Nonprofit LMSYS Org has launched the SGLang Consortium with eight founding members, including rival chipmakers NVIDIA, AMD and Intel, to help steer the open-source SGLang engine used to run AI models at scale.

The group was announced on November 12 in the opening keynote of the first SGLang Summit in San Francisco, a two-day event (November 12–13) with more than 100 speakers. Alongside the three chipmakers, the founding members are Google, Meta, Cloudflare, Modal and RadixArk.

SGLang is a full-stack inference engine, the software that serves trained models to users, and is part of the PyTorch ecosystem. According to LMSYS, it runs on hundreds of thousands of GPUs and powers production deployments at companies including Meta, Microsoft, Google, Amazon and Thinking Machines Lab.

Each founding member will name a technical representative and attend two roadmap meetings a year. Work will focus on frontier model support, reasoning and multimodal workloads, distributed inference and hardware optimization. SGLang’s maintainers keep the final say on technical decisions and code, and the project remains open to everyone. LMSYS cites a Deloitte estimate that inference will take roughly two-thirds of AI compute in 2026, up from about half in 2025.

Why it matters

Most of the ongoing cost of AI now comes from running models, not training them. A shared, open engine that performs well on NVIDIA, AMD, Intel and other hardware lowers that cost and reduces lock-in to any single vendor. That is valuable for universities, startups and public data centers worldwide, including across the Arab region, that want to host open models locally on whatever hardware they can source.

Source: LMSYS Org

By the TechMojaz team