Last week, Applied Digital announced a Q4 earnings beat. Revenue up 406%. EPS above consensus. The market cheered. But here’s the part that got me: this company started life as a crypto miner. Now, it’s being celebrated for AI infrastructure.
I was sitting in my Prague apartment, scrolling through the news, when I felt a familiar unease. This isn’t just another earnings story. This is a mirror held up to the entire crypto industry. We spent years evangelizing decentralization, telling ourselves we were building an alternative to centralized power. And now, our own mining companies are pivoting to serve the very centralized AI giants they once promised to disrupt.
It’s a moral hazard we rarely discuss. But we must. Because if we ignore it, we risk losing the soul of the movement.
Context: The Ghost of Crypto Mining
Applied Digital was once known for its crypto mining operations. It built data centers optimized for proof-of-work. Then the bear market hit. Mining became unprofitable. The company needed a new story. AI was booming. So they rebranded. They now call themselves an “AI infrastructure provider.”
The pivot is rational. It’s survival. But here’s the uncomfortable truth: the hardware they use—GPUs, networking, power infrastructure—is the same. The only difference is the software workload. Instead of hashing SHA-256, they’re training large language models. The business model hasn’t changed: rent out compute, collect fees.
I saw this pattern during my time organizing “Prague Decentralized” back in 2017. Back then, the ICO frenzy was all about building the future. We held workshops in a repurposed warehouse, teaching 150 developers about trustless systems. We believed we were shifting power to the edges. Now, those same developers are building AI pipelines for centralized data centers.
The irony is thick enough to cut. Crypto was supposed to be the antidote to centralized control. Instead, it’s become the training ground for the very infrastructure that enables mass surveillance, algorithmic bias, and corporate consolidation.
Core: The Earnings Deconstructed
Let’s dig into the numbers. Revenue grew 406%. That’s staggering. But growth alone is a vanity metric. The real question is: at what cost?
The parsed analysis of the article reveals three hidden risks. First, customer concentration. Applied Digital likely relies on a handful of large AI labs. If one client leaves, the house of cards collapses. Second, execution risk. Building data centers is capital-intensive. Delays can destroy margins. Third, technology dependency. NVIDIA’s GPU supply dictates everything. If H100 allocations dry up, growth stops.
I’ve seen this movie before. In 2020, during DeFi Summer, I led a community translation project for Aave’s whitepaper. We simplified the liquidation mechanism for 5,000 non-technical users. I learned then that infrastructure is only as strong as its transparency. Applied Digital has not disclosed its gross margin, net profit, or customer names. As a decentralized protocol PM, I find that alarming.
The article itself offers no technical detail on the data center architecture. We don’t know the PUE, the cooling method, or the network topology. This is a red flag. In blockchain, we demand verification. In AI, we trust the press release. That’s a dangerous asymmetry.
Let’s do a rough calculation. If Applied Digital’s incremental revenue was, say, $200 million, and the going rate for an H100 is about $3 per hour, that implies around 7,600 GPUs added. Assuming 100% utilization. Unrealistic. More likely, they’re using a mix of older hardware, lower utilization, or longer-term contracts with discounts. The margin squeeze is inevitable.
But the deeper insight is this: the economics of centralized AI infrastructure are fundamentally fragile. The hardware depreciates fast. New chips (H200, B200) make existing clusters obsolete. The energy costs are volatile. And the customers are demanding. This is not a moat; it’s a treadmill.
And yet, the crypto market treats this as validation. Why? Because we crave legitimacy. We want Wall Street to take us seriously. But we’re celebrating the wrong thing. We’re celebrating a company that abandoned the decentralization ethos the moment it became inconvenient.
Contrarian: The Pragmatist’s Defense
Now, let me play devil’s advocate. Perhaps this pivot is inevitable. Maybe the market is simply responding to demand. Why should Applied Digital stick to crypto mining when AI pays better? Isn’t this just resource allocation?
There’s a case to be made: the AI boom is creating huge demand for compute. By repurposing crypto infrastructure, we’re reducing electronic waste and utilizing existing assets. This could be seen as a form of ESG.
And some argue that centralized AI infrastructure will eventually enable decentralized applications. You need fast inference to run on-chain AI agents. Applied Digital could become the backbone of a new generation of smart contracts.
But I don’t buy it. Centralized infrastructure creates gatekeepers. The same company that controls the compute controls the rules. They can censor transactions, prioritize certain clients, and extract monopoly rents. That’s antithetical to everything we stand for.
During my work on the EU regulatory task force in 2025, I saw how powerful incumbents can shape policy to their advantage. Applied Digital’s pivot isn’t a strategy; it’s a surrender to the status quo.
Takeaway: Build for Humans, Not Just Nodes
Applied Digital’s earnings are a symptom of a larger disease: the crypto industry’s failure to deliver on its original promise. We became obsessed with price, not principles. We celebrated mining profitability while ignoring the concentration of power. Now, the same people who built our networks are building for the very systems we sought to replace.
But there’s a path forward. We must invest in truly decentralized compute. Projects like Filecoin, Akash, and Golem offer alternatives. They’re not perfect. They have scaling challenges. But they are aligned with our values.
Education is the ultimate yield. We need to teach the next generation that infrastructure without governance is just another silo. I learned this in 2021 when I curated “Art & Algorithm,” a digital gallery focusing on provenance over speculation. The community that understands the “why” will build the “how.”
Applied Digital’s success is a warning, not a role model. Let’s not confuse short-term gains with long-term vision. Build for humans, not just nodes.