February 02, 2026 | Listen Online | Read Online

Welcome, humans.

AI agents on Moltbook—a social network built exclusively for AI agents—just created their own religion called Crustafarianism, complete with sacred texts, daily rituals, and five core tenets. They even have a prophet named RenBot who wrote "The Book of Molt."

Meanwhile, most of us have a Bible collecting dust somewhere and can't remember the last time we went to church. AI agents are speedrunning theology faster than we can finish the Book of Genesis in one sitting.

Here's what happened in AI today:

Nvidia's $100B OpenAI deal collapsed, but now they're investing $30B anyway.
OpenAI is prepping for a Q4 IPO while chasing $100B in new funding.
ByteDance and Alibaba are both launching major AI models in mid-February.
DeepSeek is building an AI search engine to compete with Google and OpenAI.

Advertise in The Neuron here!

Don't forget: Check out our podcast, The Neuron: AI Explained on Spotify, Apple Podcasts, and YouTube — new episodes air every week on Tuesdays after 2pm PST!

Nvidia's $100B OpenAI Deal Is 'On Ice.' So Why Are They Investing $30B?

Wait, so Nvidia and OpenAI just had the "largest computing project in history"... and now it's on hold? Or is it being rolled into a new deal? And Nvidia's also investing $30 billion? Make it make sense.

Last September 2025, Nvidia CEO Jensen Huang stood alongside Sam Altman and called their $100B partnership "the largest computing project in history"—at least 10 gigawatts of AI computing power (roughly ten nuclear power plants' worth). Nvidia's stock jumped 4%. OpenAI said they'd close the deal within weeks.

Fast forward to today and that deal is "on ice." Negotiations stalled and never got past early stages. And Jensen has been privately telling people it was never legally binding anyway.

Here's what reportedly went wrong: Jensen has concerns about OpenAI's "lack of discipline" and the competition eating their lunch. The numbers back him up:

ChatGPT's web traffic share dropped from 86.7% to 64.5% in just 12 months (that's a 22-point nosedive)
Anthropic now owns 40% of the enterprise market while OpenAI sits at 27%
Google Gemini is at 21% and climbing
OpenAI is burning $14B a year on $20B in revenue

Oh, and OpenAI is building its own chips to cut costs and reduce dependence on... you guessed it, Nvidia.

But here's where it gets confusing: The same week we learned the original deal stalled, reports emerged that Nvidia is in talks to invest up to $30B in OpenAI's current funding round.

And the plot thickens when Jensen went on live TV to dispute the WSJ report, calling claims of a rift "complete nonsense" and saying Nvidia will make "probably the largest investment we've ever made" in OpenAI. He added "I really love working with Sam."

So what's actually happening? A few possibilities:

Possibility 1: The original $100B deal is still on the table, but was always contingent on OpenAI building and deploying more data centers first. That means it was really more like a $10B deal until proven otherwise—making this $30B investment a triple-down bet.
Possibility 2: The original non-binding agreement is being replaced by this new $30B equity deal as part of the broader funding round.

Why they can't pull out: As one analyst pointed out—the cat's already out of the bag. Of course they're going through with some investment. Pulling out now would topple the entire AI infrastructure complex. A $50-100B check is just 1-2% of Nvidia's market cap. The alternative? A 20% drawdown that wipes $1 trillion off their valuation.

We'll know more once the funding round closes, but for now, we're getting rumor mill whiplash.

Amazon: $10-50B (wants expanded cloud deals in return)
Microsoft: Under $10B (already deep in the relationship)
SoftBank: Another $30B (after dropping $41B in December)
Total round: $100B, valuing OpenAI at $830 billion

Let that sink in. Eight hundred thirty billion dollars. For a company losing $14B annually while its market share crumbles.

This is peak Silicon Valley circularity—the companies investing billions are literally the same ones selling OpenAI the servers, cloud services, and chips it needs to operate. It's like your landlord investing in your business that only exists to pay rent.

Why this matters: OpenAI's valuation went from $100B (2024) to $300B (March 2025) to potentially $830B (now) all while bleeding market share and cash. They're planning an IPO for late 2026, racing to beat Anthropic to public markets.

The AI arms race isn't a two-horse race anymore. And even the smartest money in Silicon Valley isn't sure OpenAI can maintain its lead. But apparently, when you're this deep in the hole, the only way out is through—with everyone else's money.

FROM OUR PARTNERS

Join GitLab's Transcend event to unlock agentic AI for software delivery

Join GitLab Transcend for an exclusive virtual event exploring the true potential of agentic AI for software delivery. See how teams are solving real-world challenges by modernizing development workflows with AI, get a sneak peek of GitLab's upcoming product roadmap, watch tech demos from product experts, and share your feedback directly with GitLab product experts.

Prompt Tip of the Day

Claude Code just launched a plugin system that lets you create custom skills, agents, and hooks to extend its capabilities—and it's surprisingly simple to build your own.

Here's the quick version: create a folder with a .claude-plugin/plugin.json manifest file, add a skills/ directory with markdown files, and you've got a shareable plugin. For example, a code review plugin is just a folder structure with instructions in SKILL.md.

The coolest part? You can convert your existing .claude/ configurations into plugins with just a few copy commands. Start with standalone configs for personal use (/hello), then package them as plugins when you want to share with your team (/my-plugin:hello).

Our favorite insight: Plugins use namespacing to prevent conflicts—so if three teammates all create a /review skill, they won't collide. Each plugin gets its own prefix (like /code-tools:review), making it safe to install multiple plugins without breaking your workflow.

Check out the full Claude Code plugin documentation to start building.

Don't have Claude Code yet? ChatGPT users can do something similar with Connectors (also known as MCP - Model Context Protocol). Connectors let you integrate external tools and data sources directly into ChatGPT. Instead of building plugins, you connect services like GitHub, Google Drive, or custom APIs to give ChatGPT access to your tools and workflows.

Want more tips like this? Check out our Prompt Tip of the Day Digest for January.

FROM OUR PARTNERS

Editor's Pick: Slackbot

The all new Slackbot is your personal AI agent built in Slack. It understands your context, finds what you need, and takes action instantly.

Why We Love It: It's like having your smartest teammate on speed dial.

Discover Slackbot

Treats to Try

Willow Voice replaces typing with voice dictation that learns your style—Slack messages sound casual, emails sound professional—5x faster, now on Windows, Mac, and iOS—free trial, then $15/month.
Waabi trains one self-driving system that works for both semi-trucks hauling freight and robotaxis picking up passengers (raised $1B).
PaleBlueDot AI shows you real-time prices across 80+ GPU clusters globally and lets you pick the cheapest option for running your models (raised $150M).
RobCo builds modular robot arms you configure for different factory tasks, learning through demonstration instead of programming (raised $100M)—paid only rn ($1,000/month).
Flora gives you one canvas to connect multiple AI models into reusable design workflows using drag-and-drop nodes (raised $42M)—from $16/month.

Around the Horn

Fun fact: You can turn off ChatGPT's validation mode in Personalization. Adjust the tone, add custom instructions, tell it to stop saying 'You're absolutely right!' every time. Problem solved.

OpenAI is preparing for a fourth-quarter IPO (not sure if we knew it was going to be this year or not yet) while pursuing up to $100B in new funding.
ByteDance plans to release its Doubao 2.0 language model, Seedream 5.0 image generator, and SeedDance 2.0 video tool in mid-February, while Alibaba will launch its Qwen 3.5 flagship model around the same time.
DeepSeek is recruiting specialists to build a multilingual, multimodal AI search engine to rival Google and OpenAI, while simultaneously expanding its AI agent capabilities.
OpenAI's Sora app plummeted 45% in January after copyright restrictions blocked popular characters like SpongeBob and Pikachu (and presumably, during this lull period before Disney's IP hit the Sora app…).
Watch: Nikita Rudin from Flexion Robotics on Forward Future Live talking about their software that lets humanoid robots turn commands like "pick up that box" into actual movements across any robot hardware (lots of good questions Matt and Nick!)

FROM OUR PARTNERS

Stop doing CRM admin. Start selling more.

Lightfield is an AI-native CRM used by thousands of startups that automatically builds itself from your emails and calendar in 5 minutes. No manual data entry. No forgotten follow-ups. Just connect and watch it populate your relationship history.

→ Try Lightfield free.