Wang Jian’s speech at the 2026 World AI Conference wasn’t a routine keynote. It was a liquidity event—for a new narrative. The former Alibaba CTO and founder of Alibaba Cloud didn’t talk about LLM scaling or GPT-5. He reframed AI as infrastructure, akin to mathematics. Then he dropped the bomb: the next AI paradigm depends on tokenizing multimodal scientific data—not text, not code, but the raw, non-discrete data of proteins, weather patterns, and molecular structures.
For anyone who’s spent years tracking narrative cycles in crypto, this is a signal. A strong one. The speaker’s pedigree (Alibaba Cloud, CSDN contributor) gives weight. But more importantly, the shift he describes directly intersects with crypto’s deepest structural weaknesses—data provenance, compute marketplaces, and incentive alignment. If he’s right, the next wave of crypto adoption won’t come from trading bots or degen NFTs. It will come from decentralized science (DeSci) and data tokenization networks.
Context: The Infrastructure Play
Wang Jian’s thesis is simple: current AI models are trained on text and code—discrete, clean sequences. Scientific data (protein folding, climate radar, astronomical images) is heterogeneous, high-precision, and non-discrete. Converting it into tokens that transformers can process is a non-trivial engineering challenge. He calls this “scientific data tokenization.” And he argues that whoever masters this will own the next decade of AI.
Crypto native projects have been dancing around this idea for years. Filecoin and Arweave store scientific datasets. Render Network and Akash provide decentralized compute. But the missing piece is a standardized tokenization layer—a way to turn raw scientific output into fungible, tradeable, and AI-compatible assets. Wang Jian’s speech effectively legitimizes this as the next frontier. For crypto, it’s a narrative bonanza. The “AI + Crypto” convergence has been stuck in agent hype. This pushes it toward real utility.
Core: The Tokenization Mechanism and Sentiment Analysis
Let’s dissect the mechanism. Wang Jian envisions a universal technical architecture that ingests all modalities—text, code, and scientific data. That’s ambitious. It requires breaking down scientific data into atomic units that can be sequenced and attended to by transformers. Existing tokenization methods (BPE, WordPiece) fail on spatial and geometric data. The solution might be a new primitive: data embeddings that capture multi-dimensional relationships.
From a crypto perspective, this creates a direct need for on-chain data ownership and verification. If a research institute tokenizes its protein folding dataset, it can license it to AI models via smart contracts. Compute providers (Render, Akash) can offer GPU time, paid in tokens, to process those datasets. Data DAOs can curate, validate, and sell scientific data. The entire value chain becomes programmable.
But here’s the sentiment reality: the crypto market is currently obsessed with AI agents running on L2s. Every week a new project launches promising autonomous trading or social media agents. I’ve seen the numbers. The daily active user counts are inflated by bots. The narrative is oversupplied. Meanwhile, projects focused on decentralized scientific computing—like Akash or even the nascent DeSci vertical—are undervalued. A quick look at token prices over the past 90 days shows Akash up only 12% while L2 agent coins are up 80-200%. The market is chasing the wrong utility.
Note: Sentiment turning bearish on L2s.
My own audit of DeFi protocols shows that L2 liquidity is thinning. Total value locked across major L2s has dropped 8% in April. The compute-intensive nature of scientific data tokenization does not favor L2 architectures with high proving costs. ZK rollups are bleeding money even at current gas prices. If scientific data tokenization requires large-scale computation (e.g., training foundation models on weather data), it will favor Layer 1s or dedicated sidechains with deterministic execution and low latency.
Note: ZK Rollup proving costs are absurdly high; unless gas returns to bull-market levels, operators are bleeding money.
Contrarian: The Blind Spot in the AI-Crypto Narrative
Every major crypto media outlet is bullish on AI agents. The reasoning: agents will drive on-chain activity, creating demand for blockspace and revenue for protocols. It’s a clean story. But it’s also a liquidity trap. Most agent projects lack sustainable revenue models. Their value is derived from attention, not utility.
Wang Jian’s thesis suggests the real value is downstream—in data infrastructure, not agent interfaces. The agents themselves are just frontends. The backend will be dominated by data tokenization networks and decentralized compute. Projects that enable scientific data to be stored, verified, and computed upon will have higher moats. They will own the supply-side assets.
This is where my experience in the DeFi derivatives crisis taught me a lesson. In 2020, I wrote a white paper arguing that order-book centralization was the only path for institutional capital. The market ignored it until the AMM bubble burst. Similarly, the market is ignoring data infrastructure today. It’s a classic second-order effect. The first order is “AI agents are hot.” The second order is “who provides the data they train on?” The second order is where the lasting value lives.
Note: Oracle feed latency is DeFi's Achilles' heel; Chainlink solving decentralization with centralized nodes is itself a joke.
The same logic applies to scientific data. If a single entity controls the tokenization standard (e.g., Alibaba Cloud), it will extract massive rents. Decentralized alternatives need to emerge. Chainlink’s oracle model could theoretically verify data provenance, but its centralized node structure undermines trust. A truly decentralized data tokenization protocol would require zk-proofs for data integrity and a distributed network of curators.
Takeaway: The Next Narrative Cycle
Wang Jian has given us a roadmap. The next crypto narrative will not be “AI agents.” It will be “Data as a Layer 1.” Specifically, scientific data tokenization. Projects that bridge DeSci, compute marketplaces, and data provenance will absorb the liquidity currently chasing agent coins. I am already shifting my portfolio toward tokens that enable this thesis: Arweave for storage, Filecoin for data marketplace, Akash for compute, and fresh DeSci tokens like Bio Protocol or Molecule.
The market is always wrong at the extremes. Right now, it’s wrong about L2s and agents. It’s underestimating the infrastructure layer. The signal from Wang Jian’s speech is clear: the next wave of AI will be built on scientific data. And crypto is the only neutral settlement layer for that data.