The Announcement That Intel and AMD Cannot Ignore
Buried beneath the headline data centre revenue figures in Nvidia's Q1 FY2027 earnings call was an announcement that may prove more strategically significant than any single quarter's results: Nvidia is aiming to become the world's leading CPU supplier. Chief Financial Officer Colette Kress confirmed that the company's new Vera central processing unit opens a brand new $200 billion addressable market for Nvidia, with every major hyperscaler and system maker partnering with the company to get it deployed. She anticipated $20 billion in total CPU revenue in the current fiscal year — a figure that, if achieved, would represent more first-year revenue from a new product category than any technology company has ever generated. Jensen Huang, speaking on the same call, pointed to the Vera CPU as a major new growth engine, describing physical AI — including robotics and autonomous systems — as the next wave of computing demand that would drive Vera adoption across a customer base that extends well beyond the data centre. The announcement represents the most significant expansion of Nvidia's competitive scope since it entered the data centre market with the Tesla GPU architecture in 2006.
Until now, Nvidia has led the AI boom with its graphics processing units that excel at the parallel mathematical operations necessary for training large language models and running inference workloads. The CPU market — historically dominated by Intel's x86 architecture in servers and AMD's EPYC processors in the cloud — is a fundamentally different competitive landscape. CPUs handle the serial processing, operating system management, and orchestration tasks that GPUs cannot efficiently perform. In a conventional AI computing cluster, a small number of CPUs direct the work of a large number of GPUs — making the CPU a control-plane device rather than a compute-plane device, with a corresponding difference in revenue intensity. Nvidia's Vera CPU changes this architecture by designing the CPU specifically for AI workload orchestration, with a memory bandwidth and interconnect specification optimised for the Blackwell and Rubin GPU generations that it will manage. The result is a CPU whose performance in AI computing contexts is designed to exceed Intel's and AMD's general-purpose server processors by margins that justify the displacement of a decades-old incumbent supply relationship.
The $200 Billion Market: How Nvidia Calculated the Opportunity
The $200 billion total addressable market figure that Nvidia's CFO cited for the Vera CPU warrants examination. The global server processor market — the primary market for high-performance CPUs — is currently approximately $25 to $30 billion annually, dominated by Intel and AMD. Nvidia's $200 billion figure is therefore not an estimate of the existing server CPU market. It is an estimate of the expanded market that AI infrastructure creates for compute orchestration: the CPUs required to manage the GPU clusters being built at every hyperscaler, the processors needed to orchestrate agentic AI workloads that interact with multiple systems simultaneously, and the computing infrastructure required for physical AI applications including robotics, autonomous vehicles, and industrial automation systems. Each of these application categories requires CPU capability that is qualitatively different from conventional server workloads — faster interconnects, more memory bandwidth, lower latency communication with accelerators — and Nvidia's Vera is designed from the ground up for these requirements rather than adapted from general-purpose server architectures.
The Vera Rubin platform — which combines the Vera CPU with the Rubin GPU — is designed to deliver up to a 10x reduction in inference token cost compared with the Blackwell platform, according to Nvidia's own benchmarks. A 10x cost reduction in AI inference changes the economics of AI deployment for every application category that is currently too expensive to deploy at scale. It makes AI economically viable for use cases — real-time fraud detection across billions of transactions, personalised medical imaging analysis, continuous quality control in manufacturing — where current inference costs make commercial deployment marginal or unviable. When inference costs fall by an order of magnitude, the number of commercially viable AI applications expands by more than an order of magnitude, creating demand growth that justifies the $200 billion market size estimate regardless of whether Nvidia captures the entire market or only a fraction of it.
What This Means for Intel and AMD: The Competitive Response Problem
Intel and AMD face a competitive challenge with Nvidia's Vera CPU that is structurally different from their historical competition with each other. When AMD competed with Intel for server CPU market share, the competition was primarily on price-performance within a shared architectural framework — both companies making x86 processors for the same workloads, with customers able to switch between them with software compatibility maintained. Nvidia's Vera CPU is not competing within the x86 framework. It is a custom ARM-based design, optimised for AI workload orchestration, sold as part of an integrated Vera Rubin compute platform that hyperscalers are already committing to deploy. The competitive advantage Nvidia brings is not price-performance on a CPU benchmark. It is the integration advantage of a CPU designed specifically to orchestrate Nvidia's own GPUs, with interconnect specifications, memory hierarchy, and software stack optimised for that specific combination.
Intel's response capability is constrained by the same structural challenges that have plagued the company for several years: manufacturing process delays that have left it behind TSMC in leading-edge node capability, a product portfolio weighted toward general-purpose computing applications that are growing more slowly than AI-specific workloads, and a financial profile that limits the R&D investment required to develop competitive AI-specific processor architectures. AMD is better positioned, with its EPYC server processors manufactured on TSMC's leading-edge nodes and its ROCm software stack providing a credible alternative to CUDA for some AI workloads. But AMD's competitive position in the AI CPU market depends on demonstrating that its general-purpose EPYC processors can orchestrate AI workloads as efficiently as Nvidia's purpose-built Vera — a comparison that Nvidia's integrated platform strategy is specifically designed to make unfavourable. The $20 billion in CPU revenue that Nvidia is targeting for its current fiscal year is, if achieved, more than AMD's entire annual data centre revenue. The CPU market just acquired its most formidable new entrant in decades.
The Physical AI Opportunity: Robots, Vehicles and the Next Computing Wave
Jensen Huang's identification of physical AI — robotics, autonomous vehicles, and industrial automation — as the demand driver for the Vera CPU beyond the data centre is the strategic framing that matters most for Nvidia's long-term competitive position. Data centre AI is, by definition, centralised: it happens in large computing facilities that can be equipped with Nvidia's complete integrated stack. Physical AI is distributed: it happens at the edge, in robots moving through factory floors, in vehicles navigating real roads, in autonomous systems operating in unstructured environments where latency and power consumption constraints make cloud-connected inference architectures impractical. The CPU that orchestrates physical AI workloads must be small enough to fit in a mobile platform, efficient enough to operate on battery power, and fast enough to process sensor data and generate control outputs in real time. Nvidia's Vera CPU, designed for the AI workload orchestration context that the data centre has created, is the platform from which the company intends to extend its AI computing dominance from centralised cloud infrastructure to distributed physical AI systems — a market expansion that, if successful, would make Nvidia's current data centre dominance look like the first chapter of a much longer story.