AI Furnace Newsroom: Yann LeCun Raises $1B+, Musk’s xAI Is Being “Rebuilt From The Foundations Up”, Amazon Launches Health AI Agent

Welcome to this week’s AI Furnace Newsroom

In today’s insights we cover:

1. Yann LeCun’s AMI Labs raises $1B to build “world models”

2. Musk’s xAI is being “rebuilt from the foundations up” as it hones in on coding tools

3. Amazon launches a health AI agent across Amazon.com

4. Anthropic takes Pentagon to court over AI blacklist

5. Nebius signs a mega compute deal with Meta that could reach $27B

Read time: 5 mins

💡 Furnace Insights

AMI

Yann LeCun’s AMI Labs Raises $1B+ To Build “World Models”

Former Meta Chief of AI Yann LeCun launched a new startup, AMI Labs (Advanced Machine Intelligence), raising about $1.03B at a $3.5B pre-money valuation, one of the biggest seed rounds Europe has ever seen. The company says it’s going after “world models” (AI that learns from reality - video, space, cause-and-effect). Based on JEPA ideas, the architecture LeCun proposed in 2022, AMI plans to open-source a meaningful portion of its research. AMI’s first sector wedge is healthcare, via a partnership with health company Nabla.  

This is a direct bet against the “just scale language models forever” consensus - and investors are funding it like a platform, not a project. World models have become the new hot category, but AMI is unusual in two ways: it’s aiming at a high-stakes vertical early (health) instead of entertainment, and it’s leaning into openness while most well-funded labs focus on locking things down. If it succeeds, it changes what “frontier” means. If it fails, it will be an expensive reminder that physics-grounded AI is a slower road than shipping a chat product. 

xAI

Musk’s xAI Is Being “Rebuilt From The Foundations Up” as it Hones in on Coding Tools

Only 2 of xAI's original 11 co-founders remain after a wave of departures and re-organizations. Elon Musk says xAI is effectively starting over and that it was “not built right the first time” amid a broader reshuffle, even as the company tries to close the gap with Claude Code and Codex. It is being framed as a serious rebuild rather than a routine re-org, with Musk publicly acknowledging the competitive lag in developer tooling.

Since coding assistants are now becoming the primary revenue driver for AI labs, falling behind becomes a problem if xAI is to seriously compete with Anthropic and OpenAI. Rebuilding can work if it tightens focus and execution, but repeated resets also burn time, morale, and product momentum, especially when rivals are compounding. With xAI now tied more closely to Musk’s broader ecosystem, the pressure isn’t just to impress users but to justify the valuation story with a coding product that developers and enterprises can get behind. 

Amazon

Amazon Launches A Health AI Agent Across Amazon.com

Amazon is rolling out its Health AI assistant beyond One Medical and into the main Amazon website and app, letting users pull in medical records, ask health questions, and take actions like booking appointments and renewing prescriptions through Amazon Pharmacy. It’s powered by Amazon Bedrock, uses a multi-agent setup with guardrails, and Amazon says it runs in a HIPAA-compliant environment. Prime members get an introductory bundle of free virtual care visits.  

In the past months, OpenAI, Anthropic and Microsoft have all launched health AI products as the race heats up. What makes Amazon different is its primary care network, pharmacy and the ability to book appointments directly from the consumer app you already use to buy other products - closing the loop from “what is this symptom?” to “here’s a clinician” to “your prescription is refilled”. That’s a huge distribution advantage over standalone health assistants. The tradeoff is trust: consumers will have to decide whether they’re comfortable linking sensitive health data to a company that also runs a giant retail marketplace, and regulators will have to keep pace as these systems move from advice to action. 

Anthropic

Anthropic Takes the Pentagon to Court Over an AI Blacklist

Anthropic has filed suit after the U.S. government labeled it a “supply chain risk” following a dispute over defense contract terms. The fight traces back to Anthropic refusing broader “all lawful purposes” language without clearer prohibitions around mass domestic surveillance and fully autonomous weapons, and the resulting designation now pressures contractors to certify they aren’t using Claude in Pentagon work.  

The bigger issue is whether “AI safety red lines” survive contact with national security. If governments can effectively blacklist an AI lab for holding boundaries, other companies get the message: don’t be the one that says no. And if AI labs don’t hold boundaries, then the limits on surveillance and autonomy end up being set by contract language and market competition instead of clear law - an uncomfortable way to decide how powerful systems get used. 

Nebius

Nebius Signs a Mega Meta Compute Deal that Could Reach $27B

European AI infrastructure company Nebius just landed a major AI infrastructure agreement with Meta that’s being described as up to $27 billion over five years, starting with $12B in dedicated AI capacity from 2027 and giving Meta the option to buy up to $15B more. The deal is also positioned as an early large-scale deployment of Nvidia’s next-gen “Vera Rubin” platform. Nebius shares jumped on the news as markets treated it as another proof point that “neoclouds” are becoming trusted suppliers to hyperscalers.  

Big Tech’s AI race is quietly creating a new tier of “infrastructure middlemen.” Instead of building every GPU cluster themselves, hyperscalers like Meta can lock in capacity through specialists like Nebius offloading parts of execution risk while still securing compute at scale. For Nebius, these huge take-or-pay style contracts are a credibility unlock, but they also create concentration risk: one or two giant customers can define growth, capex plans, and financing needs.

📈 AI Venture Deals of the Week

• Axiom, an AI math verification company, raised its $200M Series A.

• Eridu, an AI data-center networking infrastructure startup, raised its $200M Series A.

• Mind Robotics, an industrial AI-powered robotics company, raised its $500M Series A.

• Sunday, an AI humanoid household robotics developer, raised its $165M Series B.

• Nexthop AI, an AI data-center networking infrastructure provider, raised its $500M Series B.

• ORO Labs, an AI procurement orchestration platform, raised its $100M Series C.

⚒️ New AI Product Launches You Don’t Want to Miss 

• Anthropic doubled Claude usage allowance during off-peak hours.

• Google introduced new Gemini-powered tools for content creation in Workspace.

• SpinachAI, a AI meeting assistant for notes and action items.

• DexData, a AI business intelligence assistant for querying data.

• Obsidian, a linked-notes personal knowledge management app.

• Banana, a real-time voice translation app preserving your voice.

Upcoming Events 📅

• AIF Founder Supper Club @ Private Room - Tuesday, March 24

Interested in meeting the Who’s Who of AI? Join the AI Hot 100 Summit in NYC

The AI Hot 100 Summit is back for its third edition on May 6-7th, 2026 in New York City. The only AI Summit built for real connections where AI Visionaries meet Industry Leaders. Join over 500 AI Executive Leaders, Founders and Investors for a jam packed 2 days of enterprise AI lightning talks, innovation showcases, and curated networking.

Get ready to meet and greet speakers and attendees from OpenAI, NVIDIA, Haleon plc, Legora, Anthropic, Cursor, Eleven Labs, Adobe, Walmart, Insight Partners & more.

🗓️ May 6-7, 2026

📍New York City