Galaxy.ai Logo

OpenAI drops $300B on Oracle servers

·

📖4 min read

Cover Image for OpenAI drops $300B on Oracle servers

☁️ OpenAI Signs $300 Billion "Please Don't Let Us Run Out of GPUs" Deal

  • OpenAI just agreed to buy $300 billion worth of cloud compute from Oracle over 5 years. That's roughly $164,000 per day. Every day. For five years.

  • This is what desperation looks like when wrapped in a press release. OpenAI is basically saying "we'll mortgage our future to avoid another ChatGPT outage during a demo." (source)


💰 Replit Raises $250M, Promises AI That Actually Codes

  • Replit hit a $3 billion valuation and launched "Agent 3" — an AI that can supposedly code, test, and debug without having a mental breakdown every 10 minutes.

  • We're one step closer to AI replacing junior developers, and three steps closer to AI creating bugs that no human can understand. (source)


🚀 Perplexity Somehow Convinces VCs It's Worth $20 Billion

  • The "Google killer" that nobody asked for just raised $200 million at a $20 billion valuation. That's more than some actual profitable companies.

  • Either Perplexity cracked the code on search, or we're witnessing the most expensive hallucination in venture capital history. (source)


🏛️ Ted Cruz Wants an "AI Sandbox" (What Could Go Wrong?)

  • Senator Cruz introduced a bill for 2-year regulatory exemptions for AI companies. Because when you think "tech oversight," you definitely think "Ted Cruz."

  • Nothing says "innovation" like letting AI companies self-regulate while using Americans as beta testers. (source)


📜 California Actually Does Something About AI Chatbot Girlfriends

  • A new bill would regulate AI companion chatbots to protect minors. Companies would have to remind users they're talking to code, not their soulmate.

  • Finally, someone's addressing the elephant in the room: people are getting emotionally attached to autocomplete. (source)


🧠 Secret Lab Tries to Make AI Less... Random

  • Thinking Machines Lab (run by OpenAI's ex-CTO) is working on making AI responses more consistent. Revolutionary concept: computers that give the same answer twice.

  • Turns out "it depends" isn't a great answer when you're trying to run a business on AI. (source)


🎶 Stability AI Launches Corporate Muzak Generator

  • Stable Audio 2.5 can create 3‑minute tracks in seconds. Perfect for when you need elevator music but want it to sound like it was made by a robot.

  • Human musicians can now add "background music for corporate videos" to the list of jobs AI is coming for. (source)


🤖 Box Turns Your Documents Into AI Playgrounds

  • Box Automate lets AI agents loose on your business workflows. What could possibly go wrong with letting AI handle your legal reviews?

  • We're moving from "AI can write emails" to "AI can accidentally approve million‑dollar contracts." (source)


🏗️ Data Centers Get $40 Billion, Planet Gets Hotter

  • U.S. data center construction hit $40 billion in June. That's a lot of air conditioning for computers that argue with humans on the internet.

  • The AI revolution is literally concrete — and it's using more electricity than some small countries. (source)


📈 Oracle Stock Goes Brrrr, Drags Market Along

  • Oracle's stock jumped 36% in one day after reporting massive AI demand. The entire market decided this was worth celebrating.

  • Apparently, "we sell servers to AI companies" is the new "we have a blockchain strategy." (source)


📱 Apple Releases iPhone Air, Still Confused About AI

  • The new iPhone Air has an AI‑tuned processor, but analysts think Apple is still playing catch‑up in the AI game.

  • Even Apple can't just slap "AI" on a thin phone and call it innovation anymore. (source)


💸 Klarna Admits It Maybe Fired Too Many Humans

  • After going all‑in on AI cost‑cutting, Klarna's CEO admits they "probably over‑indexed" on replacing people with bots.

  • Turns out customers prefer talking to humans who understand their problems, not chatbots that suggest turning it off and on again. (source)


🤖 AI Simulates Fed Meeting, Discovers Politics Exists

  • Researchers used AI to simulate a Federal Reserve meeting and found that political pressure makes decisions messier. Shocking.

  • We've reached peak AI: using artificial intelligence to discover that humans are influenced by politics. (source)


🔧 Claude Takes a Nap, Everyone Panics

  • Anthropic's Claude had a brief outage, reminding everyone that AI tools are still just fancy websites that can break.

  • We've become so dependent on AI that a chatbot going down for 20 minutes feels like the apocalypse. (source)


Until tomorrow (assuming the data centers don't melt),
Team Galaxy.ai

P.S. $300 billion could buy a lot of things. Like... Elon Musk's next company.