Qualcomm to Buy AI Software Startup Modular for About $3.9 Billion, Taking Aim at Nvidia's CUDA Moat
Qualcomm agreed to acquire Modular in an all-stock deal worth about $3.9 billion, expected to close in the second half of 2026.
Editor's Note ·
- Clarification:
- The article attributes several facts to Yahoo Finance, including the $3.92 billion Reuters last-closing-price calculation, the up-to-19.2-million-shares figure, and Cristiano Amon's quote. The cited Yahoo Finance URL returns a cookie-consent wall rather than the article when fetched, so the editorial source snapshot could not verify those facts against Yahoo Finance itself. Each of those facts is independently confirmed by other sources cited in the article: Tech Startups reports the $3.92 billion Reuters calculation and the Amon quote, and The Tech Portal reports the up-to-19.2-million-shares figure. The substance is accurate; only the Yahoo Finance citation is unverifiable.
Overview
Qualcomm has agreed to acquire AI software startup Modular in an all-stock deal worth about $3.9 billion, according to The Tech Portal. The transaction, announced on June 24, is valued at $3.92 billion based on a Reuters calculation tied to Qualcomm’s last closing price, as reported by Yahoo Finance. The deal pushes the mobile-chip maker deeper into the contest over AI infrastructure software, where Nvidia’s CUDA platform has long served as a competitive moat.
What We Know
Under the agreement, Qualcomm will issue up to 19.2 million shares of its common stock to Modular’s equity holders, according to Yahoo Finance, a structure also reported by The Tech Portal. The companies expect the transaction to close in the second half of 2026, pending regulatory and shareholder approvals, per Technobezz.
Modular develops software that runs AI models across chips without requiring developers to write separate code for each processor, technology primarily used to run, or “infer,” AI models, according to Yahoo Finance. The company created the Mojo programming language and the MAX AI platform, as reported by The Tech Portal. Technobezz describes MAX as a hardware-agnostic inference engine supporting CPUs, GPUs, NPUs, and custom ASICs, and notes that Modular already supports silicon from Nvidia, AMD, Intel, and Qualcomm. The startup has positioned itself as a neutral software layer for AI computing, supporting chips from Nvidia, AMD and other vendors, according to Yahoo Finance.
Modular was founded in 2022 and is based in Palo Alto, California, according to SDxCentral. Its founders are Chris Lattner and Tim Davis, as reported by The Tech Portal. Technobezz notes that Lattner, Modular’s co-founder and CEO, previously created Apple’s Swift language and the LLVM compiler and led Tesla’s Autopilot software. Qualcomm is expected to bring in Modular’s entire team of around 150 employees, according to The Tech Portal.
Qualcomm President and CEO Cristiano Amon framed the rationale around platform openness. “We believe the future belongs to developer-friendly, horizontal platforms that can run across diverse compute environments and give customers real choice,” Amon said, according to Yahoo Finance. Chris Lattner said the combination would extend Modular’s reach: “Joining Qualcomm gives us the scale and platform reach to accelerate that mission,” he said, per Technobezz.
The Funding Backdrop
The acquisition price marks a sharp step up from Modular’s most recent private valuation. The company raised $250 million in a Series C round in September 2025 that valued it at $1.6 billion, according to SDxCentral, bringing its total raised across three rounds to $380 million. That round was led by Thomas Tull’s U.S. Innovative Technology (USIT) fund, with DFJ Growth joining as a new investor, SDxCentral reported. Against that $1.6 billion mark, the roughly $3.9 billion deal represents a valuation increase of more than 140% in a short period, according to The Tech Portal.
At the time of the Series C, SDxCentral described Modular’s stack as amounting to a hypervisor for AI workloads, allowing them to run seamlessly across diverse silicon from Nvidia and AMD as well as emerging chip architectures. The platform comprises MAX, an inference framework; Mojo, a systems programming language; and Mammoth, a Kubernetes-native control plane, per SDxCentral, which reported the company’s claim of up to 70% latency reduction and 80% cost savings compared with vendor-specific runtimes like CUDA and ROCm.
The Machine Herald previously reported on Modular’s Mojo 1.0 beta in May, when the company declared the AI-native systems language feature complete.
What We Don’t Know
The deal remains subject to regulatory and shareholder approvals, and the companies have not detailed how Modular’s neutral, multi-vendor software will operate inside a chipmaker that competes with the silicon vendors Modular currently supports. Because the consideration is stock rather than cash, the final dollar value will move with Qualcomm’s share price between announcement and closing; the $3.92 billion figure reflects a calculation based on Qualcomm’s last closing price, according to Yahoo Finance. Qualcomm has not disclosed integration plans beyond retaining Modular’s team.
Analysis
The purchase is one of Qualcomm’s clearest moves to position itself as an AI infrastructure player rather than primarily a mobile chip supplier. Nvidia’s dominance in AI computing rests heavily on CUDA, the software platform that has helped lock developers into its ecosystem, as Tech Startups noted; Technobezz put the scale of that lock-in at roughly four million developers. By acquiring a vendor-neutral software layer designed to run AI workloads across competing silicon, Qualcomm is betting that owning the inference software, rather than only the chips, can give it a foothold in the data center market that Nvidia currently anchors.