Qualcomm Incorporated has completed its acquisition of Modular Inc., adding AI-native software infrastructure to its portfolio of compute platforms spanning devices, edge deployments, and data centers.
Qualcomm announced the agreement on June 24 and closed on July 29, roughly five weeks later, ahead of the second-half-of-2026 window it guided to at signing. Neither company has disclosed terms in its releases. Qualcomm treated the close as a headline event rather than a footnote, listing it as one of three items at the top of its third-quarter fiscal 2026 results, published the same day.
Modular develops software intended to simplify the optimization and deployment of generative and agentic AI workloads across heterogeneous hardware. Its platform is designed to support CPUs, GPUs, NPUs, and custom silicon through a unified software approach, addressing a central challenge for enterprises deploying AI across varied infrastructure architectures.
The acquisition expands Qualcomm Technologies’ AI software capabilities alongside its existing portfolio of high-performance, energy-efficient compute. Qualcomm positions the combined organization to support personal, industrial, edge, and data center AI workloads, with emphasis on portability and performance optimization across different processor types.
Mojo, MAX, and Modular Cloud continue as products and brands with expanded investment and support through Qualcomm Technologies. Mojo is Modular’s Python-like language built for writing GPU kernels that stay portable across hardware. MAX is an inference and serving framework, and its design choice is worth noting: it does not depend on PyTorch, CUDA, or ROCm, so the same code runs on NVIDIA, AMD, and Apple Silicon without a vendor runtime to bundle, patch, or keep in sync. Modular says that the Python API, model pipelines, and GPU kernels are open source, with more than 500 models available. Modular Cloud extends the stack into managed deployment, offering shared or dedicated endpoints in Modular’s cloud or the customer’s own.
Chris Lattner, Modular co-founder and CEO, takes on the role of Executive Vice President of Advanced AI Software and Platforms. The rest of the team lands intact and, notably, Modular survives as a named organization: Modular’s own site now lists co-founder Tim Davis as SVP and GM of Modular at Qualcomm, along with Mostafa Hagog as VP of Engineering and Eric Johnson as VP of Product Management. Keeping a general manager over a named Modular unit is a stronger signal than the usual product-continuity language, since it means the brand has an owner inside Qualcomm rather than being absorbed into an existing software group.
Qualcomm said Modular will continue supporting an open, hardware-diverse AI software ecosystem rather than limiting optimization to a single class of accelerators. That approach is increasingly important as enterprises combine CPUs, GPUs, dedicated NPUs, and custom silicon to balance AI performance, power consumption, cost, and deployment flexibility.
“With Modular’s world-class engineering team, we’re enabling a new and open approach to AI software development, enabling AI to run efficiently across any hardware while maximizing performance,” said Cristiano Amon, President and CEO, Qualcomm Incorporated. “This solves one of AI’s biggest challenges, gives developers and customers genuine choice, and advances competition, innovation and resilience across the industry.”
The strategic logic sits in Qualcomm’s data center push. In its June release, Qualcomm argued that as AI scales, efficiency rather than capability becomes the constraint, since performance per watt drives inference cost, and cost determines what scales. It also said Modular would enable “optimal day-zero performance on new Qualcomm Technologies AI hardware,” which is the real prize: a software layer that lets new Qualcomm accelerators run production models on launch day rather than waiting out a framework-porting cycle. The financial context arrived the same week. Qualcomm reported third-quarter fiscal 2026 revenues of $9.9 billion and told investors it expects year-over-year growth in non-handset revenues, including data center, to accelerate from 24% in fiscal 2026 to more than 60% in fiscal 2027, on the way to $40 billion in total non-handset revenue by fiscal 2029. Buying the software layer is what makes that hardware target credible.