Qualcomm acquires Modular: Dominating AI at the edge

What happened?

Qualcomm, the US semiconductor giant based in San Diego, has announced the acquisition of Modular, a startup specializing in software for artificial intelligence chips, for approximately $3.9 billion, according to sources close to the deal. Modular, founded in 2022 by former Google engineers (Chris Lattner, creator of the LLVM compiler) and Apple (Tim Davis), had raised around $100 million in funding, with investors including GV (Google Ventures) and General Catalyst. The startup is known for developing a software ecosystem that allows AI models to run efficiently on different types of hardware, including CPUs, GPUs, and custom accelerators. Its flagship product, the MAX compiler, lets developers write code in Python and deploy it across multiple architectures without modifications, reducing reliance on proprietary platforms like NVIDIA's CUDA. The deal, expected to close in the coming months, represents a multiple of nearly 40 times Modular's total funding, reflecting high demand for AI startups from big tech companies.

Why is it important?

This acquisition is strategic for Qualcomm, which seeks to diversify beyond its traditional mobile phone chip business, which accounts for approximately 60% of its revenue. With the explosion of generative AI and edge computing, chips designed to run AI models locally (without relying on the cloud) are increasingly in demand. According to IDC data, the edge AI market is expected to grow from $12 billion in 2023 to over $60 billion by 2028. Modular brings key software that allows developers to efficiently deploy AI models on Qualcomm hardware, such as its Snapdragon processors, which already power over 2 billion devices worldwide. This positions Qualcomm to compete directly with NVIDIA in the edge AI market, which includes smartphones, laptops, cars, and IoT devices. Historically, Qualcomm has tried to diversify into servers and automotive, but with mixed results. The purchase of Modular represents a more aggressive move to capture value in the AI ecosystem, where software is as important as hardware. Additionally, Modular's technology is compatible with multiple architectures, which could allow Qualcomm to offer an open platform that attracts developers looking to avoid NVIDIA's vendor lock-in.

Market consequences

• Pressure on NVIDIA: Qualcomm strengthens its AI offering, challenging NVIDIA's dominance in AI hardware, especially in the on-device inference segment. NVIDIA has capitalized on the generative AI boom with its A100 and H100 GPUs, but its CUDA ecosystem is closed. Modular, with its open compiler, could erode that advantage if Qualcomm integrates it effectively into its chips. Moreover, Qualcomm already collaborates with Microsoft to run large language models like GPT-4 on devices with Snapdragon X Elite.
• Impact on startups: The purchase of Modular at a $3.9 billion valuation (nearly 40 times its total funding) signals big players' appetite for AI startups, which could spur more acquisitions. In recent months, companies like Snowflake, Databricks, and Amazon have made similar acquisitions in the AI space. However, it also raises concerns about market consolidation and reduced options for developers.
• Developer ecosystem: With Modular, Qualcomm can offer more powerful tools for developers to optimize their models for its chips, attracting more customers. Modular's MAX platform is already used by companies like OpenAI and Meta for testing, and its integration with Qualcomm could accelerate Snapdragon adoption in edge data centers. However, success will depend on Qualcomm maintaining Modular's open-source philosophy, which is not guaranteed.
• Regulation: The deal could face antitrust review, given Qualcomm's size and the strategic importance of AI. In 2019, Qualcomm faced fines for anticompetitive practices in the EU and US. The purchase of Modular could be seen as an attempt to control a key software layer, potentially drawing regulatory scrutiny, especially in Europe and China.

What readers should know

Modular does not manufacture chips; instead, it creates software that allows AI models to run efficiently on various hardware. Its technology, known as an 'AI compiler,' is similar to what CUDA does for NVIDIA, but open to multiple architectures. By acquiring Modular, Qualcomm gains not only talent but also a platform that can integrate with its own chips to offer complete edge AI solutions. This is crucial because the trend is to bring AI to devices, reducing latency and improving privacy. For consumers, this means smarter, faster devices with more capable AI assistants running locally. However, risks exist: integrating Modular could take time, and the startup culture might clash with corporate bureaucracy. Additionally, Qualcomm faces competition from other players like AMD, Intel, and Apple, which are also developing their own edge AI solutions. Compared to past events, such as Intel's purchase of Nervana (2016) or Arm's acquisition of Graphcore (2023), the Modular acquisition stands out for its software focus and high valuation, reflecting the urgency of big tech to secure their position in the generative AI era.

“Qualcomm's purchase of Modular is a bold move that redefines competition in edge AI. It's not just about hardware, but the software ecosystem that enables AI to run efficiently on any device.” — Analyst at TheVortiq