GPU HODL: From Mining Rig to AI Warhorse – Why Depreciation is FUD

Michael Intrator, the co-founder, chairman, president, and CEO of CoreWeave, laid out the bullish case for why GPUs are the ultimate multi-tool, far outlasting their crypto-mining phase. He argues that the hardware’s chameleon-like adaptability, the juicy economics of inference, and a hilariously misunderstood lifespan are completely reshaping the AI infrastructure game.

From Crypto to CGI to Curing Stuff CoreWeave’s own origin story reads like a degen portfolio: starting with GPUs bought for crypto mining, then pivoting to CGI rendering, batch computing, and even medical-research workloads. Intrator says this pivot “shows the versatility of GPU technology” and is a classic example of the market reacting to volatility by relentlessly chasing the next alpha, or in this case, the next compute cycle.

Strategic Tuition‑Style Investments (Or, Paying Your Dues in Silicon) “The initial GPUs were the tuition we paid to learn how to run this business,” Intrator quips. Those early, probably sweat-filled purchases gave CoreWeave the operational scars and knowledge needed to scale properly. He adds that “scaling laws were going to drive… computing de‑commoditizes at scale,” a fancy way of saying that when you're dealing with planet-sized AI models, understanding how compute scales isn't optional—it's the whole thesis.

Inference = The Money Printer Goes Brrr For Intrator, inference isn't just a technical step; it's the cash-printing machine that finally monetises all that eye-watering AI investment. CoreWeave’s compute power is constantly chewing through massive inference workloads, effectively turning the raw, depreciating asset of a GPU into a daily revenue-generating beast.

CoreWeave + Nvidia = The Tip of the Spear (And the Rest of the Spear) CoreWeave proudly positions itself as “the tip of the spear” in shoving Nvidia’s newest, shiniest architectures into commercial production at scale. This partnership underscores a simple truth: in the modern AI arms race, being first with the sharpest GPU is a competitive advantage, not a nice-to-have.

The GPU Depreciation Myth (Sponsored by Short Sellers) Intrator waves off the frantic chatter about rapid GPU depreciation as pure “nonsense,” likely fueled by traders with short positions who wouldn't know a tensor core if it bit them. In the real world, CoreWeave’s clients are locking down compute for five to six years, with an average contract length of five years. Those long-term deals are the ultimate rebuttal, proving GPUs hold commercial value far longer than any bear-market blog post would have you believe.

Lifespan Reality Check (It's Not a Smartphone) The popular notion that a GPU becomes a fancy paperweight after 16‑18 months is, in Intrator's view, “farcical.” He stresses that these workhorses remain viable and profitable for much longer, directly challenging the pervasive, and convenient, misconception that they go obsolete faster than a meme coin's utility.

Demand Fuels Competition (A Good Problem to Have) The explosive surge in demand for AI infrastructure is, naturally, attracting a swarm of new competitors. Intrator sees this not as a threat, but as a definitive sign of a healthy, obscenely profitable market. This competition just fuels more innovation and reinforces the desperate, global need for robust, scalable GPU resources—music to a cloud provider's ears.

Innovative Financing – The ‘Box’ (Where the Magic Happens) To handle the monstrous cash flow dynamics of giant compute contracts, Intrator engineered a structure he bluntly calls “the box.” This mechanism neatly governs all