Fidelity warns: The boom in AI proxies may not be a feast for public blockchains.

Original title: Will AI Proxy Systems Really Bring Prosperity to Public Blockchains? Fidelity Warns: Don't Equate the Two Tracks Too Quickly

If we turn back the clock a few years, the blockchain industry was talking more about "users going on-chain." Users would register wallets, buy tokens, conduct transactions, and then complete asset activities through DeFi, NFTs, or other on-chain applications.

The emergence of AI agents may change this way of using them.

An AI agent capable of autonomously performing tasks could theoretically handle tasks such as searching, purchasing, making payments, accessing data, and even collaborating between software programs on behalf of users. If, in the future, the agent needs to have its own wallet, access on-chain services, or perform automated settlements between machines, then blockchain could indeed become an optional infrastructure.

The problem is that there is a big difference between "can be used" and "must be used".

This is also what makes Fidelity Digital Assets' analysis noteworthy. In the past, the market easily deduced from "AI agents need payment and identity" that "AI agents need public blockchains," and further deduced that "the increase in public blockchain usage will drive the increase in token value."

But this actually involves at least three assumptions.

First, AI agents must conduct on-chain settlements; second, on-chain settlements must occur on permissionless public networks; and third, economic activities generated on public networks must be effectively transmitted to native token holders.

If any one of these steps fails, the entire investment logic needs to be re-evaluated.

First risk: AI agents may not necessarily require a public blockchain.

This was the first key question raised by Fidelity.

Imagine an AI agent within a company. It needs to access the company database, invoke cloud services, execute procurement tasks, and complete payments according to employee permissions. For enterprises, a closed system built by a large technology company or financial institution may be more likely to meet practical needs than an open, public blockchain.

The reason is not complicated.

Enterprises are typically more concerned with system speed, cost, stability, identity authentication, access control, and regulatory responsibility than with whether the network is completely open.

If a closed infrastructure can provide lower transaction costs, more stable performance, and clearer data and compliance boundaries, then there is no compelling reason for enterprises to choose a public blockchain simply because it is "decentralized".

This means that the growth of AI agents cannot automatically translate into the growth of public blockchains.

The future may even present a seemingly paradoxical situation: a significant increase in AI agents, but a considerable portion of their activity will occur within enterprise databases, private networks, consortium systems, or closed infrastructure controlled by large platforms.

For public blockchains, the real question is not "whether AI will use blockchain", but "why it is necessary to use a public blockchain".

The second risk: Increased on-chain transactions may not necessarily benefit tokens.

This is the layer that the market most easily overlooks.

In the past, the crypto market often used a simple logic: increased network usage → increased transactions → increased transaction fees → increased demand for native tokens → increased token value.

However, the payment activities brought about by AI agents may not be so direct.

Assuming a large number of AI agents complete micro-payments via blockchain in the future, the number of online transactions could indeed increase rapidly. However, if the amount of each transaction is very small, or if the network's transaction fees remain at a low level for a long period, then the massive number of transactions may not generate commensurate economic value.

More importantly, the real beneficiaries of the revenue may not be the holders of the public chain tokens.

Stablecoin issuers, payment service providers, wallets, and companies responsible for proxy infrastructure may all occupy different positions in the value chain.

This actually raises a long-standing question in public blockchain investment: how far is there between network prosperity and token value?

If AI leads to more payments, but these payments are primarily completed using stablecoins, then the growth of the AI economy may first strengthen the use of stablecoins, rather than directly strengthening the native assets of a particular public blockchain.

Therefore, when judging the opportunities of AI + blockchain in the future, simply counting the number of on-chain transactions may not be enough. It is also necessary to observe transaction fee income, asset settlement methods, and which layer captures the value.

The third risk: AI writing more code does not equate to creating more value.

AI is rapidly reducing software development costs, which is a relatively clear trend.

Functionality that previously required several engineers and weeks to complete can now be implemented in a much shorter time using AI programming tools. For the blockchain industry, this means a lower barrier to entry, allowing more teams to create wallets, smart contracts, DeFi applications, and various proxy tools.

However, Fidelity makes an important point: the number of codes and economic value are not the same concept.

If AI makes developing a blockchain application cheaper, the market may see a surge in new projects, but an increase in the number of projects does not necessarily mean a corresponding increase in user demand.

Conversely, the lower the development threshold, the easier it is for supply to become excessive.

In the past, a project could create a certain barrier to entry based on its technological development capabilities. However, as AI commercializes some of the R&D work, the same function may quickly appear in multiple versions. As a result, the focus of competition will gradually shift from "who can develop it" to "who owns users, liquidity, brand, security record, and distribution channels".

This is a significant change for entrepreneurs.

AI has lowered the cost of starting a business, but it may also lower the barriers to entry for competition.

The fourth risk: Technological advantages are becoming less scarce.

If any team can quickly generate code with the help of AI, then the scarcity of "technological leadership" itself may decrease.

In the past, a project's competitive advantage might have come from complex smart contracts, underlying infrastructure, or the capabilities of its development team. However, in the future, when similar functionalities can be quickly replicated, it will become increasingly difficult to build a long-term competitive advantage solely based on code.

This does not mean that technology is unimportant.

On the contrary, security, stability, and architectural capabilities may be more important.

Only the logic of competition has changed.

For AI+blockchain projects, what is truly difficult to replicate may be user relationships, liquidity, brand trust, ecosystem cooperation, and compliance capabilities.

This also explains why future competition between blockchain projects may increasingly resemble competition between internet platforms, rather than just a traditional technological contest.

The fifth risk: AI reduces development costs, but also reduces attack costs.

This is the risk that security professionals should be most wary of among the six.

AI can help developers write code, and it can also help attackers find problems in code.

In the past, discovering smart contract vulnerabilities might have required a professional security team to invest a significant amount of time in code auditing. However, with the improvement of AI tools, the barriers to vulnerability analysis, code understanding, and automated testing may be decreasing.

This means the industry may face two trends simultaneously.

On the one hand, AI enables more teams to develop blockchain products; on the other hand, it may also enable more attackers to analyze and find vulnerabilities.

If development speed far exceeds the speed of security audits, the risks to the entire ecosystem may increase.

This is especially important for institutional investors. When entering an emerging asset market, institutions are not only concerned with yields, but also with custody, access control, smart contract security, and the boundaries of liability in the event of system failures.

Therefore, if AI truly drives the rapid expansion of on-chain applications, whether security infrastructure can mature simultaneously may become a crucial factor in determining the speed of institutional adoption.

The sixth risk: What institutions may truly need is a "controllable blockchain".

The last question comes from regulators.

One of the biggest advantages of public blockchains is their openness and permissionlessness, but this may conflict with some of the needs of large financial institutions.

When banks, payment institutions, and large enterprises use AI agents to process assets, they need to know "who is operating," "what the agent can do," "where the data goes," and "who is responsible for any errors."

This means that identity authentication, access control, audit logs, and compliance controls may be more important than openness itself.

Therefore, a direction worth discussing has emerged: the blockchain infrastructure used by institutions in the future may not be a completely open public network, but a system with stronger access control capabilities.

This does not mean that public blockchains will necessarily be eliminated, but rather that the two architectures may coexist for a long time.

Public networks are responsible for open settlement and asset circulation, while enterprises or financial institutions control risks through permission layers, identity layers, and compliance infrastructure.

The real competition may not be as simple as "public blockchain or private blockchain", but rather who can find a more suitable balance between openness and controllability for AI agents.

What Fidelity is really reminding the market is not to simply add the two narratives together.

Ultimately, there is indeed a possibility of combining AI and blockchain.

AI agents require payment capabilities, and blockchain offers features such as global settlement, stablecoins, and programmable assets. AI can also improve software development efficiency and help users interact with complex on-chain applications. These are all real, potential needs.

However, there is still a long way to go from potential demand to real economic value.

In the past, the market liked to tell a very smooth story: AI agents increased, on-chain transactions increased, public chain usage increased, and token value increased.

Fidelity's six risks are essentially a reminder to investors that each arrow in this chain needs to be verified individually.

AI may drive the development of blockchain, or it may strengthen closed systems; on-chain payments may grow rapidly, but stablecoins and payment service providers may gain more value; AI may make development more prosperous, but it may also make products quickly homogenized; code will become cheaper, but vulnerabilities will also be easier to find.

Therefore, what truly deserves attention is not whether "AI will save public blockchains," but rather which infrastructures, once the AI economy takes shape, will be able to truly meet real demand and transform that demand into sustainable business value.

This may also be a key step in the next stage of AI and blockchain narratives moving from "storytelling" to "accounting".