through partners and may resell portions to third parties; as capacity is utilized, corresponding commitments are reduced. However, if external customer demand falls short, NVIDIA might still bear certain procurement obligations.

These arrangements reflect a shift in NVIDIA's business model: the company is no longer simply purchasing chips against orders and delivering products, but is increasingly involved in advance in allocating resources across the entire AI infrastructure. Huang excels at leveraging customer relationships and industry influence to drive project execution; his successor will need to balance sustaining growth with managing financial risks.

The Next Growth Engine?

In recent years, NVIDIA capitalized on the generative AI wave to become a primary beneficiary of global AI infrastructure construction. However, for a company whose quarterly revenue is already approaching $100 billion, maintaining high growth has become increasingly important. NVIDIA's most critical current growth source remains its data center segment. In Q2 FY2027, company revenue reached $96.221 billion, a year-over-year increase of 106%; data center revenue totaled approximately $89 billion, up 117% YoY, accounting for the vast majority of total revenue. The company provided a Q3 revenue guidance of $108 billion, indicating robust growth momentum.

Furthermore, as global tech giants continue to expand AI infrastructure investments, the Blackwell platform is undergoing mass deployment, and the next-generation Vera Rubin platform has begun entering production and shipping stages.

Additionally, NVIDIA is also exploring other new growth opportunities.

Among these, AI inference represents a key direction. As large models transition from training to scaled applications, enterprises require continuous model invocation to complete various tasks, and the widespread adoption of AI agents may further drive compute demand.

Compared to model training, inference workloads place greater emphasis on compute costs, energy consumption, and operational efficiency. NVIDIA is improving inference performance through next-generation chips and full-system architectures, hoping to further expand its market share in this segment.

Other fields such as robotics, autonomous driving, and industrial AI are also viewed by NVIDIA as future growth drivers. For instance, as more countries and regions seek to build sovereign and controllable AI infrastructure, sovereign AI is generating new chip procurement demand. Robotics and autonomous driving are expected to further expand the application scope of GPUs and related computing platforms.

However, whether these new markets can reach the scale of current data center operations remains to be validated by time.

Meanwhile, the competitive landscape facing NVIDIA is also shifting.

In the past, NVIDIA primarily competed with chipmakers like AMD. Today, major clients such as Google, Amazon, and Microsoft are developing their own AI chips to reduce compute costs and lessen dependence on a single vendor. Broadcom, for its part, participates in chip R&D for large tech companies through its custom ASIC business. While these products may not fully replace NVIDIA GPUs in the short term, they have already provided customers with additional options for certain compute workloads.

Under Huang's leadership, NVIDIA seized the growth opportunities presented by generative AI. Moving forward, amidst the constantly evolving competitive landscape of the AI industry, who will decide NVIDIA's next steps and whether they can lead the company to capture the next wave of growth opportunities will become the new test.