Nvidia’s AI Infrastructure Revolution: The Rise of Agentic AI and the Vera Processor

Artificial intelligence in business is undergoing a seismic shift. What was once a collection of isolated features—smarter search bars, fraud scores, and recommendation engines—is now evolving into a new paradigm: AI agents. These agents don’t just flag anomalies or suggest options; they take instructions, break them into actionable steps, and complete tasks autonomously. This transformation is driving a fundamental change in the compute infrastructure required to power AI, with Nvidia at the forefront of this evolution.

The Two Layers of AI Workflow

At the heart of this shift is a two-layer architecture: reasoning and execution. According to Nvidia CEO Jensen Huang, “All of the thinking happens on GPUs,” while “all of the orchestration essentially runs on CPUs.” This division of labor is critical. For example, an AI agent processing a chargeback might analyze evidence on a GPU, then pull transaction records, draft a response, and update a case log on a CPU. While the reasoning layer has been around for years, the execution layer—designed for speed and cost-efficiency—has only recently emerged as a distinct requirement.

Nvidia’s Vera processor is specifically built for this execution layer. Unlike traditional chips, which were designed to allocate processing capacity across multiple users, Vera prioritizes completing tasks as quickly and affordably as possible. “The economics of AI, of the future, is tokens per dollar,” Huang stated during Nvidia’s fiscal first-quarter earnings call. The company expects nearly $20 billion in Vera chip revenue this year, a market it has never addressed before.

Partnerships and Enterprise Demand

The demand for this new infrastructure is already evident. Nvidia’s partnership with Anthropic, a leading AI company, underscores the growing need for specialized compute. Claude models, embedded in workflows for document review, financial analysis, and compliance, are now supported by Nvidia’s infrastructure. “The amount of capacity that we’re going to bring online for Anthropic this year and next year is going to be quite significant,” Huang said, highlighting the scale of this collaboration.

This shift is also reshaping Nvidia’s business model. The company now separates its operations into two groups: one focused on hyperscale platforms and another on AI cloud providers, enterprises, governments, and industrial operators. The latter grew 31% in a single quarter, with AI cloud revenue tripling year over year. The number of large AI-specific data centers has nearly doubled in 12 months, signaling a broader industry pivot.

The Financial Implications

Nvidia’s financial results reflect this transformation. For the fiscal first quarter, the company reported $82 billion in total revenue, a 85% year-over-year increase. Data center revenue alone reached $75 billion, up 92% from the previous year. Within that, chip revenue hit $60 billion, a 77% surge. Networking revenue, driven by Nvidia’s Spectrum X platform, nearly tripled year over year.

Huang projected that AI infrastructure spending could reach $3 trillion to $4 trillion annually by the end of the decade. Hyperscaler capital expenditure on AI alone is forecast to exceed $1 trillion in 2027. “Compute is revenues. Compute is profit,” he emphasized, underscoring the profitability of this new era.

Expanding Beyond Data Centers

Nvidia’s influence extends beyond data centers. Its AI for physical operations—covering logistics robotics, warehouse automation, and autonomous vehicles—generated over $9 billion in the past 12 months. A partnership with Uber aims to deploy this technology across a robotaxi fleet in nearly 30 cities and four continents by 2028.

Meanwhile, the company’s networking business continues to grow. The Spectrum X platform now surpasses all competing ethernet networking businesses combined. On the consumer side, AI-capable laptops and workstations generated $6.4 billion in revenue, up 29% year over year, though demand for consumer devices dipped slightly due to higher memory prices.

Challenges and Considerations

Despite its success, Nvidia faces challenges. U.S. Export licenses for certain chips have been approved for Chinese buyers, but the company has excluded all China data center revenue from its forward projections due to unresolved import approvals. This highlights the geopolitical complexities shaping the AI industry.

Key Takeaways

• Nvidia’s Vera processor is revolutionizing AI infrastructure by prioritizing task execution over shared computing capacity.
• Agentic AI is transforming business workflows, with Nvidia leading the charge in providing the necessary compute power.
• Enterprise demand for AI infrastructure is growing rapidly, with AI cloud revenue tripling year over year.
• Nvidia’s financial results reflect this shift, with data center revenue surging 92% year over year.
• Geopolitical factors, such as U.S.-China trade dynamics, continue to influence the AI market.

The future of AI is not just about smarter algorithms—it’s about redefining the infrastructure that powers them. As Nvidia continues to innovate, its role in this transformation will be pivotal, shaping the next era of business and technology.