FullyPorted

If Stripe had its way, it would spend $63 billion this quarter on two seemingly unrelated acquisitions: PayPal and OpenRouter Here's why Stripe wants both, and why the two deals are really one thesis

IMHO there's a more nuanced way to look at this. The monetization function doesn't actually matter that much. Whether you're hashing SHA-256 or running LLM inference, you're ultimately converting electricity into a monetary instrument. Once that conversion becomes reliable, capital markets follow. But lumping all crypto mining together is a bit of a gross generalization. Bitcoin (ASIC dominated) and GPU mining taught two very different lessons. Bitcoin proved you could monetize stranded or underutilized energy almost anywhere in the world. It created the first truly global buyer of electricity that was location agnostic, interruptible, and almost infinitely composable. That fundamentally changed how people thought about energy infrastructure. GPU mining solved a very different problem. Companies like CoreWeave, early on, became exceptionally good at rapidly deploying, orchestrating, and reallocating fleets of GPUs. What we used to call speculative mining, constantly redirecting compute toward the highest value PoW workload on behalf of clients, looks surprisingly similar to modern AI infrastructure, albeit at a fraction of today's scale. AI workloads are fundamentally different operationally. Crypto miners, by design, can shut down in seconds. In fact, we built both physical and economic systems around that capability. The economic consequence is borne by the miner. AI infrastructure typically carries customer workloads, SLAs, storage, networking, and stateful processes. Reliability requirements are in an entirely different class. That's why Crusoe starting on flare gas is so interesting (well timed pivot). Variable power is perfectly acceptable for Bitcoin mining. It's much harder to run reliable AI infrastructure on intermittent generation. For NVIDIA specifically, crypto absolutely accelerated GPU demand, CUDA adoption, and developer mindshare. It also created years of inventory volatility. It's difficult to argue NVIDIA simply benefited or simply suffered. Both are true. The GeForce 10 series and the years that followed are probably the best example of that tension (long live the 1080ti). I agree with you in general however, the real bridge wasn't crypto itself, It was the operational knowledge the industry accumulated: how to finance compute, negotiate power, source sites, deploy hardware at scale, manage thermals, optimize utilization, and ultimately treat compute as an asset class rather than just IT infrastructure. While I was running North America, Europe, and South America for Binance Pool in 2020-2021, I pitched hashrate futures and secondary markets for compute capacity. At the time, they were considered too early. Maybe the market simply wasn't ready. Looking back, I think those markets eventually emerged, just inside AI infrastructure, inference, and financing before filtering up into public capital markets. Or perhaps they're still in the process of emerging. I think I can speak for many people who were deeply involved in mining. At some level, we knew we weren't just building for Bitcoin. We were building systems for monetizing compute and electricity. AI simply became the next buyer, and for now, the more profitable taker.