The Ledger Was Right

In 2017, a broad consensus emerged across major consulting firms: blockchain technologies would transform enterprise systems. Forecasts emphasized applications in supply chain management, logistics, procurement, and inter-organizational data exchange.

This expectation was not merely theoretical. IBM and Maersk jointly developed TradeLens, a blockchain-based platform designed to streamline global shipping operations. The initiative attracted substantial investment, amounting to hundreds of millions of dollars and was developed over a period of seven years. TradeLens was discontinued in 2022.

The official explanation cited "insufficient adoption". Subsequent peer-reviewed analysis has provided a more granular account of the underlying causes, identifying several interrelated factors: limited stakeholder engagement, concerns regarding data confidentiality, ambiguous governance structures, and persistent reluctance among firms to share operational data with competitors on a shared ledger.

These findings point to a central issue.

Magnetic Core was among the consulting firms involved in these efforts, working across sectors with stakeholders such as shipbuilders, beverage suppliers, cobalt producers, tokenized workspace initiatives, digital identity programs, sports organizations, and government institutions worldwide. The firm participated directly in blockchain developments and observed everything in practice.

The primary constraint was not technological feasibility, but institutional compatibility.

Decentralisation and Institutional Misalignment

Enterprise blockchain initiatives were premised on a specific architectural logic: decentralisation. In practice, this entails the distribution of control across participants, the maintenance of a shared and often transparent ledger, and the reduction of reliance on centralized intermediaries.

However, such architectures impose implicit requirements on participating actors. Most notably, they require firms to relinquish certain forms of informational asymmetry in exchange for collective trust and system-wide coordination.

This requirement is in tension with the operating logic of most corporations.

Firms derive competitive advantage from controlled access to information. Supply chain opacity can function as a source of bargaining power; proprietary data contributes directly to margins and strategic positioning. From this perspective, participation in a shared ledger, particularly one characterized by immutability and transparency, may entail a perceived loss of advantage rather than a gain in efficiency.

The reluctance to adopt enterprise blockchain systems can therefore be understood not as resistance to innovation per se, but as a rational response to misaligned incentives.

Empirical Pattern: Where Distributed Ledgers Succeed

Nearly a decade after the peak of enterprise blockchain enthusiasm, the empirical pattern is relatively clear. Distributed ledger technologies have proven viable in domains where their architectural properties align with participant incentives.

These domains include financial technologies, payment systems, and decentralized finance (DeFi). In such contexts, trustless settlement mechanisms reduce counterparty risk, and no single participant derives sustained advantage from informational opacity. Network effects are intrinsic to the value proposition, and transparency functions as a feature rather than a liability.

Where these conditions are absent, adoption has been limited or unsustainable.

Implications for Contemporary AI Infrastructure

A parallel dynamic is observable in the current development of artificial intelligence infrastructure. Leading AI systems are increasingly integrated with, and dependent upon, a small number of large-scale cloud providers.

OpenAI operates in close alignment with Microsoft. Google develops and deploys models within its own cloud ecosystem. Anthropic is similarly aligned with Amazon.

As a result, the computational and infrastructural layers underpinning advanced AI systems are becoming concentrated within a small number of hyperscale providers.

This trend toward centralisation is occurring despite the availability of decentralised alternatives at the level of network architecture. The contrast with earlier blockchain discourse is notable: whereas enterprise actors previously rejected decentralisation due to perceived losses in competitive advantage, current AI development appears to embrace centralisation as a condition for scale and performance.

The AI community is, in effect, building an increasingly centralised information infrastructure, while framing this consolidation as progress.

The Return of Distributed Coordination

The limitations of centralised architectures do not disappear; they compound with scale.

As AI systems expand, a set of structural requirements becomes unavoidable: verifiable outputs, agent coordination, and mechanisms for exchanging value across independent actors.

At the same time, AI is becoming local. Agents can increasingly run on individual machines, outside centralised platforms. This changes the economic model: if intelligence can be owned and deployed independently, paying continuous rent for access becomes unnecessary.

The problem shifts from access to coordination.

How do independent agents discover each other, interact, verify results, and exchange value without a central intermediary?

Centralised systems address this by reinforcing control: gating access, mediating interactions, and embedding extraction into the infrastructure itself. Even emerging forms of "AI tokenisation" within these environments remain platform-bound: credits, API tokens, and usage units are issued, priced, and controlled by the same central actors, limiting interoperability and preserving rent.

A distributed approach reframes this. Utility tokenisation of AI services — where tokens represent compute, inference, data access, or agent capabilities — allows these resources to be exchanged across participants rather than locked within platforms. When combined with decentralised finance primitives, this enables open markets for AI infrastructure: agents paying agents, models paying for data, and services priced dynamically through interaction rather than fixed by providers.

In this model, payment is not for access to a system, but for value received from other participants.

This is the missing layer.

Coordination without rent requires shared, neutral infrastructure. Distributed ledger systems provide this layer: enabling trust, verification, and exchange between actors that do not rely on a central authority.

Blockchain was not solving enterprise workflows. It was solving coordination between autonomous participants.

As AI becomes agent-based and locally deployable, that problem re-emerges, now as an infrastructural requirement rather than a speculative use case.

Conclusion

The experience of enterprise blockchain provides a clear analytical result: decentralisation fails where incentives oppose it — but centralisation fails where systems scale beyond the capacity of controlled trust.

Distributed ledger systems demonstrated this with unusual clarity. By removing single points of control, they revealed both the technical feasibility of distributed coordination and the institutional resistance to it.

Enterprise actors rejected decentralisation not because it was ineffective, but because it redistributed power. That rejection did not eliminate the underlying problem. It deferred it.

Centralised systems continue to concentrate information, control, and dependency. This concentration produces efficiencies, but it also generates structural vulnerabilities: single points of failure, asymmetries of power, and susceptibility to capture. These are not edge cases. They are inherent properties of centralised design.

The distributed ledger made this visible. It showed, in operational terms, why centralised control is fragile across complex, interdependent systems.

The current trajectory of AI infrastructure reproduces these same conditions at a significantly larger scale.

This is the critical point: decentralisation is not adopted when it is technically possible. It is adopted when centralisation becomes unsustainable.

AI has not yet reached that threshold. But it is moving toward it.

As dependence on AI systems deepens, and as the costs of centralised control accumulate, economically, politically, and operationally, the pressure for verifiable, distributed coordination will increase.

When that pressure becomes binding, decentralised layers will not be optional.

They will be required.

The ledger did not fail. It established the boundary conditions. We are approaching them again.

The missing layer is not intelligence. It is coordination without rent.