Modal shifts AI sandboxes to Nvidia’s Vera CPU, claims double throughput
AI cloud startup Modal announced plans to transition its Sandboxes product from x86 chips to Nvidia’s Arm-based Vera CPU, citing double the throughput in internal tests. The company, which recently raised $750 million at a $15.75 billion valuation, aims to optimize its infrastructure for AI agent workloads.
Editor, Lazyfounder
30 SEC SUMMARY
- Modal, an AI cloud startup, announced plans to migrate its Sandboxes product from x86 chips to Nvidia’s Arm-based Vera CPU, claiming double the throughput in internal tests.
- The company recently raised $750 million at a $15.75 billion valuation, highlighting strong investor confidence in its AI infrastructure.
- Nvidia’s Vera CPU, designed for AI agents, features 88 cores and is already being tested by companies like Anthropic, OpenAI, and Perplexity.
- Modal’s Sandboxes product generates $300 million in annual revenue and is used to run isolated AI agent workloads.
- The transition to Vera requires infrastructure changes, as the chip uses Arm’s design instead of the traditional x86 architecture.
TABLE OF CONTENTS
KEY HIGHLIGHTS
- Modal plans to transition its Sandboxes product from x86 chips to Nvidia’s Arm-based Vera CPU, claiming the new chip doubles throughput in internal tests.
- The company raised $750 million at a $15.75 billion valuation, reflecting significant investor interest in AI infrastructure.
- Nvidia’s Vera CPU, unveiled in March, is designed for AI agents and features 88 cores.
- Modal’s Sandboxes product generates $300 million in annual revenue and runs isolated AI agent workloads.
- Modal received its first Vera server last week and tested it against chips from Intel, AMD, and Amazon, reporting superior performance in container-packed scenarios.
Modal’s shift to Nvidia’s Vera CPU
Modal, an AI cloud startup, announced plans to migrate its Sandboxes product from x86 chips to Nvidia’s Arm-based Vera CPU. According to The Next Web, the company claims the new chip doubles throughput compared to Intel servers in its internal benchmarks.
The Sandboxes product, which generates $300 million in annual revenue, provides isolated computing environments for AI agents to run code. Modal’s tests showed Vera outperformed chips from Intel, AMD, and Amazon, particularly when servers were packed with containers.
This transition marks Modal’s first move away from x86 architecture, requiring changes to its infrastructure and software image-building system. The company did not disclose when Vera-powered Sandboxes will be available to customers.
Nvidia’s Vera CPU and its role in AI infrastructure
Nvidia unveiled the Vera CPU in March as a chip designed specifically for AI agents. It features 88 cores and uses Arm’s architecture, diverging from the traditional x86 design. Nvidia has cited companies like Anthropic, OpenAI, and Perplexity as early adopters of the chip.
Modal’s decision to adopt Vera aligns with a broader industry shift toward specialized hardware for AI workloads. The company’s internal tests found that Vera maintained consistent performance even when servers were densely packed with containers, a scenario where rival chips hit performance ceilings.
The chip’s high memory bandwidth is credited for its efficiency in handling background coding agents and reinforcement learning tasks, where CPU performance can bottleneck GPU utilization.
Funding and market positioning
Modal recently raised $750 million at a $15.75 billion valuation, underscoring strong investor confidence in its AI infrastructure strategy. The funding round reflects growing interest in cloud-based AI agent platforms and the hardware that powers them.
The company’s Sandboxes product competes in a crowded market, with rivals like Baseten expanding their own sandbox offerings. Baseten acquired Blaxel in September to strengthen its position in this space.
Developing: what is not yet confirmed
The following is reported but has not been independently confirmed.
Modal’s claims about Vera’s performance improvements have not been independently verified. The company’s benchmarks were conducted internally, and while the results appear promising, external validation is needed to confirm their accuracy.
The company built a benchmark of 50 coding tasks to test the chip, including software compilation and file searches. While Vera often outperformed competitors, the margins were not always significant when tasks were run individually.
What this means
Lazyfounder analysis — our interpretation, not reported fact.
Modal’s move to Nvidia’s Vera CPU signals a broader trend in the AI infrastructure space: the shift toward specialized hardware optimized for agentic workloads. For founders and operators in the AI space, this highlights the importance of evaluating hardware not just for raw performance but for its ability to handle unpredictable, containerized workloads efficiently.
The decision to abandon x86 in favor of Arm-based chips also reflects growing confidence in Arm’s architecture for high-performance computing. However, the lack of independent verification for Modal’s performance claims means operators should approach these benchmarks with caution until third-party validation emerges.
For startups in the AI cloud space, Modal’s $15.75 billion valuation and $750 million funding round demonstrate that investors are still willing to bet big on infrastructure plays—provided the technology aligns with long-term trends like agentic AI and reinforcement learning.
Key takeaways
- AI infrastructure startups are increasingly adopting specialized hardware like Nvidia’s Vera CPU to optimize performance for agentic workloads.
- The shift from x86 to Arm-based chips requires significant infrastructure changes, including adjustments to software image-building systems.
- Investors continue to show strong interest in AI infrastructure, as evidenced by Modal’s $15.75 billion valuation and $750 million funding round.
- Internal performance claims should be treated with caution until independently verified, especially in competitive markets like AI cloud computing.
- For AI startups, evaluating hardware for unpredictable, containerized workloads is becoming as critical as assessing raw performance metrics.
FAQ
What is Modal’s Sandboxes product?
Sandboxes is Modal’s product that provides isolated computing environments for AI agents to run and execute code. It generates $300 million in annual revenue.
Why is Modal switching from x86 chips to Nvidia’s Vera CPU?
Modal claims Nvidia’s Vera CPU doubles throughput in its internal tests, particularly in scenarios where servers are packed with containers. The chip’s Arm-based design is optimized for AI agent workloads, which are often unpredictable and resource-intensive.
When will Modal’s Vera-powered Sandboxes be available to customers?
Modal has not disclosed a timeline for when the Vera-powered Sandboxes will be available to customers.
How does Nvidia’s Vera CPU compare to competitors like Intel and AMD?
According to Modal’s internal tests, Vera outperformed chips from Intel, AMD, and Amazon, particularly in container-packed scenarios. However, these claims have not been independently verified.
Related on Lazyfounder
Sources
- The Next Web · 2026-10-19
Modal says Nvidia’s Vera CPU doubles throughput for its AI sandboxes
This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.
About the author
Editor, Lazyfounder
Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.
More stories by Tarun MottliaGet the LazyFounder Brief
Startup, funding and AI news in a five-minute read. Join the early-access list.
