AI Briefing

Today's AI news is about where compute lives, who controls it, and how fast it's leaving Earth. Google is reportedly talking to SpaceX about literally launching AI data centers into orbit, while at home it's previewing 'Gemini Intelligence' on Android plus a new Googlebook laptop class to outflank Apple. Meanwhile Anthropic's controversial Mythos model crosses the Pacific into Japan's top three banks, IBM puts open-weights inference on Cloud, and a voice AI startup that quietly took over Amazon Ring's call queue just hit a $500M valuation.

01IndustryTechCrunch ↗

Google and SpaceX in Talks to Launch AI Data Centers Into Orbit

The Wall Street Journal reports Google is negotiating with SpaceX to put its first prototype orbital data centers in space, part of Google's 'Project Suncatcher' moonshot to scale ML compute beyond Earth. SpaceX is using the pitch to anchor its upcoming $1.75 trillion IPO around the idea that orbital data centers will be the cheapest place to run AI within a few years. Google is reportedly also shopping the launches to other rocket providers, and aims to fly its first TPU-equipped satellites by early 2027.

02ProductTechCrunch ↗

Google Unveils 'Gemini Intelligence' on Android 17 and a New Googlebook Laptop Class

At its virtual Android Show: I/O Edition, Google rebranded its on-device AI as 'Gemini Intelligence' and previewed a new laptop category called Googlebook, built with Acer, Asus, Dell, HP, and Lenovo around Gemini at its core. The pitch is to turn Android from an operating system into an 'intelligence system' with proactive agents that can book classes, make purchases, and draw context from whatever's on screen. Google also confirmed that Gemini will power Apple's upcoming Siri overhaul, beating Apple's WWDC reveal by nearly a month.

03PolicyNews.Az / Kyodo ↗

Japan's Three Megabanks to Gain Access to Anthropic's Restricted Mythos Model

MUFG, Sumitomo Mitsui, and Mizuho are expected to receive Claude Mythos access by the end of May, becoming the first Japanese institutions admitted to Anthropic's tightly controlled Project Glasswing. The deal followed a meeting between the bank chiefs and U.S. Treasury Secretary Scott Bessent during his visit to Tokyo earlier this week, framed as a cybersecurity hedge against the same offensive capabilities that make Mythos too dangerous to release publicly. Japan's regulator has also stood up a public-private working group focused specifically on Mythos-class threats to the financial system.

04ProductIBM Newsroom ↗

IBM Brings Red Hat AI Inference and OpenShift Virtualization to IBM Cloud

IBM launched two new managed services: Red Hat AI Inference on IBM Cloud, powered by vLLM with a model catalog that already includes Granite 4.0, Mistral Small 3.2, Llama 3.3 70B, GPT-OSS-120B, and Nemotron-3-Nano; and Red Hat OpenShift Virtualization Service for migrating and running VMs at scale. The inference service exposes OpenAI-compatible APIs, ships with IBM Cloud IAM, audit logging, and SLA guarantees, and is designed to let enterprises swap open and custom models without managing GPUs themselves. Red Hat AI Inference goes generally available May 22, 2026.

05FundingTechCrunch ↗

Voice AI Startup Vapi Hits $500M Valuation After Quietly Taking Over Amazon Ring's Call Queue

Vapi raised a $50M Series B led by Peak XV at a roughly $500M post-money valuation, after Amazon Ring evaluated more than 40 voice-AI vendors and ended up routing 100% of its inbound customer calls through Vapi's platform. The company says agents on its platform have now handled over a billion calls, with more than a million developers building on it, and enterprise ARR up 10x since its last round. M12, Kleiner Perkins, and Bessemer also participated, signaling that enterprise voice agents are now eating contact-center workloads at industrial scale.