We Didn’t Build Blockchains for Sovereign AI Kings
We didn’t need another sign that the future of intelligence would be centralized. But there it was: Temasek, the Singaporean sovereign wealth fund, announced plans to pour $75 billion into AI by 2030, alongside an $8 billion private credit platform. For those of us who spent years arguing that decentralized, permissionless systems are the only ethical path forward, this felt like a punch in the gut. The very institutions we built crypto to bypass are now the ones funding the plumbing of the next era. And they’re not investing in your decentralized AI agent—they’re investing in walled gardens, proprietary models, and closed infrastructure.
Context: Temasek’s move is not just a capital allocation—it’s a structural bet. The fund currently has roughly $250 billion in total assets under management, and this $75 billion AI target represents a tripling of its existing AI-related exposure. The $8 billion private credit platform is particularly telling: it allows Temasek to lend to AI startups at high interest rates while taking warrants, essentially replicating the upside of equity with the downside protection of debt. This is a classic sovereign fund playbook—low cost of capital, patient time horizon, and deep portfolio synergies. But in a bull market where crypto AI tokens are flying, the real money is flowing to entities that would never put a DAO in charge of their model weights. Temasek’s credit platform is the financialization of AI control, and it’s a direct challenge to the DeFi ethos of trustless, transparent lending.
Core: Let’s break down what this means for the crypto-AI stack—data, compute, model, and verification. First, compute. Temasek is already the largest owner of data centers in Southeast Asia through its stakes in Mapletree and other REITs. That $75 billion will overwhelmingly go toward expanding GPU clusters, locking NVIDIA’s latest chips into sovereign-controlled facilities. During the 2022 bear market, I audited the smart contracts of a dozen decentralized compute protocols—Akash, Render, Gensyn, and others. The code was elegant; the liquidity was not. What they lacked was the kind of backstop that Temasek provides to its portfolio companies. Contrast that with Temasek’s approach: they can pre-order entire supplies of H100s, negotiate volume discounts, and write off losses over decades. Decentralized compute networks rely on token incentives that fluctuate with market sentiment. When the bull market ends, those incentives dry up. Temasek doesn’t care about market cycles—it cares about strategic control.
We didn’t design blockchain to compete with state-backed capital. But here we are, watching sovereign funds outspend the entire crypto AI ecosystem by orders of magnitude. The $8 billion credit platform is especially deceptive because it looks like a financial innovation—but it’s a centralized version of what Compound or Aave do on-chain. Temasek’s credit decisions are made behind closed doors, based on relationships and geopolitical alignment. There’s no code enforcing loan conditions, no liquidation mechanism transparent to the public. This is the ultimate irony: while we struggle to onboard institutional capital into DeFi lending, they are creating parallel systems that are less transparent but more capitalized.
Now consider the model layer. Temasek will likely invest in proprietary foundation models—think OpenAI, Anthropic, or perhaps a Singaporean national champion. These models are closed weights, API-gated, and subject to government oversight. Any crypto project that tries to run a decentralized AI agent on top of these models will be at the mercy of API pricing changes, censorship, or outright shutdown. The only path to real AI sovereignty is through open-source models that can be verified on-chain. But who will pay to train those models? The crypto community can crowdfund modest amounts, but we’re talking about billions for frontier models. Temasek’s $75 billion could train a model ten times the size of GPT-5. The decentralization narrative collapses under that weight unless we rethink our approach entirely.
We didn’t anticipate that the biggest threat to decentralized AI wouldn’t be technical, but financial. During the NFT identity crisis of 2021, I co-founded Canvas Chain and watched speculative capital distort artist incentives. Now the same thing is happening at a macroeconomic scale: sovereign funds are pouring into AI not because they believe in ethical innovation, but because they see a strategic necessity. Temasek’s investment is part of a broader trend—SoftBank, Mubadala, and even China’s CIC are all racing to own AI infrastructure. The result is a winner-take-all dynamic where the largest pools of capital capture the most scarce resources: compute, talent, and data.
From a game theory perspective, Temasek’s move exacerbates the coordination problem for crypto AI projects. If I’m an entrepreneur building a decentralized inference network, I can either raise $50 million from a VC with a 5-year fund life, or I can take a $200 million loan from Temasek’s credit platform with fewer strings attached. The latter is easier, but it ties me to sovereign interests. Most founders will choose the money, and the crypto AI ecosystem will bleed talent and ambition to centralized giants. The only way to counter this is to create a liquidity flywheel that makes decentralized capital more attractive—not just as a speculative store of value, but as productive credit.
Contrarian: But maybe the contrarian view is that Temasek’s move actually validates the need for decentralized AI. If they’re pouring $75 billion into centralized infrastructure, they’re admitting that trust in a single entity is unsustainable. The market will eventually fracture. The real opportunity for crypto is not to compete on scale, but on trustlessness. Sovereign funds are exposed to geopolitical black swans—sanctions, confiscation, diplomatic breakdowns. A decentralized network of compute and models, even if smaller, offers a guarantee of availability that no single state can match. During the Bear Market Refinement of 2022, I studied failed DeFi protocols and learned that incentive alignment matters more than capital size. Temasek’s $75 billion buys a lot, but it cannot buy the property of censorship resistance. In the long run, the ability to run an AI agent without permission—without a credit check or a license—will become a non-negotiable civil liberty.
Perhaps the contrarian truth is that we need sovereign funds to fund the compute, while crypto funds the coordination. The two might be complementary, not adversarial. Temasek could, in theory, open-source the models it funds, or allocate part of its credit platform to decentralized collateral. But that would require a shift in worldview—from control to abundance. Until then, the responsibility falls on the crypto community to build the alternative, not by matching their capital, but by outlasting it.
Takeaway: The next five years will decide whether AI becomes a public utility or a private empire. Temasek’s $75 billion is a bet on empire. Our job is to build the alternative—not by matching their capital, but by outlasting it. We didn’t build blockchains for sovereign AI kings. We built them for the unbanked, the uncensored, the unowned. And that fight is more urgent now than ever.
We didn’t need Temasek to remind us of the asymmetry—but now that we have the reminder, let’s code the response.