Key facts
- OpenAI and Broadcom jointly unveiled a custom AI chip
- Qualcomm to acquire Modular for $3.9 billion
- Move reduces OpenAI's reliance on Nvidia GPUs
- Modular specialises in AI development infrastructure
OpenAI and Broadcom have unveiled a custom AI chip, a landmark step in OpenAI's long-running effort to build its own semiconductor infrastructure rather than remaining wholly reliant on Nvidia's dominant GPUs. The announcement marks one of the most concrete moves yet by a leading AI software company into dedicated hardware design, a frontier that could fundamentally alter cost structures and competitive dynamics across the industry.
In the same news cycle, Qualcomm announced it will acquire Modular for $3.9 billion, a deal that signals the chip giant's intent to deepen its software and AI infrastructure capabilities. Modular, known for its work on AI development tools, represents a strategic complement to Qualcomm's existing hardware portfolio as the company positions itself for the next phase of AI deployment at the edge and in data centres.
The twin announcements come at a moment of intense global investment in AI-specific silicon. Custom chips tailored to the mathematical operations underpinning large language models promise significant efficiency gains over general-purpose processors, and every major technology company — from Google with its TPUs to Amazon with Trainium — is now pursuing this path. OpenAI's entry into the field, backed by Broadcom's manufacturing and design expertise, is seen as a direct challenge to Nvidia's near-monopoly on AI training hardware.
For India, which has flagged semiconductor self-sufficiency as a national priority and hosts a growing base of AI research and deployment activity, the rapid consolidation of AI chip capabilities among a handful of global players raises both opportunity and strategic urgency.