Nvidia's Vera CPU Lands Modal Order, Challenging Intel and AMD in AI ComputeIT之家 · IT业界T2Infra

IT之家 · IT业界 · 10/20/2026, 11:50:40

Nvidia's Vera CPU Lands Modal Order, Challenging Intel and AMD in AI Compute

By IT之家Original title: 英伟达 Vera CPU 获 AI 初创公司 Modal 订单，杀入英特尔、AMD 主场

78AI Score

Executive Summary

Cloud compute startup Modal announced its first adoption of Nvidia's Vera CPU to host AI agents, citing a doubling of task throughput in early tests. The company will deploy a hybrid cluster running Nvidia chips alongside existing Intel and AMD hardware for optimized scheduling. This move marks Nvidia's entry into the general-purpose computing market dominated by traditional x86 vendors, targeting the high-performance needs of long-running AI agent tasks.

SOURCE COVERAGEFull Original Report

Nvidia's Vera CPU Wins Order from AI Startup Modal, Entering Intel and AMD's Home Turf "960x640"

Cloud compute startup Modal provides compute rental services to AI developers. The company announced plans to adopt Nvidia’s new Vera central processing unit (CPU) for the first time to host AI agents—systems capable of executing long-running tasks on behalf of humans. Previously, businesses like Modal ran exclusively on Intel and AMD chips.

Modal completed a $750 million funding round led by Accel last week (Note: approximately 5.02 billion RMB at current exchange rates), reaching a post-money valuation of$ 15.75 billion (approximately 105.425 billion RMB). The company projects revenue could reach $1 billion (approximately 6.694 billion RMB) next year. Modal reported that its service usage has doubled, with year-over-year revenue growth exceeding 15 times.

On Monday local time, Modal stated that early tests showed Nvidia’s new chip could double task processing throughput for AI agents. Consequently, the company decided to integrate Nvidia’s chip solution for the first time, requiring deep modifications to its underlying software stack.

However, CEO Erik Bernhardsson specifically emphasized that the company will not abandon its existing suppliers. Modal plans to adopt hybrid compute clusters, deploying Nvidia chips in parallel with current Intel and AMD devices to schedule different tasks onto the most suitable hardware.

Bernhardsson explained that the benchmarks for this test were against commercially available Intel and AMD chips. Both of Nvidia’s competitors are developing next-generation products expected to narrow the performance gap with Nvidia.

“We will conduct tests once the new products launch,” he added, noting that the hybrid hardware architecture means the company can “deploy the best-performing hardware at any time.”

Nvidia has informed investors that its target for Vera chip sales this fiscal year is at least $20 billion (approximately 133.873 billion RMB). This marks Nvidia’s entry into a market valued at roughly$ 200 billion (approximately 1.34 trillion RMB). Nvidia previously stated that AI startups such as OpenAI, Anthropic, Oracle, and Perplexity also plan to adopt the chip.