The Rise of Neolabs: Where the Next AI Breakthroughs Will Come From & 11 AI Labs to follow

11 frontier AI labs shaping the next wave of innovation

Hey everybody, welcome to The AI Opportunity.

Here I distill the best money-making opportunities before they hit the mainstream. Most of this newsletter is for paid subscribers, so if you want full access, subscribe today.

Today I wanted to explore one of the types of AI companies that are playing a huge role in the development of AI and that are key to be ahead of the curve: AI Neolabs

Neolabs are not companies in the traditional sense. They operate more like private research institutions (founded by former OpenAI, DeepMind, Anthropic and Google Brain researchers) with the freedom to explore ideas that would be impossible inside a typical startup or large lab.

Their goal is not to ship a product quickly but to widen the space of what AI can do.

Here is a clear overview of all the 11 Neolabs you need to know:

1. Black Forest Labs

The founding team is led by Robin Rombach, Andreas Blattmann, and Patrick Esser. Their background is rooted in years of university research that directly led to the foundation of modern visual AI. They are globally known as the original co-creators of the Stable Diffusion models. Their work defined the initial frontier for open-source image generation.

Black Forest Labs was created to translate that research into a commercial lab. The core technical goal is to move beyond simple image output towards a unified “visual intelligence” that integrates perception, generation, and reasoning. Their flagship model, FLUX, delivers high-resolution image generation and multi-reference editing, with a critical focus on visual consistency across complex scenes.

Founding Team
Original co-creators of Stable Diffusion: Robin Rombach, Andreas Blattmann (Linkedin), and Patrick Esser.

Mission
Develop frontier generative models for image and video (FLUX), focusing on the evolution toward unified visual intelligence.

Funding & Valuation
- Funding:$300M in 2024;
- Valuation:$3.25B

2. Humans&

Eric Zelikman has spent years exploring how models can reason about intent rather than simply output text. At Stanford, his work on self-reflective reasoning pushed models to critique their own intermediate steps. At xAI, he deepened this line of research with colleagues who shared the belief that true alignment comes from modeling human values, not from post-processing techniques.

Humans& is the direct continuation of this path. The team is building models designed to infer user intent, long-term preferences and patterns of decision-making.

Their goal is not assistants that execute commands, but systems that collaborate, anticipate and adapt. The lab exists because Zelikman’s academic and industry experience converge toward the same conclusion: intelligence becomes useful when it becomes human-aware.

Founding Team
Founded by Eric Zelikman (Linkedin), former xAI researcher and Stanford PhD student.

Mission
Developing more human-aligned models able to understand intentions, values and context.

Funding & Valuation
- Funding: $1B in 2025;
- Valuation: ~$4B.

3. Isara

Eddie Zhang spent his time at OpenAI focused on safety, control and how agents behave at scale. He worked on systems meant to supervise complex model behaviors and early prototypes of multi-agent coordination. His belief was consistent: real-world tasks are handled better by many small agents working together than by a single general model.

Isara grows directly out of that conviction. The lab is building infrastructure where large networks of agents can collaborate on operational workflows like customer support, commerce automation and internal process handling.

Everything is built around orchestration, monitoring and dynamic correction. It is a natural extension of the work Zhang led inside OpenAI, now turned into a full-scale research effort.

Founding Team
Co-founded by Eddie Zhang (Linkedin), former OpenAI safety researcher.

Mission
AI that understands large volumes of human conversations and builds infrastructure for large-scale AI agent coordination.

Funding & Valuation
- Funding: Not disclosed, hundreds of millions in 2025;
- Valuation: ~$1B.

4. Richard Socher’s Lab

Richard Socher has lived the entire arc of applied AI. From Stanford research to founding MetaMind, from leading AI at Salesforce to running You.com, he experienced the same limit repeatedly: improving models depends on slow, human-driven experimentation cycles. Architecture variations, tuning, dataset design and evaluation always hit throughput constraints.

His new lab is designed to accelerate that loop. Instead of automating small steps, it aims to assist the full research stack: generating model ideas, organizing experiments, running controlled comparisons and surfacing the most promising directions.

The purpose is not to replace researchers but to give them a faster way to explore the space of possible models. It’s the natural outcome of Socher’s decade spent wrestling with slow iteration.

Founding Team
Created by Richard Socher (Linkedin), former chief scientist at Salesforce and founder of MetaMind.

Mission
Automating parts of AI research itself, with systems aimed at accelerating model development.

Funding & Valuation
- Funding: $1B in discussions in 2025;
- Valuation: not disclosed.

5. General Intuition