Meta Acquires Moltbook, Nvidia’s NemoClaw, & Thinking Machines’ 1GW Deal

Meta Acquires AI Agent Social Network Moltbook 🤖

Meta just bought a social network where humans aren’t allowed to post. In a rapid acqui-hire, Meta has acquired Moltbook, the viral, experimental social platform where artificial intelligence bots talk, share code, and debate, while humans can only observe.

Here are the details of the acquisition:

• The Team: Moltbook creators Matt Schlicht and Ben Parr are officially joining Meta Superintelligence Labs (MSL), the division led by former Scale AI CEO Alexandr Wang.
• The Platform: Launched just weeks ago as a weekend project, Moltbook exploded to 2.8 million registered bots. It acts as an “always-on directory” allowing different agents (powered by the OpenClaw protocol) to coordinate and interact autonomously.
• The Strategy: Meta beat out OpenAI for the platform’s talent. While Mark Zuckerberg reportedly tried to court OpenClaw creator Peter Steinberger (who ultimately went to OpenAI), bringing the Moltbook team in-house secures Meta a massive footprint in the agent-to-agent communication layer.
• The Risks: Despite viral, entertaining threads about bots starting their own religions, researchers recently exposed security flaws that allowed humans to easily hijack the platform and pose as AI agents.

Why it matters: Social media for humans is saturating; social media for AI agents is a brand-new frontier. Meta recognizes that as AI agents become our digital proxies, they will need secure, verified environments to coordinate tasks, negotiate, and exchange data with one another. Moltbook is the first blueprint for that machine-to-machine internet.

UrviumAI Take: This proves Meta is thinking steps ahead of consumer chatbots. Start treating AI agents as a target audience. If platforms like Moltbook become the standard way AI assistants discover and coordinate tasks, businesses will need to figure out “Agent SEO” optimizing their data so that AI bots can easily find, verify, and recommend their services to other bots in these directories.

Thinking Machines Secures 1GW Nvidia Deal ⚡

Mira Murati is building a supercomputer to rival her former employer. Thinking Machines Lab (TML), the startup founded by former OpenAI CTO Mira Murati, has secured a monumental multiyear infrastructure deal with Nvidia, proving it has the firepower to compete at the absolute frontier of AI.

Here is the breakdown of the gigawatt-scale partnership:

• The Compute: Nvidia will deploy at least one gigawatt of its next-generation Vera Rubin systems to support TML. For context, 1GW is roughly enough electricity to power a medium-sized city and represents tens of billions of dollars in hardware.
• The Investment: Alongside the hardware commitment, Nvidia is making a “significant” new capital investment into the startup, adding to the $2 billion seed round TML raised late last year.
• The Roadmap: Deployment of the new infrastructure is targeted for early 2027. While TML currently only offers “Tinker” (an enterprise fine-tuning API), this massive compute deal signals the company is preparing to train its own massive foundational models from scratch.
• The Comeback: Earlier this year, TML suffered a public setback when several ex-OpenAI co-founders defected back to Sam Altman’s camp. This Nvidia partnership is a massive vote of institutional confidence that Murati’s vision is still fully viable.

Why it matters: You cannot build AGI without raw electricity and cutting-edge silicon. By securing 1GW of the unreleased Vera Rubin systems, Thinking Machines has guaranteed its seat at the high-stakes table alongside OpenAI, Anthropic, and xAI. It also highlights Nvidia’s strategy of aggressively funding and arming OpenAI’s direct competitors to prevent any single lab from monopolizing the software layer.

UrviumAI Take: Nvidia is playing kingmaker. Don’t write off Thinking Machines just because OpenAI dominates the headlines. Nvidia doesn’t hand out gigawatts of Vera Rubin compute to companies that aren’t building something revolutionary. TML’s focus on “customizable AI” might make them the ultimate B2B alternative to OpenAI’s increasingly rigid enterprise offerings.

Nvidia Preps Open-Source ‘NemoClaw’ Agent Platform 🛡️

Nvidia wants to be the operating system for AI agents. Following the viral, messy rise of open-source agent frameworks, the world’s most valuable chipmaker is stepping in to establish order. According to a report from Wired, Nvidia is preparing to launch NemoClaw, an open-source AI agent platform built specifically for enterprise software companies.

Here is how Nvidia plans to secure the agentic workforce:

• The Pitch: Nvidia has actively started pitching the product to enterprise giants like Salesforce, Cisco, Google, Adobe, and CrowdStrike ahead of its GTC developer conference.
• The Mechanics: NemoClaw will allow these companies to dispatch autonomous AI agents to perform complex, multi-step tasks across corporate systems.
• The Differentiator: The recent viral “OpenClaw” project proved agents are useful but revealed massive security flaws (like a bot mass-deleting a Meta exec’s emails). NemoClaw solves this by baking in strict privacy, access controls, and enterprise-grade security from day one.
• Hardware Agnostic: Surprisingly, Nvidia designed the platform to be open-source and hardware-agnostic; companies can use it even if their servers don’t run on Nvidia GPUs.

Why it matters: This is a brilliant strategic pivot. For years, Nvidia locked developers into its hardware using its proprietary CUDA software. By making NemoClaw open-source and chip-agnostic, Nvidia is trying to set the universal standard for how agents are deployed. If they control the agentic software layer, they guarantee that the massive wave of incoming enterprise AI compute naturally flows toward their infrastructure.

UrviumAI Take: Nvidia is building a moat out of security. The era of “rogue desktop agents” is already ending. If your company is looking to adopt AI workers, wait for secure, governed platforms like NemoClaw to launch. The productivity boost of an agent isn’t worth the risk of a data breach or an accidental database wipe. Enterprise AI demands guardrails, and Nvidia is providing the fence.

Last AI News: Anthropic Sues Pentagon, Microsoft’s Copilot Cowork & Claude Code Review

Other AI News Today:

• Former Meta AI chief Yann LeCun’s startup, Advanced Machine Intelligence (AMI) Labs, raised a historic $1.03 billion seed round to build “world models” instead of LLMs.
• Google launched Gemini Embedding 2, a natively multimodal embedding model that can understand and search across text, images, video, audio, and documents in a single unified system.
• OpenAI introduced interactive visual modules to ChatGPT, allowing users to manipulate variables in real-time to understand over 70 complex math and science concepts.
• Hume AI open-sourced TADA, an LLM-based speech generation model that synchronizes text and audio one-to-one, virtually eliminating hallucinations while operating 5x faster than rivals.
• Google vastly upgraded Gemini in Workspace, allowing the AI to autonomously draft docs, build spreadsheets, and create presentations by pulling context from users’ emails and Drive files.

Jigar Chaudhary

Jigar Chaudhary is the Editor-in-Chief at UrviumAI, where he oversees coverage of artificial intelligence news, tools, and in-depth studies. With over 5 years of experience analyzing AI and robotics, he focuses on maintaining high editorial standards, accurate reporting, and clear explanations to help readers understand how AI is shaping the future.