Nvidia Expands Its CPU Strategy as AI Agents Gain Momentum
Artificial intelligence is moving beyond chatbots that answer questions. A growing number of AI systems can now write and test code, search for information, use software tools, and complete multi-step tasks with limited human intervention. These systems, commonly called AI agents, are creating new demands on the computing infrastructure that powers them.
On October 19, 2026, Reuters reported that Modal, a cloud-computing startup that provides computing resources to AI developers, plans to introduce Nvidia's Vera CPU to host AI agents. The move marks a new opportunity for Nvidia to expand into a processor market traditionally dominated by Intel and AMD.
Modal previously relied on Intel and AMD processors for this type of workload. Its decision to add Vera suggests that specialized CPUs could become increasingly important as AI applications evolve from generating responses to carrying out longer-running tasks.
Read the original Reuters report.
What Is Nvidia's Vera CPU?
Nvidia's Vera is a central processing unit (CPU) designed for the computing tasks that support modern AI systems, particularly agentic AI and reinforcement learning.
While Nvidia is widely known for its graphics processing units (GPUs), which accelerate AI training and inference, Vera addresses another part of the process: executing code, coordinating tools, managing data, and keeping multiple computing environments running efficiently.
According to Nvidia's official announcement, Vera features 88 custom-designed Olympus CPU cores and memory bandwidth of up to 1.2 terabytes per second (TB/s). Nvidia says the processor can deliver up to 50% faster performance and twice the efficiency of traditional rack-scale CPU infrastructure under its stated comparison.
These figures are based on Nvidia's own performance claims and should not be interpreted as guarantees for every application. Actual results depend on the workload, system configuration, and comparison hardware.
Vera can operate in standalone CPU servers and work alongside Nvidia GPUs in accelerated computing systems. This flexibility allows cloud providers to use it for AI-related tasks that need substantial CPU resources rather than relying exclusively on GPU acceleration.
Why Do AI Agents Need More CPU Power?
An AI agent does more than generate text. It may need to interpret a request, create a plan, call external tools, execute code, inspect the results, and decide what to do next.
Consider an AI coding agent asked to fix a software bug. It might inspect a codebase, edit several files, run tests, analyze errors, and repeat the process until the issue is resolved.
The AI model handles much of the reasoning and generation, but the surrounding infrastructure must execute programs, manage files, coordinate tasks, and return results to the model.
This is where CPUs become important.
Traditional CPUs remain capable of running these tasks, but cloud providers serving thousands of concurrent AI workloads must also consider throughput, response time, memory bandwidth, and energy consumption.
A processor optimized for these workloads could help reduce delays between the individual steps of an AI agent's workflow. At large scale, even modest improvements in task execution can make a meaningful difference to the overall service.
Modal's Decision Gives Nvidia an Important Opportunity
Modal's decision to incorporate Vera is a notable development because it provides a real-world cloud deployment opportunity beyond Nvidia's traditional GPU business.
According to Reuters, early tests of Vera showed roughly double the task performance compared with the Intel and AMD processor models used in Modal's evaluation. Modal plans to use Vera for hosting AI agents while continuing to work with its existing CPU suppliers.
This distinction matters. Modal is not abandoning Intel or AMD. Instead, its approach reflects how cloud infrastructure companies can select different processors for different workloads, depending on performance and operating requirements.
Reuters also reported that Modal recently raised $750 million in funding at a $15.75 billion valuation. The company said usage of its services had tripled and its revenue had grown more than 15 times year over year.
The expansion of AI infrastructure creates an opportunity for companies like Modal to offer computing resources tailored to developers building increasingly complex AI applications.
For Nvidia, winning business from a cloud provider that previously relied on competing CPU platforms could help establish Vera as a serious option for agent-focused computing.
Vera vs. Traditional CPUs: What Is Different?
The key difference is not that traditional CPUs cannot support AI agents. They can. The question is whether a processor designed around the needs of agentic workloads can execute those tasks more efficiently.
Vera combines custom CPU cores with a high-bandwidth memory subsystem and connectivity designed to work with Nvidia's broader computing platform.
Its main advantages are intended to include:
Faster task execution: CPU-intensive work such as compiling code, running software tools, and analyzing results may complete more quickly.
High memory bandwidth: Moving data efficiently can help keep processor cores busy during demanding workloads.
Concurrent workload support: Cloud providers can run numerous isolated computing environments for different AI tasks.
CPU-GPU integration: Vera can coordinate work alongside Nvidia GPUs, helping the overall system process AI workloads more efficiently.
Nvidia's technical overview of Vera explains how CPU execution, memory access, and the coordination of multiple environments affect agentic AI performance.
However, the benefits will vary. A task that spends most of its time waiting for an external website or API may not become twice as fast simply because it runs on a different CPU. Performance improvements depend on which part of the workflow is actually limiting progress.
How Vera Could Affect AI Developers and Cloud Computing
If specialized CPUs prove valuable for AI agent workloads, cloud providers could begin offering more infrastructure options designed around the complete AI execution process.
Developers may benefit from faster code execution, more responsive tool use, and better performance when running many agents simultaneously. Businesses deploying AI agents for research, customer support, software development, or data analysis could also benefit if infrastructure improvements reduce operating costs or waiting times.
Cloud providers, meanwhile, must balance processor performance against hardware prices, power consumption, cooling, software compatibility, and workload requirements.
Vera's success will therefore depend on more than raw processor speed. Its performance in real applications, availability, and total cost of ownership will help determine whether customers adopt it at scale.
Nvidia is also positioning Vera as part of a broader infrastructure strategy rather than an isolated chip. The processor can be paired with Nvidia GPUs and integrated into data-center systems, potentially giving customers a more unified platform for managing AI workloads.
What Does This Mean for Intel and AMD?
Intel and AMD have long supplied CPUs for servers, cloud platforms, and enterprise computing. Nvidia's expansion into this market introduces another competitor, particularly for workloads associated with AI infrastructure.
The competitive pressure could encourage processor makers to improve performance, memory capabilities, and energy efficiency for AI-related computing.
However, Vera's emergence does not mean traditional CPUs will become obsolete. Intel and AMD processors continue to serve a broad range of applications, and different workloads may favor different hardware platforms.
Modal's decision to retain its existing suppliers while adding Vera illustrates a likely reality of modern data centers: companies will choose hardware based on the requirements of each workload rather than adopting a single processor for everything.
The broader competition is likely to focus on delivering the best combination of performance, efficiency, reliability, and cost for increasingly demanding AI services.
Challenges Nvidia Must Overcome
Despite the potential, several questions remain.
Cost and efficiency: Specialized hardware must demonstrate that performance improvements justify the cost of acquiring, deploying, and operating it.
Software compatibility: Developers and cloud operators need software tools, runtime environments, and deployment processes that work reliably with the new architecture.
Real-world performance: Results from selected benchmarks do not automatically translate into equivalent improvements across every AI agent or application.
Competition: Intel and AMD can continue improving their processors, while cloud providers may also develop custom silicon for their own workloads.
Availability and adoption: Long-term success will depend on whether more cloud providers and enterprise customers deploy Vera and achieve measurable benefits.
These factors will determine whether Vera becomes a widely adopted option for AI infrastructure or remains one of several specialized choices available to cloud operators.
The Bigger Picture: AI Infrastructure Is Evolving
The growing interest in AI agents is changing how technology companies think about computing resources. AI systems that perform multiple actions require infrastructure capable of supporting both model inference and the many operations surrounding it.
GPUs remain central to many demanding AI workloads, but CPUs still handle essential tasks such as executing code, managing data, coordinating tools, and running isolated environments.
Final Thoughts
Nvidia's Vera CPU is entering the market at a time when AI developers are building systems that do much more than answer questions. By focusing on the computing tasks behind AI agents, Nvidia is seeking to expand its role in the infrastructure required to run these systems at scale.
Modal's planned adoption is a meaningful development, but it is still too early to conclude that Vera will replace traditional server CPUs. Its long-term impact will depend on real-world performance, operating costs, and broader adoption by cloud providers.
For now, the key takeaway is clear: as AI agents become more capable, the computing infrastructure behind them is evolving too. The next phase of AI competition may be shaped not only by more powerful models, but also by the processors that help those models turn decisions into actions.
