Last quarter, listed companies collectively raised over $30 billion to fund AI infrastructure, according to recent filings. Meanwhile, the entire decentralized compute sector—including Akash, Golem, and io.net—captured less than $200 million in equivalent token-based capital raises. The divergence is not an accident; it's a reflection of a deep structural misalignment between traditional capital markets and the values of decentralization. And if you think this doesn't affect your crypto portfolio, you're missing the signal in the noise.
Context: AI infrastructure spending has reached record levels, driven by the insatiable demand for GPU compute to train and run large language models. The pattern is familiar: companies race to raise capital, lock in long-term contracts with NVIDIA, and build massive data centers. This creates a self-reinforcing cycle—more capital leads to more chip orders, which boosts NVIDIA's stock, which makes it easier for these companies to raise even more capital. It's the same playbook we saw during the ICO boom, except now the 'tokens' are shares of public companies and the 'blockchain' is a centralized supply chain.
But crucially, this cycle is not just an AI story—it's a crypto story. Every dollar flowing into centralized AI infrastructure is a missed opportunity to build permissionless compute markets. Based on my 2017 Ethereum Foundation audit experience, where I discovered that 60% of token projects relied on flawed logic rather than technical bugs, I see a similar pattern today: the market is prioritizing speed and familiarity over resilience and sovereignty.
Core: What interests me as a decentralized protocol product manager is not the scale of the spending—it's the underlying assumption that centralized infrastructure is the only viable path. Let's break down the capital logic. These listed companies are essentially buying assets (GPUs, real estate, power contracts) that they hope will generate future revenue. But they are also creating a single point of failure: if the NVIDIA supply chain hiccups, if power grids get overloaded, or if a more efficient chip architecture emerges, the entire capital base depreciates overnight.
Decentralized compute networks, by contrast, distribute risk across thousands of independent providers. They use token incentives to align behavior, and their code is open for anyone to verify. During the DeFi Summer of 2020, I launched 'DeFi for Humans' to onboard 5,000 users by focusing on narrative over yield. The same principle applies here: the story we tell about compute matters. Today's centralized AI infrastructure is building walled gardens; decentralized alternatives can offer something radically different—a global, permissionless compute market where anyone can contribute resources and anyone can buy them.
Yet adoption remains negligible. Why? Because the user experience is still clunky, latency is higher, and there's no capital to subsidize growth. The capital supercycle has created an asymmetry: centralized players can buy their way into the market, while decentralized protocols must bootstrap through community and patience.
Contrarian: Now, the contrarian angle—and I say this as someone who has been an evangelist for decentralization since my first manifesto on code ethics. It's tempting to dismiss the centralized approach as short-sighted. But the market is voting with its money, and it's voting centralized for a reason: efficiency. Traditional capital markets are incredibly efficient at allocating resources to proven business models. Decentralized compute protocols, despite their ideological appeal, have not yet proven they can deliver lower-cost, lower-latency compute at scale. The pragmatist in me recognizes that for many applications, a centralized data center is simply better today.
However, what the capital supercycle overlooks is the long-term fragility it creates. I've seen this movie before—during the 2017 ICO craze, when everyone believed that centralized exchanges would always be sufficient. Then the 2022 collapses hit, and suddenly decentralization wasn't a nice-to-have, it was a survival need. The same will happen with AI compute. The first major data center outage, the first geopolitical crisis that disrupts chip supplies, the first evidence of censorship on cloud GPUs—these events will trigger a flight to trustless alternatives.
Takeaway: The capital supercycle is not the end of decentralized compute; it's the necessary tension that will force the sector to mature. When I started my current role at a decentralized compute protocol in 2026, I realized that our biggest competitor isn't AWS—it's the narrative that centralization is inevitable. The next bull run in crypto will be fueled not by memes or yield farming, but by the backlash against AI infrastructure monopolies. The protocols that will thrive are those that can bridge the gap between capital efficiency and decentralization—offering trustless access to compute without sacrificing privacy, autonomy, or ethical governance. That's the story I'm betting on.