Author: River AI

Compiled by: Deep Tide TechFlow

Deep Tide Guide: As mainstream AI models become more closed, River AI hands back the power of training and controlling intelligence to enterprises and individuals with an open-weight model and a customizable API that can be run in just 15 minutes. This not only breaks the monopoly of giants but also hints at a viable path for decentralized AI infrastructure — something worth closely watching for any investor and developer focused on the next wave of AI.

River AI has raised a total of $1.1 billion in seed and Series A funding

Today we announce the completion of a total of $1.1 billion in seed and Series A funding, led by General Catalyst and AMP PBC, with strategic investments from NVIDIA and AMD Ventures, as well as participation from Y Combinator and Temasek. This funding will accelerate the realization of our mission: to build a powerful personal AI that understands you enough and works for you, and to return the ownership of this intelligence to the people and organizations that use it. It all starts with providing developers and companies tools to train, fine-tune, and truly own their own AI models.

Today, most companies using AI run general models trained on the entire internet, aimed at as wide an audience as possible. While powerful, they are not tailored for any specific organization. Previously, building a custom model required a dedicated infrastructure team, specialized hardware, and months of work — beyond the capability of most companies.

River API completely disrupts this. Any enterprise can complete a complex reinforcement learning task in 15 to 20 minutes without the need for an infrastructure team, and at a cost that is two to four times lower than closed-source alternatives. We provide cutting-edge LoRA fine-tuning and reinforcement learning for state-of-the-art open-weight models, with the platform handling the underlying complexity — rapid weight transfer, sampling training consistency, and elastic computing — so developers can focus on improving the model, not managing the infrastructure. The trained models can be deployed instantly into production, with precise billing based on the token usage for training and inference, completely eliminating idle GPU costs.

“The way AI is built today is not its future form. AI should be open, freely available, and truly affordable. It should feel like it serves the user themselves, not the lab that trained it. We founded River to enable individuals and companies to own their own intelligence.”

Our ambition goes beyond the API. We are building a powerful personal AI — one that learns from you, is truly controlled by you, and is familiar enough with you to act in your best interests. We are not aligning a model to billions of users but aligning AI directly to each individual. Today, the API puts this ownership in the hands of developers and enterprises; over time, we will extend the same control to every individual. We elaborate on the reasoning behind this in our article "Introducing River AI".

Reaching this vision means building an end-to-end full-stack system: making it easy for any developer to fine-tune training infrastructure, creating products centered around personalization and continuous learning, and developing new hardware that allows personal AI to operate close to you rather than living in someone else's data center.

“The U.S. leadership in AI urgently needs to ensure its leadership in open-weight models while maintaining its advantage in closed frontier models. Igor and the River AI team have the experience to make it happen, and we see this agenda as a priority for enhancing the resilience of the United States. The core idea of handing intelligence ownership to users will be proven correct by history in the open-weight ecosystem.”

Our founding team is on the frontier of deep learning and reinforcement learning, with practical experience from xAI and Tesla, and possesses the rare ability to execute at high speed across the entire artificial intelligence stack. Before co-founding xAI, Igor worked on generative modeling and reinforcement learning research at Google DeepMind, and led large-scale training efforts at OpenAI.

“There still exists a gap between what AI can do and what most companies actually experience. Until now, enterprises have lacked a cost-effective way to train, fine-tune, and own custom AI models. River bridges this gap, helping any company build models based on their own data, tailored to their actual workflows.”

This funding will accelerate the construction of every layer of this full-stack system. If you want to participate in building an AI that people truly own, join us and help build.

— River AI Team