Nvidia’s quarterly results once again exceeded expectations, but the most important strategic detail may be a product that sits outside its familiar GPU story. Chief executive Jensen Huang says the company’s Vera central processor opens a $200 billion market beyond the $1 trillion opportunity Nvidia already forecasts for Blackwell and Rubin AI GPUs from 2025 to 2027.
Huang expects Vera revenue to reach $20 billion by the end of the current fiscal year and become Nvidia’s second-largest sales contributor. The ambition gives the company a new position as the industry shifts from training the largest models toward serving them economically at scale.
Inference creates a second competitive front
Training frontier models remains a strong Nvidia domain, but inference—the continuous generation of answers for real users—offers competitors more room. Cost, latency and energy efficiency matter across billions of routine requests.
Google, Amazon and Microsoft are investing heavily in custom silicon while simultaneously purchasing vast amounts of Nvidia hardware. Google’s TPUs and Amazon’s accelerator programmes demonstrate why hyperscalers do not want one supplier to determine the economics of every AI service.
AMD and Intel also present processors as alternatives for selected inference workloads. Nvidia therefore needs to compete across the system rather than rely on GPU demand created during the training boom.
Vera is designed for that opening. The processor incorporates technology licensed from inference specialist Groq through a reported deal worth about $17 billion. Later in 2026, the full Vera Rubin platform is expected to combine the CPU with Rubin GPUs.
A platform strategy protects Nvidia’s position
Nvidia’s advantage has long extended beyond individual chips into software, networking and complete computing systems. Adding a proprietary CPU lets the company optimise more of the data path and offer customers an integrated architecture for training and inference.
That integration can improve performance and simplify procurement, but it also deepens customer dependence. Large cloud operators will compare the operational benefit of a complete Nvidia platform with the bargaining power and specialised efficiency offered by internal chips.
The company’s first-quarter revenue reached $81.62 billion, ahead of analyst expectations of $78.86 billion. It guided the following quarter to about $91 billion, again above forecasts. Those results provide the resources to enter adjacent markets aggressively.
Nvidia also announced an $80 billion share repurchase and increased its quarterly dividend. Investors nevertheless sent the shares down in extended trading, reflecting how consistently strong performance is already embedded in expectations.
Supply may decide how quickly the strategy works
Huang expects the Vera Rubin platform to remain supply-constrained throughout its life. Nvidia’s supply commitments rose from $95.2 billion to $119 billion in one quarter as the company prepared for demand and a tightening memory market.
That commitment expresses confidence, but it also creates risk. Processors, advanced packaging, memory and networking components must arrive together. A shortage in one part can limit revenue for the entire system.
Customers building data centres also face electricity, cooling and construction constraints. Chip availability alone does not guarantee that projected capacity can enter production on schedule.
The larger investor question is whether AI infrastructure spending remains durable into 2027 and 2028. Hyperscalers are expected to invest more than $700 billion in 2026, while a growing group of AI-specific cloud providers is expanding rapidly. Custom silicon and efficiency improvements could still change where that money goes.
Vera is Nvidia’s answer to a market where inference becomes the dominant daily workload and buyers seek alternatives. It turns the company from a GPU supplier into a broader processor platform at the moment its largest customers are attempting the same integration for themselves.
The opportunity may indeed reach $200 billion, but capture depends on more than benchmark performance. Nvidia must deliver the full system in volume, prove attractive inference economics and persuade powerful customers that an external integrated platform remains more valuable than their own silicon independence.
