AI and Web3 Investment Opportunities: Prospects and Opportunities from an Investor's Perspective (Part 2)
TechFlow Selected TechFlow Selected
AI and Web3 Investment Opportunities: Prospects and Opportunities from an Investor's Perspective (Part 2)
People have been exploring the intersection of AI and Web3.
Navigating Web3 tides with focused insightsAuthor: Lao Bai, Research Partner at ABCDE
In the previous article, the author introduced one of the six Web3+AI models—Bot/Agent/Assistant assetization. This article will further explore the remaining five models: computing power platforms, data platforms, generative AI, DeFi trading/auditing/risk control, and ZKML.
Computing Power Platforms
There are fewer projects in the computing power platform space compared to bot model assetization, but they are relatively easier to understand. Everyone knows AI requires massive computational power, and BTC and ETH have proven over the past decade that there exists a method—spontaneous, decentralized, coordinated under economic incentives and game theory—to mobilize vast amounts of computational power for cooperative and competitive tasks. Now, this approach can be applied to AI.
The two most prominent projects in the industry are undoubtedly Together and Gensyn: one raised tens of millions in its seed round; the other secured $43 million in Series A funding. The reason these two raised such large sums is reportedly to fund the initial training of their own models, after which they plan to operate as computing platforms providing training services for other AI projects.
In contrast, inference-focused computing platforms tend to raise significantly less capital, as their core function is essentially aggregating idle GPU and other computing resources and supplying them to AI projects needing inference support. RNDR aggregates rendering compute power—these platforms do the same for inference compute. However, technical barriers remain blurry, and I even wonder if RNDR or some Web3 cloud computing platforms might eventually expand into inference compute platforms.
Compared to model assetization, the computing power platform direction is more tangible and predictable. It's almost certain this sector will see demand and give rise to one or two dominant players—it just remains to be seen who will emerge. The only current uncertainty is whether separate leaders will arise for training and inference, or if a single leader will dominate both.
Data Platforms
This one isn't hard to grasp either, since at the foundation, AI rests on three pillars: algorithms (models), computing power, and data.
If we already have "decentralized versions" for algorithms and computing power, data naturally won't be left out. This is also the direction most favored by Dr. Lu Qi, founder of MiraclePlus, when discussing AI and Web3.
Web3 has always emphasized data privacy and sovereignty, with technologies like ZK ensuring data reliability and integrity. Therefore, AI trained on Web3 on-chain data should differ from that trained on off-chain Web2 data. This narrative makes sense overall. Currently within the ecosystem, Ocean is considered representative of this track, and in the primary market, there are already projects building specialized AI data markets based on Ocean.
Generative AI
Simply put, this involves using AI for image generation or similar creative tasks serving other use cases, such as creating NFTs or generating in-game maps and NPC backstories. The NFT path seems challenging because AI-generated content lacks sufficient scarcity. GameFi might be a better fit, and there are teams in the primary market attempting this.
However, recently I heard news that Unity (which, together with Unreal Engine, has long dominated the game engine market) has launched its own AI generation tools, Sentis and Muse. These are still in closed beta and expected to officially launch next year. Honestly, I feel Web3-based gaming AIGC projects may face a devastating blow from Unity...
DeFi Trading/Auditing/Yield/Risk Control
Projects are emerging across all these categories, with relatively low homogenization.
-
Trading—this is somewhat tricky. If a trading strategy works well, the more people use it, the less effective it becomes over time, necessitating constant shifts to new strategies. Also curious about how successful AI trading bots will be in the future and what tier they'll occupy among retail traders.
-
Auditing—likely helpful for quickly identifying known common vulnerabilities, but probably ineffective against novel or logical flaws. That would require AGI-level capabilities.
-
Yield—it's easy to understand. Just imagine an AI-powered YFI: deposit funds, and the AI automatically stakes, forms LP pairs, mines, etc., according to your risk preferences.
-
Risk control—feels odd as a standalone project. More sensible as a plugin serving various lending or DeFi platforms.
ZKML
A rapidly growing area within the community, combining two cutting-edge technologies: ZK from the crypto world and ML (machine learning, a narrow branch of AI) from outside.
Theoretically, integrating ZK can provide ML with privacy, integrity, and accuracy. But when pressed for concrete use cases, many project teams struggle to name any—so they're building infrastructure first and figuring out applications later...
Currently, the only real, urgent need exists in certain medical machine learning applications where patient data privacy is critical. Use cases like on-chain gaming integrity or anti-cheating mechanisms feel somewhat forced.
At present, there are only a few standout projects in this space—such as Modulus Labs, EZKL, Giza—each highly sought after in the primary market.
Well, it's understandable—there are very few people globally who understand ZK, and even fewer who understand both ZK and ML. Hence, this sector has much higher technical barriers and lower homogenization compared to others.
Lastly, most ZKML applications focus on inference rather than training.
That's all for now regarding trends in AI + Web3. If you've come across promising hybrid projects or new ideas outside the six categories above, feel free to message me.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Lao Bai
