The Forge of Trust: When Bitcoin Miners Become AI Cloud Providers

From the chaos of 2017, we forged a compass. It was a time when ICOs promised Utopia, and I, a 21-year-old cryptography PhD candidate at UCL, audited 15 whitepapers, finding structural flaws that prioritized speculation over utility. That experience taught me that trust is not a metric; it is a memory we share. Now, in August 2025, that memory is being tested again—not by a token sale, but by the quiet pivot of Bitcoin miners into AI cloud providers. On Tuesday, the Neocloud sector strengthened in early US trading: IREN up over 5%, Nebius and Coreweave each rising more than 3%. The catalyst was Nebius’s Q2 earnings—a 454% revenue surge to $582.3 million, EBITDA turning positive at $236.2 million. This is not just a stock rally; it is a signal that the infrastructure built for proof-of-work is being repurposed for the next wave of compute. But as I watch this transformation, I cannot help but ask: are we building a bridge to decentralization, or are we paving a road to a new form of centralization? Let me ground this in context. The three companies—IREN (formerly Iris Energy), Nebius (NBIS), and Coreweave—represent different paths into the same territory: AI cloud computing. IREN is a Bitcoin miner that has pivoted to high-performance computing, leveraging its existing power infrastructure and land assets. Nebius is an AI-native cloud born from the ashes of Yandex’s European operations, now listed on Nasdaq with a founder who was once under EU sanctions. Coreweave is a pure-play GPU cloud provider, deeply tied to NVIDIA, with a history of leasing GPUs for Ethereum mining before pivoting to AI. All three are now part of a narrative that the market is buying: the intersection of crypto capital and AI demand. The market’s reaction—Nebius jumping 34% the day before—shows that investors see this as a fundamental shift, not a hype cycle. But I have seen this before. In 2020, during DeFi Summer, I founded a community to help non-technical users understand smart contract risks. I manually verified 200 protocols, and the lesson was clear: when everyone rushes to the same narrative, the technical details matter most. Now, let us dive into the core. The technical architecture here is fascinating. IREN’s strategy is what I call the ‘dual-rent model’: it uses the same electricity contracts, substations, and real estate that once powered ASICs to now host NVIDIA GPUs. The capital efficiency is real—a Bitcoin miner can deploy a GPU cluster at a fraction of the cost of a greenfield data center because the land and power are already secured. But this is not a paradigm shift; it is a gradual innovation. I have audited similar transitions in the mining space, and the bottleneck is always the same: GPU supply. IREN, like Core Scientific and Hut 8, competes for the same limited NVIDIA H100 and B200 chips. Nebius, on the other hand, has a different edge. Its Q2 numbers—$582.3 million in revenue, $236.2 million in adjusted EBITDA—suggest that it has achieved operational scale. The EBITDA margin of roughly 40.6% implies that the unit economics work: the cost of renting a GPU is well above the cost of running it, including amortization. This is a signal that the infrastructure layer can be profitable, but only if you have the capital to build before the demand arrives. Coreweave, with its deep ties to Microsoft and NVIDIA, is the third pillar. But here is the hidden truth: all three companies are essentially renting compute from the same upstream supplier. Their differentiation is not in technology—GPU clusters are standardized—but in access to capital and power. This is where the contrarian angle emerges. The contrarian view, which I have held since my 2022 thesis ‘Resilience in Code,’ is that the Neocloud sector is a mirage of decentralization. These companies are building centralized compute farms, not trustless networks. IREN’s pivot from Bitcoin mining to AI cloud is a move away from the permissionless ethos of PoW. Bitcoin mining, despite its energy consumption, is a decentralized security layer. AI cloud, by contrast, is a rent-seeking infrastructure that concentrates power in the hands of those who control the GPUs. The risk is not just ‘effective hash rate inflation’—where too many GPUs dilute per-unit returns—but a deeper erosion of the values we fought for. Trust is not a metric; it is a memory we share. The memory of 2017 taught us that speculative capital can corrupt technology. The memory of 2022 taught us that misaligned incentives collapse ecosystems. Now, in 2025, we are seeing the same pattern: capital is flooding into GPU buildouts, but the underlying protocol—the AI models themselves—remain opaque and centralized. The companies that succeed will be those that lock in the largest contracts, not those that build the most open platforms. This is a recursion of the old web2 playbook, dressed in blockchain’s discarded clothes. Let me be specific. The market’s euphoria over Nebius’s 34% jump ignores the elephant in the room: customer concentration. A $582 million revenue quarter likely comes from a handful of large AI labs, and if one of those contracts is not renewed, the growth story collapses. Similarly, IREN’s dual-rent model sounds elegant, but it carries the risk of ‘double-beta’: if Bitcoin prices fall and AI demand softens simultaneously, the stock could drop twice as hard. And Coreweave’s reliance on NVIDIA for both hardware and financing (via collateralized loans) creates a systemic risk that echoes the 2022 crypto credit crisis. The writers of the original analysis note that the sector’s long-term competitiveness depends on GPU supply and capital access—not on technical innovation. This is a red flag for anyone who believes in the transformative power of Web3. The true innovation in this space is not the GPU clusters themselves, but the verification layer that ensures trust. In my work on the ‘Human-Centric AI Ledger’ initiative, I have developed cryptographic protocols to verify AI decision-making origins. That is the real frontier—not just renting compute, but proving that the compute is used ethically. So, what is the takeaway? The Neocloud sector is a mirror of our own values. We cheered when Bitcoin miners survived the 2022 crash by pivoting to AI, but we forgot that the pivot is a survival mechanism, not a moral victory. The infrastructure built for proof-of-work can be repurposed, but only if we ensure that the new use case serves the same principles of decentralization and trust. From the chaos of 2025, we must forge a new compass—one that points not just to profit, but to sovereignty. The question is not whether IREN, Nebius, and Coreweave will make money; they probably will. The question is whether the memory of 2017—the memory of why we built this—will survive the rush to GPU gold. Trust is not a metric; it is a memory we share. And that memory is fragile.