Jensen Huang stakes $200B on Vera, Nvidia's first agentic-AI CPU
Nvidia's CEO told an earnings call that a new CPU built only for AI agents opens a $200 billion market. Indian IT shops and sovereign clouds are listening.
The News
On Wednesday's earnings call, Jensen Huang told analysts that Nvidia has identified a fresh $200 billion opportunity sitting outside its dominant GPU franchise. The vehicle is Vera, a new CPU that Huang described as "the world's first CPU, purpose-built for agentic AI". Vera is already running at a $20 billion standalone annual sales clip this year, before the broader enterprise pitch fully lands.
The framing arrived as Nvidia posted $81.6 billion in revenue for the past quarter and guided to roughly $91 billion for the next. Those numbers already make Nvidia the largest chipmaker by sales in the industry's history. Huang argues that agent workloads will drag general-purpose silicon demand along behind every GPU cluster. "We're going to need a lot more CPUs," he told the call.
The wedge into the CPU market puts Nvidia squarely against Intel and AMD, the incumbents that have split data-centre processor revenue between them for two decades. Vera is positioned not as a thinking chip but as the executor: while GPUs run the model, the agents that call tools, draft emails and orchestrate workflows mostly live on CPUs.
Why It Matters
Huang is telling investors that the agent boom is not solely a GPU story. The mental model from 2023 onwards has been compute equals training equals Nvidia. Vera reframes deployment as a second leg of the same trade: every agent that Marc Benioff sells through Salesforce or Andy Jassy ships through AWS will, on Nvidia's reading, push CPU consumption higher rather than collapse it into a single chip.
That argument has teeth because the last time a Big Tech vendor tried to wedge a new CPU category into the data centre, it took years and a defensible workload to stick. AWS's Graviton needed half a decade and a cost story to crack double-digit share inside Amazon's own fleet. Vera starts from a different position, with software demand pulling the silicon rather than the other way around.
For Intel and AMD, the message is harsher than the headline number suggests. A $200 billion TAM is roughly four times the entire 2025 server CPU market they shared. If even a quarter of that materialises, Nvidia turns from a partner into a direct competitor inside the rack.
Indian Angle
For Indian IT services, Vera lands at a precise moment. TCS, Infosys, Wipro and HCL have built agentic-AI practices around the assumption that GPU capacity is the binding constraint. Huang's pitch flips that: client engagements running thousands of concurrent agents for procurement, claims processing and helpdesk pipelines will sit mostly on CPU cycles. That changes both the bill of materials and the procurement conversation with Nvidia's India sales team.
The other Indian audience is the sovereign-compute build-out. Yotta, Tata Communications and CtrlS have been buying Nvidia GPU kit under the IndiaAI mission, with the central government's compute fund already disbursing across approved bidders. A Vera-class CPU adds a new SKU to those tenders and could pull total ticket sizes upwards just as MeitY weighs the next phase of allocations. Indian startups in the agent layer, including Sarvam and Krutrim, will face the same choice when they scale beyond pilot deployments.
Currency makes the maths sharper. With the rupee around 84 to the dollar, every additional CPU sub-system in a Vera-anchored cluster lands as a heavier INR line item than the equivalent Intel or AMD build did two years ago. Indian CFOs signing off on agentic-AI rollouts are now buying into Nvidia's full stack whether they intended to or not.
FAQ
How big is Vera's existing run rate?
Huang told analysts the standalone Vera business is already on a $20 billion sales pace this year, before the wider agentic push lands. That places it ahead of several of Nvidia's earlier non-GPU bets, including networking, on a like-for-like timing basis.
How does the $200 billion claim compare to Nvidia's current revenue?
Nvidia reported $81.6 billion last quarter and guided to roughly $91 billion for the next. The $200 billion figure is a long-dated total addressable market for the Vera category, not a near-term revenue forecast, but it sits at roughly half of Nvidia's expected annualised run rate.
What does this mean for Indian IT services firms?
Engagements built around large agent fleets will shift cost towards CPU procurement. Firms with deep Nvidia partnerships, including TCS and Infosys, may see short-term margin pressure and a wider product stack to resell over a 12 to 24 month window.
Where can I read the original report?
TechCrunch's Julie Bort filed the full earnings-call write-up with Huang's framing of Vera and the revenue figures cited here.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.