From the ashes of 2022, we planted seeds for 2030—Vera Rubin, Blackwell, gigawatt-scale cloud. The soil we chose was fertile: community-owned, permissionless, resilient. Yet as I watch Naver, NVIDIA, and Brookfield break ground on a 200MW factory in Sejong—with plans for 1GW across Korea and the US—I feel the ground shift beneath my feet. This is not a failure of our vision, but a reminder of the scale we still need to reach.
I remember 2017, a 19-year-old finance student in Manila, staying up late reading Golem’s whitepaper. I believed in a world where anyone could rent out their GPU cycles to train the next great AI model. It felt like the natural evolution of the web—from centralized servers to a mesh of contributors. By 2020, during DeFi Summer, I contributed my first salary to Compound and Uniswap, testing the promise of permissionless financial sovereignty. The P&L was painful in 2022—my portfolio down 85%—but the lesson stuck: infrastructure, not hype, survives the bear.
Now, in 2025–2026, I watch the other side of the coin. Naver, Korea’s digital gateway, joins hands with the world’s largest chipmaker and the planet’s oldest infrastructure asset manager to build a fortress of compute. The announcement is celebratory: “gigawatt-scale,” “latest platforms,” “lighthouse project.” But as a Web3 community founder who has mentored 50 women through wallet setup and NFT minting, I can’t ignore the shadow this casts on our decentralized dreams.
Context: The Goliath in the Room Naver’s Sejong facility will consume 200MW by 2028, powered by NVIDIA’s Vera Rubin and Blackwell architectures. Brookfield brings the capital discipline of a REIT-like model. The stated goal: “enhance global AI cloud infrastructure.” The unstated reality: this is state-backed centralization. Korea’s government sees AI sovereignty as a national security matter, and Naver—already the owner of search, commerce, and payment ecosystems—becomes the chosen champion. It’s a pattern we saw with AWS, Azure, and GCP: massive capital expenditure creates lock-in, and lock-in kills permissionless innovation.
Core: The Technical and Ethical Architecture of Control Let’s break down the technical decisions. Naver’s infrastructure will be built on NVIDIA’s proprietary CUDA ecosystem, with no mention of AMD’s MI300 or Intel’s Gaudi. The roadmap ties directly to Vera Rubin, a chip that is still a slideware promise. This creates a single point of failure: if NVIDIA’s roadmap slips, Naver’s capacity timeline slips. But more importantly, it entrenches a hardware monoculture. In DeFi, we fight for composability; here, the composability is dictated by one company’s interconnect fabric. The 200MW and 1GW are not just power numbers—they represent the energy required to cool thousands of H100/B200/Vera Rubin GPUs. Liquid cooling becomes mandatory, and the patent wall around that cooling is high.
But the deeper issue is governance. Who decides which models run on these clusters? Naver. Who profits from the training data? Naver and its partners. Who controls the API? Not you, not me. This is the antithesis of blockchain’s promise. In 2020, when I documented my yield farming journey for a Filipino audience, I stressed that DeFi offered “unstoppable access.” Today, AI compute access is becoming more stoppable than ever, gated by fiat, nationality, and corporate relationships.
Contrarian: The Pragmatic Defense Let me be fair. Not everyone sees this as a threat. Some argue that such massive infrastructure is necessary for breakthrough AI—think climate modeling, drug discovery, or fusion research—and that a single sovereign entity can enforce ethical standards better than a global pool of anonymous miners. Brookfield’s involvement signals that this is a long-term, stable asset class, which could attract more responsible capital. And yes, Naver might offer cheap GPU time to Korean startups, accelerating local AI ecosystems.
But here’s the contrarian within me: scale alone does not guarantee access. When CoreWeave emerged as NVIDIA’s pet cloud, its rates were not lower than AWS—they were higher, justified by superior performance. Naver’s infrastructure will be built for its own hyper-scale AI model (HyperCLOVA X), not for the public good. The “AI cloud” label is a wrapper for private compute. And if history repeats—as it did with AWS in 2010—the early customers will be locked into proprietary services, not liberated.
Takeaway: The Seeds Must Grow Taller From the ashes of 2022, we taught ourselves to build in the bear, to prioritize resilience over hype. Naver’s gigawatt bet is a wake-up call. Decentralized compute networks like Akash, Golem, and Livepeer need to scale from megawatts to gigawatts. They need institutional partnerships, capital efficiency, and real-world use cases beyond NFT minting. The technology gap—Vera Rubin vs. consumer GPUs—will widen, but the value gap can be closed if we focus on sovereignty, privacy, and censorship resistance.
I look at my community, “Decentralized Hearts,” the 50 women I mentored through wallet setup. They are building NFT projects, DAOs, and local DeFi apps. They don’t need 200MW for their art. But they need a future where compute is a human right, not a rental from a corporate fortress. The question we must answer: Will we let Naver- NVIDIA-Brookfield become the new gatekeepers, or will we build our own infrastructure—slowly, painstakingly, but truly ours?
Visionaries plant trees they never sit under. But those trees must be rooted in community soil, not concrete. Let’s ensure the forest we grow is open, not walled.
The next step? - If you are a developer, contribute to Akash or Filecoin’s compute market. - If you are a founder, negotiate data sovereignty clauses in your cloud contracts. - If you are a writer, tell stories that celebrate small but resilient, not big but captive.
From the ashes of 2022, we planted seeds for 2030. The sun is rising, but the soil needs more of us. Stay jagged. Stay authentic. Stay web3.