Realtime AI News
AI Innovators Adopt NVIDIA Vera — A New CPU Category Built for the Agentic AI Era
NVIDIA has positioned Vera as a new class of ‘max single-threaded CPU at scale’ designed for agentic AI workloads, and AI innovators including Perplexity have adopted it. The CPU optimizes for tool calling, code execution, and data processing in AI factories, with a roadmap extending to the Rosa CPU and Rigel core.
NVIDIA VP Ian Buck published a blog post announcing that NVIDIA Vera represents a new category of CPU — the max single-threaded CPU at scale — built specifically for the agentic AI era. AI innovators including Perplexity have adopted Vera for their AI factory workloads.
In the creation and deployment of agentic systems, the CPU sits on the critical path for reasoning, response time, and learning. CPUs execute the work AI models command: tool calling, code execution, data processing, KV-cache management, and result analysis. For AI factories, GPU utilization is the most valuable resource, and any time spent waiting for CPU tasks to complete constrains revenue.
Traditional data center CPUs have evolved toward high core counts optimized for cloud economics, minimizing cost per rentable core at the expense of per-core performance. The move to chiplet architectures further introduced a ‘chiplet tax,’ where cores cannot access the full memory performance of the chip. Vera was designed differently.
Vera delivers: strong per-core performance under load, enough memory bandwidth per core to keep active cores supplied with data, and predictable latency. Every core can finish its task at full performance without interference from other cores.
Agentic work differs fundamentally from traditional workloads. Traditional CPU work is intermittent and user-driven. Agentic work is persistent and parallel — swarms of agents running continuous loops where each step depends on the previous result. Per-core speed determines how fast the loop advances; more cores cannot make any single task run faster.
NVIDIA has also outlined its CPU roadmap, including the NVIDIA Rosa CPU with its Rigel core, signaling long-term commitment to this new CPU category.
The adoption by Perplexity and other AI innovators validates market demand for CPUs architected specifically for agentic AI workloads, as CPU performance increasingly becomes a bottleneck for AI factory efficiency.
Why it matters
NVIDIA Vera establishes a new CPU category purpose-built for agentic AI, potentially reshaping data center CPU design and directly impacting AI factory GPU utilization and agent response times.
Nearby Updates
All07/07, 23:00
First AI agent and risk prediction model unveiled for precision diabetes management
Researchers have introduced the first AI agent combined with a risk prediction model specifically designed for precision diabetes management. The system integrates autonomous AI decision-making with personalized risk forecasting to provide dynamic, individualized care recommendations for diabetes patients.
07/07, 22:56
Russia's Sber Releases Open-Source MoE LLM GigaChat 3.5 Ultra
Sber, Russia's largest bank, has released GigaChat 3.5 Ultra, a new open-source large language model built on the Mixture of Experts architecture. The release continues Russia's push into open-source AI despite ongoing chip export restrictions.
07/07, 22:50
Nvidia Shares Dip Nearly 2% on Report DeepSeek Is Building Its Own AI Inference Chip
Shares of Nvidia fell nearly 2% after a report from International Business Times revealed that Chinese AI company DeepSeek is developing its own AI inference chip. The market reaction underscores growing investor concern about shifts in the AI chip competitive landscape.
07/07, 22:43
Trump administration puts brakes on OpenAI's newest AI model
The Trump administration has imposed restrictions on OpenAI's latest AI model, effectively halting or blocking its release or deployment. This marks the first time the U.S. federal government has directly intervened at the pre-release stage of a major AI company's product launch.