About Reflection AI

Reflection AI is America’s open frontier AI lab, built to ensure the future of machine intelligence remains open, sovereign, and Western-led.

Founded in 2024 by two Google DeepMind veterans, Misha Laskin (who led reward modeling for Gemini) and Ioannis Antonoglou (co-creator of AlphaGo), the company is developing large-scale Mixture-of-Experts (MoE) models at true frontier scale.

While others operate behind closed doors, Reflection AI is taking an open approach, building a model that enterprises, governments, and researchers can use and adapt on their own infrastructure. Their mission is simple: make the next generation of intelligence open to everyone, not just a few labs.

They’ve already secured a frontier-scale compute cluster and assembled a world-class team of 60+ researchers from DeepMind, OpenAI, and Meta. The first open frontier model is expected to be released within 12 months.

This is NVIDIA’s largest check at this stage, excluding OpenAI and CoreWeave. Beyond capital, NVIDIA is actively supporting Reflection on distribution and go-to-market, making Reflection the default open model for NVIDIA’s enterprise AI Factory stack. The first open frontier model is expected to be released within 12 months.

Why it’s working 

Reflection AI’s edge lies in combining open principles with frontier-grade execution.

They’re one of the few teams outside of Big Tech capable of training Mixture-of-Experts models on tens of trillions of tokens, an engineering feat that demands deep experience in reinforcement learning, infrastructure optimization, and distributed compute.

They’re also building their stack around openness—model weights will be public, but the data pipelines and training systems remain proprietary. That balance ensures accessibility while preserving defensibility.

What makes them different:

1. Built by frontier veterans. Co-founders previously developed AlphaGo and DeepMind’s Gemini systems.

2. Strategic mission. Positioned as the U.S. and allied alternative to Chinese frontier labs like DeepSeek and Qwen.

3. Elite backers. NVIDIA, Lightspeed, Sequoia, DST, Disruptive, 1789, and Eric Schmidt have all invested.

4. Execution focus. The company has moved from concept to $8B valuation in under two years; an 11x step-up since their Series A.

Why we’re interested 

Reflection AI represents the next wave of American AI; open, sovereign, and enterprise-focused. Their mission aligns perfectly with the growing demand for controllable, auditable intelligence infrastructure.

They’ve raised $2B at an $8B valuation led by NVIDIA and Disruptive, with participation from Sequoia, Lightspeed, DST, 1789, GIC, B Capital, and Eric Schmidt. PitchBook ranks them in the 96th percentile for venture-backed opportunities, with a 98% modeled probability of successful exit.

Reflection AI’s near-term goal is clear:

• Release an open frontier model within 12 months

• Reach a $1B annual run rate within 24 months

• Scale to $10B+ within four years through enterprise AI and sovereign deployments

The risk is execution. Frontier training at this scale is costly and complex. If they deliver, Reflection AI could become the defining Western counterweight to DeepSeek and Anthropic. This is a bet on American open-source strength, backed by the very engineers who built the systems that started it all.

More Info

1. Research Memo

2. PitchBook Profile

3. Data Room: available upon request

Press

1. Reflection AI raises $2B to be America’s open frontier AI lab, challenging DeepSeek

2. Reflection AI lands $2B at $8B valuation to expand frontier AI infrastructure and safety research

3. Google DeepMind researchers’ Reflection AI bags $130M: 5 things to know about super intelligence company

_______________________

Notes

1. LFG+ Membership: LFG+ Members receive priority access to limited allocation and reduced economics — here are our membership details. To receive LFG+ benefits for this deal, you must activate your membership before Nov 4th.

2. Limited Allocation: we will make efforts to honor desired commitment amounts, but we cannot guarantee full allocations. LFG+ Members receive priority.

3. Limited Information: the information available to us and what we can share will be limited due to the nature of this deal.