Written by: Ekko an, Ryan Yoon

Translated by: Chopper, Foresight News

TL;DR

• In the context of the booming artificial intelligence industry, we need to evaluate the blockchain sector from the demand side: what problems does it solve that existing systems cannot, and what unique capabilities does it bring?
• Decentralized computing power and decentralized storage have reasonable logic regarding data sovereignty and cost advantages, but they have not yet formed absolutely persuasive technical advantages that are sufficient to make enterprises, deeply bound to traditional cloud service providers, bear the risk of switching.
• Model validation and privacy encryption technologies cannot resolve the urgent business pain points of enterprises, and companies will not proactively adopt them on a large scale; the demand in this space is likely to lag behind the introduction of regulatory policies, as seen in the case of the EU AI Act: standards must be established before market demand follows.
• The bottleneck of the underlying infrastructure track for AI intelligence does not lie in technology. Currently, mainstream enterprises focus on internal process automation, while blockchain projects are developing the underlying facilities for the next stage, and market demand maturity cannot keep up with the pace of technological development.
• AI intelligent payments are the only track where blockchain and traditional finance are on the same starting line, with both sides failing to properly address industry pain points, making it currently the only segment with direct competitive conditions.
• Overall, the dilemma of the blockchain + AI track is not about a contradictory logic of their combination but a serious mismatch between supply and demand. Each of the four sub-tracks has unique demand gap issues, with only the AI intelligent payment track having conditions for direct market competition at this time.

AI fully erupts, while the blockchain track is far behind

The AI industry is experiencing an unprecedented surge of capital and infrastructure investment, with large technology companies building large model ecosystems that fully penetrate public life and industrial production. The crypto industry is also rapidly iterating, trying to find technical integration points with AI.

Early exploration focused on supplementing and replicating traditional AI industrial chain segments: decentralized GPU computing power supply, data rights verification, and cryptographic model validation. Recently, the industry focus has shifted to solving pain points that centralized architectures struggle to address, including on-chain interactions for AI entities and real-time settlement between machines.

Using "AI + blockchain" as a general term for the entire track would only obscure the real differences in the subfields; we need a rigorous demand-side analysis: what problems does each sub-track target? Can native blockchain solutions provide truly differentiated solutions?

Four sub-tracks

Decentralized computing power

The current cloud market is highly reliant on a few leading technology companies controlling computing power resources. The difficulty and high cost of procuring high-performance GPUs put an extremely high entry barrier for AI startups and research institutions that lack the ability to build large-scale infrastructure.

Centralized platform resources tend to favor large clients, and there is a lack of neutral channels to allocate the massive idle GPU computing power in the market.

Decentralized computing power solves the issues of resource concentration and inefficiency through two models. The sharing economy model aggregates idle graphics card resources from individuals and small data centers, building a unified computing power network that bypasses the monopolies of tech giants to create a flexible supply system.

The distributed computing model allows users to rent computing power globally, not relying on the hardware of a single service provider, thus improving the utilization of idle hardware and lowering the threshold for using high-performance computing power.

Decentralized storage

The existing data storage system is almost completely dependent on centralized cloud service providers like Google and Meta. Once users upload data, the actual ownership of the data transfers to the platform, monopolizing AI training data for a long time by the giants. Centralized architectures also carry operational risks: policy changes, service interruptions, or platform failures may result in data becoming inaccessible or even permanently lost.

Decentralized storage addresses these structural issues through two methods. The sharing economy model represented by Filecoin and Arweave channels the idle storage space of various participants into a network that can replace existing centralized clouds.

The permanent storage model backs up data across multiple distributed nodes, unaffected by the operational status of a single server, thereby reducing dependence on any one platform.

On-chain data trading market

AI research and development requires massive amounts of training data, but the existing data circulation market is highly closed, with Hugging Face and major cloud providers monopolizing revenue and pricing power. Data creators earn little, and the incentive mechanism for data contribution lacks transparency.

On-chain trading markets utilize smart contracts to eliminate intermediaries and establish transparent trading rules. In direct trading models like Ocean Protocol, data owners and AI developers trade directly via smart contracts, with compensation allocated transparently. In contribution reward models like Grass, individuals connect their idle bandwidth to AI data collection and receive corresponding rewards based on their contribution value.

Model inference verification and privacy protection

Traditional AI is a black box system where external validation of compliance with model calculations and secure handling of sensitive user data is impossible.

Zero-knowledge machine learning (ZKML) overlays a cryptographic verification mechanism onto the AI inference layer, achieving both privacy protection and audit traceability. Model calculations still occur off-chain, but the process generates cryptographic credentials to prove that the entire process strictly follows preset rules.

Such proof is recorded on-chain, rather than the underlying data. For example, in auto-claim scenarios for health insurance, the hospital only uploads the AI processing compliance certificate, without needing to upload the complete patient medical record; the insurance company can verify the authenticity of the certificate to complete the claim, without ever accessing the original private medical data.

AI entity framework

AI entities are gradually becoming the core of traffic and value creation, evolving from tools to autonomous economic agents. The existing financial system is designed based on human consumption behavior, making it inherently incompatible with machine-led payment scenarios.

The agent economy requires millisecond-level high-frequency microtransactions and real-time cross-border settlements, which traditional financial infrastructure cannot support.

On-chain agent infrastructure addresses this issue through two mechanisms. The autonomous execution and control mechanism allocates a unique wallet and identity to the AI agent, enabling it to directly sign transactions and set configurable spending limits and security measures to prevent accidental behavior.

The protocol-based settlement mechanism employs stablecoin payment protocols (e.g., x402) for real-time settlement of microtransactions and high-frequency payments, bypassing currency conversion and approval processes.

Differences between blockchain + AI and traditional AI industry chain

The capital logic of the traditional AI industry revolves around "breaking development bottlenecks." As AI demand expands, memory, electricity, and data transmission bandwidth have become limitations, and companies that can quickly address these constraints (such as high-bandwidth memory manufacturers and electric infrastructure companies) will reap significant funding and market value increases. The market is willing to pay high valuations for solutions that break growth bottlenecks.

Blockchain + AI projects do target real industry pain points, but they have never managed to gain the same level of market attention. If these issues were genuinely urgent, large-scale transitions would have already occurred.

Even though decentralized computing power, data rights verification, and other tracks have reasonable value, they struggle to attract mainstream capital. The core conflict lies in the serious disconnection between the technology suppliers and the procurement parties holding the funds.

The pace of development in the AI industry is rapid, and buyers (mainly large tech companies and enterprise clients) tend to invest massively in solutions that can most quickly resolve their current operational bottlenecks. They do not spend time evaluating unproven infrastructures. Their primary concerns are computational performance, infrastructure reliability, and measurable returns on investment.

For example: when data transmission speeds become the bottleneck for model training, a large amount of capital flows into fiber-optic infrastructure to replace copper cables. When memory bandwidth becomes the main limiting factor, SK Hynix and Samsung Electronics have solved this issue by providing high-bandwidth memory, thereby gaining significant recognition globally. This model remains consistent: capital will follow those companies that can eliminate constraints and promote progress.

The fundamental issue in the blockchain + AI track is a deviation in positioning. Enterprises with substantial budgets only care about short-term performance improvements and cost reductions, while the blockchain AI projects focus on secondary long-term issues in the eyes of enterprises. The technological vision on the supply side does not match the current operational needs on the demand side.

The technological vision on the supply side does not match the current operational needs on the demand side.

Insufficient technical hard power

Many projects demonstrate the potential and design concepts of decentralized infrastructure through benchmarking tests but have not achieved disruptive technological breakthroughs that can shake the market's deep-rooted centralized cloud providers (e.g., AWS, GCP).

Centralized cloud platforms already possess vast amounts of capital and mature infrastructures. For new technologies to capture market share, they must have overwhelming performance advantages that justify enterprises accepting switching costs. For instance, Apple's switch from Intel chips to its self-developed M1 chips required bearing the substantial risk of software compatibility failures, but it was supported by a threefold improvement in energy efficiency, the benefits of which were enough to offset the costs of transitioning.

However, blockchain + AI currently cannot provide sufficiently persuasive benefit logic for enterprises requiring PB-level data synchronization and ultra-low latency, and enterprises are unwilling to bear migration risks.

Structural mismatch in supply and demand

Some decentralized computing power projects offer service level agreements to reduce enterprise risks, but companies remain cautious; the underlying issue does not lie in the contracts but in the structural foundation: leading cloud service providers can offer exclusive isolated server rooms, while blockchain networks rely on decentralized, anonymous nodes for computing power.

Once a certain node drops offline, it interrupts model training worth hundreds of millions, and token refunds or cash compensation cannot compensate for the time cost and business opportunity lost by enterprises. In a fiercely competitive industry, system stability is a non-negotiable bottom line. Even with accompanying risk-hedging tools, enterprises have no incentive to accept the uncertainties inherent to decentralized networks.

Market demand has not matured

The blockchain agent framework is aimed at a mature ecosystem of multi-agent collaborative autonomy, but mainstream market development has yet to reach this vision.

While companies like Microsoft and Salesforce are accelerating the adoption of AI agents, they currently focus entirely on automating internal processes. The infrastructure built by blockchain projects serves the next stage: autonomous agents operating independently across external networks. Currently, the vast majority of enterprises are still working on the stability and return on investment of their existing AI systems, and cross-network multi-agent collaboration is nowhere near the priority list of enterprise infrastructure planning.

The current dip in demand is a cyclical development issue rather than a technical flaw. The blockchain agent infrastructure is better positioned as a long-term infrastructure layout for the future agent economy rather than a short-term monetization business.

Regulation

Zero-knowledge proofs and privacy encryption technologies are core solutions for building trustworthy AI, but in the early stages of AI adoption, the proactive need for enterprises to implement privacy infrastructure is very low. It is difficult to rely on voluntary enterprise initiatives to drive large-scale adoption; demand in the industry is likely to stem from regulatory standards, with technology subsequently aligning with compliance requirements.

Global regulatory frameworks like the EU AI Act continue to be refined, providing benefits for the track. When data traceability and data security become rigid legal requirements, the verification capabilities of blockchain will transition from an optional feature to a compliance necessity for enterprises implementing AI.

Enhanced regulation is not an industry constraint but a catalyst created by the market. Clear regulations reduce industry uncertainties, opening stable pathways for blockchain + AI in institutional markets.

No benchmark landing cases

Multiple structural contradictions have generated the core obstacle: there are no convincing large-scale benchmark cases that prove commercial value. The traditional AI industry has relied on ChatGPT to create a growth flywheel, with a standout product that everyone can see attracting massive capital and talent for continuous iteration.

To date, the blockchain + AI track lacks equivalent products that match the market. Aside from early community interest, no project has penetrated enterprise production or everyday consumer scenarios enough to capture the attention of traditional institutional capital. The lack of benchmark landing cases poses the greatest barrier to discourage conservative institutional funding and delay industry adoption.

Does blockchain + AI have long-term value?

Setting aside the short-term market hype, blockchain + AI has yet to establish a foothold in the mainstream AI industry chain, but this does not mean that their combination lacks value.

The core reason for the cooling of the track is not a contradictory logic of the technical combination but rather the misalignment between mature industry demands and the direction of technical supply in each sub-track.

The core requests of the traditional AI industry are clear: short-term performance enhancement, cost optimization, and extreme infrastructure stability; whereas the vast majority of blockchain AI solutions focus on data ownership, computational transparency, and decentralization.

These are not the bottlenecks urgently requiring solutions in the industry; implementation often requires sacrificing performance, making the cost-benefit ratio difficult to convince enterprises.

Before the surge of artificial intelligence, electric infrastructure companies were generally categorized as mature, slowly growing firms. The explosive growth in power demand driven by data centers changed this status, attracting significant market attention. Current indifference towards blockchain artificial intelligence may also reflect a similar lag effect; the value of infrastructure may not become fully apparent until a new paradigm emerges.

During this transitional period, it is crucial how the industry responds to actual market demands.

The path forward splits into two directions: 1) Actively adapting to mature AI industry chain standards to address short-term performance gaps; 2) Sticking to existing technical routes and continuously laying out infrastructure that suits the large-scale landing of the next generation of AI in the long term.

The ultimate direction of blockchain + AI depends on which path aligns best with future real market needs.