The young full stack AI startup River AI secured $1.1 billion in Series Seed and Series A funding at a roughly $5 billion valuation, led by General Catalyst and AMP PBC, to advance its open weight training API and personal AI ownership vision.
River AI raised $1.1 billion across its Series Seed and Series A, led by General Catalyst and AMP PBC, with strategic investments from NVIDIA and AMD Ventures, plus participation from Y Combinator and Temasek. The company, headquartered in Palo Alto, is positioned as a full stack AI firm focused on personal AI that users and organizations own and shape. The round accelerates development of tools for training, tuning, and serving custom models based on open weight systems, with a longer term vision of individualized AI running close to the user (including new hardware). Reports place the valuation at roughly $5 billion. Babuschkin is said to be contributing up to $100 million of his own capital. The company was incorporated in Nevada on April 20, 2026, and publicly launched in June 2026; its training API is already live.
River AI’s founder Igor Babuschkin previously worked on generative modeling and reinforcement learning at Google DeepMind (including contributions linked to projects like AlphaCode), led large scale training efforts at OpenAI, and co-founded xAI, where he played a central role in infrastructure and scaling. He left xAI in 2025 amid broader talent departures. River’s founding team includes others from xAI and Tesla with deep expertise in deep learning and reinforcement learning, emphasizing execution speed across the full AI stack.
River argues that current AI is dominated by a few labs offering general purpose models trained on broad internet data and optimized for mass audiences. These are powerful but not tailored to specific organizations or individuals, and building custom models has historically required dedicated infrastructure teams, specialized hardware, and months of effort. River’s alternative centers on ownership: users control the hardware, the data the model learns from, and the intelligence itself. The long term vision is AI that aligns tightly with individual values, learns continually from the user, and acts in their interest, described as more like a “guardian angel” than a corporate chatbot.
The first product is the River API, which supports state of the art LoRA fine tuning and reinforcement learning on frontier open weight models ranging from ~35B to 1T parameters. Key claimed advantages include:
- Completing complex RL training runs in 15–20 minutes with no dedicated infrastructure team.
- 2–4× cost savings versus closed source alternatives.
- Handling of underlying complexity (fast weight transfers, sampling-training consistency, elastic compute).
- Instant deployment of trained models to production.
- Strictly token-metered billing for training and inference (no idle GPU charges).
The company is building a full stack: accessible training infrastructure, personalization and continual learning product layers, and new hardware intended to keep personal AI local (e.g., home or small business servers) rather than reliant on remote data centers.
Recommended: VibeIQ Raises $22.5 Million In Funding Led By Volition Capital
General Catalyst CEO Hemant Taneja framed the investment around American leadership in open weight models alongside closed frontier systems, calling ownership of intelligence by users a priority for resilience and historically correct for the open-weight ecosystem. Managing Director Marc Bhargava highlighted the gap between AI’s potential and what most companies experience, positioning River as closing that gap with cost efficient custom model tools on proprietary data.
Strategic participation from NVIDIA and AMD Ventures aligns with the hardware and compute elements of the stack. AMP PBC (a public-benefit corporation that itself raised significant capital earlier for pooled AI compute) co-led, reinforcing infrastructure themes. Y Combinator and Temasek add early stage and sovereign scale capital perspectives.
This is an unusually large raise for a company only a few months old that is still early in product shipping. It fits a pattern of high valuation “neolab”-style bets on pedigreed AI researchers pursuing ambitious, ownership oriented or fundamental agendas rather than purely productized near term applications. The emphasis on open weights, user ownership, local/edge execution, and enterprise custom models differentiates it from pure frontier closed model labs while still competing in the high capital training and personalization space. Claims around speed, cost, and accessibility target the practical barriers that have limited custom AI adoption outside well resourced organizations.
The funding directly supports accelerating every layer of the intended stack, from the already-live API to personalization products and hardware, while the founding team’s track record in scaling training systems at major labs underpins the credibility of rapid execution claims.
Please email us your feedback and news tips at hello(at)dailycompanynews.com
