AI × Web3: Building Infrastructure for a New Era

The true transformation of technological paradigms is often accompanied by a frenzy rather than a mature system. The current wave of AI is precisely this.

As a primary investor, I have always believed that the transformative power of betting on the deepest changes in the industry is far more valuable than chasing superficial narratives. Over the past year, I have encountered a large number of projects exploring the integration of real-world systems and on-chain systems. An obvious trend is that regardless of the project path, there is ultimately a need to incorporate AI collaborative logic to enhance competitiveness and efficiency.

For example, RWA projects need to consider how to utilize AI for risk control optimization, off-chain data verification, and dynamic pricing; consumers or DeFi projects require AI to complete user behavior prediction, strategy generation, and incentive distribution. Whether it is asset digitization or experience optimization, these seemingly independent narratives will ultimately converge on the same technological logic: if the infrastructure lacks the capability to integrate and support AI, it will not be able to sustain the complex collaboration of the next generation of applications.

The future of AI is not just about becoming more powerful and widely applied; the real paradigm shift lies in the reconstruction of collaborative logic. Just like the early transformations of the internet were not just about inventing DNS or browsers, but for the first time allowing everyone to participate in content creation and turn ideas into products, thus giving rise to an entire open ecosystem.

AI is following a similar path: intelligent agents will become co-creation partners for everyone, helping to transform expertise, creativity, and tasks into automated productivity tools, and even realize monetization. This is a question that is difficult to answer in the current Web2 world, and it is also the underlying logic of my focus on the AI + Web3 track: making AI collaborative, transferable, and profit-sharing is the system that is truly worth building.

Today, I want to discuss the only project so far that attempts to systematically build the underlying operations of AI from a chain-level structure: Sahara.

The essence of investment is worldview, recognizing the value system of choices.

My investment logic is not simply to combine the narrative of public chains with AI and then choose well-backed teams to bet on. Investment is essentially a choice of worldview, and I am always asking a core question: Can the future of AI be owned by more people together?

Can it leverage blockchain to reconstruct the value attribution and distribution logic of AI, allowing ordinary users, developers, and other different roles to have the opportunity to participate, contribute, and continuously benefit? I only believe that such projects could become disruptors when this logic emerges, rather than just another public chain project.

To find the answer, I researched almost all the AI projects I could access until I encountered Sahara. The response given by Sahara's co-founder was: to build an open, participatory ecosystem that everyone can own and benefit from.

This simple answer precisely hits the soft spot of traditional public chains: they often serve developers in a one-sided manner, and the token economic design is mostly limited to gas fees or governance, making it difficult to truly support the positive cycle of the ecosystem, and even harder to sustain the development of emerging sectors.

I am well aware that this road is full of challenges, but precisely because of this, it is a revolution that cannot be refused - and this is the reason for my firm investment.

As I emphasized when I previously discussed the "evolution from Web2 to Web3": the true paradigm shift is not about creating a single product, but about building a supportive system. Sahara is precisely one of the most anticipated cases I predicted at that time.

From investment to 8x valuation follow-up heavy investment

If I initially invested in Sahara because it is realizing the true leading mission of AI in my heart—building an AI economy and infrastructure system, then what made me rush to invest at eight times the previous round valuation in just six months is the rare strength I felt in this team.

The two co-founders are the youngest tenured professor at the University of Southern California and a former investment director. Their professional backgrounds, self-discipline, and full commitment to the project are impressive. I feel fortunate to have met them, as it has also changed my own lifestyle.

When someone says that Sahara gained the favor of capital due to luck, I will unapologetically add, "The pursuit of capital is an inevitable result." I clearly remember the difficulty of primary financing in this market, but Sahara was being chased by investors in the primary market.

In addition to well-known investment institutions, some funds and national banks that focus on AI have also become supporters of Sahara. What you can see is a group of more traditional technology and industry resource institutions that are quietly laying out in the AI × Web3 field because of Sahara.

Capital will only pay for a certain direction and execution capability - this is a positive feedback on the depth of Sahara technology, team background, system design, and execution ability.

This also explains why it can produce some real and solid structural indicators: over 3.2 million accounts have been activated on the test network, with over 200,000 data platform annotators (millions are in queue), serving clients that include several leading technology companies, and achieving revenue levels in the tens of millions of dollars.

On this infrastructure chain, at least from "who will do it" to "can it be done", Sahara has gone deeper and steadier than 99% of the "AI Narrative projects".

The ultimate challenge of public chains: to ensure that all contributors continue to benefit and drive a positive economic cycle.

Returning to our initial judgment logic: In a system where AI and blockchain are combined, is there really a mechanism that allows every contributor to be seen, recorded, and continuously rewarded?

Model training and data optimization rely on a large amount of labeled and interactive support; conversely, if there is a lack of user contributions, the project itself has to invest more funds to acquire data and outsource labeling, which not only increases costs but also weakens the value-driven community co-construction.

Sahara is one of the few Web3 AI projects that allows ordinary users to "participate in data construction from day one." Its data annotation task system operates daily, with a large number of community users actively participating in annotation and prompt creation. This not only helps the system improve but also invests in the future with data.

Through the mechanism of Sahara, not only is the quality of the model improved, but it also allows more people to understand and participate in this decentralized AI ecosystem, linking data contribution with rewards and forming a true virtuous cycle.

A typical example is a voice project on a certain public blockchain, which quickly built a high-quality dataset covering multiple languages and accents by leveraging Sahara's decentralized data collection and human-machine collaborative annotation, significantly enhancing the training efficiency of its voice model. This also propelled its open-source project to receive thousands of GitHub stars and over 2 million downloads.

At the same time, users participating in data labeling also received token rewards issued by the project, creating a two-way incentive loop between developers and data contributors.

Sahara's "permissionless copyright" mechanism safeguards the rights of all participants while ensuring the open circulation and reuse of AI assets—this is the underlying logic driving the explosive growth of the entire ecosystem.

Why is it said that this is a scenario with long-term value support?

Imagine if you want to build an AI application, naturally you would hope your model is more accurate and closer to real users than others.

The key advantage of Sahara is that it connects you to a vast and active data network—hundreds of thousands, and in the future, millions of annotators. They can continuously provide you with customized, high-quality data services, allowing your models to iterate faster.

More importantly, this is by no means a one-time transaction. Through Sahara, you are connecting to a potential early user community; and these contributors are likely to become your product's real users in the future.

This connection is not a one-time buyout; through Sahara's smart contract system and rights confirmation mechanism, it can achieve a long-term, traceable, and sustainable incentive system.

Regardless of how many times the data is called, contributors will receive continuous profit sharing, with earnings dynamically linked to usage behavior.

But this is not just a revenue model for data labeling and model training phases. Sahara builds an economic system that covers the entire lifecycle of AI models, with built-in profit-sharing mechanisms at every stage of model deployment, invocation, combination, and cross-chain reuse, allowing value to be captured over a longer period.

Model developers, optimizers, validators, and computing power contribution nodes can now continuously benefit at different stages, rather than just relying on a one-time transaction or buyout.

This system brings a compound effect for model combination calls and cross-chain reuse. A trained model, like building blocks, can be repeatedly called and combined by different applications, with each call generating new revenue for the original contributor.

For this reason, I agree with Sahara's underlying belief: a truly healthy AI economic system cannot just be about the plundering of data and the buyout of models, nor can it merely allow a few people to reap all the benefits. It must be open, collaborative, and mutually beneficial—where everyone can participate, every valuable contribution can be recorded, and continuous rewards can be gained in the future.

The closer we get to the real structure, the more challenges there are.

Although I am optimistic about Sahara, I will not cover up the challenges that the project will face because of my investment position.

One of the major advantages of the Sahara architecture is that it is not limited to a specific chain or a single ecosystem.

Its system was designed from the very beginning to be open, full-chain, and standardized: supporting deployment on any EVM-compatible chain, while also providing standard API interfaces that allow Web2 systems—whether e-commerce backends, enterprise SaaS, or mobile apps—to directly invoke Sahara's model services and complete on-chain settlements.

However, despite the extreme scarcity of this architectural design, it also carries a core risk: the value of the infrastructure lies not in "what it can do," but in "who is willing to do what based on it."

To become a trusted, adopted, and composable AI protocol layer, the key for Sahara lies in how ecosystem participants assess its technological maturity, stability, and future predictability. Although the system itself has been built, whether it can truly attract a large number of projects to implement based on its standards remains uncertain.

Undeniably, Sahara has achieved key validation: providing data-related services to multiple leading tech companies and addressing some of the industry's toughest data demand issues, becoming an early signal of the feasibility of this system.

However, it should be noted that these collaborations mainly come from the Web2 world. What truly determines the long-term development of Sahara is still the maturity and penetration of the entire Web3 AI track. Sahara benefits from the broader trend of Web3 AI, but to truly unleash the value of its infrastructure, it still requires more native Web3 AI products and technological solutions to be implemented and improved.

But don't forget, Sahara is currently "the only one of its kind."

In the native blockchain infrastructure track designed for AI, although there are many imitators proposing conceptual frameworks, only Sahara has achieved a complete closed-loop with real income from on-chain confirmation, off-chain execution, and cross-chain invocation, receiving validation from actual customers.

This not only gives Sahara a "monopoly advantage," but also brings structural risks: once successful, it will define the entire Web3 × AI Infra industry benchmark; however, if it fails, it may also lead to AI Layer1 being seen as a premature layout.

Since it is now the only option in this field, the market's judgment on it must naturally be stricter and more rational — it must withstand the test of time and ecology.

Finally, a message to all builders and observers｜Seize the window of the building period, rather than regretting it after it has taken shape.

For me, the core of every level of investment decision-making boils down to three things: the depth of understanding of the world, the dimensions of trend judgment, and the willpower of the team to navigate through cycles. Products and features are certainly important, but they often merely represent the concrete manifestation of these underlying cognitions.

Web3 is not lacking in ideas, nor in stories, what it lacks is the ability to implement them.