Nvidia 2026 Shareholder Meeting Preview: Can Stock Price Hit New Highs? How Blackwell, Vera Production Ramps Will Determine Future Revenue?

TradingKey - This Wednesday (June 24), NVIDIA (NVDA) will hold its 2026 annual meeting of stockholders online. The focus of this meeting will be the production ramp-up of Blackwell and the brand-new Vera architecture chips, the commercialization progress of the AI ecosystem, and capital return plans for its massive cash flow.

Looking back at last year's NVIDIA annual meeting of stockholders, the event conveyed several key takeaways: NVIDIA is entering the beginning of a 'decade-long AI infrastructure buildout cycle'; AI and robotics will be the two largest growth opportunities; and the era of robotics and autonomous driving has arrived. On the day of the meeting, NVIDIA's stock price surged 4.3% to close at a record high of $154.31, while its total market capitalization maintained the top spot.

Previously, NVIDIA announced that it would release a new generation of AI chips every year: launching the Blackwell architecture in 2024, an upgraded Blackwell Ultra in 2025, and a brand-new architecture platform featuring the Vera CPU and Rubin GPU in 2026, to meet the growing demand for model inference and training.

Resolving Blackwell and Vera Shortages: Key Signals of Capacity Ramp-Up

The Blackwell series is Nvidia's current flagship AI chip product line, with its key advantage being outstanding GPU matrix computing power. The product is the core computing platform the company is bringing to market for 2024–2025, serving as the most mainstream and powerful AI training and inference hardware used by tech giants globally. However, constrained by production capacity, the Blackwell series remains in short supply; although Nvidia has secured TSMC's (TSM) nearly 60% of its packaging capacity, it still cannot fully meet Nvidia's order volume.

Unlike Blackwell, Vera is a self-developed data center CPU chip. In terms of R&D progress, it is currently only in the comprehensive production ramp-up and trial production phase, with the first shipments to the initial batch of core customers expected in the second half of 2026. Jensen Huang previously noted at the Computex conference that Nvidia's Vera CPU will be even more popular than its GPU because of its critical role in processing information, stating, 'The Vera CPU will be our new major growth driver.'

These two chips will undoubtedly determine Nvidia's short-term prospects for the next few years. Jensen Huang predicted at the GTC developer conference that Blackwell and Rubin alone are expected to generate $1 trillion in revenue in 2026 and 2027. Currently, Blackwell is the absolute pillar supporting Nvidia's revenue. Its first-quarter fiscal 2027 results showed that data center revenue reached $75.2 billion, up 92% year-on-year and 21% quarter-on-quarter. This growth was primarily driven by the widespread adoption of Blackwell 300 products, InfiniBand, and Nvidia Spectrum-X Ethernet (supporting NVLink). Therefore, amid capacity constraints, close attention should be paid to the production ramp-up of both chips.

Nvidia CFO Colette Kress stated during the first-quarter fiscal 2027 conference call that despite extremely robust demand, physical shortages across the entire supply chain—such as HBM and advanced packaging—will remain the core constraint the entire industry must face over the next 18 months. Consequently, Nvidia's current revenue depends entirely on the speed of capacity expansion by upstream suppliers. In addition to HBM and advanced packaging shortages, Nvidia's current capacity constraints are also partly driven by shortages of other key components, such as optical elements and liquid cooling systems.

At this shareholder meeting, watch for several signals that could indicate potential capacity improvements: whether TSMC's CoWoS-L advanced packaging yield has stabilized; the initial wafer test yield for Vera and Rubin on the 3nm process; whether the waiting period from order to delivery has shortened; and whether Nvidia is considering or has already secured partnerships with suppliers beyond TSMC and SK Hynix, such as allocating some orders to other packaging and testing facilities in Taiwan.

Nvidia's AI Commercialization: AI Bubble or Commercial Closed-Loop?

Nvidia's AI commercialization essentially depends entirely on the capital expenditures of its customers, particularly hyperscale cloud service providers. Looking at specific business segments, given expectations that Nvidia's revenue growth over the next few years will be driven primarily by Blackwell and Rubin, attention should be focused on the potential revenue growth this segment may generate.

AI computing power is used for model training and inference, and the market now needs to focus on revenue guidance for the inference business. Although Nvidia's two core products can also be used for training, they are primarily designed for inference. Training can be simply understood as the process of AI companies developing models, while inference refers to the stage where models are actually deployed and utilized by customers—the former represents a continuous cost sink, while the latter generates revenue.

In Q4 of fiscal 2024, the company's management disclosed that over the past year, approximately 40% of Nvidia's data center revenue was derived from AI inference. This means that 40% of total data center revenue was generated by customers purchasing hardware and software for AI inference, while the remaining 60% came from AI training.

If the share of inference revenue is too low, it may indicate that customers are engaged in an AI arms race with no consumers ultimately paying for it. This is not a sustainable business model and would prove that AI is a massive bubble, which would eventually backfire on Nvidia. Conversely, a high share of inference revenue would suggest that Nvidia's customers have figured out a commercial closed loop in the AI industry, which would bring sustainable revenue to Nvidia.

Beyond hardware, Nvidia's pure software offerings will also play a key role in its AI commercialization process. Nvidia's software ecosystem is anchored by the CUDA software stack, NIM inference microservices, and the Omniverse industrial platform. Although software accounts for a small portion of total revenue, its high gross margins also contribute to the company's AI commercialization.