Qualcomm said Wednesday it will acquire AI software startup Modular in an all-stock deal valued at nearly $4 billion, moving to strengthen the software layer behind its push into artificial intelligence systems for data centers, edge devices and heterogeneous computing environments.

The San Diego-based chipmaker said it reached an agreement to buy Modular Inc., a Silicon Valley AI software infrastructure company whose platform is designed to help developers build and deploy AI models across CPUs, GPUs, NPUs and custom accelerators without rewriting code for each chip architecture. Qualcomm said the acquisition is intended to deepen its software foundation for generative and agentic AI and support a broader strategy that stretches from personal devices to large-scale infrastructure.

In a Form 8-K filed with the U.S. Securities and Exchange Commission, Qualcomm said it entered into the definitive agreement on June 21 and expects to issue up to 19.2 million shares of common stock to Modular’s equity owners. Reuters calculated the transaction value at about $3.92 billion based on Qualcomm’s previous closing price. The filing said the shares would be issued through a private placement at closing, subject to customary conditions.

Qualcomm said the transaction is expected to close in the second half of 2026, pending applicable regulatory approvals. The company did not provide detailed revenue projections for Modular or specify immediate product-integration milestones, but framed the deal as a way to make AI deployment more efficient across a fragmented hardware market.

Modular was founded in 2022 by Chris Lattner and Tim Davis, engineers with backgrounds in major AI and compiler infrastructure projects. The company has drawn attention for its “write once, run anywhere” approach to AI software, including tools intended to improve portability across hardware vendors. WIRED reported that Modular has about 150 employees and that its team, including the cofounders, is expected to join Qualcomm.

The acquisition gives Qualcomm a more direct answer to one of the most important competitive advantages in AI computing: software lock-in. Nvidia’s CUDA platform has helped make its GPUs the default choice for many AI developers by giving them a mature programming environment and ecosystem. Qualcomm is not acquiring Modular simply for another developer tool; it is buying a potential bridge between its own silicon ambitions and a developer market that increasingly wants flexibility across chips, clouds and deployment locations.

Qualcomm’s announcement emphasized performance-per-watt, inference efficiency and hardware choice. Those themes reflect the changing economics of AI. Training large models remains compute intensive, but running those models at scale — inference — is becoming a major cost center for cloud providers, enterprises and device makers. Qualcomm argues that more efficient software can help customers translate hardware performance into lower operating costs, particularly as AI systems move across data centers, edge networks, PCs, vehicles and mobile devices.

“We believe the future belongs to developer-friendly, horizontal platforms that can run across diverse compute environments and give customers real choice in how and where they deploy AI,” Qualcomm Chief Executive Cristiano Amon said in the company’s announcement. Amon said the deal combines Qualcomm’s energy-efficient data center technologies with an open ecosystem approach.

Lattner, Modular’s cofounder and chief executive, said in the same announcement that Modular was created around the idea that AI needs a more open and efficient software foundation across hardware and deployment environments. He said joining Qualcomm would give Modular greater scale and platform reach.

The timing is significant for Qualcomm. The company remains best known for smartphone processors and wireless technology, but it has been trying to expand into automotive systems, PCs, connected devices and data center infrastructure. AI has become the organizing theme across those efforts. In smartphones and PCs, Qualcomm is promoting on-device AI. In data centers, it is seeking a role in inference workloads, custom silicon and energy-efficient compute systems.

That strategic shift is not without risk. Qualcomm is entering a market where Nvidia holds a dominant position, major cloud providers are building internal chips, and rival semiconductor companies are racing to offer their own AI accelerators and software stacks. Software portability is attractive to customers, but developer adoption can be difficult to change once workflows, libraries and model pipelines are built around established platforms.

Modular’s appeal is that it is not tied to a single chip vendor. Qualcomm said the startup’s unified platform supports AI workloads across multiple architectures, giving developers a way to build once and deploy across varied environments. Reuters reported that Modular has positioned itself as a neutral software layer that supports chips from Nvidia, AMD and other vendors. That neutrality could be valuable if customers want to avoid being locked into one supplier, but it may also require Qualcomm to maintain trust with developers who expect broad hardware support even after the acquisition.

The transaction also shows how acquisition strategies in AI are moving beyond model developers and chip designers into infrastructure software. For semiconductor companies, owning more of the software stack can help improve performance, reduce friction for customers and make hardware road maps more attractive. For startups, the AI infrastructure boom has created a path to large exits if their tools solve bottlenecks around deployment, efficiency or portability.

Qualcomm’s regulatory filing included the usual cautions that the acquisition could face timing uncertainties, regulatory conditions, purchase price adjustments and integration risks. It also noted that the final number of shares issued could be affected by closing purchase price adjustments. Those caveats are standard for a transaction of this size, but they matter because the deal is being funded with stock and is tied to a fast-moving market where product road maps can shift quickly.

For Silicon Valley’s AI ecosystem, the deal marks another sign that the infrastructure layer remains highly strategic. Modular’s exit would place a young startup inside one of the most important U.S. chip companies just as AI development is spreading across cloud data centers, enterprise systems and edge devices. Qualcomm is betting that the next phase of AI competition will reward companies that can combine efficient silicon with software developers actually want to use.