--- title: "Nvidia's free agent runtime covers the isolation Modal just priced at $15.75 billion" description: "At GTC Berlin, Nvidia's partner conference, the company launched OpenShell, a free open-source runtime that contains, monitors and enforces AI agents, and named Modal's four closest rivals as integration partners. Six days earlier, Modal had priced its paid version of that isolation at $15.75 billion." publisher: "The Inference" section: "Platforms" published: 2026-10-22T16:11:48.516Z modified: 2026-10-22T16:11:48.516Z canonical: https://theinference.org/article/nvidia-s-free-agent-runtime-covers-the-isolation-modal-just-priced-at-15-75-billion language: en keywords: "Nvidia, Modal, AI Agents, Agent Infrastructure, Cloud Computing, Security" --- # Nvidia's free agent runtime covers the isolation Modal just priced at $15.75 billion > At GTC Berlin, Nvidia's partner conference, the company launched OpenShell, a free open-source runtime that contains, monitors and enforces AI agents, and named Modal's four closest rivals as integration partners. Six days earlier, Modal had priced its paid version of that isolation at $15.75 billion. [Nvidia](https://theinference.org/markets/companies/nvidia) handed out for free the layer [Modal](https://theinference.org/markets/companies/modal) priced at $15.75 billion six days earlier [1][3]. At GTC Berlin, Nvidia's partner conference, the company launched OpenShell on October 21: free, open-source software that contains, monitors and enforces AI agents [1][3]. Six days before that, Modal, a US cloud platform that rents GPU capacity for AI workloads, [raised $750 million at a $15.75 billion valuation](https://cloudnews.tech/modal-tops-1-billion-in-annualized-revenue-raises-750m/), with Sandboxes, isolated environments where AI agents run code, as its fastest-growing product [2]. The runtime itself was bundled into Nvidia's Open Agent Safety Platform on September 28; what arrived on October 21 is [the full-stack reference architecture](https://decrypt.co/381570/nvidia-ai-agent-containment-tools-full-stack-blueprint) built around it [3]. ## Nvidia named Modal's rivals and skipped Modal [Nvidia's announcement](https://nvidianews.nvidia.com/news/openshell-reference-architecture) names [Baseten](https://theinference.org/markets/companies/baseten), Deepinfra, Fireworks and [Together AI](https://theinference.org/markets/companies/together-ai), inference providers that like Modal rent out compute, as building OpenShell Reference Architecture integrations [1]. Modal appears nowhere in the release [1]. The absence can be checked against the partner list; the reason for it cannot. What the record does show is Modal [building on Nvidia's hardware](https://theinference.org/article/modal-moves-300-million-of-ai-agent-work-to-nvidia-s-vera-cpu-after-its-own-tests-showed-up-to-2x-throughput): on October 19 it [published benchmarks of its Sandboxes on Nvidia's new Vera CPUs](https://modal.com/blog), two days before the OpenShell launch [4]. [A Form D](https://www.tradingview.com/news/reuters.com,2026:newsml_FWN4661PX:0-modal-labs-says-it-has-raised-601-9-mln-in-equity-financing-from-total-offering-amount-of-800-mln/), the SEC filing a private company makes for an equity offering, shows $601.9 million already sold [6]. The filing covers an $800 million offering, so the round is still filling [6]. ## Sandboxes are Modal's fastest-growing product Modal says Sandboxes alone earn $300 million a year [2]. It counts nearly 3 billion sandboxes created so far [2]. It says the product can scale to a million running in parallel [2]. All of that sits inside an annualized run rate, revenue projected from a recent period, that the company puts above $1 billion and has not audited [2]. The valuation works out to roughly 15 times that self-reported figure [2]. *Chart: **Sandboxes alone now earn what Modal's whole run rate did in May** Bar chart of Modal's company-reported annualized run rate of $300 million in May 2026 and above $1 billion in October 2026, next to the reported $300 million a year its Sandboxes product alone earns in October.* *Company-reported annualized run rate, $ millions per year, as stated by Modal in May and October 2026. All figures are self-reported and unaudited. Source: CloudNews. [2]* ## Free runtime at the top, paid storage underneath The money story is not the launch but where the value sits. OpenShell's own sandbox component does the same job Modal's Sandboxes sell: isolating untrusted agent code [1][3]. What it does not do is orchestrate. Modal's rebuilt scheduler, file system and capacity pool aggregated from more than 20 clouds stay Modal's [2]. If OpenShell becomes the default agent runtime, isolation risks becoming a commodity handed out by the GPU vendor, and the premium has to come from Modal's scheduling layer or move down a level. [VAST Data](https://theinference.org/markets/companies/vast-data), an enterprise storage company, [shipped with Nvidia an Agent Sandbox Foundation Stack](https://www.zawya.com/en/press-release/companies-news/vast-data-introduces-new-foundation-stack-for-secure-production-ready-ai-agent-sandboxes-with-nvidia-1843667) that sells exactly that layer below: zero-copy clones for sandbox workspaces, durable agent memory and audit records [5]. Enterprises get free governance. The storage landlord gets paid. ## The premium now rests on what OpenShell leaves out Two checkable things settle it. One: whether Modal joins a later OpenShell partner list, as its four rivals already have [1]. Two: how fast enterprises adopt it; Nvidia says the architecture is available now, but its Sentry hardware-enforcement framework is listed as arriving early next year [1][3]. The round's premium was priced against a self-reported revenue figure, not audited accounts. Free runtime now sits at the top, paid storage at the bottom, and the middle layer has to earn its keep. ## Takeaway Nvidia's free OpenShell runtime makes agent isolation a commodity and names Modal's rivals as partners, so the premium behind Modal's $15.75 billion round has to come from scheduling and storage instead. ## Sources 1. [NVIDIA Newsroom, NVIDIA Launches OpenShell Reference Architecture for Safe AI Agents, 21 October 2026](https://nvidianews.nvidia.com/news/openshell-reference-architecture) 2. [CloudNews, Modal Tops $1 Billion in Annualized Revenue, Raises $750M, 21 October 2026](https://cloudnews.tech/modal-tops-1-billion-in-annualized-revenue-raises-750m/) 3. [Decrypt, Nvidia Unveils Its Full-Stack Blueprint to Contain and Monitor AI Agents, 21 October 2026](https://decrypt.co/381570/nvidia-ai-agent-containment-tools-full-stack-blueprint) 4. [Modal blog, NVIDIA Vera CPUs double throughput for Modal Sandboxes, 19 October 2026](https://modal.com/blog) 5. [Zawya (VAST Data press release), VAST Data Introduces New Foundation Stack for Secure, Production-Ready AI Agent Sandboxes with NVIDIA, 22 October 2026](https://www.zawya.com/en/press-release/companies-news/vast-data-introduces-new-foundation-stack-for-secure-production-ready-ai-agent-sandboxes-with-nvidia-1843667) 6. [Reuters via TradingView, Modal Labs Says It Has Raised $601.9 Mln In Equity Financing From Total Offering Amount Of $800 Mln, 22 October 2026](https://www.tradingview.com/news/reuters.com,2026:newsml_FWN4661PX:0-modal-labs-says-it-has-raised-601-9-mln-in-equity-financing-from-total-offering-amount-of-800-mln/) --- Nvidia's free agent runtime covers the isolation Modal just priced at $15.75 billion — The Inference. Canonical: https://theinference.org/article/nvidia-s-free-agent-runtime-covers-the-isolation-modal-just-priced-at-15-75-billion