News
White Paper
Why Enterprise AI Fails at Scale
Four months after walking away from Meta, Turing Award-winning AI scientist Yann LeCun has closed one of the largest seed rounds in startup history. AMI Labs — Advanced Machine Intelligence — announced today that it has raised $1.03 billion at a $3.5 billion pre-money valuation.
The round, which LeCun confirmed on X, calling it "probably the largest [seed round] for a European company," is a major vote of confidence for a startup that has no product, no revenue and, by its own admission, no near-term prospect of either. Yet investors — Nvidia, Bezos Expeditions, Samsung, Temasek and Toyota Ventures, alongside prominent angels such as Eric Schmidt, Mark Cuban and the web's inventor Tim Berners-Lee and his wife Rosemary — appear willing to wait.
Unveiling our new startup Advanced Machine Intelligence (AMI Labs).
— Yann LeCun (@ylecun) March 10, 2026
We just completed our seed round: $1.03B / 890M€, one the largest seeds ever, probably the largest for a European company.
We're hiring!
[the background image is the Veil Nebula - a picture I took from my… https://t.co/voyuA5nwiL
Table of Contents
- Why AMI Is Rejecting the AI Mainstream
- From the ICU to the Assembly Line
- A European Counter to Silicon Valley
- Open Science in a Closed World
Why AMI Is Rejecting the AI Mainstream
LeCun spent twelve years building Meta's AI research operation, FAIR, into one of the most respected in the world. He also spent much of that time as one of the industry's most outspoken internal critics. Large language models, he argued publicly and repeatedly, are an architectural dead end: powerful at generating plausible text, but structurally incapable of genuine reasoning, planning or understanding physical reality.
He publicly stated that one reason for leaving Meta was because "they... became so LLM-piled."
LeCun's alternative to LLMs a name — JEPA, the Joint Embedding Predictive Architecture, which he first proposed in 2022. Rather than predicting the world token by token, as generative models do, JEPA learns abstract representations of how the world works. The goal is AI that can simulate cause and effect, plan for future states and maintain persistent memory — capabilities that today's chatbots lack.
"Real intelligence does not start in language," AMI's mission statement reads. "It starts in the world."
From the ICU to the Assembly Line
AMI's first disclosed commercial partnership is with Nabla — a clinical AI assistant platform founded and led by Alexandre LeBrun, AMI Labs' CEO. LeBrun still serves as chairman at Nabla.
The collaboration will give Nabla early access to AMI's world models as part of an effort to develop what both companies describe as FDA-certifiable agentic AI systems for clinical workflows. World models need real-world data and real-world evaluations, and healthcare offers both in abundance, according to LeBrun.
But the ambition stretches beyond hospitals. AMI's official mission targets applications where "reliability, controllability, and safety really matter" — including robotics, industrial process control, automation and wearable devices. Several of the round's corporate investors, including Toyota Ventures and Samsung, indicate that industrial and consumer hardware applications are very much in view.
A European Counter to Silicon Valley
AMI Labs is headquartered in Paris, with additional hubs in New York (where LeCun teaches at NYU), Montreal and Singapore. LeCun has been explicit that AMI is designed as a European, specifically French, alternative to the AI giants of the United States and China.
That framing resonates with European investors and policymakers alike. Pierre-Éric Leibovici, founder and managing partner of French VC firm Daphni — one of the round's participants — said, "AMI Labs could be the first European company to reach the scale of the GAFAM companies."
Whether that ambition is achievable from a standing start, with no product and a research timeline measured in years, remains open. But AMI's funding suggests that, for now, the combination of LeCun's scientific credibility, LeBrun's operational track record and a growing sense that the LLM paradigm has real limits is more than enough to open wallets.
Related Article: Yann LeCun Warns Meta Is Betting on the Wrong AI Future
Open Science in a Closed World
Another defining element of AMI that distinguishes it from the current generation of AI frontier labs: its stated commitment to openness. In a time when leading research is increasingly conducted behind closed corporate doors, AMI says it will publish its findings and release substantial portions of its code as open source.
The promise is notable — and will be tested.
OpenAI was founded in 2015 explicitly as a non-profit committed to open research. But it quickly closed off its work as it scaled. GPT-2 was released with staged rollouts, citing safety concerns. GPT-3 and beyond were never open-sourced. In fact, Elon Musk sued the tech company in late 2024, arguing it had abandoned its founding mission. That lawsuit was eventually settled.
Google DeepMind published landmark research openly for years — AlphaGo, AlphaFold — but has become notably more guarded as its models became commercially valuable. AlphaFold's data was released, but the underlying model weights for its most capable systems have not been.
And in another case, Mistral AI initially positioned itself strongly on openness, releasing early models fully openly. But its more capable models have moved to a commercial licensing model with restrictions.
Webinar
Mar
25
Content Leaders Collective: Navigating Content Decisions at Scale
Discover how content leaders are modernizing content operations, avoiding costly missteps and preparing for scale and AI.
Register
Webinar
On demand
Content Strategy Leaders Live: Scaling for Speed, Complexity and AI in High Tech
A candid roundtable on how high-tech leaders are rethinking content at scale.
Watch Now
Webinar
On demand
Do More with Less: Modernizing the Cloud Contact Center for 2026
Learn how to leverage cloud platforms without adding a single hire to personalize every customer interaction.
Watch Now
Webinar
On demand
Cut the Noise: Deploying AI That Actually Moves the Needle
Learn how to turn AI experimentation into concrete revenue operations.
Watch Now
Webinar
On demand
Ditch the Desk Phones: How Modern Teams Drive AI-First Communications
Find out how one team finally pulled the plug on a legacy phone system. And built something smarter.
Watch Now
Webinar
On demand
Rebrand. Migrate. Optimize. How to Do It All (Without Slowing Down)
Cresta leveled up site speed, design flexibility and marketer sanity (in record time). Find out how.
Watch Now
The broader pattern is pretty consistent: openness tends to be easiest to maintain early, when the research is exploratory and commercial stakes are low. Once a model becomes genuinely competitive and monetizable, the incentive structure shifts hard toward closure. AMI's $1 billion war chest arguably makes that pressure arrive sooner rather than later.