I. Arithmetic Without a Remainder
The cryptocurrency market is built on one fundamental property its participants prefer not to say aloud: it produces nothing.
An Apple share is a fractional ownership stake in a company that designs, manufactures, and sells physical devices to hundreds of millions of people. The dividends a shareholder receives are a derivative of profit earned on a real market. A bond is a debt obligation backed by a government's tax base or a corporation's cash flow. Even an oil futures contract is ultimately tied to a barrel of crude that is burned to move turbines and trucks.
Bitcoin has no cash flow. No dividends. No revenue. No product. Neither does the vast majority of altcoins. When one market participant sells a token for more than they paid, the price difference is money paid by another participant. When that other participant sells at a profit, they take money from a third. The chain works for exactly as long as there is a next buyer.
Economists call this a zero-sum game. In 2022, researchers from Brazil and the United Kingdom published a paper in the Review of International Political Economy formally demonstrating that Bitcoin functions as a digital commodity that has value but creates no added value. Mining does not generate new purchasing power — it redistributes existing purchasing power. At the micro level someone gets rich. At the macro level it is a rearrangement of terms in which the sum does not change.
The counterargument is well known: the stock market is also speculative, and people lose money on shares. That is true. But the crucial difference is that behind a share sits a productive asset that generates profit independently of stock exchange trading. If all the world's exchanges closed tomorrow, Apple would keep selling phones. If all crypto exchanges closed tomorrow, the holder of most tokens would be left with a blockchain record backed by neither factory, nor patent, nor contract.
II. Twenty Million Experiments, Half of Them Dead
The scale of the problem is visible in the numbers. According to CoinGecko data, from mid-2021 to the end of 2025 approximately 20.2 million tokens appeared on the GeckoTerminal platform. Of these, 53.2% stopped trading and are classified as "dead." 11.6 million projects ceased to exist in 2025 alone — 86.3% of all failures over five years. In the fourth quarter of 2025, 7.7 million tokens died in three months — more than a third of all recorded project deaths in the entire history of observations. The catalyst was a liquidation cascade on October 10, 2025, when $19 billion in leveraged positions was wiped out in 24 hours — the largest single-day deleveraging event in the history of the crypto market.
The reason for this mass extinction is simple: the entry barrier for token creation has dropped to zero. Launchpads like Pump.fun on Solana allow a token to be created in minutes with no audit, no product, no team. Most of these tokens are memecoins: projects with no functionality, built entirely on the hope that someone will buy after you. In 2024, investor losses from memecoin scams exceeded $500 million according to Merkle Science data. In 2025, total losses from rug pulls reached $6 billion according to DappRadar. The number of rug pulls itself fell by 66% — each individual incident became larger and more destructive.
Half of the market's derivatives volume consists of futures and perpetual contracts — pure bets on price movement. The typical crypto market participant of 2024–2025 does not use blockchain to pay for goods, does not interact with decentralized applications, and does not store data in a distributed ledger. They trade volatility.
There is nothing criminal in this, but it is not a technological revolution either. It is a casino with an advanced interface.
III. Why Crypto Has Not Died Yet
If things are so bad, why has the market existed for eighteen years? The answer has two parts.
The first is structural. Bitcoin solved one genuine engineering problem: how to transfer value between two people without an intermediary that both must trust. Before 2009, any digital transaction required a third party — a bank, payment system, or clearing house. Satoshi Nakamoto showed that the double-spend problem could be solved cryptographically, and this achievement will remain significant regardless of the price of BTC.
The second is economic. In a world where interest rates in developed countries trended toward zero for decades, then shot up sharply creating trillions of dollars in unrealized bond losses, people are looking for alternative assets. Crypto is one of them. It is volatile, risky, and largely speculative, but it is liquid, global, and accessible to anyone with a smartphone. For a resident of Nigeria, Argentina, or Turkey whose national currency loses purchasing power faster than they can earn, a USDT stablecoin on the Tron blockchain is a real instrument for preserving value. Not perfect, not safe, but functional.
These two factors — technical innovation and economic demand under conditions of currency instability — have allowed the crypto market to survive several cycles of total collapse and recover each time. But neither of them resolves the fundamental problem: the majority of activity on the market still represents a redistribution of money between participants, not the creation of new value.
IV. Machines Need Money
This is where a different story begins.
In 2025–2026, something happened that the crypto industry had been waiting fifteen years for without realizing it: a class of economic agents appeared for whom traditional financial infrastructure is physically inaccessible.
An AI agent — an autonomous program capable of planning, making decisions, and executing multi-step tasks — cannot open a bank account. It has no passport, social security number, or physical address. It will not pass KYC verification. It cannot sign a paper contract. The entire system of global finance is built on the assumption that the participant in a transaction is a human being, and this assumption has become a technological dead end.
Crypto does not make this assumption. A blockchain wallet is a pair of keys. No citizenship is required to create one. A transaction is confirmed by a cryptographic signature, not an identity document. A stablecoin can be sent in fractions of a second, for fractions of a cent, without bank approval, without weekends, without borders.
For an AI agent that needs to pay for an API request, purchase compute time, gain access to a dataset, or settle with another agent — this is the only viable financial layer.
And the world's largest technology companies are already building infrastructure for this scenario.
In September 2025, Google introduced the Agent Payments Protocol (AP2), backed by more than 60 organizations. Visa released the Trusted Agent Protocol — a cryptographic standard for identifying and authorizing AI agents in payment systems. PayPal and OpenAI announced a partnership for agent payments through ChatGPT. Coinbase and Cloudflare developed the x402 protocol, reviving the unused HTTP 402 "Payment Required" status code for stablecoin micropayments directly through web requests. BNB Chain implemented the ERC-8004 standard for verifiable on-chain identifiers for AI agents. Mastercard acquired stablecoin infrastructure company BVNK for $1.8 billion — the largest deal in the history of the stablecoin industry.
In April 2026, Ant Digital Technologies (the blockchain division of Ant Group) introduced the Anvita platform, consisting of two products: Anvita TaaS — tokenization of real-world assets for institutional clients — and Anvita Flow — a platform where AI agents register, find one another, coordinate tasks, and settle in real time using stablecoins. According to the Solana Foundation, the Solana network has already processed more than 15 million on-chain transactions involving agents.
MarketsandMarkets estimates the AI agent market at $7.84 billion in 2025, with projected growth to $52.62 billion by 2030 at a compound annual growth rate of 46.3%. If autonomous agents genuinely become a mass class of economic participants, the volume of transactions passing through crypto rails could exceed everything the market has seen to date. On select L2 networks in early 2026, spikes in agent transactions were already being recorded with growth exceeding 10,000%.
V. Stablecoins as the Settlement Layer of the New Economy
Stablecoins — crypto assets pegged to the dollar — turned out to be the first segment of the crypto market to find application beyond speculation.
According to a16z estimates, stablecoin transaction volume for 2025 was approximately $46 trillion. For comparison: this is more than 20 times larger than PayPal and approximately 3 times larger than Visa. Stablecoin volume is approaching ACH — the electronic payment network through which the bulk of interbank settlements in the United States passes. Stablecoin market capitalization exceeded $300 billion in 2025, and analysts project growth to $420 billion in 2026.
An important caveat: a significant portion of this volume is bot activity. According to CEX.IO data, in 2024 approximately 70% of stablecoin transaction volume came from bots, and on the Solana and Base networks this share reached 98%. This means the real human transaction volume is substantially lower than the headline figures. But the very fact that machines are already generating the overwhelming majority of stablecoin traffic confirms the thesis: digital settlement instruments are critically in demand by automated systems, and this demand is organic.
For an AI agent, a stablecoin solves a fundamental problem. The volatility of Bitcoin or Ether makes them unsuitable for autonomous financial decisions: an agent executing hundreds of micropayments per hour cannot risk a 5% fluctuation in the base asset's exchange rate. The peg to the dollar eliminates price risk while preserving the speed, programmability, and minimal fees of blockchain settlement.
VI. Tokenization: When Crypto Becomes the Ledger of Reality
The second front on which crypto is acquiring a function beyond speculation for the first time is the tokenization of real-world assets (RWA).
The principle is simple: any asset — a bond, a share, a fund stake, a loan obligation, a piece of real estate — can be represented as a token on a blockchain. The token carries ownership rights, dividend streams, and redemption terms encoded in a smart contract. It trades 24/7, can be divided into shares of any size, and settles in seconds instead of days.
The market for tokenized real-world assets (excluding stablecoins) grew from $5 billion in 2022 to $24 billion by mid-2025 — a 380% increase in three years. BlackRock launched the tokenized BUIDL fund, which became the largest in the category with a market share of approximately 45%. Franklin Templeton, Ondo Finance, Hamilton Lane, and KKR have issued tokenized shares in their investment instruments. Swift — the global interbank network serving 11,000 financial institutions — completed the design of a shared ledger for cross-border settlements using tokenized deposits, with the participation of JP Morgan, HSBC, Deutsche Bank, and Bank of America.
Projections vary: McKinsey expects $2 trillion in tokenized assets by 2030, BCG expects $16 trillion, and some analysts speak of $50 trillion. The actual figure will be determined by the speed of regulatory adaptation. But the direction is unambiguous: the world's largest financial institutions have already moved tokenization from the pilot stage to industrial scaling.
This is the first case where crypto infrastructure serves an economy that produces value rather than merely redistributing it.
VII. The Agent Economy: When Software Starts Earning
The convergence of two trends — autonomous AI agents and crypto settlement — creates the contours of an economic system with no precedent in history.
Imagine: an AI agent receives a task — plan and book a business trip for an employee. To do this it needs to query flight data through an API (a micropayment in stablecoin), pay for access to a hotel aggregator (another micropayment), compare options using another specialized agent (an agent-to-agent settlement), book and confirm payment — all without a single human involved in the payment cycle. Every step is a transaction, and every transaction is an economic event generating demand for a settlement instrument.
Now multiply this by millions of agents working simultaneously. Agents trading compute capacity on decentralized marketplaces. Agents buying and selling data. Agents optimizing supply chains and settling with one another in real time. Agents managing portfolios of tokenized assets, rebalancing positions every second.
Each of these transactions creates real demand for settlement infrastructure. And for the first time in the history of the crypto market, this demand has a utilitarian nature — it arises from economic activity, not from expectations of price appreciation.
VIII. Honest Limitations
It would be intellectually dishonest to end on an optimistic note without identifying the unknowns.
We do not know whether AI agents will become a mass economic class or remain a niche tool for automating routine tasks. The MarketsandMarkets projections ($52.62 billion by 2030) are extrapolation, not fact. The AI market is subject to the same hype cycles as crypto, and winter may arrive before the agent economy reaches critical mass.
We do not know whether these agents will choose crypto rails. Visa, PayPal, and Mastercard are building their own protocols for agent payments that may operate on top of traditional card and banking networks. Crypto is one possible settlement layer, but its victory is far from guaranteed.
We do not know how regulators will respond to a world in which autonomous programs handle money without human oversight. The question of legal accountability — who is responsible if an agent executes a fraudulent transaction? — has not been resolved in any jurisdiction.
And finally, we do not know whether the agent economy will turn out to be yet another narrative the crypto industry uses to attract capital. An industry with a history of ICOs, DeFi Summer, the NFT boom, and memecoin mania has conditioned the market with a persistent reflex: each new cycle brings a new promise of real utility, and each time a significant portion of that promise goes unfulfilled.
IX. Where the Logic Leads
Nevertheless, the structural shift of 2025–2026 differs from previous cycles in one fundamental property: for the first time, the infrastructure is being built not by crypto-garage startups but by Google, Visa, Mastercard, JP Morgan, and Ant Group. For the first time, demand for blockchain settlement is being driven by technological necessity rather than speculative appetite. For the first time, the crypto market has a chance to become not an arena of redistribution but a layer on which machines create, exchange, and account for value.
For fifteen years cryptocurrencies existed as a financial instrument stripped of function. Technically sophisticated, ideologically ambitious, economically empty. Every dollar of profit in this system was a dollar of someone else's loss. Every triumphant portfolio screenshot had a mirror image — someone's liquidated deposit.
If the agent economy, real-world asset tokenization, and stablecoin settlement genuinely scale, crypto will receive for the first time what it has critically lacked: an economic foundation that does not depend on whether someone will buy at a higher price tomorrow. The blockchain will cease to be a venue where some people take money from others and become infrastructure through which machines serve people.
This has not happened yet. But for the first time in the entire history of the market, there are grounds to believe it might.