I think people are slowly discovering that buying $2B worth of GPUs and actually turning $2B worth of GPUs into revenue are two very different businesses. You can have warehouses full of Blackwells, half finished data centers, transformers arriving sometime before the heat death of the universe, cooling systems that aren't ready, networking that doesn't work, and lenders suddenly discovering what credit risk is. Congratulations, you've built the world's most expensive collection of paperweights. Meanwhile the guy across the street with functioning infrastructure is renting out older GPUs at rising prices. Apparently this is what a compute glut looks like now. Everyone saw the shortage and rushed to build. That rush overloaded the same electrical supply chains, construction capacity, engineering talent and financing markets needed to solve it. Now the bottlenecks themselves are preventing new supply from arriving. The beautiful irony is that the more projects get delayed or stranded, the more valuable the capacity that actually works becomes. Scarcity doesn't just raise prices. It starts selecting winners and losers. The real moat is coordinating several moving parts into something that actually produces billable compute before the hardware becomes yesterday's technology. That's the refining analogy we've been hammering on. Owning crude doesn't make you a refinery. Owning silicon doesn't make you an AI factory. The value is in converting a bunch of constrained inputs into something customers can actually use. And every month you screw that up, your competitor collects the scarcity rent while your debt collects interest. Lovely business. Now here's where 2027 potentially gets nasty. Some of these guys won't die because AI demand disappeared. They'll die because they were right about demand and catastrophically wrong about the time and capital required to supply it. Stranded GPUs, unfinished shells, expensive debt, angry customers. And who comes shopping? The operators with functioning infrastructure, liquidity and the ability to commission capacity. Suddenly those failed projects become someone's discounted expansion pipeline. That's where $IREN and $NBIS could become interesting . they might graduate from fighting for market share to consolidating the industry. Maybe through acquisitions, partnerships, distressed hardware deployments, customer funded infrastructure, whatever. But that's conditional on execution. $IREN isn't immune to the same physics just because we like the ticker. My bet is that 2027 is about the brutal sorting of industrial competence. The industry may end up simultaneously oversupplied with purchased silicon and undersupplied with working compute. Some shareholders get vaporized, some lenders discover religion, and the survivors inherit infrastructure for pennies on the dollar.