The sky over Chengdu is clear, but the digital horizon is clouded with ambition. Last week, the city unveiled its "AI+" action plan, promising to inject 260 billion yuan into a future where smart terminals and agents saturate 70% of industries by 2027. The target is bold, the timeline aggressive—but as a Web3 community founder who has watched centralized promises collapse in Cape Town, I see a familiar trap: a top-down vision that ignores the very architecture that could make AI trustworthy.
Let’s start with the numbers. The plan aims for 260 billion yuan in core AI industry scale, with a compound annual growth rate exceeding 30%. That is twice the national average. To get there, it plans to incubate 100 innovative products and 100 demonstration scenarios, releasing 20 flagship scenes each year. Sounds impressive—until you read the fine print. There is no mention of decentralized compute, no token incentives for data providers, no on-chain verification for outputs. The plan is built on the old playbook: government subsidies, state-owned enterprise contracts, and centralized data silos. Vibes > Algorithms? Here, the vibes are all top-down control.
Context matters. Chengdu is no stranger to tech booms. It hosts the Tianfu Software Park, a manufacturing hub for Foxconn and Intel, and boasts two supercomputing centers: the National Supercomputing Center (100 PetaFLOPS) and the Tianfu Intelligent Computing Center (targeting 1,000 PetaFLOPS by 2025). These are powerful assets—but they are also single points of failure. In a bear market where survival matters more than gains, the question every builder should ask is: can this infrastructure survive a chip ban or a power grid collapse?
My experience in the 2021 NFT cultrual renaissance taught me that centralized hype fades. When we launched "AfricanCode" in Cape Town, we generated $80,000 in 48 hours, but the project stagnated because we relied on a single curator and a single smart contract. Diversity of validators and compute providers is the only resilience in Web3. Chengdu’s plan, by contrast, leans on state-backed chips (Huawei Ascend) and a closed ecosystem. Code is law, but people are truth—and truth requires transparency. Without a blockchain layer, how do you audit whether those 100 demonstration scenarios actually deliver on their promises?
Let’s dig into the core technical gaps. The plan targets "new generation smart terminals and agents" but never defines them. Are they edge AI devices running local large language models? Or are they centralized cloud agents that phone home? The difference matters because the latter creates a surveillance honeypot. Embrace the volatility, find the signal—the signal here is that the plan lacks a decentralized data provenance system. Every AI output, every decision by an agent, should be recorded on-chain to ensure accountability and prevent bias. Instead, the policy is silent on ethics, security, or even basic algorithm registration.
Consider the cost. The plan assumes 260 billion yuan will materialize through government spending and private investment. But if 90% of AI compute happens on centralized clouds (Alibaba, Huawei, Tencent), the cost structure will bloat. My own DeFi liquidity trap in 2020 taught me that chasing high yields without managing composability risk leads to exhaustion. Here, the composability risk is between expensive cloud compute and low-margin local manufacturing. A decentralized alternative like Akash Network or Render Network could cut costs by 40–60% through idle capacity pooling. But the plan ignores this, likely because regulators fear losing control.
Build in public, live in truth. That phrase is my mantra, and it is the antidote to closed-door policy-making. The contrarian angle is this: even a top-down plan can succeed by embracing bottom-up tech. The city could issue a stablecoin-based reward token for data labeling and model training, creating a local economy that aligns incentives. It could mandate that all demonstration scenarios use zero-knowledge proofs to protect user privacy while proving compliance. It could open its supercomputing centers to a decentralized compute layer, renting out idle cycles to global AI startups. None of this is in the current text.
Why does this matter? Because the future-back view shows that trust in AI will erode without transparency. If Chengdu’s smart terminals become ubiquitous—smart gates, AI cameras, healthcare bots—the data they collect will be a gold mine for hackers and a nightmare for civil liberties. The EU AI Act and China’s own generative AI regulations require traceability and audit trails. Blockchain is the only viable infrastructure for that. A city that spends billions on AI without a cryptographic backbone is building a beautiful castle on sand.
My 2017 Cape Town DAO experiment collapsed because we chose ideology over technical rigor. We raised $120,000 in ETH, but gas fees during the November congestion killed our community voting system. The lesson: infrastructure matters more than philosophy. Chengdu’s plan has philosophy—catchy slogans about "empowering thousands of industries"—but it lacks the infrastructure roadmap for decentralization. It talks about "agents" but not about Agent-to-Agent protocols on-chain. It talks about "scale" but not about verifiable inference.
So here is my takeaway: The plan is a 10x opportunity if it pivots from centralized to decentralized. If it refuses, it will be a monument to old thinking, unable to adapt when the next Dencun upgrade or rollup fee hike shakes the digital economy. The choice is clear: build a walled garden or a public commons. I’m betting on the commons. Will Chengdu?