At some point between 2022 and 2023, we stopped writing about the blockchain hype cycle. Not because we published a reckoning, or concluded the experiment had failed, or even noticed we were doing it. The critical coverage simply stopped. If you search the NEXT Insights archive today, you’ll find a rich seam of material running from 2016 to late 2022 – and then silence. That silence is worth examining, because we think it’s about to repeat.

The promise: data democracy over data monarchy

We were early believers. At NEXT16, Jamie Burke told us that blockchain would do to Web 2.0 what Web 2.0 had done to Web 1.0 – dismantle the platform layer and rebuild it with distributed ownership. The promise was explicit:

So, in theory, we could dismantle the big platforms of Web 2.0 and then effectively rebuild them on new blockchain platforms – but with more distributed ownership and wealth.

A year later, Shermin Voshmgir gave NEXT17 its most memorable formulation. The internet we had was broken, she argued. Web 2.0 had created a “peer to peer economy, with one player in the middle, who controls the data and the transactions.” Blockchain offered a way out:

Blockchain is a protocol of trust. It enables the move from the data monarchy – where data is stored at a centralised point – to data democracy.

She also issued a warning that, in retrospect, deserves more attention than we gave it at the time: “If we don’t get this right, we could create the centralised control machine.”

By NEXT18, Joseph Lubin was framing blockchain as nothing less than a new organising principle for civilisation. He had learned a lesson from the internet’s history – and thought blockchain would correct it:

The internet ended up enabling centralising technologies and traditional financial markets. This technology should give people more autonomy and disempower intermediaries.

The irony embedded in that sentence is worth sitting with. Lubin named the exact failure mode of the previous technology wave, and proposed this one as the antidote. He was right about the diagnosis. The cure didn’t work.

The peak: NFTs as Medici-era status games

Between 2016 and 2021, blockchain coverage was earnest and infrastructural – about protocols, smart contracts, identity systems, and the architecture of a decentralised web. In 2021, something shifted. The conversation moved from infrastructure to culture, from protocols to tokens, from building to collecting.

David Mattin captured the moment well in May 2022. He traced NFTs back to their real origin – not in code, but in the ancient human drive for status:

This is a very familiar, ancient status play: elite art ownership. It’s about more than showing off your wealth, it’s also about saying that you’re cultured, a patron of the arts. And that goes right back to the Medici family in the 15th century – and probably before that.

It was good analysis. It explained why NFTs had cultural traction. But it just didn’t ask whether cultural traction was enough to sustain a technology revolution. Ana Andjelic published a Web3 playbook for brands. DAOs were theorised as new forms of community ownership. The discourse expanded, the ambition inflated, and the infrastructure question – whether any of this was actually being built – became harder to find.

The crack: caveats we didn’t follow

In August 2022, we published something we should probably have written earlier. “The cultural dimension of Web3” offered a sober reassessment that now reads as a pivot point:

We shouldn’t take anything we currently associate with Web3 for granted. It could well be that neither blockchains nor NFTs, neither VR nor AR will have anything to do with Web3.

The piece also, almost in passing, raised the constraint that would quietly kill the cycle: energy. The global energy crisis – the war in Ukraine, the climate emergency – was identified as a potential blocker, not just a delay. It was the first time we had named physical infrastructure as a limiting factor rather than an enabling one.

By the time Trevor McFedries spoke at NEXT22, the mood had shifted. Our post-event write-up noted that the blockchain discourse had become dominated by investment narratives, that assets were collapsing in value, and that perhaps only a necessary correction would allow us to see the real potential behind the technology.

And then – nothing. No post-mortem. No reckoning with what had been promised. The topics simply ceased to appear in our coverage, which is itself the most honest verdict we could have delivered.

What actually worked – and why

The blockchain hype cycle didn’t end in total failure. Some applications survived the crash and continue to function. But the pattern of what survived is more instructive than the survival itself.

Liquid staking protocols, lending platforms with automatic liquidation, binary prediction markets like Polymarket – these all share a common characteristic. They handle fully self-contained transactions with no real-world ambiguity. If you deposit X, you receive Y. The smart contract executes because all the relevant conditions live inside the blockchain environment.

Tokenisation of real-world assets has also grown – bonds, treasury bills, money market funds managed by BlackRock, JPMorgan and Franklin Templeton now move across blockchain rails. But the underlying legal framework remains entirely intact. Blockchain is the settlement mechanism, not a replacement for law.

The pattern is consistent: what worked was automation of the settlement layer, not the judgement layer. Every blockchain application that thrived handles the “if X then Y” part of a transaction – the bit that was always mechanically straightforward. None of them replaced the human decision-making about whether X should happen in the first place. The decentralised identity system never arrived. The data democracy never came. Smart contracts didn’t displace lawyers. The “centralised control machine” Voshmgir feared is more powerful today than when she issued the warning.

A 2017 footnote we overlooked

There is a detail from our 2017 coverage that deserves retrospective attention. In a piece published that November, we noted – almost as an aside – that blockchain’s tokenisation promises bore a resemblance to patterns from twenty years earlier, and that we might see people “make the exact same mistakes they made almost 20 years ago.” In the same piece, almost in the same breath, we raised the question of AI-driven job displacement.

We were already holding both ideas simultaneously in 2017. The parallel between technology hype cycles. The connection between new technologies and labour displacement panic. We just didn’t pursue either thread far enough.

The pattern we’re now watching repeat

We are not arguing that AI is equivalent to blockchain. The capabilities are categorically different. The underlying technology is more substantial, the adoption more genuine, the economic value more real.

But the pattern of how we talk about it is strikingly familiar. The promises are recognisably similar, if in a different register. Where blockchain promised to dismantle the platform layer and redistribute structural power, AI promises something more modest: democratised access to expertise. Virtual tutors. Personal AI teams. The knowledge of a doctor or lawyer available to anyone with a phone. Sam Altman framed the stakes plainly in his 2024 essay The Intelligence Age: if the infrastructure isn’t built, AI becomes “a very limited resource that wars get fought over and that becomes mostly a tool for rich people.” He named, precisely, the failure mode he was trying to avoid.

The AI promises are less structurally radical than blockchain’s were. There is no claim that platforms will be dismantled, no protocol for redistributing ownership, no Voshmgir-style demand for data democracy. And yet the concentration dynamic is playing out identically – arguably more severely, because the infrastructure required is more capital-intensive, more energy-hungry, and more dependent on a handful of sovereign-scale actors than anything blockchain ever demanded.

The gap between announcement and reality

And the infrastructure question is being handled the same way – raised, acknowledged, and then quietly set aside as the cultural and investment conversation takes over. When blockchain re-entered our coverage last year, it was through the frame of AI and fintech disruption – the same promises of disintermediation and democratised access, refreshed under a new banner. The cycle doesn’t just repeat. It compounds.

The Stargate project was announced at a White House press conference as a $500 billion commitment to build the infrastructure that AI requires. More than a year later, the joint venture had not hired staff or broken ground, the partners were in dispute over ownership, and SoftBank scrambled to find the cash for its year-end commitment by selling positions and drawing on margin loans. The gap between the announcement and the physical reality is not a footnote. It is the story.

The settlement layer and the judgement layer

What survived the blockchain hype cycle were the applications that automated the settlement layer while leaving the judgement layer intact. The honest question about AI is: Which layer are we actually talking about when we make the claims we do?

The settlement layer, for AI as for blockchain, is for work where success is verifiable and self-contained – where output can be checked against an objective standard without requiring anyone to exercise discretion. Code that passes its test suite. Documents classified correctly. Contracts drafted from structured inputs. These are tasks where right and wrong can be established without reference to context, relationships, or values.

The judgement layer is different in kind, not just degree. Should we hire this candidate? Should this loan be approved? Is this patient’s risk profile sufficient to escalate? The issue isn’t whether AI can produce an answer – it can, fluently – it’s that the legitimacy of those decisions depends on someone being accountable for them. Accountability can’t be automated. It requires a person or institution that can be questioned, challenged, and held liable.

Smart contracts failed to displace lawyers not because they couldn’t encode rules, but because the legitimacy of legal agreements derives from an enforcement and interpretation system that exists outside the contract itself. AI faces the same structural constraint. The evidence so far suggests the same answer will emerge: the settlement layer will be transformed, genuinely and durably. The judgement layer will remain stubbornly human – not because AI lacks capability, but because the judgement layer is where someone must be responsible for being wrong.

The warning we’ve already heard

Voshmgir warned us in 2017 that there was a narrow window of time to get this right, or we would create the ultimate control machine. She was right. And we didn’t get it right. The platforms are more powerful than ever. The data monarchy is more entrenched, not less.

The same warning is being issued about AI. We should take it more seriously this time – and ask, honestly, whether the technology will deliver the distribution it promises, or whether it will concentrate value at the infrastructure layer just as every previous wave has done.

We already know how this story goes – because we covered it ourselves.

Photo 721699984 by MZ on Adobe Stock