In 2025, global corporate investment in artificial intelligence reached $581.7 billion, according to Stanford’s 2026 AI Index. That’s more than double the previous year’s total.
A few years earlier, the industry’s most ambitious predictions were centered on something quite different.
Web3 promised to reorganize the internet around decentralized ownership. The blockchain would allow people to hold digital assets independently of platforms, participate in tokenized economies, and capture more of the value they generated online.
Today, the dominant promise is almost the opposite.
Artificial intelligence is making digital production dramatically cheaper. Software, images, articles, music, and videos can be generated at a fraction of their historical cost. Meanwhile, developing frontier AI systems requires enormous concentrations of capital, computing infrastructure, and specialized expertise.
Within a few years, we went from engineering digital scarcity to celebrating digital abundance; from distributing ownership to concentrating computational power; and from trying to eliminate intermediaries to building increasingly capable intermediaries between people and information.
Of course, crypto hasn’t disappeared, and AI isn’t exclusively centralized. But the narrative contrast raises a question that goes well beyond either technology.
How do you build a durable company when the industry’s prevailing vision of progress can change faster than your business can adapt?
The future of ownership and the digital commons
Web3 emerged partly as a reaction against the economic structure of the modern internet.
Web2 had produced extraordinary services, but much of the value generated by users accumulated inside a relatively small number of platforms:
- Social networks controlled distribution.
- Marketplaces controlled access to customers.
- App stores determined which software could reach users and under what conditions.
In its early days, crypto was exclusively focused on the development of untraceable digital currencies, free from the banking system’s manipulation and the State’s restrictions. But, as the space evolved, the idea of Web3 started gaining traction. Web3 was powered by the Ethereum blockchain, a programmable blockchain that allowed people to develop their own decentralized experiences. Let’s leave it at that, lest we get super technical.
Web3 offered an exciting alternative to the hyper-centralized social-media driven Web2: an internet where ownership could be established and transferred without relying entirely on centralized institutions. Scarcity was essential to that vision.
A trip down memory lane: Remember NFTs?
One of Web3’s core promises was the traceable ownership and monetization of digital assets. Digital files are ordinarily easy to reproduce, but the blockchain could provide us with a mechanism to turn digital files into assets that could be authenticated, bought and sold.
It’s worth clarifying a potentially confusing fact about blockchain-backed digital assets – it was confusing back then (in 2022), so it may as well be confusing now: The blockchain never hosted the digital asset. It didn’t store a .jiff containing an artwork, or a digital house on a metaverse. It just held the asset’s ownership track record.
NFTs (non-fungible tokens) could represent digital assets – but also access rights, governance privileges, or financial claims, depending on their design.
Art NFTs were particularly popular in the heyday of Web3, mostly due to their speculative potential. But this space had far more to offer than short-term frenzy. If new modes of digital ownership were possible, entirely new markets and organizational structures could emerge.
Often, during the NFT craze, speculative potential was prioritized over artistic interest. But, regardless of whether investors actually liked the art they were buying, one of the ecosystem’s narrative fortes was the idea of paying digital artists for their work. This work’s value was boosted by artificial scarcity and community building mechanics.
Generative AI approaches digital economics from another direction. In just a couple of years, we went from artificially scarce, human-made digital assets to machines that can provide us with endless volumes of content, at a couple of cents per piece. If we’re agnostic to quailty, we can affirm that the cost of content production has plummeted to $0.
Cheap, fast – and good?
Stanford’s 2025 AI Index found that the cost of querying a model achieving approximately GPT-3.5-level performance on the MMLU benchmark fell from $20 per million tokens in November 2022 to $0.07 by October 2024. That’s a reduction of more than 280 times in roughly two years.
The comparison measures the cost of achieving a particular benchmark performance, not the price of running an identical model. Still, the direction is unmistakable: capabilities that once required substantial expenditure are becoming inexpensive components of other products.
But the fact that content production’s getting cheaper doesn’t mean that everything’s worthless. Otherwise, I wouldn’t be manually editing this piece.
A photograph documenting a historical event isn’t interchangeable with a synthetic image. Proprietary research isn’t equivalent to a generated summary. A reliable enterprise application isn’t necessarily interchangeable with software assembled in an afternoon. What AI commoditizes is often a particular capability or production process rather than the final product.
As those capabilities become abundant, scarcity moves elsewhere: toward trust, distribution, proprietary information, customer relationships, and the ability to coordinate complex systems. AI progress doesn’t eliminate scarcity, it moves scarcity up market.
So, are we right to diagnose the shift from crypto to AI as a shift in the perceived value of digital content?
Did the technology industry change its mind?
It’s tempting to interpret the transition from crypto to AI as an ideological reversal.
But that assumes the technology industry operates as a coherent intellectual movement, collectively committed to a particular vision of progress. It doesn’t.
It’s a collection of investors, founders, researchers, institutions, and customers with overlapping but frequently conflicting interests.
What looks like a change in collective beliefs may actually be a change in where these actors see the greatest opportunity.
Capital flows = narrative flows
Venture capital plays a crucial role in all of this.
We know that nvestment finances technological development – but what does “technological development” entail? Capital flows determine which companies can recruit talent, which problems receive attention, and which predictions about the future acquire credibility.
Once a technological narrative attracts sufficient capital, an entire commercial ecosystem begins forming around it. Startups adopt its vocabulary, investors develop theses around it, conferences organize panels about its implications, and consultants explain how established companies should respond.
Eventually, companies begin positioning themselves around the narrative, sometimes because the technology genuinely improves their business and sometimes because the association itself has become commercially valuable.
The enormous scale of AI investment illustrates this process, although it doesn’t prove that capital moved directly from crypto into AI. Stanford’s $581.7 billion figure includes corporate investment categories such as private financing and mergers and acquisitions, not just venture capital.
There’s also an important counterpoint: the industry’s attention may have shifted, but crypto development hasn’t stopped.
Is crypto dead?
Electric Capital’s live developer data counted approximately 26,700 monthly active crypto developers in October 2026, including nearly 10,900 established developers with at least two years of experience. That established cohort grew 30% over the preceding year. Speculative enthusiasm and productive activity aren’t necessarily the same thing. A technology can lose its privileged position in the industry’s imagination while developers continue building around it.
On a personal sidenote: Even during the crypto spring, when the ecosystem was in love with the idea of “mass adoption”, my crypto futurology was more aligned with B2B applications – crypto’s greatest adopters would be institutional.
The numbers are beginning to support that intuition. Research from Artemis put annualized B2B stablecoin payments at $36 billion in early 2025, growing 315% year over year. By September 2026, Visa was reporting more than $20 billion in annualized stablecoin settlement volume, over 15 times its year-earlier level. Meanwhile, the market for tokenized US Treasuries had reached nearly $13 billion by March 2026.
The interesting part is that none of these applications requires consumers to care about the blockchain. Banks need faster settlement, companies need to move money across borders, and financial institutions need more efficient ways to manage assets. Gone are the days of mass adoption, crypto might find its most durable applications in infrastructure that ordinary people never notice.
Crypto isn’t dead, it just grew up and got a real job.
Infrastructure and the decentralization paradox
There’s another problem with treating crypto and AI as ideological opposites: both movements promise to redistribute power, but end up creating new power centers.
Crypto sought to reduce dependence on centralized institutions, yet much of its commercial ecosystem came to rely on exchanges, stablecoin issuers, infrastructure providers, and other intermediaries. The protocols might be decentralized while the services most people actually use remain concentrated. AI presents a different version of the same tension.
Open-weight models and accessible development tools distribute capabilities to millions of people. At the same time, developing frontier systems can require enormous amounts of computing infrastructure, specialized hardware, electricity, and capital.
According to Synergy Research Group, Amazon, Microsoft, and Google accounted for 63% of global enterprise spending on cloud infrastructure services in the third quarter of 2025. That statistic measures the broader cloud market, not AI infrastructure specifically. Nevertheless, it illustrates how digital capabilities can become widely accessible while the infrastructure supporting them remains concentrated.
Nvidia provides another supporting perspective: the company reported $215.9 billion in revenue for fiscal 2026, an increase of 65% year over year. While developers gain access to increasingly inexpensive AI capabilities, the companies supplying critical infrastructure can capture enormous economic value.
A decentralized protocol doesn’t automatically produce a decentralized market. An accessible technology doesn’t guarantee a competitive industry.
The relevant questions are structural: Who controls distribution? Who owns the infrastructure? Where do switching costs accumulate? Who captures the margins?
Those questions remain useful regardless of which technology dominates the conversation.
What survives when the narrative changes?
Consider two hypothetical software companies.
The first is built around the premise that a particular technology will become the dominant organizing principle of the internet. Its positioning, product architecture, fundraising strategy, and customer acquisition depend on that prediction becoming reality.
The second uses the same technology to solve a persistent customer problem. Its customers care about the outcome, not the technological philosophy behind it.
Both might grow rapidly while the narrative is popular, but when expectations change, their positions become very different. The first company may discover that demand for its product was inseparable from enthusiasm for the underlying technology. The second has a better chance of adapting its implementation while preserving the customer relationship.
This suggests a useful distinction between narrative-dependent and narrative-resilient businesses. Narrative-dependent companies require a particular story about the future to remain persuasive. Narrative-resilient companies can benefit from technological change without requiring the industry’s prevailing ideology to remain intact.
Neither category guarantees success. A company solving a persistent problem can still be displaced by a competitor with better technology or distribution. And some of the most consequential companies are built precisely because their founders commit to a future that doesn’t yet exist.
You should take risks and embrace what could be emergent categories. Innovation is risky, but it’s important not to confuse enthusiasm for a technology with durable customer demand.
The companies that accumulate capabilities
Nvidia offers an interesting example. Its position in AI didn’t emerge from suddenly deciding to become an AI company when generative models attracted mainstream attention. It developed over years of investment in accelerated computing, software tooling, developer relationships, and specialized hardware. AI dramatically increased the strategic importance of capabilities the company had already accumulated.
That doesn’t make Nvidia invulnerable. Competition, changing computing architectures, customer concentration, and technological substitution remain meaningful risks. But its trajectory illustrates the difference between attaching a company to a fashionable narrative and developing capabilities that become useful across technological transitions.
Mercado Libre offers a different example. Its business spans commerce, payments, logistics, and financial services. Those activities reinforce one another through customer relationships, distribution, infrastructure, and operational knowledge. New technologies can improve those systems without requiring the company to redefine its entire purpose whenever the industry’s attention shifts.
These examples don’t establish a universal formula for durability. Both companies face their own competitive risks, and successful businesses are easy to rationalize in hindsight.
They do, however, suggest a useful hypothesis: companies that accumulate capabilities serving persistent needs may be better positioned to benefit from successive technology waves.
Distribution matters. Customer relationships matter. Institutional knowledge matters. So do proprietary data, operational excellence, network effects, and the ability to adopt new technologies faster than competitors.
None of these advantages lasts forever. But they’re generally harder to reproduce than a fashionable product description.
Building for a future you can’t predict
The conventional advice would be to ignore the hype and focus on fundamentals. That’s comforting, but insufficient.
Technological shifts can genuinely destroy established business models. Companies that dismiss emerging technologies because they appear fashionable may discover that their supposedly durable advantages depended on capabilities that have become commodities. On the other hand, quite frequently, a shift in beliefs is enough to hurt a business model. The idea that AI can replace a core capability can be extremely destructive, even if it’s not true.
Caution can go a long way:
- For founders, that means asking whether customers are paying for a persistent outcome or for participation in a technological movement.
- For established companies, it means identifying which capabilities should remain proprietary and which can become inexpensive external services.
- For investors, it means separating adoption driven by measurable customer value from adoption driven primarily by speculative expectations.
Consider a software company whose product depends heavily on a capability that becomes dramatically cheaper through AI. That company might lose its pricing power, or it might use the same technological development to improve its margins and expand its market. The outcome depends on where customers perceive value and how effectively the company can adapt. The same technological shift can destroy one business model while enabling another.
Three questions worth asking
Before betting big on the next tech bubble, leadership teams should be able to answer three questions:
- What would remain valuable if our core technology became widely accessible tomorrow? – This identifies the capabilities, relationships, and assets that aren’t easily replicated when production costs collapse.
- Are we accumulating advantages that become stronger as the technology evolves? – A company might be building proprietary data, customer trust, distribution, operational expertise, or network effects. Alternatively, it might simply be accumulating exposure to a fashionable category.
- What would we do differently if today’s dominant narrative lost its privileged position within five years? – The answer reveals how much of the company’s strategy depends on expectations that haven’t yet materialized.
What happens after AI?
The next technological narrative may contradict important assumptions underlying today’s AI boom.
Perhaps computing becomes substantially more decentralized. Perhaps regulation changes the economics of foundation models. Perhaps new interfaces alter how people interact with software. Or perhaps AI becomes so deeply embedded in ordinary products that the category itself loses much of its commercial meaning. These are potential scenarios, not predictions.
What seems more certain is that the industry will eventually find another compelling story about the future. There will be new terminology, new investment theses, new categories, and new companies promising to reorganize the economy. Some will succeed. Many won’t.
The companies that endure won’t necessarily be those that predicted the next narrative correctly. They’ll be those that understood which customer needs, relationships, capabilities, and economic advantages were worth preserving as the narrative changed.
Crypto promised a world where digital assets could become scarce. AI promises a world where anyone can produce complex digital content for cents. Neither changes the fundamental challenge of building a business: creating something valuable, reaching people who need it, and capturing enough of that value to keep doing it.
The story of the future will change again. The question is whether your company needs that story to remain true.
By Aaron Marco Arias
Frequently asked questions
Does AI make digital products worthless?
No. AI can dramatically reduce the cost of producing certain digital outputs, but production cost and economic value aren’t equivalent. Distribution, reliability, trust, proprietary information, and customer relationships can remain scarce even when generating content or software becomes inexpensive.
Is crypto becoming irrelevant because of AI?
The evidence doesn’t support that conclusion. Electric Capital’s live developer data exposes the fact that established crypto developers continue operating, even as the total developer population declined. The two technologies also address different problems and can coexist.
What makes a technology company resilient to hype cycles?
There is no universal formula. A useful starting point is to identify which customer needs persist independently of a particular technology and which competitive advantages the company can continue developing as production costs, infrastructure, and market expectations change.
Should startups avoid building around emerging technologies?
Not necessarily. Emerging technologies can create entirely new markets. The important distinction is whether a company understands the assumptions behind its technological bet and has a credible path toward durable customer value.
Methodological note: Investment figures refer to different financing categories and should not be interpreted as evidence of a direct transfer of capital between crypto and AI. Inference-cost comparisons refer to equivalent benchmark performance, not identical models. The narrative-dependency framework is an editorial hypothesis, not an empirically validated predictor of company performance.