Beyond Brief Daily — I'm Michael Benatar. AI, tech, business. Let's get into it.

Oracle just executed what might be the largest layoff in company history. Twenty to thirty thousand employees — eighteen percent of their entire workforce — got termination emails yesterday across the US, India, Canada, and Mexico. The kicker? Oracle posted a ninety-five percent jump in net income last quarter. Six point one three billion in profit. They filed a two point one billion dollar restructuring plan with the SEC to fund AI infrastructure their balance sheet apparently can't handle comfortably.

This is the new math. Oracle isn't struggling — they're profitable as hell. But the AI infrastructure race demands so much capital that even a company pulling in six billion quarterly is cutting nearly thirty thousand jobs to stay competitive. Every major tech company is making this choice. Human costs versus compute costs. And compute is winning every time. The restructuring filing makes it explicit — they're eliminating tens of thousands of people to build the data centers and buy the chips they need for AI. When profitable companies start mass layoffs to fund technology transitions, that's your signal the transition is real and expensive.

Amazon just bought thirteen hundred acres in Oregon for what could become an eight to twelve billion dollar data center campus. Exascale computing. One gigawatt of power demand — that's enough to power seven hundred thousand homes. Sixteen to twenty data center buildings on Columbia River land. The scale is insane. Amazon's not building a data center, they're building a small city that happens to run AI workloads.

The Oregon purchase shows how AI is reshaping geography. These aren't just buildings — they're infrastructure investments that rival airports or highways. One gigawatt changes regional power grids. Exascale computing means processing speeds we've never seen commercially. And Amazon's doing this while every other hyperscaler is making similar moves. The land grab for AI-suitable real estate is happening right now, and most people don't realize how much it's going to reshape where technology gets built.

I break this infrastructure race down every morning in the newsletter — theBeyondbrief.com.

Qwen dropped their 3.5-Omni model yesterday and the specs are wild. Full omnimodal — text, images, audio, audio-visual content. It processes over ten hours of audio input and four hundred seconds of 720P video at one frame per second. Trained on more than one hundred million hours of audio-visual data. Speech recognition in one hundred thirteen languages, generation in thirty-six. These aren't incremental improvements — this is AI that actually understands multimedia at human scale.

Defense tech just had its biggest funding day ever. Saronic closed a one point seven five billion dollar round at a nine point two five billion valuation for autonomous ships. Autonomous ships! Defense contractors are now venture capital darlings. The military-industrial complex is getting the startup treatment, and investors are writing billion-dollar checks. Autonomous systems aligned with national security priorities are first-tier venture bets now.

Meanwhile, the AI chip race heated up with Rebellions raising four hundred million in South Korea. Everyone's hunting for Nvidia alternatives, and Asia's becoming a real player in AI hardware. Most chip startups won't make it, but capital keeps flowing to teams that can pitch performance and regional relevance in a market desperate for more supply.

Runway launched a ten million dollar fund and builders program for AI, media, and world simulation startups. They're evolving from a single creator product to infrastructure for other companies. Free API credits, ecosystem building — classic platform play. The AI video leaders are realizing being a destination isn't enough. You need to be the rails other products run on.

Speechify launched a native Windows app using local models for transcription and dictation. Local processing, not cloud-dependent. Lower latency, better privacy, faster workflows. This matters because not every AI task should hit the cloud. On-device processing is the next battleground, and companies are starting to realize local models solve real problems cloud models can't.

The wildcard story is Starcloud raising one hundred seventy million at a one point one billion valuation for space-based data centers. Eighty-eight thousand satellite constellation running AI workloads in orbit using continuous solar power. Sounds like science fiction, but AI's power constraints are real. Land, cooling, energy — these are the actual bottlenecks now, not just chips. When terrestrial infrastructure hits limits, orbital computing starts looking like a long-term infrastructure bet.

We're watching the AI economy reshape everything — labor, geography, infrastructure, even space. Oracle cutting thirty thousand profitable employees to fund compute. Amazon buying cities worth of land for exascale data centers. Billion-dollar defense tech rounds. Space-based computing getting real funding. The companies that understand this isn't about better software anymore — it's about rebuilding the entire stack from power generation to orbit — those are the ones that survive the next phase. Everyone else is optimizing for a game that's already over.

That's your brief. Follow the show on Instagram @thebeyondbrief, find me on X @MichaelBenatar, and if you want this in your inbox every morning — theBeyondbrief.com. I'm Michael Benatar. See you tomorrow.