The Wire Weekly
Episode 8 · May 25, 2026 · 00:08:08.18

This Week's Stories

Full Transcript

DeepSeek just raised more money than OpenAI's last round while openly telling investors that turning a profit is... not really the priority right now. Anthropic hit a nine hundred billion dollar valuation without taking a single defense contract, then immediately turned around and bet forty-five billion dollars that Elon Musk can keep their servers running. And SoftBank's founder, the guy who lost sixteen billion on WeWork, just went ALL IN on Sam Altman with sixty billion reasons this time will be different. That's the world we're living in. I'm Travis Wright, this is The Wire Weekly, and we need to talk about what happens when the money gets so big it stops making sense. Here's the thing nobody's saying out loud. We just watched the entire AI funding landscape go completely vertical in the span of seven days. DeepSeek raising ten billion. Anthropic closing thirty billion at a valuation that puts them ahead of Boeing. SoftBank dumping sixty billion into OpenAI like Masa Son just discovered a new religion. And the wildest part? These aren't even the same KIND of bets anymore. DeepSeek is literally telling investors... we're gonna give this stuff away. Open source. Free. The whole business model is building AGI first, figuring out revenue never. Meanwhile Anthropic, the company that publicly said no thanks to defense contracts because of safety concerns, just committed to paying SpaceX one point two five BILLION dollars a month for the next four years. Let me say that again. One point two five billion. Per month. That's not a typo. That's forty-five billion dollars to rent compute from Elon Musk's Colossus data centers in Tennessee. The same Elon who's apparently now trying to buy Cursor, the AI coding tool that somehow hit three billion in annual revenue before most people even knew it existed. This is not normal scaling. This is not even normal Silicon Valley crazy. This is something else entirely. And I've been in crypto for nine years, I've seen some absolutely ridiculous capital flows, but this feels different. Because the people writing these checks are not dumb. They're not retail investors FOMOing into dog coins. They're sophisticated institutions making bets so large that if they're wrong, the collateral damage reshapes entire industries. So what's actually happening here? Three things this week that look unrelated but are absolutely part of the same pattern. First, Nvidia. Jensen Huang stood up in front of investors and said the quiet part out loud. The next trillion-dollar companies won't be BUYING Nvidia chips. They'll be built ON TOP OF them. Nvidia's not pivoting to software, they're positioning themselves as the infrastructure layer for the entire AI stack. Rent the GPUs, build the ecosystem, take a piece of everything that runs on the metal. And here's the kicker. Same week, Jensen publicly told Super Micro, one of his biggest server partners, to get their compliance house in order. That doesn't happen unless the supply chain risk is REAL. Taiwan just detained three Super Micro employees. Nvidia's CEO doesn't burn political capital on a public callout unless he's genuinely worried about operational risk breaking the entire pipeline. Second story. Exa Labs. You probably haven't heard of them. They just raised two hundred fifty million at a two point two billion valuation to build search infrastructure for AI agents. Not search for humans. Search FOR agents. Google didn't even get a chance to compete here because Google is still building for people typing questions into boxes. Exa is building the pipes for machines that need to retrieve structured data at scale. Andreessen Horowitz led the round. And if you connect this to the Cursor story, the SpaceX story, the DeepSeek story, you start to see it. The market is bifurcating. There's the AI we USE. Chat interfaces, coding assistants, image generators. And then there's the AI INFRASTRUCTURE that makes all of that possible. The search layers. The compute farms. The retrieval systems. The companies building infrastructure are raising at valuations that make the 2021 crypto bull market look quaint. Third piece. OpenAI just accidentally proved why building on rented land is still a terrible idea, even when your landlord is worth a hundred and fifty-seven billion dollars. A team spent months building Critterz, the first AI-generated feature-length film, using OpenAI's Sora. Got it ready for Cannes. Marketing locked in. Press lined up. And then OpenAI shut down the tool. Just... turned it off. The film never screened. The whole thing became a cautionary tale in forty-eight hours. And look, I get it. OpenAI's gotta do what's best for OpenAI. But if you're building a business on top of someone else's model and they can just pull the rug because of safety concerns or strategic pivots or whatever, you don't have a business. You have a very expensive hobby. Now let's talk about the story that's just... pure chaos. SoftBank. Masayoshi Son. The guy who famously lost sixteen billion dollars on WeWork, who's had more bad bets than a Vegas regular with a drinking problem, just committed SIXTY BILLION DOLLARS to OpenAI. And apparently his own people are spooked. Bloomberg reported that insiders at SoftBank are nervous. Because Masa's not treating this like an investment. He's treating it like a religious conviction. He genuinely believes Sam Altman is building the most important company of the century and he's willing to bet the entire fund on it. Here's what makes this hilarious and terrifying at the same time. Masa might be RIGHT. That's the thing. WeWork was a real estate company pretending to be a tech company. OpenAI might actually be building AGI. The difference is... we won't know until it's way too late to adjust the position. And if Masa's right, sixty billion looks like the bargain of the century. If he's wrong, this makes WeWork look like a rounding error. Okay. Wild card time. Let's talk about what happens when you're DeepSeek. You're a Chinese AI lab. You embarrassed Silicon Valley earlier this year by building comparable models for a fraction of the cost. You're open-sourcing everything. You're telling the West that the moat isn't the model, it's the application layer. And now you're raising ten billion dollars while openly saying... yeah, profitability is not the goal here. This is functionally a state-backed moonshot with venture capital aesthetics. And the West is paying attention because if DeepSeek is right, if you CAN build AGI-level systems without commercializing every layer, then the entire venture model for AI in the US is built on a flawed assumption. OpenAI charges for API access to fund research. Anthropic raises at insane valuations to justify the compute spend. DeepSeek just says... we'll give it away and figure out the business model later. Or never. That's not a bug. That's a feature. And it's going to force every Western AI lab to answer a question they've been avoiding: are you building a product or are you building a public good? Alright. What to watch. Three things. One. Anthropic's spending forty-five billion on SpaceX compute over four years. If that deal gets renegotiated, restructured, or breaks for ANY reason, it tells you something about how confident they really are in the scaling laws. Watch that contract like a hawk. Two. Cursor's revenue trajectory. Three billion annualized in April. If Musk acquires them and integrates it into the SpaceX or X ecosystem, that's a signal that AI coding tools are about to go from niche productivity hack to core infrastructure. And if the acquisition falls through, that's ALSO a signal. Three. Watch what DeepSeek actually does with ten billion dollars. If they keep open-sourcing everything, that's ideological. If they start closing parts of the stack, that's commercial. The delta between those two paths is the difference between a research lab and a competitor. We're not in the AI hype cycle anymore. We're in the AI COMMITMENT cycle. The money's too big to pull back. The bets are too public to quietly exit. And the infrastructure layer is getting built whether or not the application layer ever makes sense. Forty-five billion dollars says Elon can keep the lights on. Ten billion says you can give away AGI research and still win. Sixty billion says Sam Altman is the chosen one. And somewhere in the middle of all that, Nvidia's just selling shovels and telling everyone else to figure out where to dig. I'm Travis Wright. That's The Wire Weekly. See you next Tuesday.