Hook
When a company that builds nothing but ‘AI-optimized data centers’ files for a $3 billion IPO, the crypto community should pause and ask: where is the decentralization in artificial intelligence? Nscale, a relatively unknown player in the infrastructure space, has announced its intention to raise $3 billion through an initial public offering, positioning itself as a direct challenger to AWS, Azure, and Google Cloud. The news is being celebrated by mainstream tech media as a sign of the AI boom’s vitality. But for those of us who have spent years advocating for decentralized, user-owned networks, this IPO is a flashing red warning light. It signals that the most valuable compute resource of the coming decade—AI training and inference power—is being consolidated into the hands of a few centralized entities. Every line of code is a hand extended in trust, and right now, that trust is being placed in a handful of data centers.
This is not a story about Nscale’s technology. In fact, the company has revealed almost nothing about its technical architecture, its GPU inventory, or its energy efficiency. What we know is that it plans to spend $3 billion on building more data centers, and that it aims to steal market share from the cloud giants. As an open source evangelist who has spent years auditing smart contracts and building community-driven DeFi education, I see a pattern here. The same forces that centralized the internet, that put our data in the hands of a few corporations, are now centralizing the infrastructure of AI. We build bridges, not just blocks, between people—but Nscale is building a toll booth.
Context
For context, Nscale is a British company that operates AI-optimized data centers. The term “AI-optimized” typically means facilities designed to house high-density GPU clusters, with advanced liquid cooling, high-bandwidth networking (InfiniBand or RoCE), and low-latency interconnects. These are the physical assets that power the training of large language models like GPT-4, Claude, and Gemini. The demand for such infrastructure has exploded since the launch of ChatGPT in late 2022, leading to a GPU shortage that has persisted for over two years. Companies like CoreWeave, Lambda Labs, and now Nscale have emerged as specialized providers, offering compute rentals that are often cheaper and more flexible than the offerings of traditional cloud providers. Nscale’s IPO, with a target of $3 billion, is one of the largest in the AI infrastructure space. It reflects a market belief that the demand for AI compute will continue to grow exponentially, and that owning the hardware is the surest way to capture value.
But beneath the surface, this is a story about centralization. The entire AI industry is currently dependent on a narrow supply chain: NVIDIA’s GPUs, TSMC’s fabrication, and a handful of data center operators. The barriers to entry are astronomical. A single H100 GPU costs around $30,000, and a cluster of 10,000 GPUs can cost $300 million. This means that only the best-funded companies can compete. The result is a world where the most powerful AI models are trained in secret, on private infrastructure, controlled by a tiny elite. Tracing the code back to the conscience behind it, I find that the conscience of the AI industry is increasingly corporate, not communal.
Core
Let me take you through a technical analysis of what Nscale’s $3 billion actually buys, and why it matters for the blockchain community. Based on my experience auditing smart contracts in 2017, I learned that technical precision is a form of social protection. The same principle applies to AI infrastructure. The $3 billion will likely be used to purchase approximately 50,000 to 100,000 NVIDIA H100 GPUs (depending on discounts and partnership deals). These GPUs will be deployed in a handful of massive data centers, each consuming tens of megawatts of power. The cooling will be liquid, likely using direct-to-chip or immersion solutions. The networking will be InfiniBand, which provides low latency but is proprietary and expensive. The entire stack—from the hardware to the operating system to the AI frameworks—is optimized for a single purpose: maximizing the speed of training large models.
Now, compare this to a decentralized compute network like Golem or iExec. In a decentralized network, compute resources are contributed by individuals and small data centers around the world. The network is permissionless, censorship-resistant, and distributed. The trade-off is performance: decentralized networks cannot match the raw throughput of a co-located GPU cluster because of latency and bandwidth limitations. But for many AI workloads—particularly inference, fine-tuning, and small-scale training—decentralized compute is already viable. The problem is that the market is not prioritizing these alternatives. Venture capital is flowing into centralized data centers, not into decentralized compute protocols. Education is the only true decentralized currency—and we have failed to educate investors and developers about the long-term risks of centralized AI infrastructure.
During my work with NFT artists in 2021, I saw firsthand how centralized platforms can exploit creators. We built a royalty enforcement toolkit because marketplaces refused to pay artists automatically. The lesson was that power structures in technology must be designed for creators, not just speculators. Nscale’s IPO is a power play for speculators. It is a bet that the AI market will grow so fast that the returns from owning the infrastructure will dwarf any ethical concerns about centralization. But history tells us that centralized infrastructure leads to centralized control, which leads to rent extraction. The cloud giants already take 30-40% margins on compute. Nscale will likely offer lower prices to gain market share, but once it has a critical mass of customers, it will raise prices. That is the nature of the game.
Contrarian
Now, let me offer a contrarian perspective. Perhaps Nscale’s IPO is actually a good thing for decentralization. The massive capital injection will drive down the cost of AI compute, making it more accessible to researchers and small startups. It will also force traditional cloud providers to compete on price and performance, potentially accelerating the commoditization of GPU compute. When compute becomes cheap enough, it becomes feasible to run a decentralized network that aggregates spare capacity from thousands of users. In other words, Nscale’s data centers could become the “wholesale” source of compute, which decentralized networks can then resell at retail. This is similar to how AWS provides the backbone for many decentralized applications today—ironically, the very apps that aim to replace centralized services often run on centralized infrastructure.
Moreover, the IPO might be a signal that the market is overheating. During the 2022 crypto bear market, I started a “Code & Conversation” support group to help developers cope with the stress of portfolio losses and project failures. The lesson was that resilience comes from community, not from capital. Nscale’s $3 billion is capital, not community. It is a bet that the AI boom will continue unabated, but what if the boom slows? What if a more efficient chip architecture emerges, reducing the need for massive GPU clusters? What if AI regulation limits the kinds of models that can be trained? The risk is real. If the demand for AI compute plateaus, Nscale could be left with billions of dollars in underutilized hardware. That would be a disaster for its investors, but it could also be an opportunity for the decentralized community to acquire cheap compute.
Takeaway
So where does this leave us? The crypto community has a choice. We can ignore Nscale’s IPO, dismiss it as a Web2 phenomenon, and continue building our own infrastructure. Or we can see it as a call to action. The AI revolution is happening, and it is being built on centralized infrastructure. If we believe in the principles of decentralization—sovereignty, permissionlessness, and community ownership—we must actively work to make decentralized AI compute viable. That means funding research into distributed training algorithms, building user-friendly interfaces for decentralized compute networks, and educating developers about the importance of owning their own infrastructure. Open source is not a license; it is a promise. The promise that the tools we build will remain accessible to everyone, not just those who can afford to rent from a data center.
Artists own their pixels; we just hold the keys. Similarly, AI developers should own their compute; they should not be locked into a single provider. The Nscale IPO is a reminder that the battle for the future of AI is not just about algorithms and models—it is about infrastructure. And infrastructure, like code, must be built with conscience. Let us learn from the mistakes of the internet era, where we gave away our data for free. Let us not give away our compute. The code we write today is a hand extended in trust. Let’s ensure that trust is distributed, not centralized.
