Crypto Keeps Reinventing Itself. But Who Is It For?

On October 12, the Sui Foundation announced that it was investing $30 million in a project that, until recently, had been attempting to reinvent how the world owns intellectual property. The announcement included an unusual detail. The project, once valued at more than $2 billion, would be retiring the blockchain it had spent years building. [149] [150]

The project was Story Protocol, founded around an ambitious idea: that intellectual property could become programmable. Music, books, films, characters, artwork and other creative assets could be registered on a blockchain, where their ownership and licensing conditions would be available for software to interpret and execute. An artist might permit their work to be remixed under predefined terms, with royalties distributed automatically. A film studio could license characters through programmable agreements. The cumbersome business of establishing rights, obtaining permissions and distributing payments might be transformed into something resembling the open infrastructure of the internet.

Investors found the proposition compelling. In August 2024, Story raised $80 million in a round led by Andreessen Horowitz, bringing the total raised by its developer, PIP Labs, to approximately $143 million. The funding valued the company at $2.25 billion. Its founder, S.Y. Lee, had previously helped build the serialized-fiction business Radish, which was acquired by Kakao for approximately $440 million. He understood firsthand how complicated the economics of intellectual property could become when creators, platforms, distributors and rights holders all needed to be compensated. [151]

Story's proposition was particularly timely. Generative AI was creating new questions about who owned creative works, whether they could be used for training, and what compensation might be owed when models produced derivative material. A common technical system for registering rights and expressing licensing conditions appeared to offer an answer to a problem that was becoming larger and more difficult to manage.

But by June 2026, Story had discovered something awkward about the market it wanted to transform.

The owners of the most valuable intellectual property were not particularly interested in making it permissionlessly available.

Story had spent 2025 experimenting with applications in entertainment, gaming, music and creative software. Some had gained traction. The foundation highlighted Magma, a collaborative art platform with three million registered artists, and Aria, which was developing tokenized music rights involving well-known performers and catalogues. Yet these successes had not produced the broader market the organization originally imagined. Major entertainment companies, music rights holders and gaming businesses remained reluctant to surrender control over how their intellectual property could be used. [81] [83]

In an unusually candid announcement, Story acknowledged that the openness of permissionless licensing conflicted with the control these companies wanted to retain. The statement is striking because it describes a difficulty that cannot be resolved simply by improving the software. Story had built a mechanism for making intellectual property easier to license, but the owners of valuable rights were not necessarily trying to make them easier for everyone to use. Selective permission, exclusivity and contractual restrictions were part of what made those rights commercially valuable.

Story changed direction. It rebranded as the DATA Foundation, renamed its IP token to DATA and shifted its attention toward recording the origins and licensing rights of information used to train artificial intelligence. The organization argued that AI developers had a growing need for specialist, rights-cleared data, and that its infrastructure could help establish where information came from, who had contributed it and whether it could be licensed. It also delayed scheduled token unlocks by six months as the project adjusted its economic arrangements around the new direction. [81] [82] [165]

Then came the October announcement. DATA Network would become a protocol operating natively on Sui, with the Walrus data platform providing storage for the datasets being registered. The independent blockchain, once central to Story's identity, would be retired. Sui Foundation's $30 million investment would establish it as the largest institutional holder in the network, while existing token holders would migrate to the new arrangement on a one-for-one basis. The precise migration timetable had not yet been announced. [149] [150]

There were sensible reasons for the decision. Maintaining a dedicated blockchain requires validators, network security, software development and the supporting infrastructure needed to keep the system operating reliably. If DATA's business is now principally concerned with licensing and tracking AI training data, running its own chain may be an expensive distraction from the commercial problem it is attempting to solve.

But the changes also invite scrutiny. Story had raised substantial capital around the proposition of programmable intellectual property, then shifted to AI training data, one of the most heavily financed areas of technology. A new strategic investor had arrived, and the project was abandoning much of the infrastructure around which its original investment story had been constructed.

Whether the new business succeeds remains an open question. The Sui announcement says DATA's Trace registry has recorded more than 250 million contributions and processes millions of receipts daily, with applications including Kled, Numo and Oto participating in the network. It also claims that AI labs and robotics companies are licensing data through the ecosystem. These are potentially meaningful developments, but the public announcement does not provide independently audited customer revenue, the economics of individual transactions or enough detail to establish how much registered data is actually being purchased. [149] [166]

Nor does registering a dataset on a blockchain independently establish that the contributor possessed the rights claimed. A ledger can preserve a record of an assertion and when it was made. Establishing whether the assertion is true still requires evidence, contractual arrangements and verification outside the ledger.

DATA may eventually become a successful business. But the latest investment is evidence that another investor believes in the opportunity, not proof that the opportunity has been commercially established. The distinction matters because it resembles a pattern that has repeated throughout cryptocurrency's history: substantial financial commitments made around a technological possibility before the market for the resulting service has become clear.

Story is hardly alone.

In January 2026, the founders of Farcaster, one of crypto's most credible attempts to reinvent social media, transferred their protocol to another company after years spent trying to build a decentralized alternative to conventional social networks. In June, Sophon announced it would stop operating its own blockchain and build applications on Coinbase's Base instead. On October 2, Blast, an Ethereum network that once attracted more than $2 billion in deposits, announced it was shutting down because the revenue generated by the chain could no longer cover its operating costs. [35] [152] [160]

These are not the familiar frauds and financial implosions of crypto's previous cycles. They involve credible founders, functioning technology, respected investors and, in some cases, genuine communities. Their difficulties are more ordinary, and in some ways more revealing. They concern the difference between creating a technically impressive system and convincing enough people to use it for a business to survive.

It is worth contrasting this with what happened on November 30, 2022.

That was the day OpenAI released ChatGPT, through an interface consisting of little more than a text box and an invitation to ask questions. Five days later, Sam Altman announced that the product had crossed one million users. By January 2023, UBS analysts, citing Similarweb data, estimated that it had approximately 100 million monthly active users, making it one of the fastest-growing consumer applications on record. [159] [6] [7]

More important than the speed was what happened afterwards. ChatGPT changed the way ordinary people understood artificial intelligence. A technology that had spent decades developing in laboratories and specialist applications became something millions of people could interact with directly. Developers began building products around it, businesses reconsidered how work could be performed, investors committed enormous amounts of capital, and governments accelerated their efforts to understand and regulate the technology.

In February 2026, Pew Research Center found that 44% of American adults had used ChatGPT, up from 18% in 2023. The survey's wording changed in 2026, which complicates precise comparison across the years, but the increase in public reach remains considerable. [148] [161]

A separate Pew survey conducted in January found that 19% of American adults had ever invested in, traded or used cryptocurrency, little changed from the 16% recorded in 2021. The questions measure different activities, and cryptocurrency has substantial financial and infrastructure usage that these surveys do not capture. Nevertheless, the contrast between the trajectories is striking. [2] [4]

AI found a defining product that transformed public expectations of what the technology could do. Crypto has spent more than fifteen years creating financial markets, programmable networks and increasingly sophisticated infrastructure, yet has not produced a comparable transformation in how ordinary people experience it.

Why has one field had its breakthrough while the other continues to search for one? And what would crypto actually have to produce to create a moment as consequential as ChatGPT's arrival?

Artificial intelligence did not begin in November 2022. By the time ChatGPT appeared, machine learning was already embedded in an enormous range of products and industries. Recommendation systems influenced what billions of people watched, search engines relied on models to rank information, banks used machine learning to detect fraud, and researchers had made substantial advances in areas including image recognition, speech processing and protein structure prediction.

Nor was ChatGPT the beginning of serious commercial investment in AI. The field had attracted substantial capital, computing resources and talent long before its public release. Transformer architectures, large-scale pretraining and successive generations of language models had developed over years of research, with systems such as GPT-3 demonstrating increasingly general capabilities.

The public simply experienced most of those advances indirectly. A search engine might become more useful without the person using it knowing which model was responsible. A recommendation system could become more accurate without requiring anybody to understand neural networks. A scientific model could transform a particular field while remaining almost invisible outside it.

ChatGPT changed that relationship by making a powerful capability directly accessible.

Someone could ask for an explanation of a difficult subject, receive help writing software, translate a passage, explore a question through conversation or attempt a task they previously would have regarded as beyond their technical ability. The system was far from flawless. Early versions produced invented information, made reasoning errors and struggled with tasks requiring reliable specialist knowledge. But the range of things ordinary people could suddenly attempt was sufficiently unusual to change expectations almost immediately.

The significance was not simply that millions of people found a chatbot entertaining. Once the public could experience the capabilities of a large language model for itself, the commercial possibilities became much easier to imagine. Developers began creating applications around language models, established software companies incorporated them into existing products, and businesses started examining which forms of knowledge work could be accelerated or partially automated.

This process is now measurable beyond consumer surveys. The U.S. Census Bureau's Business Trends and Outlook Survey found that nearly one in five firms were using AI in at least one business function by the period ending in May 2026, with much higher reported adoption among larger companies. Methodology changes in November 2025 mean the series must be interpreted carefully, but the evidence shows AI becoming part of business operations as well as consumer experimentation. [144] [145]

AI's popularity has not eliminated questions about reliability or trust. A 2026 Gallup and Special Competitive Studies Project study found that a much larger share of Americans had used AI than trusted it to make fair decisions. Pew likewise found that many chatbot users remained skeptical about the information the systems provided. Yet usage continued to expand because the technology offered capabilities that people found useful despite those reservations. [146] [147] [148]

This is what made ChatGPT a defining moment for the industry. It did not solve every technical or commercial problem facing AI, but it transformed a large body of research into a capability that ordinary people could understand through direct experience.

Crypto has produced important technical achievements of its own. Bitcoin demonstrated that a network could maintain a shared record of digital value without a central institution deciding which transactions were valid. Ethereum extended that model into a programmable environment where software could hold assets and execute financial agreements. Subsequent developments made blockchain transactions cheaper, increased throughput and expanded the range of applications that could operate on these networks.

But the usefulness of those achievements has often been most apparent to the people already interested in cryptocurrency. They have not generated the same widespread recognition among ordinary users that an entirely new capability had become available to them.

The distinction is partly about what the two technologies require from the people adopting them. A person can open ChatGPT and attempt a useful task immediately, without persuading friends, employers, merchants or financial institutions to make a coordinated change. Many blockchain applications have a much more complicated relationship with the users they hope to attract. A payment system needs people willing to accept payments. A currency needs counterparties and liquidity. A marketplace needs buyers and sellers. A social network is more valuable when the people somebody wants to interact with are already there.

These are not merely problems of interface design. They concern the difficulty of persuading groups of people to change their behaviour together, often when existing institutions and services already meet much of their needs.

Crypto has spent years trying to solve that problem. Its history contains several moments when participants believed the solution had finally arrived.

Bitcoin's arrival in 2009 introduced the possibility of a scarce digital asset that could be transferred through a network without relying on a single bank to maintain the ledger. Ethereum subsequently demonstrated that shared networks could execute software, allowing developers to build financial applications around programmable agreements. The initial coin offering boom of 2017 suggested that issuing tokens could become an alternative way to finance technology projects, while decentralized finance demonstrated that trading and lending could take place through smart contracts rather than conventional intermediaries.

Each development was associated with a new theory of how crypto would enter the mainstream. Sometimes those theories produced valuable technology. Sometimes the excitement surrounding them proved much larger than the market for the products that emerged.

The NFT boom offered one of the clearest examples.

During 2021 and early 2022, digital collectibles became a cultural phenomenon. Artists, celebrities, athletes and major brands entered the market. Images that could be copied perfectly were bought and sold for extraordinary amounts because ownership of the associated token was scarce, even if the underlying media could be reproduced.

The technical proposition was not entirely frivolous. A blockchain could maintain a public record of transfers and ownership claims, while creators could experiment with new ways of selling access, collectibles and commercial rights. But the intensity of the market was inseparable from the expectation that desirable tokens might become more valuable.

OpenSea, the dominant NFT marketplace during that period, recorded approximately $4.85 billion in monthly trading volume in January 2022. By April 2024, its weekly trading volume was below $20 million, a dramatic contraction in activity. The comparison uses different time periods, but the decline from the peak was unmistakable. [33] [34]

The collapse did not invalidate every proposed use of digital ownership. But it complicated the assumption that a large speculative market for collectibles demonstrated equally large demand for blockchain-based property rights more generally.

The metaverse boom encountered a similar problem. Projects such as Decentraland and The Sandbox proposed virtual worlds in which land and other digital assets could be owned through tokens. The price of those assets became central to the investment narrative, even as the number of people actively spending time inside the environments remained difficult to establish.

In October 2022, DappRadar reported only 38 daily active wallets interacting with Decentraland's smart contracts. The Decentraland Foundation disputed the resulting portrayal of its audience, arguing that many genuine users did not perform on-chain transactions and that its own measure showed around 8,000 daily users. The disagreement was partly about measurement, but it also revealed how poorly blockchain activity alone could describe whether people were actually using a consumer product. [28] [29]

Gaming produced an even more revealing experiment.

Axie Infinity allowed players to acquire digital creatures, breed them and compete against other players, earning tokens that could be sold for money. During the 2021 boom, the game attracted millions of participants. In countries where the potential earnings represented meaningful sums relative to local wages, the prospect of making a living through play became a powerful attraction. Its developer, Sky Mavis, raised $152 million in October 2021 in a round led by Andreessen Horowitz, at a reported valuation close to $3 billion. [22]

Yet Axie's popularity depended on more than entertainment. The game created a relationship between people earning tokens and those willing to purchase the assets and rewards being produced. New participants bought assets to join the economy, while existing players earned tokens that could be sold into the market.

That arrangement could appear extraordinarily successful while recruitment remained strong and token prices stayed sufficiently high. But if new participation slowed, the value of the rewards could decline, reducing the financial reason to continue playing. It was an economic structure in which the activity could be genuine while the durability of demand remained uncertain.

Axie suffered a devastating security breach in March 2022, when attackers compromised the Ronin network and stole assets valued at more than $600 million. The breach involved control of five of the network's nine validator keys, including one whose access permissions had not been properly revoked. But user activity had already been declining before the hack. Data reported at the time indicated that daily active users had fallen substantially from the November 2021 peak. [23] [24] [25]

STEPN applied a similar model to exercise, rewarding people for walking and running. Its GMT token traded above $4 at its April 2022 peak but had fallen by more than 99% by October 2026. Monthly activity declined sharply after the early excitement, with third-party analysis reporting a drop from more than 700,000 monthly users at the May 2022 peak to fewer than 50,000 by January 2023. Those figures are not independently audited user counts, but they describe the scale of the retreat. [26] [27]

The usefulness of exercise had not changed. What had changed was the financial reward for participating in this particular application.

It is easy to dismiss these projects with hindsight, but their difficulties expose something more consequential than a handful of poor investment decisions. Crypto had developed a mechanism for attracting participants by offering them financial exposure to the future of a network. That mechanism could help a useful network overcome the difficulty of attracting early users. It could also make speculation look remarkably similar to genuine demand.

The catastrophic collapses of 2022 demonstrated the consequences when this financial machinery was combined with leverage, fraud or unsustainable economic structures. Terra's UST stablecoin relied on a mechanism involving its sister token, LUNA, and the Anchor lending protocol attracted deposits by offering yields around 19.5%. When confidence failed, the mechanism intended to defend the peg accelerated the collapse. Do Kwon later pleaded guilty to fraud and was sentenced in December 2025 to 15 years in prison. [14] [15] [16]

FTX revealed a more conventional fraud. The U.S. Department of Justice said Sam Bankman-Fried had misappropriated more than $8 billion of customer funds, in addition to defrauding investors and lenders. He was sentenced to 25 years in prison. Celsius and Three Arrows Capital demonstrated other ways in which leverage, poor governance and interconnected financial obligations could inflict enormous losses throughout the ecosystem. [18] [19] [20] [21]

These collapses have been examined in criminal trials, bankruptcy proceedings and countless investigations. They explain why so much money was lost and why confidence in cryptocurrency deteriorated.

But they offer a limited explanation for why many technically credible projects have also struggled to reach mainstream users. Fraud can destroy a successful business, yet the absence of fraud does not establish that a useful business exists.

That distinction becomes much clearer when examining the generation of projects that followed.

When Farcaster raised $150 million in May 2024 at a reported valuation of $1 billion, it was pursuing a problem familiar to almost everyone who uses social media.

The major platforms had accumulated enormous control over the people and communities operating on them. A creator could spend years building an audience, yet the ability to reach that audience ultimately depended on a company that owned the platform. Algorithms could change, accounts could be suspended and the distribution of content could shift dramatically following decisions over which users had little control.

Farcaster's founders, former Coinbase executives Dan Romero and Varun Srinivasan, believed the relationship could be organized differently. Rather than having each application maintain a closed record of identities and social connections, Farcaster would provide an open protocol that allowed different applications to use the same underlying network. A user could move between applications without losing the relationships established elsewhere.

The idea resembled the openness that had helped earlier internet protocols succeed. Email works across different providers because the standards allow competing services to communicate. Farcaster proposed a similar arrangement for social media, separating the underlying network from the companies building applications on top of it.

Unlike many projects whose ambitions never extended beyond presentations and token launches, Farcaster produced substantial software. Its Warpcast application developed an active community, while features such as Frames allowed interactive experiences to appear inside social feeds. Developers experimented with social applications, payments and new forms of online coordination. [35] [36]

Yet Farcaster faced a problem that the quality of its technology could not solve on its own.

Social networks derive much of their usefulness from the people already using them. Someone joining a new network wants to follow friends, creators and communities. A more open architecture can become valuable when a person needs to leave one application for another, but it is less obvious as the reason to join in the first place.

Farcaster therefore needed to achieve two things simultaneously. It had to establish that a shared social protocol offered advantages over existing platforms, and it had to attract a sufficiently large audience to make applications built on that protocol worth using.

By December 2025, Romero was acknowledging the difficulty. After four and a half years pursuing a social-first strategy, the founders had concluded that it had not worked for them. Their attention shifted toward a wallet application, and in January 2026 Neynar, a company that already provided infrastructure to much of the ecosystem, announced it would acquire Farcaster's protocol, clients and associated products, including Clanker. [35] [36]

Romero subsequently said Merkle Manufactory, the company behind Farcaster, intended to return the full $180 million it had raised from investors. He emphasized that Farcaster was not shutting down, citing approximately 250,000 monthly active users in December 2025 and more than 100,000 funded wallets. Those were founder-reported numbers, and independent aggregators using different activity definitions have reported much smaller figures, making a clean comparison difficult. [37] [157] [38]

It would be misleading to call Farcaster dead. The protocol continues under Neynar, and its founders preserved enough capital to announce a full repayment to investors, an unusually orderly outcome for a heavily funded project changing direction.

But the original ambition of building a mass-market decentralized social network had not been realized. The project demonstrated that such a network could be built; it did not demonstrate that ownership of the social graph was enough to persuade an enormous audience to abandon the services it already used.

The same problem appeared in a different form among the networks intended to support blockchain applications themselves.

Ethereum's congestion and transaction costs created a genuine need for scaling technology. Layer-2 blockchains emerged to process activity more cheaply while relying on Ethereum for settlement and security. The technical approach was sound, and some networks attracted substantial activity.

But the existence of a need for cheaper transactions did not mean the market needed every new network being financed.

Running a blockchain incurs costs. Developers must maintain the software, provide transaction sequencing, support bridges and other infrastructure, attract applications and maintain enough economic activity to justify continued operation. In a market where numerous networks offer broadly similar services, attracting users can become expensive.

Blast attempted to solve that problem through yield and incentives. Before its mainnet launched in February 2024, users had deposited more than $2 billion in assets. For a new network, the figure was extraordinary, and it appeared to provide evidence of strong early demand. [39] [40]

Yet the assets deposited into a protocol are not the same as revenue earned by the protocol. They remain the property of users, who may withdraw them when the incentives become less attractive.

This distinction has always complicated the interpretation of total value locked, a metric widely used to describe activity in decentralized finance. A network can attract an enormous amount of capital without developing a business capable of sustaining its operations.

By October 2026, Blast's total value locked had fallen to approximately $32 million. Its team announced that the network would close because operating costs exceeded revenue and there was no credible route to sustainability. Independent reporting confirmed that users would have until October 26 to withdraw through the normal interface before needing to interact with bridge contracts directly. [41] [42] [160]

A network that had once attracted billions in deposits could no longer generate enough revenue to justify keeping its infrastructure operating.

Blast was not alone. Kinto, another Ethereum layer-2 project, shut down in September 2025 following a security exploit of approximately $1.6 million. Its token fell sharply following the announcement, and the episode exposed how vulnerable smaller independent chains could be when security incidents combined with already difficult economics. [43] [44]

Sophon reached a related conclusion before suffering the same fate. In June 2026, the project announced that it would close its own layer-2 network and move onto Base, Coinbase's Ethereum scaling network. Sophon had raised approximately $70 million, but its co-founder said maintaining the blockchain and its associated infrastructure cost around $3.4 million annually. The move was expected to reduce annual expenditure by approximately $3 million, allowing the project to concentrate on consumer applications. [152]

It is difficult to imagine an ordinary software company attracting much attention for deciding that running its own infrastructure was unnecessarily expensive. Businesses routinely rent computing capacity, use existing databases and build on platforms maintained by other companies.

But crypto had spent years treating the launch of independent networks as evidence of ambition and commercial opportunity. Sophon's decision suggested that even some of the people who had financed those networks were beginning to question whether owning the underlying infrastructure created enough value to justify its costs.

Story's decision to migrate to Sui belongs in the same broader pattern. These are not identical cases, and abandoning a standalone blockchain is not necessarily evidence of failure. It may be the commercially sensible decision that allows a viable application to flourish.

What makes the examples interesting is how often expensive infrastructure has been built before the enduring demand for the applications underneath it has been established.

Crypto's professional token markets provide another window into the problem. Memento Research examined 118 token generation events in 2025 and found that 84.7% were trading below their launch fully diluted valuation by December 20. The median decline was 71.1%. Among 28 launches that began at valuations of at least $1 billion, none was above its initial valuation. [9]

A fully diluted valuation multiplies the current token price by the total supply that could eventually circulate. When only a small percentage of tokens are initially available for trading, that price can imply an enormous valuation for assets that have not yet entered the market.

Over time, allocations held by founders, investors and other participants become available for sale. Unless demand increases sufficiently, that additional supply can place considerable pressure on prices. The mechanism allows a project to achieve a billion-dollar headline valuation before its services generate meaningful revenue, but it also creates a financial structure in which early investors and later token buyers can have very different economic positions.

The more speculative end of the market demonstrates how far the creation of tradable assets has become separated from building useful businesses.

CoinGecko found that 11,564,909 tokens stopped trading during 2025, accounting for 86.3% of all such failures it identified between 2021 and 2025. Its methodology counted tokens that had traded at least once and subsequently ceased trading, not millions of substantial businesses that had filed for bankruptcy. Nevertheless, the numbers are a remarkable illustration of how inexpensive it has become to create speculative financial assets. [8]

Platforms such as Pump.fun made launching a token possible in minutes, while the resulting assets could trade on markets dominated by speculation and rapid changes in attention. The platform has faced a class-action lawsuit alleging that its business facilitated a large-scale speculative casino, allegations its defenders dispute and which should not be treated as established findings. But even without accepting the allegations, the extraordinary number of short-lived tokens illustrates a financial ecosystem capable of producing tradable assets far more quickly than enduring applications. [53] [54]

There have also been cases where the financial incentive structure was accompanied by serious security and governance problems. Chainalysis estimated that more than $3.4 billion in cryptocurrency was stolen through hacks in 2025, including approximately $1.5 billion in the Bybit compromise. The same firm's work on scams and sanctions evasion describes billions more moving through fraudulent or illicit activity, although its categories and methodologies should not be added together as if they measured the same losses. [11] [12] [55]

Speculation itself is not illegitimate. Financial markets exist because people want to exchange risk, hedge positions and take views on future prices. Crypto has created substantial trading businesses around those activities.

The problem emerges when the demand for trading an asset is used as evidence that the underlying project has found a market for its services.

The two can coexist. A useful network can have an overpriced token, and an almost useless project can have a token that appreciates dramatically. But they are different economic propositions.

For much of its history, crypto has treated them as though they were interchangeable.

The newest attempt to find a compelling application is unfolding around artificial intelligence. Given the scale of investment and commercial activity in AI, it is hardly surprising that crypto entrepreneurs have begun proposing decentralized networks to supply computing power, data, storage and coordination mechanisms for AI systems.

Some of the ideas are technically serious. AI requires substantial computing resources, and certain workloads can be divided among independent machines. Inference, rendering and some data-processing tasks do not always require the tightly connected clusters used for frontier model training. A marketplace capable of organizing spare computing capacity could provide a useful service if it offered attractive economics and sufficient reliability.

The difficult part is determining where a token or blockchain materially improves the service, rather than simply providing a way to finance it.

Bittensor is one of the more ambitious experiments. Its network consists of specialized subnets in which participants compete to provide computing resources, AI outputs and other services. Validators assess the contributions, while token emissions reward participants according to the network's rules.

The concept is attractive because it proposes an alternative way of allocating resources to the development of machine intelligence. Rather than relying on one company to decide which services to finance and how to compensate contributors, a network can create incentives for participants to develop useful capabilities.

The economics, however, contain two separate sources of value: rewards distributed within the token system and revenue paid by customers outside it.

In September 2026, a revenue index published by SubConnect estimated that 24 Bittensor subnets were collectively generating between $28 million and $35 million in annualized revenue from external customers. The figures were based on varying combinations of public dashboards, company disclosures and on-chain information, with stronger confidence in some estimates than others. They were not audited financial statements, but they provided a useful attempt to measure customer activity independently of token emissions. [61] [64]

That external revenue is encouraging. It suggests some subnets have found customers willing to pay for the services being produced. But the payments remain modest when considered against the broader value of tokens being distributed across the ecosystem.

An analysis of Chutes, one of Bittensor's larger inference-related subnets, estimated that it was receiving token emissions worth tens of millions of dollars annually while generating only a small fraction of that amount from external customers. The analysis was based on public metrics and assumptions about token prices, rather than audited company accounts, so its precise figures should be treated cautiously. Nevertheless, the comparison illustrates why distinguishing subsidies from customer revenue matters. [65]

There is also a genuine technical achievement that deserves more attention than the token economics alone.

In March 2026, researchers associated with the Templar project published Covenant-72B, a language model with 72 billion parameters trained through an open network of participants communicating over the internet. Large-scale AI training normally depends on accelerators connected by exceptionally fast networks, allowing the machines involved to exchange information continuously during optimization. Distributing that work across independent machines using ordinary internet connections introduces severe problems involving latency, bandwidth and coordination.

Covenant's researchers used a communication-efficient optimization approach to demonstrate that a substantial training run could be completed through an openly participating network. The published paper reports training on approximately 1.1 trillion tokens and compares its results with centralized models trained under related compute budgets. The model was not competitive with the most capable frontier systems of 2026, but the experiment established something technically meaningful: training a large language model through distributed, permissionless participation was possible at a scale previously difficult to achieve. [154] [66] [67]

Whether that capability can become a commercial alternative to conventional training infrastructure remains unanswered. Frontier model development involves much larger compute requirements, and customers choosing between centralized and distributed services care about performance, reliability, cost and security, not simply whether the underlying architecture is decentralized.

Render and Akash offer variations on a more conventional idea: organize distributed computing resources and sell their capacity to customers. Rendering workloads can be naturally parallel, while certain inference workloads can be completed without tightly coupled accelerator clusters.

Both markets are real, but they are also competitive. Specialized GPU providers and major cloud companies already sell computing resources, often with substantial advantages in networking, reliability, support and access to hardware.

Render operates a burn-and-mint mechanism intended to connect network usage with token economics. Fees generated by jobs are associated with token burns, while emissions compensate participants providing resources. But the value of tokens burned through activity remains small relative to the market valuations implied during periods of enthusiasm. Public dashboards also measure different categories of fees and network activity, making comparisons with conventional cloud revenue difficult. [68] [69]

Akash's first-quarter 2026 report, published by Messari, recorded approximately $253,000 in quarterly lease revenue, down 45% from the previous quarter, with average GPU utilization of 33.7%. Other promotional material has cited much larger measures of compute spending, but those figures are not directly reconcilable with the lease revenue presented in the report. [71] [72]

Other decentralized infrastructure projects face variations on the same commercial challenge. Filecoin has an established distributed-storage network, but its protocol revenue is small compared with the market capitalization of its token. io.net has claimed more than $20 million in annualized on-chain revenue, but the figure is company-reported rather than audited, and independent analyses have questioned how reported capacity compares with the number of devices visibly active on its network. [73] [74] [76] [77]

The technical possibilities are therefore more interesting than the financial evidence currently allows us to conclude. Distributed computing may become commercially important in particular niches, especially when it can access underused resources or provide capabilities that customers cannot obtain conveniently elsewhere. But the presence of a large token market does not establish that the underlying service can compete economically without its subsidies.

The governance record of decentralized AI has also produced some less encouraging developments.

Ocean Protocol withdrew from the Artificial Superintelligence Alliance in October 2025, following disagreements that later became the subject of litigation. Fetch-related entities alleged that community token allocations had been improperly sold, accusations disputed by Ocean. The allegations remain allegations, but the dispute exposed how complicated the governance of supposedly aligned token communities can become when substantial financial interests are involved. [78] [79]

The AI16Z token, associated with the ElizaOS agent framework, provides another revealing example. At the height of excitement around AI agents, the token reached a reported valuation of approximately $2.4 billion. By August 2026, its successor had collapsed and the project's founder was publicly describing the token as finished, while indicating that the underlying open-source agent software would continue to be developed. [80]

It is an almost perfect illustration of the distinction running through the industry. A financial asset associated with a technological project can fail while the technology remains useful enough for its developers to continue working on it.

World, the iris-scanning identity project co-founded by Sam Altman, approaches the intersection of AI and crypto from another direction. Its aim is to establish that an online participant is a unique human being rather than a bot, using biometric verification and cryptographic credentials. The problem is becoming more important as AI-generated identities and automated activity proliferate.

But World's reliance on specialized scanning hardware has complicated adoption, while its handling of biometric information has attracted regulatory scrutiny in multiple countries. Kenya's High Court ordered the deletion of unlawfully collected biometric data in May 2025, and regulators elsewhere have taken action against aspects of the project's operations. World has continued developing new verification methods, including alternatives intended to reduce dependence on the Orb devices used in its original approach. [84] [86] [87] [88]

These are distinct businesses with distinct technical problems. They should not be dismissed simply because tokens are associated with them. But neither should the enormous demand for AI be taken as evidence that every decentralized system promising to serve it has discovered a viable market.

DATA, Story's latest incarnation, faces the same question.

AI companies need specialist information, and there are genuine opportunities for businesses capable of acquiring, licensing and supplying valuable training data. But the existence of that market does not establish that a blockchain registry is essential to serving it, or that DATA has already developed the necessary customer relationships and commercial economics.

After moving from programmable intellectual property to AI data, Story has found another large market and another investor willing to finance the opportunity.

Whether it has found a lasting business remains to be demonstrated.

There is one part of the industry where the question of demand is becoming easier to answer.

Stablecoins offer a proposition that is almost the reverse of Bitcoin's original monetary ambition. Rather than asking people to adopt a new currency whose value fluctuates with market demand, they provide digital tokens intended to maintain the value of an existing currency, usually the U.S. dollar.

The largest issuers, Tether and Circle, maintain reserves against the tokens they issue, principally in dollar-denominated assets such as short-term Treasury instruments, cash and other liquid holdings. Eligible customers can redeem through the issuer's arrangements, while tokens can be transferred between compatible wallets and services using blockchain networks.

The appeal is straightforward in markets where ordinary dollar banking is expensive, slow or difficult to access. A business paying an overseas supplier may prefer a transfer that can settle outside conventional banking hours. Someone exposed to an unstable domestic currency may want access to dollar-denominated assets. A company operating across several countries may find programmable transfers easier to integrate into its software than traditional banking arrangements.

These are not necessarily advantages for every user. In countries with reliable instant-payment systems and inexpensive card networks, a stablecoin may offer little practical improvement for ordinary domestic purchases. But where existing payment infrastructure imposes meaningful friction, the benefits can be substantial.

Stablecoins also avoid one of the most persistent problems associated with speculative cryptocurrency applications: the user does not need the token's price to rise for the service to be useful. In fact, the entire proposition depends on the token remaining worth approximately one dollar.

That difference has produced businesses with measurable revenue.

Tether reported approximately $1.5 billion in net operating profit for the second quarter of 2026, with total assets of $187.75 billion and liabilities of $183.64 billion, leaving excess reserves of approximately $4.11 billion. Its figures were included in an attestation prepared by BDO, which is not equivalent to a full public-company audit. [91] [92]

Circle, whose financial statements are filed with the SEC, reported $701 million in revenue and reserve income for the second quarter, with $48 million in net income from continuing operations. USDC circulation stood at $73.3 billion at quarter end. Circle's fiscal 2025 figures had previously shown $2.75 billion in revenue and reserve income, but also substantial distribution costs, including approximately $1.4 billion paid to Coinbase under their commercial arrangements. [93] [94] [158]

These results reveal an unusual business model. Stablecoin issuers can earn income on the financial assets backing their tokens, particularly when interest rates are sufficiently high. The user obtains access to a dollar-denominated digital asset, while the issuer earns income on the reserves held against it.

But that makes the business sensitive to Treasury yields and distribution economics. When interest rates fall, reserve income can decline unless other factors compensate. And companies controlling access to customers, such as Coinbase in Circle's case, can capture a substantial share of the economic value. [93] [94]

Stablecoin transaction figures require equal care.

Circle reported $14.8 trillion in USDC on-chain transfer volume during the second quarter of 2026. The figure is extraordinary relative to the amount of USDC actually outstanding, but it cannot be interpreted as $14.8 trillion in purchases of goods and services. Stablecoins move repeatedly between exchanges, liquidity providers, lending protocols and automated trading systems. A single dollar of stablecoin supply can support many transfers without ever leaving the financial machinery of the crypto market. [94] [95]

A study by Boston Consulting Group and Allium Labs attempted to distinguish those movements from economically meaningful transactions. Starting with more than $62 trillion in annual stablecoin transfers, the researchers estimated that approximately $4.2 trillion represented adjusted economic activity after excluding categories such as internal movements, intermediary routing and transaction patterns associated with bots.

Within that adjusted total, they estimated between $350 billion and $550 billion of observable bilateral payments for goods and services during 2025, across roughly 650 million transactions. Their methodology excluded certain off-chain payment categories, so these figures should be interpreted as conservative estimates rather than a complete account of every stablecoin payment. [153] [163]

The distinction is remarkable. Only a small fraction of the headline transfer volume corresponds to payments in the wider economy. Yet even that smaller figure represents hundreds of billions of dollars in activity for which customers are using stablecoins to perform a recognizable financial task.

The study also found that business-to-business payments represented approximately 40% of real-economy stablecoin payment value, while consumer-to-consumer transfers accounted for around a quarter. The largest opportunities were concentrated in places where conventional payment systems imposed delays, foreign-exchange costs or restrictions on access to dollar liquidity. [153]

There is an additional complication in measuring adoption. Much of the demand for stablecoins remains tied to cryptocurrency trading itself. Exchanges use them as trading pairs and collateral, while market makers and other participants rely on them to move liquidity across venues. That is real financial activity, but it does not establish that blockchain technology has transformed ordinary commerce.

Nevertheless, the real-world payment use cases have attracted increasingly important institutions.

In February 2025, Stripe completed its $1.1 billion acquisition of Bridge, a company developing stablecoin infrastructure for businesses. Stripe was already one of the largest providers of payment infrastructure for internet companies, giving it a direct interest in technologies that could reduce the cost and complexity of moving money across borders. The acquisition was followed by the development of Tempo, a payments-focused blockchain backed by Stripe and Paradigm, which raised $500 million at a reported $5 billion valuation in October 2025. [98] [99] [156]

Visa has been expanding its own stablecoin settlement programmes. In April 2026, the company announced that its pilot had reached an annualized settlement run rate of approximately $7 billion, up 50% from the previous quarter. It was also adding support for five blockchains, bringing the total supported by the programme to nine. This remains small relative to Visa's conventional payment volumes, and a pilot run rate is not the same as full-scale commercial deployment, but it is a tangible example of a major payment network using blockchain infrastructure for an operational purpose. [155] [167]

Mastercard has also announced settlement support for several dollar stablecoins, while PayPal's PYUSD demonstrates both the opportunities and limitations of established distribution. PYUSD benefited from PayPal's customer reach and periods of incentive-driven growth, but its supply declined by approximately 31% during the second quarter of 2026 as incentives tapered. Distribution alone does not guarantee continued demand. [100] [102] [103]

These developments point toward a plausible route to adoption that looks very different from many of crypto's earlier ambitions.

A merchant using Stripe might eventually receive faster settlement without knowing which blockchain processed the transfer. A business receiving international payments might interact with an ordinary financial application while stablecoins move through the infrastructure underneath. Banks and payment providers could incorporate blockchain settlement into familiar products without asking customers to become cryptocurrency enthusiasts.

The technology could become more widely used precisely because customers do not need to change very much about how they think of money.

There is an irony in this outcome.

Bitcoin emerged partly from distrust of governments and conventional financial intermediaries. Stablecoins, the industry's strongest commercial application, are primarily digital representations of government-issued money, backed by assets held within the existing financial system. Their value depends on issuers, custodians, reserve management and redemption arrangements.

They are also centralized in ways that complicate claims about financial independence. Tether has demonstrated its ability to freeze particular USDT balances by blacklisting addresses. In January 2026, it froze approximately $182 million across five Tron wallets, and subsequent incidents have shown how issuer decisions can affect the usability of tokens moving across otherwise public blockchain networks. [109] [110] [111] [112]

Regulation is bringing stablecoins closer to conventional financial oversight. The GENIUS Act was enacted in the United States in July 2025, creating a framework for payment stablecoin issuance. Its provisions are scheduled to become effective by January 18, 2027, unless final implementing regulations trigger an earlier date. The Office of the Comptroller of the Currency and Treasury have been developing rules addressing reserves, redemption and the obligations of issuers. [104] [106] [107] [108]

Stablecoins therefore present an awkward challenge to the industry's original mythology. Their strongest appeal is not necessarily that they replace the dollar or eliminate financial intermediaries. It is that they may allow dollars to move through more flexible infrastructure, increasingly incorporated into the businesses of established payment companies.

For users seeking faster transfers, easier access to dollar-denominated assets or more programmable settlement, that may be entirely sufficient.

For an industry that once promised to reinvent money, it is a more complicated kind of success.

Stablecoins are not the only evidence that crypto has produced economically useful services.

Bitcoin has established itself as a widely traded scarce digital asset, with access through regulated exchange-traded funds making it easier for institutions and retail investors to acquire exposure using conventional brokerage accounts. Its price remains highly volatile, and its long-term role as a store of value is contested, but the market for holding and trading Bitcoin is substantial and cannot simply be dismissed as a failed payments application. [113] [115] [116]

Ethereum and Solana support applications that generate recurring transaction fees, although comparing their economics requires attention to how fees, revenue and settlement costs are defined. Some measurements refer to gross transaction fees paid by users, while others describe the revenue retained by validators, operators or protocols. The resulting figures are not always directly comparable. [117] [118] [119]

Decentralized financial applications have developed recognizable businesses. Aave operates lending markets in which participants can deposit assets and borrow against collateral. Uniswap enables trading through liquidity pools, while Hyperliquid has attracted substantial activity in perpetual futures, a market that already exists in conventional financial systems. These protocols can generate meaningful fees because customers want the financial services they offer, although much of their activity remains closely tied to cryptocurrency speculation. [126] [127] [128]

Chainlink provides another important category of infrastructure. Smart contracts frequently depend on information that exists outside a blockchain, including asset prices, exchange rates and evidence that an external event has occurred. Systems providing data and communication between networks address genuine technical requirements, even if the amounts of value they claim to secure are not equivalent to revenue earned by the underlying businesses. Chainlink has highlighted relationships and experiments involving organizations such as Swift, DTCC, Euroclear and UBS Asset Management, although some institutional deployments remain exploratory and company-reported. [123] [124] [125]

Tokenized real-world assets are a related area of experimentation. Representing Treasury bills, money-market funds or other securities on a blockchain could reduce settlement friction, improve recordkeeping and allow assets to be used in new financial workflows. But putting a security onto a blockchain does not automatically make it liquid or remove the legal restrictions governing who may own it. The market remains much smaller than conventional securities markets, and estimates of its size vary depending on which asset categories and tokenized representations are included. [129] [130]

Coinbase's financial statements illustrate how these different activities coexist within an established crypto business. In 2025, its transaction revenue was approximately $4.06 billion, while subscription and services revenue reached $2.83 billion. Stablecoin revenue, primarily reflecting the company's share of income associated with USDC reserves, accounted for approximately $1.35 billion. The business was therefore earning money both from people trading cryptocurrencies and from the financial infrastructure supporting dollar-denominated tokens. [120]

Its second-quarter 2026 results also demonstrated the cyclicality of the business, with Coinbase reporting a net loss of approximately $359.5 million on quarterly revenue of $1.22 billion. Exchanges can become major businesses, but their profitability remains exposed to market activity, asset prices and changes in trading behaviour. [121] [122]

Taken together, these examples suggest that crypto has produced some durable categories of economic activity: holding and trading financial assets, moving dollar-denominated value, lending, settlement and supplying infrastructure to financial applications.

Those achievements should not be minimized. But they also fall short of the enormous transformation promised by many of the industry's more ambitious narratives.

There is a difference between establishing a substantial market for cryptocurrency-related financial services and making blockchain technology part of everyday life in a way that changes public expectations of what computers and networks can do.

That is the distinction the comparison with ChatGPT forces us to confront.

There is a temptation to explain ChatGPT's success and crypto's slower mainstream adoption through differences in interface design. ChatGPT is accessible through ordinary language, whereas blockchain applications often require wallets, private keys, network selection, transaction fees and a willingness to accept financial risks.

Those barriers are real. They impose costs on users who may have little reason to understand why the underlying technology requires them. But simplifying an interface is not sufficient to create a compelling product.

ChatGPT offered people a capability that they could recognize immediately. Someone who had never written software could obtain help programming. A person struggling with an unfamiliar subject could ask for an explanation and follow-up questions. A worker facing a long document could use the model to explore its contents.

The system did not need to be perfect to change what those people believed was possible.

Many cryptocurrency applications, by contrast, have asked people to adopt different arrangements for activities they could already perform reasonably well. A decentralized social network offers greater portability and control, but its users may still prefer the platform where their friends are. A blockchain-based intellectual-property marketplace offers programmable licensing, but valuable rights holders may prefer bespoke agreements and selective permission. A new payment network can offer faster settlement, but only if the people and institutions involved are willing and able to use it.

The technology may be superior in a particular dimension without creating an immediate reason for an individual to change behaviour.

This does not make the comparison between AI and crypto entirely fair. AI can deliver value to an individual through a single application, whereas many blockchain technologies concern coordination between multiple parties. A better payment or settlement system may require institutions to change their procedures before its advantages become widely available.

Yet the difficulty of changing an existing market cannot explain every failure. Crypto has also developed economic incentives that make it unusually easy to mistake early financial participation for evidence of enduring demand.

Artificial intelligence is hardly immune to speculative excess, of course. Companies developing models and computing infrastructure have attracted extraordinary investment, and many will undoubtedly struggle to justify their valuations. Training frontier models requires immense amounts of capital, while inference costs, cloud commitments and competition create uncertainty about eventual profitability.

The headline revenue figures themselves require care.

In October 2026, Axios reported that OpenAI's annualized revenue was closer to $50 billion than a previously circulated figure of $70 billion, with much of the difference arising from the treatment of partner revenue. Anthropic was reportedly reaching an annualized revenue run rate of approximately $65 billion by the end of July, but the two companies use different treatments of some cloud-related sales, complicating comparisons. These figures are run rates rather than audited full-year revenue, and neither establishes sustainable profitability. [138] [139] [140] [143] [165]

AI companies may ultimately prove less profitable than their investors expect. An enormous amount of capital could be misallocated, and some of today's most celebrated businesses may disappear.

But the adoption of AI products would remain a real development even if the financial returns disappointed.

ChatGPT demonstrated that people wanted to use the capability independently of whether they could invest in the company providing it. Businesses are paying for models through APIs, enterprise software and applications that perform recognizable tasks. That doesn't settle the question of price or profitability, but it provides a clear relationship between the product and the customer.

Crypto's financial structures often complicate that relationship. A person can buy a token because they believe the network will eventually become valuable, without having any intention of using the network themselves. The asset can appreciate because more people share that expectation, even while demand for the associated service remains limited.

The industry's greatest strength as a financing mechanism can therefore become a weakness when measuring adoption.

Financial markets are perfectly capable of assigning enormous values to possibilities. The existence of those valuations does not establish that the possibilities have become realities.

This distinction helps explain why the industry's repeated booms have not necessarily produced a defining consumer breakthrough.

It also makes stablecoins particularly interesting. Their usefulness does not depend on persuading people to adopt a new currency or speculate on its future price. They take something that people already want, dollar-denominated value, and provide an alternative way of moving it.

But the fact that stablecoins are becoming useful financial infrastructure does not automatically mean they will create a moment resembling ChatGPT's arrival.

They may become widespread precisely because their users never need to understand the technology underneath.

On October 7 and 8, more than 25,000 attendees were expected at TOKEN2049 Singapore, one of the industry's largest gatherings. The organizers advertised participation from more than 7,000 companies across over 160 countries, with hundreds of speakers and exhibitors. The agenda included institutional finance, tokenization, prediction markets, AI agents and the continuing evolution of blockchain infrastructure. [131] [132]

The scale was a reminder that cryptocurrency remains capable of attracting substantial capital, technical talent and commercial attention.

But the timing was almost too revealing. Blast had announced its shutdown just days before the conference. Farcaster's original founders had already handed their project to Neynar. Sophon had decided it no longer needed its own chain. And shortly afterwards, Story's successor announced that it would move onto Sui.

In the conference halls, the industry was discussing new financial infrastructure, AI and the next generation of applications. Outside them, some of the projects that had embodied earlier ambitions were quietly changing direction.

Coverage of the event suggested a more subdued mood than earlier speculative peaks, with greater emphasis on institutional participation and financial infrastructure. The conference also featured new announcements and proposals whose commercial significance remained uncertain. As ever, there was a considerable difference between a roadmap and a product people were already using. [132] [134] [135]

It would be easy to interpret all this as the familiar story of an industry that promised a revolution and delivered something smaller. But technological progress rarely follows the expectations established during the earliest period of investment. Many important technologies have emerged from years of unsuccessful commercial experiments, with the successful applications arriving somewhere quite different from where their original advocates expected.

The internet itself offers plenty of examples of technology whose economic significance became much clearer only after large amounts of capital had been lost pursuing less useful ideas.

Yet crypto has one unusual complication. Its ability to create tradable financial assets has allowed expectations about future adoption to become an economic product in their own right. A network can attract investors, liquidity and attention long before anyone has established why ordinary people should want the services it provides.

That does not mean the technology cannot succeed. It means that the usual evidence of financial enthusiasm tells us relatively little about when a genuine breakthrough might arrive.

The next important crypto application might concern payments, where faster settlement and easier access to dollar-denominated assets could change how businesses and individuals move money. It might emerge from AI agents, which could eventually require systems for making small, frequent and programmable transactions between software services. Or it might come from some new form of digital coordination that has not yet been made useful or accessible enough for ordinary people to adopt.

None of these possibilities is established merely because the market is large or the technology sounds plausible. A blockchain-based payment system would still need to outperform conventional alternatives. An agent-payment network would need to solve authorization, security and dispute problems rather than simply demonstrating that software can transfer tokens. A decentralized data market would need to show that customers genuinely value its guarantees enough to pay for them.

There is also the question of what the breakthrough would look like when it happened.

ChatGPT was a visible application. People could experience its capabilities directly, share the experience with others and imagine new uses almost immediately. Its impact on public expectations was part of what made the event so transformative.

A successful financial infrastructure technology may spread very differently. Stablecoin settlement could eventually become part of products operated by Stripe, Visa, banks and other institutions without consumers recognizing that a blockchain was involved. The technology could become economically important while remaining largely invisible to the people benefiting from it.

That would be meaningful adoption, but it would not necessarily produce the cultural moment the cryptocurrency industry has spent years anticipating.

It is also possible that a genuinely new consumer application will emerge, one that uses blockchain technology to provide a capability sufficiently compelling that people are willing to change their behaviour in order to obtain it. If that happens, its success will need to be explained by the experience offered to users, not merely by the ownership structure of the network or the financial rewards available for participating.

This is where the comparison with ChatGPT becomes most useful.

The defining achievement of ChatGPT was not that OpenAI raised enormous amounts of money, developed sophisticated infrastructure or persuaded investors that AI would become important. Much of that work had been taking place for years. Its achievement was making the accumulated capability immediately understandable through a product that ordinary people wanted to use.

For crypto, the equivalent would require more than another wave of token appreciation, a successful fundraising round or a new blockchain attracting billions in deposits. Those are developments the industry has already experienced repeatedly.

The breakthrough would have to change what people believe the technology makes possible.

Perhaps stablecoins are an early version of that transition, even if the change is happening primarily inside the financial system. Perhaps decentralized infrastructure will enable applications that become valuable for reasons unrelated to the prospect of token appreciation. Perhaps the necessary product has yet to be invented, and the industry's existing focus on financial incentives is making it harder to discover.

After more than fifteen years, crypto has demonstrated that it can create financial assets, attract enormous investment, sustain global trading markets and support increasingly sophisticated technical systems. It has also demonstrated how difficult it can be to distinguish people who want those systems from people who want to speculate on their future.

Artificial intelligence faced its own long period of research, investment and unmet expectations before ChatGPT made its potential tangible to an enormous audience. The moment did not resolve the industry's economic uncertainties, but it fundamentally changed the public's relationship with the technology.

Crypto is still looking for an equivalent transformation.

The uncomfortable question is whether its next breakthrough will come from building a better version of the financial systems people already use, from creating something that genuinely requires decentralized networks, or from an application whose appeal has little to do with the ideas the industry has spent fifteen years promoting.

And perhaps the most interesting question is what would have to become possible for someone with no interest in cryptocurrency, no intention of buying a token and no particular enthusiasm for decentralization to encounter a crypto application and immediately understand why they wanted it.

When ChatGPT appeared, millions of people discovered a use for artificial intelligence before they had any reason to care about the industry behind it.

What would have to happen for crypto to do the same?