The Neocloud Mirage: Mining Pivots, GPU Capitalism, and the Hashtag Inflation Trap

The market is euphoric about Neocloud. Nebius posts a 454% revenue surge, EBITDA turns positive, and the sector rallies. IREN jumps 5%, Coreweave follows. The narrative is seductive: bitcoin miners are pivoting to AI, repurposing their cheap power and land into GPU clouds. It sounds like a win-win. But as someone who has spent the last decade auditing the gap between narrative and infrastructure, I see a different story. Chaos is data in disguise. The real data lies not in the stock price jumps, but in the capital economics of GPU deployment. Let's start with the hook. August 13, 2025. Neocloud stocks are up across the board. Nebius (NBIS) leads with a 34% single-day surge after its Q2 report, then another 3% the next day. IREN and Coreweave ride the wave. The headlines scream "AI Cloud Boom Validated." But validation of what? A 454% revenue increase from $105 million to $582 million sounds staggering. Yet the adjusted net loss is still $33 million, though narrowed 64%. The EBITDA margin of 40% is impressive, but it masks the capital intensity of the business. In my experience, when a capital-intensive infrastructure company reports such a dramatic revenue jump, the first question is not "how much did they earn?" but "how much did they spend to earn it?" Follow the liquidity, ignore the hype. Context: The Neocloud sector is a catch-all term for companies that provide GPU compute for AI workloads, distinct from the hyperscalers (AWS, Azure, GCP). The three companies in focus – IREN (formerly Iris Energy), Nebius, and Coreweave – each represent a different path to the same destination. IREN is a bitcoin miner that is repurposing its existing power infrastructure for HPC. Nebius is a spin-off from Yandex, aiming to be the AI-native cloud for Europe and the Middle East. Coreweave is a pure-play GPU cloud, deeply tied to NVIDIA. The common thread: all are capital-intensive, all depend on NVIDIA's supply chain, and all are betting that AI demand will outrun the colossal GPU supply being deployed. Now, the core of my analysis. The real innovation here is not technical – it's financial. These companies are effectively creating a new asset class: GPU compute as a service, funded by equity and debt. But the economics are fragile. I've seen this playbook before in the DeFi lending protocols of 2020. Over-collateralization was supposed to be safe until underpriced risk collapsed the system. Here, the collateral is the GPU itself. When a company like Coreweave issues debt backed by its NVIDIA H100s, it's essentially a leveraged bet on GPU rental prices staying high. But what happens when the next wave of Blackwell chips makes H100s obsolete, or when AI demand plateaus? Let's use a framework from my years in crypto: the tokenomics of compute. Think of each GPU as a unit of supply. The total supply of GPU compute in the Neocloud sector is growing exponentially. Nebius alone is adding thousands of GPUs per quarter. If you map this to a token model, the "inflation rate" of compute is staggering. The question is whether the demand for AI training and inference can absorb this supply without rental prices collapsing. In my fund, I track a metric I call "effective hashrate inflation" – the percentage increase in total GPU compute available for rent. If this inflation exceeds the growth in AI workloads, the unit economics of each GPU will deteriorate. Nebius may have a 40% EBITDA margin today, but that margin is sensitive to utilization rates. If utilization drops from 90% to 60%, that margin evaporates. The contrarian angle: The market is treating these stocks as proxies for the AI boom, but they are also proxies for the crypto mining cycle. IREN is a prime example. Its stock is still correlated with bitcoin, because its mining division provides a floor. But the pivot to AI is a double-edged sword. In a bull market, the dual narrative amplifies returns. In a bear market, both legs can break simultaneously. The algorithmic has no conscience. The same capital that funded GPU expansion during cheap money will be repriced when rates stay high. Remember, Coreweave's debt is variable-rate, and IREN's power contracts are often indexed to energy prices. The bears are not in the open – they are in the footnotes of the 10-K. Moreover, the regulatory landscape is shifting. Nebius carries a Russian overhang. Its founder, Arkady Volozh, was under EU sanctions until 2024. The company is now a Dutch entity, but any geopolitical turbulence could reignite scrutiny. Meanwhile, the US is tightening export controls on AI chips. If the next round of restrictions limits GPU supply to non-US allies, Nebius's access to NVIDIA's latest hardware could be constrained. The market is pricing in a frictionless global GPU market. That is a fantasy. Volatility is the price of admission. Now, let's zoom out to the macro picture. The Neocloud sector is a microcosm of a larger shift: the industrialization of compute. We are moving from the era of retail crypto mining to institutional AI cloud. But the capital cycle is the same. In 2017, ICOs raised money to build networks that never achieved product-market fit. Today, Neocloud companies are raising billions to build data centers that may or may not be fully utilized. The difference is that the underlying demand for AI is real – but the supply is being built by multiple players simultaneously. The risk of overbuilding is high. I've seen this movie before: the dot-com bubble, the shale oil boom, the crypto mining arms race. Each time, the early movers win, but the latecomers get crushed. What does this mean for the reader? If you are an investor, do not confuse the Nebius revenue jump with a sustainable moat. The moat in GPU cloud is not technology – it's capital access and supply chain relationships. Nebius has a head start in Europe, but Coreweave has the Microsoft contract. IREN has cheap power, but its AI cloud is still nascent. The real test will come in 2026, when the current wave of CapEx hits the balance sheet. That is when we will see which companies have the unit economics to survive a downturn. My takeaway: The Neocloud sector is a fascinating experiment in the convergence of crypto infrastructure and AI demand. But the hype is ahead of the fundamentals. The liquidity is flowing into these stocks because the market is starved for AI exposure. However, the next leg of the market will be determined not by revenue growth, but by capital efficiency. Follow the cash flow, not the headlines. The algorithm has no conscience. The only question is whether the GPU inflation will be absorbed by AI demand before the capital cycle turns. I am positioning my fund for a scenario where the sector consolidates, and the survivors are those with the lowest cost of capital and the strongest customer relationships. That is the only way to profit from the chaos. Chaos is data in disguise. The data right now screams: enjoy the rally, but watch the leverage.