Google Cloud doubles and Anthropic buys a one-person biotech

Two of the biggest companies on earth reported the same quarter and the market read them in opposite directions. Alphabet doubled its cloud business and lost 9% of its value after hours; Microsoft grew less and was rewarded. Meanwhile Anthropic quietly bought a biotech with exactly one employee, DeepMind released a world model you can rewrite while it runs, and the FCC cleared the satellite deal behind your iPhone's emergency SOS.

Google Cloud grew 103% and the stock fell anyway

Alphabet reported $127.58 billion in Q3 revenue, up 25% year over year, with Google Cloud at $30.71 billion versus $15.16 billion a year earlier — a 103% jump. Operating income rose 30% to $40.58 billion.
Capex hit $51.36 billion in the quarter, against $23.95 billion a year ago. Profit available to shareholders fell 25% to $26.17 billion, dragged down by an $8.66 billion loss on equity stakes in other companies (versus a $10.73 billion gain last year). Class A shares traded at $294.20 after hours, 9.4% below the $324.89 close.

Why it mattersDoubling a $15 billion business in a year is extraordinary, and it still wasn't enough: the market has stopped paying for growth and started pricing in whether $51 billion a quarter of capex ever comes back.

Azure grew 48% and the OpenAI stake finally added money

Microsoft reported $92.3 billion in revenue for its fiscal first quarter, up 19%, with Microsoft Cloud at $63.9 billion (up 30%) and Azure growing 48%. Net income was $36.6 billion, up 32%, and diluted EPS $4.92.
The OpenAI investment flipped sign: net gains added $207 million to net income and $0.03 to EPS this quarter. A year earlier, losses from that same stake cut $3.1 billion off net income and $0.41 off earnings per share.

Why it mattersSame quarter, same AI bill, opposite verdict. Microsoft can point to a 48% growth line customers are already paying for; Alphabet is asking investors to wait.

Anthropic bought a biotech with one employee

Mithril Medicine had exactly one full-time employee: founder Cody Tranbarger, who had been posting under the pseudonym "Professor Oak" on X. Anthropic closed the acquisition in July 2026 and disclosed it on October 27; terms were not revealed. Using Claude, Tranbarger developed candidate treatments for two rare forms of muscular dystrophy — LAMA2-related congenital muscular dystrophy and alpha-dystroglycanopathy — in under 18 months.
In a six-week mouse study of a severe LAMA2 form, the highest antibody dose roughly doubled grip strength and muscle force versus untreated animals; four of eight untreated mice died between two and six weeks of age, against one of eight treated. Anthropic calls the model "Pipeline-in-a-Person", aims to pick formal candidates in 2027, and is working with six partners: Bristol Myers Squibb, BigHat Biosciences, Generate Biomedicines, Divergence Labs, Enveda and The Column Group.

Why it mattersThese are preliminary mouse results and nothing has been tested in humans — trials are penciled in for late 2028 at the earliest. The real news is the cost structure: if one person plus a model can run a rare-disease pipeline, the economics of diseases too small to be profitable flip completely.

Genie 4: worlds you can rewrite while they run

Google DeepMind put Genie 4 into research preview on October 27. It generates photorealistic interactive worlds that hold a consistent scene for up to two minutes — double Genie 3 — and it now produces matching audio by itself: music, ambient noise, interactive dialogue. You type prompts mid-simulation to change the weather, move the camera, or add people and objects. Everything carries a SynthID watermark.
Gemini acts as a "director" that plans steps and checks frames while Genie renders. In a Gemini Robotics experiment, tasks were broken into steps Genie 4 simulated in real time. Access is limited to trusted testers, with a rollout coming for Google AI Ultra subscribers in Project Genie and early-access partners of the Gemini API and AI Studio. Limits: detail drifts over longer runs, audio accuracy is unsolved, and text inside the world is still illegible.

Why it mattersThe cheapest place to train a robot is anywhere that isn't the real world. A simulator you can edit by talking to it is how robotics stops being bottlenecked by hardware time.

The FCC cleared Amazon's $11.5B buy of the satellite behind your iPhone's SOS

On October 27 the FCC approved transferring Globalstar's licenses to Amazon, finding the deal serves the public interest with no competitive harm. Globalstar becomes a wholly owned Amazon subsidiary operating as an affiliate of Kuiper Systems (Amazon Leo) while keeping its name. The transaction, announced in April, is valued at about $11.5 billion and covers Globalstar's satellite operations, infrastructure and L-band and S-band spectrum; closing was originally expected in 2027.
Globalstar is the network behind Apple's Emergency SOS on supported iPhones and Apple Watches, and Apple held roughly 20% of the company before the deal. Amazon filed in April for a direct-to-device constellation of up to 5,105 satellites, and the FCC recently authorized Globalstar's next-generation C-3 system: 48 low-orbit satellites plus six on-orbit spares, built by MDA Space.

Why it mattersSatellite-to-phone is a coverage story, and Amazon just bought the spectrum to fight Starlink over it — with Apple's emergency traffic already riding on the network it acquired.

The big picture

Four of today's five stories are the same story: capital meeting physical reality. Alphabet spent $51 billion in one quarter and was punished; Microsoft spent too but can show a 48% growth line customers are already paying for, and was rewarded. Anthropic's one-person biotech and DeepMind's editable simulator are the same bet — that a model can replace an expensive apparatus, a lab, a test track. And the governance layer is scrambling to catch up: Delaware's AI Commission is drafting a corporate category called the Artificial Intelligence Company, where an AI agent would direct day-to-day business and the single member would be liable mainly for keeping it funded. It isn't law yet, and legislation could be introduced as early as 2027, but on October 26 House Oversight chairman James Comer sent Governor Matt Meyer a formal document request, worried that investors could keep the upside while blaming an algorithm for the downside.

Go deeper

Sources & further reading

Topics:AI Big Tech Space