NVIDIA, AMD, Intel, Google and Meta — rivals that rarely share a stage — have joined as founding members of the SGLang Consortium, a new group led by LMSYS Org to shape the roadmap of SGLang, the open-source inference engine that runs some of the largest AI deployments in the world. Why it matters: inference — running trained models to answer real requests — is where most AI compute is now spent, and the software layer that serves models decides how fast and how cheaply every GPU, TPU and accelerator actually performs.
November 13, 2026
What happened
LMSYS Org, the nonprofit behind SGLang, announced the consortium on November 12 in the opening keynote of the inaugural SGLang Summit in San Francisco, held November 12–13 at the Fort Mason Center for Arts & Culture. The founding members are NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk. Each member will name a technical representative and join two roadmap meetings a year, contributing engineering expertise, deployment feedback, testing resources or joint development. SGLang’s own maintainers keep responsibility for technical decisions and code acceptance, and the project stays open to the whole community.
The details
SGLang is a full-stack inference engine for frontier AI and part of the PyTorch ecosystem. According to LMSYS, it runs on hundreds of thousands of GPUs and powers some of the largest production inference deployments in the world, for companies including Meta, Microsoft, Google, Amazon, Thinking Machines Lab, Cloudflare, Reflection and Modal. The consortium’s joint work will focus on frontier model support, reasoning and multimodal workloads, distributed inference and hardware optimization. LMSYS cites Deloitte’s expectation that inference will account for roughly two-thirds of all AI compute in 2026, up from about half in 2025.
“The teams running SGLang at the largest scale see what inference needs next before anyone else does,” said Ying Sheng, co-author of the original SGLang paper and co-founder and CEO of RadixArk, speaking on behalf of the SGLang team. NVIDIA’s Ian Buck, vice president of hyperscale and high-performance computing, said better software helps developers get more performance from NVIDIA compute and that lower inference costs make AI more accessible.
Why it matters
When the three big chip rivals and the two biggest open-model publishers agree to coordinate around one open-source serving stack, that stack stops being a side project and becomes shared infrastructure. For everyone buying inference — which is now most AI companies — a better SGLang means the same models run faster on whatever hardware they already have, without waiting for new data centers. That is the “everything AI, tested” angle worth watching: efficiency gains in software are the cheapest capacity the industry can add this year, and this consortium is where many of them will be negotiated in the open.
Frequently asked questions
What is SGLang?
SGLang is an open-source, full-stack inference engine for frontier AI models, developed by LMSYS Org as part of the PyTorch ecosystem. It serves models in production at companies including Meta, Microsoft, Google and Amazon.
Who founded the SGLang Consortium?
LMSYS Org launched it on November 12, 2026, with founding members NVIDIA, AMD, Intel, Google, Meta, Cloudflare, Modal and RadixArk.
Who controls SGLang’s roadmap now?
Members advise through a standing forum and two roadmap meetings a year, but SGLang’s maintainers retain responsibility for technical decisions and code acceptance.
Sources: Business Wire (LMSYS Org announcement), LMSYS Org