Reactor Hits $74M Total as NVIDIA and Sapphire Join Series A

Reactor said NVIDIA and Sapphire Ventures joined its Series A, bringing total funding to $74 million, according to the company’s October 5, 2026 Business Wire release and Fortune’s same-day report. The Series A is led by Lightspeed Venture Partners; the October tranche size was not disclosed.

The numbers on the wire

• $74 million total funding after the October investor additions (Fortune). Business Wire does not restate a round size for the October joins.
• New investors named: NVIDIA (via NVentures in Fortune’s telling) and Sapphire Ventures.
• Series A lead: Lightspeed. Cap table also lists WndrCo, Amplify Partners, Sky9 Capital, FPV Ventures and others.
• Performance claims on the release: run models at 60+ FPS and sub-40ms latency via a proprietary inference engine, global GPU cloud and low-latency streaming layer.
• No valuation disclosed.

What Reactor sells

Reactor positions itself as the developer platform for real-time generative video and world models: SDK/API, managed GPU capacity, and infrastructure for interactive inference rather than batch generation. The release says Hollywood studios, advertising platforms and streaming services have active projects on the platform, and that robotics is a second vector — running policies in the cloud while streaming observations and actions, or evaluating policies inside closed-loop generated worlds.

Co-founders Alberto Taiuti (CEO) and Bryce Schmidtchen (CTO) are former Apple Vision Pro technical leads; Taiuti previously co-founded Luma AI and served as CTO there.

Why NVIDIA is on the wire

NVIDIA already ships Cosmos world models, Isaac robot simulation tools and Jetson edge computers. Fortune frames the Reactor stake as a bet on the middleware that makes interactive world models runnable at production latency — the same picks-and-shovels pattern that surrounded LLM serving.

What to watch

• The size of the October checks. $74 million is cumulative; neither primary names the incremental dollars.
• Named paying customers beyond the Sapphire blog’s Overworld / Visko / Moonlake references.
• Independent latency and cost benchmarks against general-purpose GPU clouds for interactive world-model workloads.