Reactor serves live, interactive video from servers on Nebius infrastructure in Kansas City. In a recorded demonstration, CEO Alberto Taiuti identifies the Nebius-hosted servers delivering the stream and says Reactor chose the location for low latency and proximity to users.
Nebius separately states that Reactor’s custom world-model inference stack runs on its infrastructure. Its webinar announcement outlines planned discussion of inference kernels, networking, managed Kubernetes and failure recovery. The webinar page places Reactor’s serving stack on Nebius compute and managed Kubernetes, with Reactor handling the application-facing runtime, streaming and session lifecycle. It describes model-context management, state snapshots and restoration, and scheduling execution across users.
Sapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius. It describes deployments as live or coming online across the US, Europe, Japan and Korea, collectively across those providers. Nebius also names Reactor among teams running AI workloads on its infrastructure, and its event schedule labels Reactor a customer.
Scaling world models and real-time video inference
14 Oct 2026 · nebius.com · Web page
Nebius places Reactor’s serving stack on its GPU compute and managed Kubernetes. Reactor handles the application-facing runtime, streaming and session lifecycle. The page describes model-context management, state snapshots and restoration, scheduling across users, WebRTC transport and routing informed by live network conditions.
Scaling world models and real-time video inference14 Oct 2026 · nebius.com · Web pageNebius places Reactor’s serving stack on its GPU compute and managed Kubernetes. Reactor handles the application-facing runtime, streaming and session lifecycle. The page describes model-context management, state snapshots and restoration, scheduling across users, WebRTC transport and routing informed by live network conditions.
Nebius Reactor CEO · live video deployment
16 Apr 2026 · Nebius · Video
At 2:28 and 7:03, Alberto Taiuti says the live video demo streams from Reactor servers on Nebius in Kansas City. He explains choosing Nebius for latency and proximity to users.
Nebius Reactor CEO · live video deployment16 Apr 2026 · Nebius · Video · Starts at 2:28At 2:28 and 7:03, Alberto Taiuti says the live video demo streams from Reactor servers on Nebius in Kansas City. He explains choosing Nebius for latency and proximity to users.
The Real-Time Inference Layer Powering World Models: Why We're Thrilled to Back Reactor | Sapphire Ventures Twitter | X Icon LinkedIn Icon Youtube Icon Menu close Sapphire Ventures
14 Oct 2026 · sapphireventures.com · Web page
Sapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius for its real-time inference platform. Deployments are live or coming online across the US, Europe, Japan and Korea, collectively across those providers.
The Real-Time Inference Layer Powering World Models: Why We're Thrilled to Back Reactor | Sapphire Ventures Twitter | X Icon LinkedIn Icon Youtube Icon Menu close Sapphire Ventures14 Oct 2026 · sapphireventures.com · Web pageSapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius for its real-time inference platform. Deployments are live or coming online across the US, Europe, Japan and Korea, collectively across those providers.
Nebius Reactor cloud-use announcement
28 May 2026 · Nebius · Web page
Nebius calls Reactor a partner and states that Reactor is building on Nebius. It describes Reactor’s platform as letting developers stream frontier world models into apps globally in real time in under 10 lines of code. Reactor’s embedded announcement describes pixels, audio and actions generated in real time in response to user inputs and the environment.
Nebius Reactor cloud-use announcement28 May 2026 · Nebius · Web pageNebius calls Reactor a partner and states that Reactor is building on Nebius. It describes Reactor’s platform as letting developers stream frontier world models into apps globally in real time in under 10 lines of code. Reactor’s embedded announcement describes pixels, audio and actions generated in real time in response to user inputs and the environment.
Reactor provides a developer platform for running real-time generative video and world models—AI systems that generate environments responding to user inputs. Its SDK and API handle model deployment, computing resources and interactive streaming, allowing developers to build applications without managing the underlying infrastructure. The platform serves world-model developers and teams working in media, entertainment, gaming and robotics. Robotics teams can also use it to run generated simulation environments for training and evaluating robot policies. Access is priced by usage and model type.
FastH3-Omni is available through Reactor’s API in preview. The model generates video with synchronized audio and supports image references, audio references, and first and last frame conditioning.
NVIDIA and Sapphire Ventures join Reactor’s Series A
NVIDIA and Sapphire Ventures joined Reactor’s Series A led by Lightspeed Venture Partners. The investments support its infrastructure for running generative video and world models in real time. Investment amounts and valuation were not disclosed.
Reactor made Vidu’s S2-Editing available at launch. The real-time video-to-video model supports character replacement, style changes, and virtual outfit and hairstyle changes.
Reactor serves live, interactive video from servers on Nebius infrastructure in Kansas City. In a recorded demonstration, CEO Alberto Taiuti identifies the Nebius-hosted servers delivering the stream and says Reactor chose the location for low latency and proximity to users.
Nebius separately states that Reactor’s custom world-model inference stack runs on its infrastructure. Its webinar announcement outlines planned discussion of inference kernels, networking, managed Kubernetes and failure recovery. The webinar page places Reactor’s serving stack on Nebius compute and managed Kubernetes, with Reactor handling the application-facing runtime, streaming and session lifecycle. It describes model-context management, state snapshots and restoration, and scheduling execution across users.
Sapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius. It describes deployments as live or coming online across the US, Europe, Japan and Korea, collectively across those providers. Nebius also names Reactor among teams running AI workloads on its infrastructure, and its event schedule labels Reactor a customer.
Scaling world models and real-time video inference
14 Oct 2026 · nebius.com · Web page
Nebius places Reactor’s serving stack on its GPU compute and managed Kubernetes. Reactor handles the application-facing runtime, streaming and session lifecycle. The page describes model-context management, state snapshots and restoration, scheduling across users, WebRTC transport and routing informed by live network conditions.
Scaling world models and real-time video inference14 Oct 2026 · nebius.com · Web pageNebius places Reactor’s serving stack on its GPU compute and managed Kubernetes. Reactor handles the application-facing runtime, streaming and session lifecycle. The page describes model-context management, state snapshots and restoration, scheduling across users, WebRTC transport and routing informed by live network conditions.
Nebius Reactor CEO · live video deployment
16 Apr 2026 · Nebius · Video
At 2:28 and 7:03, Alberto Taiuti says the live video demo streams from Reactor servers on Nebius in Kansas City. He explains choosing Nebius for latency and proximity to users.
Nebius Reactor CEO · live video deployment16 Apr 2026 · Nebius · Video · Starts at 2:28At 2:28 and 7:03, Alberto Taiuti says the live video demo streams from Reactor servers on Nebius in Kansas City. He explains choosing Nebius for latency and proximity to users.
The Real-Time Inference Layer Powering World Models: Why We're Thrilled to Back Reactor | Sapphire Ventures Twitter | X Icon LinkedIn Icon Youtube Icon Menu close Sapphire Ventures
14 Oct 2026 · sapphireventures.com · Web page
Sapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius for its real-time inference platform. Deployments are live or coming online across the US, Europe, Japan and Korea, collectively across those providers.
The Real-Time Inference Layer Powering World Models: Why We're Thrilled to Back Reactor | Sapphire Ventures Twitter | X Icon LinkedIn Icon Youtube Icon Menu close Sapphire Ventures14 Oct 2026 · sapphireventures.com · Web pageSapphire Ventures reports that Reactor secured hundreds of NVIDIA chips through AWS and Nebius for its real-time inference platform. Deployments are live or coming online across the US, Europe, Japan and Korea, collectively across those providers.
Nebius Reactor cloud-use announcement
28 May 2026 · Nebius · Web page
Nebius calls Reactor a partner and states that Reactor is building on Nebius. It describes Reactor’s platform as letting developers stream frontier world models into apps globally in real time in under 10 lines of code. Reactor’s embedded announcement describes pixels, audio and actions generated in real time in response to user inputs and the environment.
Nebius Reactor cloud-use announcement28 May 2026 · Nebius · Web pageNebius calls Reactor a partner and states that Reactor is building on Nebius. It describes Reactor’s platform as letting developers stream frontier world models into apps globally in real time in under 10 lines of code. Reactor’s embedded announcement describes pixels, audio and actions generated in real time in response to user inputs and the environment.
Reactor provides a developer platform for running real-time generative video and world models—AI systems that generate environments responding to user inputs. Its SDK and API handle model deployment, computing resources and interactive streaming, allowing developers to build applications without managing the underlying infrastructure. The platform serves world-model developers and teams working in media, entertainment, gaming and robotics. Robotics teams can also use it to run generated simulation environments for training and evaluating robot policies. Access is priced by usage and model type.
FastH3-Omni is available through Reactor’s API in preview. The model generates video with synchronized audio and supports image references, audio references, and first and last frame conditioning.
NVIDIA and Sapphire Ventures join Reactor’s Series A
NVIDIA and Sapphire Ventures joined Reactor’s Series A led by Lightspeed Venture Partners. The investments support its infrastructure for running generative video and world models in real time. Investment amounts and valuation were not disclosed.
Reactor made Vidu’s S2-Editing available at launch. The real-time video-to-video model supports character replacement, style changes, and virtual outfit and hairstyle changes.
