Dream Labs raises $90M seed round for robot foundation model API
Startup launches Dream-1 model and public API, aiming to simplify multi-task robotic training through video prediction methods.
Key points
- Dream Labs secured $90 million in seed funding led by B Capital and Kindred Ventures.
- The company launched a public API for its Dream-1 universal robot foundation model.
- The model treats robotic skills as video prediction problems to enable multi-tasking.
Dream Labs announced a $90 million seed funding round on October 19, 2026, coinciding with the launch of its public application programming interface. The startup, which previously operated in stealth mode, is introducing Dream-1, a universal robot foundation model designed to provide intelligent control for diverse robotic hardware.
The funding round was jointly led by B Capital and Kindred Ventures. Other participants included A.Capital, Unusual Ventures, Wing VC, SV Angel, Scribble VC, and angel investor Adrian Aoun. This capital injection supports the commercialization of the Dream API, which is now live for all users.
Context and Background
Dream Labs aims to address the limited intelligence of current industrial robots. According to co-founder and Chief Executive Joel Jang, there are more than five million industrial robots currently operating in factories and warehouses globally. While these machines perform useful work, Jang argues they are restricted to single tasks and lack the ability to transfer learned knowledge to other units.
The core issue, according to the company, is not hardware cost but software complexity. Robotic arms with high dexterity now sell for a few thousand dollars, yet they ship without inherent intelligence. Buyers must spend months on tedious, specialized programming to teach each arm a specific job. When new hardware is introduced, this entire training process must be repeated from scratch.
Who Is Affected
The launch of Dream-1 targets developers and enterprises seeking to reduce the time and expertise required for robotic training. The model is designed to control bimanual robots it has never encountered, on tasks it was not explicitly shown, in environments it has never visited. Dream Labs states the API access costs $3 per hour.
Dream-1 approaches robotic skills as simple video prediction problems. This methodology allows the model to act as a universal brain, potentially eliminating the need for manual reprogramming when deploying robots in new settings. As Threat Times reported on October 19, 2026, the startup secured this funding specifically to advance this video-trained foundation model.
What should you do?
Organizations evaluating robotic automation should assess whether their current hardware supports API integration with external foundation models. Developers should review the Dream API documentation to understand latency and control limits for bimanual systems. Security teams must evaluate the data privacy implications of sending video feeds to third-party AI models.
What to do and how to stay safe: Dream Labs
- Review current robotic hardware specifications to determine compatibility with external foundation model APIs.
- Evaluate data privacy risks associated with streaming operational video feeds to third-party AI services.
- Assess the total cost of ownership when comparing manual programming hours against API usage fees.
- Monitor vendor documentation for updates on model capabilities and supported robotic form factors.
Step-by-step guide: Mobile Device Security: How to Protect Data Beyond Passwords
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
How much funding did Dream Labs raise?
Dream Labs raised $90 million in seed funding, according to the October 19 announcement.
What is the cost of using the Dream API?
The Dream API costs $3 per hour to access, as stated by the company.
How does Dream-1 train robots?
Dream-1 treats robotic skills as simple video prediction problems rather than traditional programming.
