Nebius' Q2: The Neocloud Inflection Point the Market Is Sleeping On

Let’s cut straight to the tape. Nebius just dropped a Q2 that shouldn’t be possible. Revenue up 454% to $582.3 million. Adjusted EBITDA flipped positive to $236.2 million. Net loss narrowed by 64%. The market rewarded it with a 34% single-day rip. And as I sit here, watching IREN and Coreweave ride the coattails with 5% and 3% gains, I’m asking myself: is this real, or just another narrative pump ready to fade? I’ve been doing this long enough to know that headline numbers can hide a multitude of sins. In 2018, I manually executed 50+ swaps on Uniswap’s testnet, documenting every slippage failure. That taught me that theoretical promises mean nothing until you see the data. Pain is just data you haven’t decoded yet. So let’s decode Nebius. First, the context. The Neocloud sector is the intersection of AI infrastructure and the legacy of blockchain mining. These are companies that either started as crypto miners (IREN) or pivoted from that space (Coreweave) to offer GPU-as-a-service. Nebius is the odd one out—it’s a spin-off from Yandex, built on a foundation of AI-native cloud services. But all three are capital-intensive beasts, burning cash to build data centers filled with NVIDIA’s latest chips. The narrative is simple: AI training demand is insatiable, and these companies are the picks-and-shovels suppliers. The question is execution. Nebius’ Q2 is the first hard evidence that the model works at scale. A 454% revenue jump isn’t just a good quarter; it’s a structural break. To put it in perspective, I’ve seen DeFi protocols claim similar growth, but they were often inflated by token incentives and wash trading. Nebius is a Nasdaq-listed company with audited numbers. The EBITDA margin of ~40.6% suggests they’re past the break-even point on their GPU clusters. When I backtested my ETF integration strategy in 2024, I learned that institutional buying pressure shows up in volume and margin expansion. This is textbook. But let’s drill into the core. The revenue growth implies massive GPU deployment. Nebius’ AI Studio and LLM inference services are seeing real demand. I’ve run my own AI-trading agent experiments in 2026—I know how quickly compute costs can eat into profits. The fact that Nebius is generating positive EBITDA means they’re either running at high utilization or have secured favorable pricing from NVIDIA. Either way, it’s a signal that the unit economics are healthy. Now, contrast with IREN and Coreweave. IREN’s 5% gain is purely sympathetic—their financials haven’t been released. IREN is a Bitcoin miner transitioning to AI, leveraging their power infrastructure. I’ve tracked this miner-to-AI pivot since 2023. It’s a compelling thesis, but the execution gap is wide. IREN doesn’t have the same scale as Nebius yet. Coreweave is a pure GPU cloud player, but they’re private (or IPO’d recently, depending on the timeline). The lack of disclosed data means we’re trading on story, not substance. Market noise is just fear wearing a suit. Right now, the noise is bullish. But the contrarian angle is what keeps me awake. Nebius’ revenue growth is incredible, but it hides two risks. First, client concentration. If a single customer (like a large AI lab) accounts for a significant chunk of that $582 million, the revenue is fragile. Second, CapEx. To sustain growth, Nebius must pour billions into new data centers. If they over-leverage, a downturn in AI demand or a GPU supply shortage could crush them. I survived the Terra-Luna collapse in 2022 by refusing to panic-sell; instead, I executed flash loans to preserve capital. That crisis taught me that balance sheet health matters more than growth rates. And then there’s the geopolitical wildcard. Nebius’ founder, Arkady Volozh, was previously sanctioned by the EU. The sanctions were lifted, but the Russian connection lingers. If the political climate shifts, Nebius could face regulatory scrutiny that no amount of EBITDA can fix. The market is pricing this risk at zero. The candlestick doesn’t lie, but your bias might. For IREN, the risk is different. As more miners pivot to AI, the GPU rental market could face a supply glut. I’ve seen this in crypto mining—when too many ASICs come online, the hashprice crashes. The same logic applies to GPU compute. If IREN’s AI business doesn’t scale fast enough, they’ll be stuck with excess capacity. And their Bitcoin mining leg is still exposed to BTC price volatility. I call this the “dual beta” trap—when both sides of the business move in the same direction, it’s a rocket. When they diverge, it’s a freefall. So what’s the forward-looking takeaway? The Neocloud thesis is validated by Nebius’ numbers. But chasing a 34% day is not my style. I need to see the next CapEx guidance and the client concentration breakdown. If Nebius can maintain EBITDA margins above 30% while doubling capacity, this is a multi-year trend. For IREN, I’d wait for their own Q2 report to confirm the AI revenue stream. Coreweave is a story I’ll play only if I get a clear entry after their IPO lockup expires. Risk is not the enemy; ignorance is. The market is cheering, but I’m watching the tape. The candlestick doesn’t lie, but your bias might. I’ll position on the next pullback, not the breakout.