The numbers are staggering: $82 billion in quarterly revenue, up 85% year-over-year. But Nvidia’s latest earnings reveal something far more profound than another record-breaking quarter. We’re witnessing the complete overthrow of how computers actually work — and most people are missing it entirely.
Agentic AI has arrived, and it’s not just another incremental upgrade. It’s a fundamental rewiring of the relationship between thinking and doing in digital systems.
The Two-Layer Revolution That Changes Everything
For decades, computing followed a simple model: humans think, machines execute. AI added a layer of pattern recognition, but humans still made the decisions. That era is over.
Modern AI agents operate on two distinct hardware layers that mirror how human cognition actually works. The reasoning layer runs on GPUs — the same chips that have powered Nvidia’s dominance. But the execution layer, powered by Nvidia’s new Vera processors, is entirely different. As CEO Jensen Huang explained, “All of the thinking happens on GPUs. All of the orchestration essentially runs on CPUs.”
This isn’t just technical architecture — it’s a complete reimagining of what computers do. An AI agent processing a chargeback doesn’t just flag an anomaly and wait. It investigates, pulls records, drafts memos, and routes for review. The machine completes the entire workflow.
Consider the historical parallel: the shift from mainframes to personal computers didn’t just make computing faster — it fundamentally changed who could use computers and how. Similarly, agentic AI doesn’t just make existing processes more efficient; it eliminates entire categories of human involvement.
The $20 Billion CPU Coup
Here’s where the story gets truly remarkable. Nvidia disclosed $20 billion in expected revenue from Vera CPU sales this year — their first year selling CPUs at scale. To put this in perspective, AMD generated roughly the same amount in total CPU revenue over the past 12 months.
“This means NVIDIA, in its very first year of selling CPUs, is already on track to become the world’s largest CPU vendor — mirroring exactly what it did in networking, where it went from zero to larger than all competitors combined in just two years.” — @Aaronwei3n
This mirrors the IBM-to-Intel transition of the 1980s, when a new computing paradigm created entirely new market leaders overnight. But this transition is happening in months, not decades.
The key insight: agents don’t rent capacity like traditional cloud workloads. They need tasks completed as fast as possible at the lowest cost per action. Previous chips were designed for the economics of shared computing resources. Vera processors are built for the economics of completed work.
The Infrastructure Spending Explosion
Nvidia now separates its business into two groups, and the numbers tell a striking story:
- Hyperscale platforms: The traditional cloud providers enterprises rent from today
- AI-native infrastructure: AI cloud providers, on-premises deployments, governments, and industrial operators
The second group grew 31% in a single quarter. AI cloud revenue within it more than tripled year-over-year. The number of large AI-specific data centers has nearly doubled in 12 months.
Huang projects that AI infrastructure spending could reach $3-4 trillion annually by the decade’s end. Hyperscaler capital expenditure on AI alone is forecast to exceed $1 trillion in 2027. As he put it: “Compute is revenues. Compute is profit.”
This spending isn’t speculative anymore. When an AI agent completes a chargeback investigation that previously required two hours of human work, that’s immediate ROI. Infrastructure supporting agents doing real work is no longer discretionary spending.
Beyond the Data Center: Physical World Domination
The earnings revealed another crucial dimension: AI for physical operations generated more than $9 billion over the past 12 months. A partnership with Uber will deploy robotaxi technology across nearly 30 cities and four continents by 2028.
This echoes the Tesla playbook — start with a focused application (autonomous driving), then expand the underlying technology across entire industries. Nvidia is positioning itself as the computational backbone for any system that needs to think and act in the real world.
The Anthropic Signal and Enterprise Reality
The partnership with Anthropic provides concrete evidence of enterprise demand. Claude models are embedded in document review, financial analysis, and compliance workflows across financial services, legal, and commerce sectors. Critically, Nvidia had essentially no infrastructure relationship with Anthropic before this year.
“The amount of capacity that we’re going to bring online for Anthropic this year and next year is going to be quite significant.” — Jensen Huang
This represents a greenfield market expansion — new revenue from entirely new use cases, not just upgrades to existing systems.
What History Teaches Us About Computing Transitions
Every major computing transition follows a similar pattern:
- 1960s-1980s: Mainframes to minicomputers to PCs
- 1990s-2000s: Desktop to web to mobile
- 2010s-2020s: Cloud to edge to AI inference
- 2020s-2030s: AI inference to agentic execution
Each transition doesn’t just improve performance — it eliminates entire categories of human work while creating new ones. The companies that recognize and capitalize on these shifts become the dominant platforms of the next era.
The evidence suggests we’re in the early stages of the agentic computing era. Nvidia’s $82 billion quarter isn’t just a financial milestone — it’s confirmation that the infrastructure for autonomous digital workers is being built at unprecedented scale and speed.
The age of computers that think and act independently has begun. The only question now is how quickly enterprises will adapt to this new reality.
Published in Stream · Dispatch #360 · May 21, 2026 · 5 min read.
Reply to paolo@mont3.ch - every email gets a human answer within 24h.