NVIDIA completed its Run:ai acquisition that it announced in Q2 2024. Along with closing the deal, and bringing the Run:ai team into NVIDIA, the companies discussed some future goals including supporting more accelerators and even open-sourcing the platform. For NVIDIA, this gives the company a broader platform to focus its efforts on a larger set of customers since it helps with the orchestration of AI clusters.
NVIDIA Completes Run:ai Acquisition
For those unfamiliar with Run:ai, the company helps with the orchestration of AI clusters. It is more advanced than the Inspur AIStation that we looked at in 2020 (and pre-trade exclusions.) There are a number of competing AI orchestration platforms, but with NVIDIA’s acquisition, Run:ai has had a leadership role.
As part of the release posted on the Run:ai site, the company said that the Run:ai products would continue under the NVIDIA brand. The company also plans to open-source the software at a later date. By open-souring the software, it should allow for other GPUs and accelerators to be enabled on the platform, albeit someone will have to do the work.
For NVIDIA, this makes a lot of sense since controlling the control plane tends to be a very profitable spot in the computing ecosystem. NVIDIA’s message that it is no longer selling just a single GPU, and instead is now selling vast clusters aligns with the desire to also own the control plane for those vast clusters. One of NVIDIA’s biggest advantages is that it has a massive software team, and this will help those teams tackle the challenges of orchestrating even more AI accelerators as they come online.
Final Words
There were a few large sites out there that proclaimed that NVIDIA had already open-sourced Run:ai. Those are incorrect. NVIDIA only plans to open-source it at a future date. Still, for NVIDIA, this could end up being really interesting if it ever finds that it is more profitable in the enterprise space if it plans to expand its NVIDIA Enterprise and NVIDIA DGX Cloud offerings.
Another idea would be that Nvidia merge this into Bright (which they bought a few years back)