The Wire Weekly
Episode 10 · June 8, 2026 · 00:08:35.48

This Week's Stories

Full Transcript

Alphabet just admitted it can't fund the AI war with search revenue alone. The most profitable ad company on Earth is selling equity for the first time in years. Warren Buffett is writing checks. And Google, Microsoft, and the rest are racing to build something that might actually make their cash cows obsolete. That's the world we're living in. I'm Travis Wright, this is The Wire Weekly. Eighty billion dollars. Let me say that again. EIGHTY. BILLION. DOLLARS. That's what Alphabet is raising right now, and they're not doing it because they're broke. They're doing it because the AI infrastructure war has gotten so capital-intensive that even Google's money printer can't keep up. And when Berkshire Hathaway, Warren Buffett's fortress of conservative investing, decides to cut a check for AI infrastructure... you know something fundamental just shifted. This isn't venture capital. This isn't some Sand Hill Road fund taking a flyer on the future. This is the endgame moving into view. Anthropic just filed confidentially for an IPO. They're beating OpenAI to the public markets. Both of them are looking at TRILLION dollar plus valuations by fall. Not million. Not billion. Trillion. With a T. We're watching the entire AI stack go public while it's still being built. That's like selling tickets to a rocket launch while you're still welding the fuel tanks. And here's what's wild. Anthropic's move isn't happening because they figured everything out. It's happening because the capital requirements are so insane that even the deepest-pocketed VCs can't keep writing checks. Claude's good. Really good. But training runs now cost hundreds of millions of dollars. The next generation? Billions. Per model. You can't fund that on venture terms anymore. You need public markets. You need sovereign wealth funds. You need Warren Buffett. So what does this actually mean? It means the age of scrappy AI startups is basically over. If you're not raising nine figures minimum, you're not playing the foundation model game. Period. The barrier to entry just went from high to damn near impossible. And that's probably by design. Now here's where it gets interesting. Three things happened this week that look unrelated. They're not. SoftBank is writing an eight hundred million dollar check to Agile Robots. Robotics companies have raised twenty three BILLION dollars in twenty twenty six so far. Sam Altman, while everyone's watching OpenAI fight chatbot wars, just backed a nine month old stealth startup called Alfred that's building software to accelerate manufacturing of physical goods. Not software tools. Not productivity hacks. Physical manufacturing. The pattern? Silicon Valley just remembered that atoms exist. For the last three years it's been all language models, all the time. Write better emails. Summarize documents. Generate code. And don't get me wrong, that stuff matters. But you can't eat a chatbot. You can't drive a chatbot to work. You can't have a chatbot assemble your iPhone. The entire Valley spent ten trillion dollars, give or take, teaching AI to write really good text. Now they're pivoting HARD into teaching it to fold laundry, build cars, and run factory floors. SoftBank betting on robotics again might make you nervous. They famously lit billions on fire in WeWork and other disasters. But here's the difference. Last time they bet on robotics, the AI wasn't ready. Boston Dynamics made incredible demos that couldn't scale. Now? The models work. You can actually train a robot to adapt to new tasks without hardcoding every movement. That changes everything. SoftBank's not early this time. They might actually be late. And Altman's bet on Alfred tells you where his HEAD is. Yes, OpenAI's fighting Google and Anthropic in the chatbot arena. But Altman's personal money is going into companies that make STUFF. Physical stuff. That's not a hedge. That's a conviction play. He thinks the next trillion dollar company isn't the one with the best chatbot. It's the one that brings AI into the real world first. Which brings me to the story that made me laugh out loud this week. Zip, a procurement software company you've probably never heard of, just launched five AI agents and they're marketing it as the kill switch for your finance team's shadow AI problem. Here's the setup. Finance teams across the world are copy-pasting contracts, vendor agreements, and budget spreadsheets into ChatGPT because it's faster than their internal tools. Which is great until you realize they're leaking confidential data into a third party system with zero governance. Oops. So Zip built agents that live INSIDE the procurement workflow. Need to summarize a contract? Agent does it. Need to compare vendor pricing? Agent does it. Need to draft a purchase order? Agent does it. All without leaving the system. And they're calling it a "kill switch" for shadow AI, which is fantastic branding for what is essentially "we added ChatGPT to our software before your employees decided to use the free version." But here's why this matters. Shadow AI isn't going away. Employees will ALWAYS use the fastest tool available. If your internal systems suck, they'll route around them. Zip figured out the only way to win is to make the compliant tool faster and better than the rogue one. That's the game now. You can't ban ChatGPT. You have to out-compete it. And while we're talking about infrastructure that actually works, Perplexity just quietly shipped something that might be more important than any single model release this year. They built a router. Not a chatbot. A ROUTER. It decides, mid-query, whether your question gets answered on your device or in the cloud. You don't choose. The system does. Ask it something simple, something private? Stays local. Ask it something complex, something that needs real compute? Hits the cloud. Seamlessly. You don't even notice. And THAT is the future. Not local versus cloud. Both. Simultaneously. Intelligently. We've been having this dumb debate for two years about whether AI should run on your phone or in a data center. Perplexity just said "yes" and built the thing that makes the question irrelevant. This is the same move Apple's gonna make with on-device versus cloud AI in iOS. This is the same move every enterprise AI vendor is gonna have to make. Because users don't care WHERE the AI runs. They care that it's fast, private when it needs to be, and powerful when it needs to be. Perplexity's router is the first real implementation of that vision I've seen actually ship. Oh, and one more thing. AI agents can now swipe Visa cards. Crossmint just launched an API that gives agents tokenized payment credentials with spending controls. Which sounds boring until you realize what it actually means. Your AI can now buy things. Book travel. Pay invoices. Subscribe to services. Without you. With guardrails, sure. But still. The wallet wars just became the AGENT wallet wars. Whoever controls the payment layer for autonomous agents controls a trillion dollar market that didn't exist six months ago. So what do we watch now? First, Microsoft Build starts this week. Satya Nadella's keynote is June second. The agenda looks like a checklist for automating developer jobs. If you write code for a living, this is the keynote that tells you which parts of your job disappear first. Pay attention. Second, Anthropic's IPO timing. They filed confidentially, which means the S-1 is coming in the next few months. If they price north of a trillion and the market holds, every AI company with a pulse is going public this year. The window is open. Watch who jumps through it. Third, physical AI deployments. Not demos. Deployments. SoftBank's writing checks. Altman's writing checks. Nvidia's shipping humanoid robot blueprints. Someone's gonna put a meaningful number of AI-powered robots into a real factory or warehouse this year. When that happens, and it actually works at scale, the narrative shifts permanently. The age of AI as a cool software trick is over. The age of AI as industrial infrastructure is here. And the companies that can't fund it with venture capital are going public. The companies that CAN fund it are printing equity. And the ones that figure out how to make it work in the physical world first? They're the next trillion dollar giants. The infrastructure's being built in public now. The bets are on the table. And the money's too big to fail quietly. See you next week.