Qualcomm announced on June 24, 2026, that it will acquire Modular, an AI infrastructure software startup, for approximately $3.9 billion in an all-stock deal. The acquisition was unveiled at Qualcomm’s Investor Day in New York alongside a new chip announcement, and is expected to close in the second half of 2026 pending regulatory approval.
What Modular Built
Modular was founded by Chris Lattner — the engineer behind Apple’s Swift programming language and the LLVM compiler infrastructure — to tackle a core problem: every time a team wants to run an AI model on a different chip, they must effectively rewrite their inference code from scratch.
Modular’s answer is the MAX inference engine and the Mojo programming language, a Python-compatible language designed for high performance that runs on CPUs, GPUs, NPUs, and custom AI accelerators without modification. The platform is often described as a write-once, run-anywhere layer for AI — similar to what Java promised enterprise software in the 1990s, but built from the ground up for modern AI workloads.
A Bet Against Nvidia Lock-In
By combining Qualcomm’s semiconductor manufacturing scale with Modular’s vendor-neutral software layer, the merged company aims to give enterprises a credible alternative to Nvidia’s tightly integrated hardware and software stack.
Qualcomm CEO Cristiano Amon called the deal a pivotal moment for the AI industry, combining Qualcomm’s silicon reach with an open ecosystem approach. Lattner, who joins Qualcomm, said the acquisition provides Modular the scale and platform reach to accelerate that mission.
At the same Investor Day, Qualcomm also unveiled the Dragonfly C1000, a data center processor with more than 250 cores and speeds above 5 GHz designed for agentic AI workloads. Meta has committed to deploying the chip beginning in the second half of 2028 under a multi-generation agreement — the first major public customer for the new design.
Why It Matters for Georgia
Modular’s hardware-agnostic approach is especially relevant for developers and research teams in countries that lack access to expensive Nvidia GPU clusters. Georgian AI teams — including researchers at universities and startups supported by the Georgia Artificial Intelligence Association (GAIA) — could run AI models on more affordable or locally available hardware using Mojo and MAX, without the costly step of rewriting inference code for each target chip. If Modular’s core components remain open source, as has been the case historically, the platform could meaningfully lower the cost of AI deployment for Georgian institutions and regional cloud operators.