Nvidia fuels OpenAI's compute chase

PLUS: Use GPT-5 in Microsoft 365 to analyze emails

Good morning, AI enthusiasts. OpenAI’s compute ambitions just got a major boost — with Nvidia promising to invest up to $100B in the company.

The deal will put millions of Nvidia GPUs into OpenAI’s multi-gigawatt data centers and power its next-gen AI models. But the big question remains: is this the real path to superintelligence, or just an endless money loop between major AI power players?

In today’s AI rundown:

  • Nvidia fuels OpenAI’s compute chase with $100B

  • Meta brings AI into the dating experience

  • Use GPT-5 in Microsoft 365 to analyze emails

  • Google to tackle AI’s shutdown resistance

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

OPENAI & NVIDIA

Image source: Reve / The Rundown

The Rundown: Nvidia and OpenAI just announced the “biggest AI infrastructure project in history” — with the Jensen Huang-led company intending to invest up to $100B in OpenAI as it uses its advanced GPUs for training and running next-gen AI.

The details:

  • The companies have signed a letter of intent to deploy 10 GW worth of Nvidia systems — representing millions of GPUs — to power OpenAI’s AI infrastructure.

  • With each GW, Nvidia will progressively invest in OAI to support its deployment, including data center and power needs, with the total going up to $100B.

  • The first gigawatt of the project is expected to come online in the second half of 2026, using Nvidia’s next-gen Vera Rubin platform.

  • OAI said it will treat Nvidia as a “preferred strategic compute and networking partner,” as it maintains ties with Microsoft and Oracle for the infra push.

Why it matters: With this deal, Nvidia is giving OpenAI the capital boost it needs to secure compute and power capacity for the next era of intelligence — and locking in a major customer for its GPU business. However, many also see it as an infinite money loop, with capital simply cycling between OpenAI and partners Oracle and Nvidia.

TOGETHER WITH SANA

The Rundown Sana’s AI agents connect with 100+ tools—Gmail, Outlook, Teams, Slack, Salesforce, Google Drive, BigQuery, and more—so every workflow is powered by instant answers and automation. No more silos, no more wasted time. When AI is everywhere you work, your team moves faster, stays focused, and delivers more.

Use Sana Agents to:

  • Orchestrate actions across all apps and data

  • Instantly search, summarize, and automate without switching context

  • Slash operational overhead by 10-20x

META

Image source: Meta

The Rundown: Meta is bringing AI into the Facebook Dating experience with two new features — dating assistant and Meet Cute — that help users find perfect matches without the dreaded “swipe fatigue.”

The details:

  • The dating assistant is an AI chatbot that helps users find relevant matches from text prompts and suggest dating ideas and tips to level up their profile.

  • With text prompts, the assistant focuses on unique interests and preferences, going beyond basics like height or education to make the search more tailored.

  • Meet Cute, on the other hand, will be entirely automated and give a “surprise match” every week, based on Meta’s personalized matching algorithm.

  • Meta says the feature is for anyone who’s “tired of swiping” but notes that users can opt out of it any time they want.

Why it matters: Tackling swipe fatigue with AI could give Facebook Dating serious ammunition against Tinder and Bumble's dominance. But advanced AI in dating demands careful monitoring — you definitely don't want the algorithm hallucinating wildly off-base matches and creating cringeworthy experiences.

AI TRAINING

The Rundown: In this tutorial, you will learn how to leverage GPT-5 through Microsoft Copilot to automatically search your email history, analyze complex threads, and generate personalized replies that perfectly match your writing style.

Step-by-step:

  1. Open Microsoft Edge and click the Copilot ribbon (top right) — sign in with your Microsoft account for free access.

  2. Enable “Smart” mode in Copilot to connect your Outlook data.

  3. Prompt: “Summarize my most recent 10 emails with bullet points on what needs replies today, then draft responses in my usual tone.”

  4. GPT-5 analyzes your entire email history, extracting key decisions, recent developments, and your typical communication patterns.

  5. Review the AI-generated reply and refine with prompts like “Make this more formal” or “Add timeline details.”

Pro tip: Create context-aware templates by prompting “Analyze my email patterns with executives vs. team members, then draft this using my appropriate tone.”

PRESENTED BY SYNK

The Rundown: Snyk, a founding partner of The AI Security Summit, brings together leaders and security teams to build trust for AI initiatives and bridge the “AI Security chasm.” Join The AI Security Summit  live in San Francisco on Oct. 22–23 for hands-on learning, community building, and sessions from leading experts.

Agenda highlights include:

  • Agentic Identify & Security Panel with leading agentic AI companies

  • Deep Dive Talk: AI Code Review In Depth

  • Panel Discussion: Building AI Native Applications

  • Happy hour with Vibe Coding DJ (even security people need to unwind)

AI SAFETY

Image source: Reve / The Rundown

The Rundown: Google DeepMind just released Frontier Safety Framework 3.0, expanding its AI risk monitoring efforts to cover emergent AI behaviors like shutdown resistance and persuasive ability that could complicate human oversight.

The details:

  • The updated framework will track whether frontier AI resists attempts to turn them off or modify their operations — a risk flagged in recent external studies.

  • It will also monitor models for unusually strong influence on human beliefs and behaviors, which could potentially lead to harm in high-stakes contexts.

  • DeepMind also sharpened its Critical Capability Level definitions to specifically identify critical threats warranting immediate governance and mitigation efforts.

  • To address CCL’s risks, the company will conduct safety reviews before external launches and even track its internal deployments made for R&D.

Why it matters: DeepMind’s move underscores a broader shift, where AI leaders, including Anthropic and OpenAI, are not just flagging current risks but also tightening protocols to brace for what could happen in the future. As models gain unpredictable behaviors, these efforts will be the key to building truly safe superintelligent systems.

QUICK HITS

  • 🧠 Sembly - The only AI Notetaker with a “brain” that unlocks meeting intelligence*

  • 📷️ Photo Genius - Genspark’s voice-controlled AI photo editor

  • 📧 Email Assistant - Perplexity’s AI assistant to automate email workflows

  • 🤖 Qwen3-Omni - AI that unifies text, image, audio, and video into one

*Sponsored Listing

Perplexity launched an Email Assistant that automates tasks like scheduling meetings, drafting replies, and adding labels in Gmail/Outlook, available to Max users.

Alibaba’s Qwen team dropped three new open-source AI models, including Qwen3 Omni, Qwen3 TTS, and Qwen-Image-Edit-2509.

Nvidia announced an investment in the UK-based AI voice startup ElevenLabs, just days after the U.S. state visit to the UK.

Google announced it is starting the rollout of Gemini for TVs, a move that will take its AI to over 300M active Google TVs and Android TV OS devices.

The U.S. General Services Administration added Llama to its list of approved AI tools for federal agencies, following models from Google, OpenAI, and Anthropic.

Ice cream giant Magnum is set to use Chilean startup NotCo’s AI to reformulate its products and launch new plant-based ones.

COMMUNITY

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Philip W. in Bossier City, Louisiana:

“I am using Midjourney + HeyGen to advertise my book, The Piper. Using Midjourney, I generated the image of one of my characters, Lucia, then uploaded it to HeyGen as an Avatar. I then added speech to the Avatar and had her talk about the book from her perspective. I posted it to Facebook. This helps creatively advertise my book.”

How do you use AI? Tell us here.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown

Reply

or to participate.