OpenAI is spending $750b on compute anthropic can't match

I use Claude daily and OpenAI's stuff on the side, and the compute gap between these two companies is starting to get me angry... WSJ reported last week OpenAI now expects to spend $750B on infra by 2030, up 25% from their estimate earlier this year. They're also building a 3.2GW data center in Georgia, $30B price tag. It looks insane. Now their valuation is scraping $1T ahead of IPO and it looks like maybe they are right ? Anthropic have a different strategy... They've been leasing compute through a pile of deals, SpaceX, Amazon, Microsoft, Google, and now Meta and AMD too. They do have one build-out, $50B with Fluidstack for data centers in Texas and New York, but it's small next to what OpenAI's doing. So, Claude users hit rate limits and outages way more than GPT users do. Fable 5 launched gated to Max and Team Premium only, at 50% of normal weekly limits. That's not a pricing decision, that's a "we don't have enough chips" decision. Apollo's head of thematic investing put it well: "in a compute-constrained world, access itself becomes a competitive moat." Owning the stack is a leverage over your own roadmap. Not saying OpenAI's approach is automatically the winning bet, an analyst I trust pointed out both companies are probably burning cash faster than revenue can catch up, and we won't really know who made the right call until the IPO filings force real numbers into the open. But right now, one of them is compute-rich and shipping, and the other is rationing its own flagship model to its paying users. Here, i'm tired of Claude's limits to be honest...