The Central Macroeconomic Thesis
Our analysis starts with a proposition that is more consequential than any single Bitcoin price target: the economic transmission mechanism itself may be changing. The traditional macro framework assumes that monetary conditions influence financial assets through recognizable channels. Central-bank policy affects borrowing costs, borrowing affects investment and consumption, those flows affect employment and nominal income, and the resulting earnings determine asset valuations. That framework remains relevant, but the acceleration of artificial intelligence introduces a competing transmission mechanism in which software, compute, autonomous agents, and tokenized financial infrastructure can generate economic activity with substantially less dependence on human labor.
This distinction matters because an economy can simultaneously experience tighter financial conditions and accelerating productivity. The Federal Reserve's 2026 Jackson Hole remarks explicitly identify AI as a potentially new factor of production and raise the question of whether AI-driven productivity can become large and persistent enough to alter monetary-policy analysis. The same remarks point to extraordinary growth in token usage and the enormous capital flows into AI infrastructure. r1
I therefore would not frame the underlying thesis as simply “Bitcoin rises because AI rises.” That is too narrow. The more important hypothesis is that autonomous software could create a new class of economic participants. If agents can search, negotiate, purchase, sell, allocate capital, invoke software, manage workflows, and eventually transact continuously, the number of economically active decision-makers can increase without a proportional increase in human headcount. That has implications for productivity, labor income, corporate margins, payments, liquidity, financial intermediation, and ultimately asset valuation.
That is a much larger macroeconomic proposition than a conventional crypto-cycle thesis. It also explains why the most interesting evidence is not necessarily a chart comparing Bitcoin with M2. The crucial question is whether technological adoption is beginning to alter the relationship among capital, labor, transactions, and productivity.
The First Regime Shift: From Labor Productivity to Machine-Augmented Production
Traditional productivity statistics were built around an economy in which humans supplied most of the economically meaningful cognitive and physical labor. If a company increased output while retaining the same number of employees, measured labor productivity improved. If software eliminated a task, the statistics could capture some of the resulting output or cost reduction. But autonomous AI agents create a more difficult measurement problem because the “worker” may no longer appear in employment statistics at all.
This is not merely theoretical. An IMF working paper published in July 2026 estimated that time currently saved through observed AI usage has a labor-cost equivalent of approximately $2.7 trillion annually, or about 3.4 percent of global GDP. The researchers also found substantial variation in AI adoption and concentration across countries, implying that the benefits are already measurable but unevenly distributed. r2
The investment implication is important. If AI substitutes for portions of knowledge work, the immediate macroeconomic effect does not have to be a collapse in aggregate demand. A company can simultaneously spend less on labor, produce more, widen margins, and lower prices. Consumers can then receive more purchasing power from lower costs even if employment growth is weaker. The resulting economy can therefore become more productive without following the historical pattern in which higher output mechanically requires proportionally more workers.
This creates a statistical tension. GDP can remain useful, but employment, wage growth, hours worked, and even conventional capital-to-labor ratios may become less informative about the productive capacity of the economy. The Federal Reserve has explicitly identified this uncertainty, asking whether AI's token-intensive architecture will complement or compete with labor and where the resulting surplus will accrue among technology firms, semiconductor producers, energy companies, cloud providers, businesses, consumers, and workers. r3
Our macro framework therefore has to separate three concepts that were historically intertwined: economic output, human employment, and monetary transactions. AI makes it possible for those variables to diverge.
Why the Liquidity Narrative Requires Reassessment
The conventional crypto interpretation of liquidity is straightforward: more money and easier financial conditions tend to support risk assets, while tighter monetary conditions tend to pressure them. Bitcoin has often been incorporated into this framework because its price has exhibited periods of strong sensitivity to global liquidity, real yields, dollar conditions, and investor risk appetite.
But I would distinguish liquidity from economic velocity. M2 measures a stock of monetary assets; it does not tell us precisely how frequently those assets circulate through productive transactions. A financial system can contain enormous quantities of money while still suffering from low transactional velocity. Conversely, technological infrastructure can increase the velocity and usefulness of existing capital without requiring an equivalent expansion of the monetary base.
This is where tokenization becomes economically significant. The IMF describes tokenization as a structural change in financial architecture rather than merely a faster version of existing finance. Programmability, shared ledgers, and atomic settlement can compress execution, clearing, settlement, and collateral processes that historically occur sequentially. r4
That distinction gives us a different conception of liquidity. Instead of asking only how much money exists, we should increasingly ask how much economic collateral can be mobilized, how rapidly it can move, what transactions it can support, and how much idle capital can become programmable collateral.
The phrase “dormant liquidity” is useful as an analytical concept, although claims about the precise amount of dormant global assets should be treated cautiously. Tokenization does not magically create wealth. What it can potentially do is reduce frictions surrounding ownership, collateralization, settlement, and transfer. The economic consequence could be higher capital utilization rather than simply more money.
Tokenization Is the Financial Infrastructure Layer
Tokenization deserves separate treatment from speculative cryptocurrency activity because its macroeconomic mechanism is fundamentally different. When a financial claim becomes programmable and transferable on a shared ledger, the transaction architecture changes. Collateral can potentially be posted and released automatically. Payments can be conditional. Securities can settle atomically. Compliance rules can become embedded in transaction logic. Liquidity management can become continuous rather than dependent on batches of institutional reconciliation.
The IMF's 2026 research emphasizes exactly these characteristics. It argues that tokenization can move risk away from traditional institutional balance sheets and toward software, infrastructure, governance, and operational systems. That does not eliminate risk; it changes where the risk resides. r5
This is why the convergence between AI agents and tokenized finance is economically more interesting than the simplistic proposition that “AI is bullish for crypto.” An AI agent needs an economic environment in which it can act. If the agent can identify a supplier but cannot authorize payment, verify identity, access collateral, settle a transaction, or manage permissions, the agent remains an advanced recommendation engine rather than an autonomous economic actor.
Tokenized financial infrastructure potentially supplies that missing layer. An agent can operate within programmable limits, access approved assets, execute conditional transactions, and interact with other software systems without requiring a human to manually intervene at every step.
The IMF has now explicitly connected these two developments. Its 2026 work on agentic AI argues that autonomous systems could shift payments from human-initiated instructions toward agent-mediated decisions, while its analysis of stablecoins notes that programmable settlement could make them particularly suited to an economy in which AI agents transact with one another. r6
Stablecoins May Be the Missing Monetary Interface
If agents become economically meaningful, the payment instrument they use becomes strategically important. Stablecoins are currently the most obvious candidate because they combine digital transferability with a unit of account linked primarily to the U.S. dollar.
The scale should nevertheless be kept in perspective. IMF analysis published in August 2026 put total stablecoin market capitalization at roughly $300 billion, with approximately 99 percent denominated in U.S. dollars. The IMF also noted that reported stablecoin transaction volumes can dramatically overstate genuine payment activity because much of the activity remains within crypto markets and includes bot-driven trading and arbitrage. BIS estimates cited by the IMF put payment-related stablecoin flows in 2025 at roughly $390 billion. r7
That distinction is essential. High transaction volume does not automatically prove that stablecoins have become a parallel global monetary system. The economically relevant variable is genuine settlement demand from households, businesses, financial institutions, and eventually autonomous agents.
There is also a macroeconomic trade-off. A June 2026 BIS working paper finds that widespread stablecoin adoption could affect both bank lending and government financing conditions. Stablecoin demand can pull deposits away from banks, potentially increasing bank funding costs and reducing lending, while stablecoin issuers' demand for short-term government securities can lower sovereign borrowing costs. The model's calibrated U.S. results find that the long-run output effect is modestly negative under its baseline assumptions, although the short-run fiscal-space channel can be positive. r8
In other words, stablecoins should not be treated as a frictionless economic free lunch. Their adoption can redistribute liquidity between banks, sovereign debt markets, payment networks, and digital asset markets. That redistribution itself could become macroeconomically important.
Bitcoin's Role Is Different From the Role of the Agentic Economy
The long-term Bitcoin thesis should therefore be separated from the much larger universe of crypto assets. Bitcoin has characteristics that are materially different from application-specific tokens, experimental blockchain networks, or emerging agent-payment protocols. Its scarcity, liquidity, market history, institutional recognition, and use as collateral give it a different monetary and financial role.
But the agentic thesis does not automatically imply that every crypto asset appreciates. In fact, technological disruption normally produces intense competition. If hundreds of applications can be created around similar models, interfaces, or blockchain functions, economic rents can migrate rapidly toward the layers with durable network effects, scarce infrastructure, trusted settlement, distribution, or proprietary data.
That is why I would resist treating the entire crypto market as a single macro asset. The investment question is increasingly architectural: which layer captures value when autonomous agents become users of financial infrastructure?
| Economic layer | Primary function | Potential macro significance |
|---|---|---|
| AI models | Reasoning, generation, prediction, planning | Raises productive capacity and changes the cost of cognitive labor |
| Agents | Autonomous execution of tasks | Creates new economic actors capable of continuous decision-making |
| Compute and semiconductors | Physical execution of AI workloads | Captures capital expenditure and productivity investment |
| Stablecoins | Digital settlement and dollar-denominated payments | Potentially changes payment velocity and financial intermediation |
| Tokenized assets | Programmable ownership and collateral | Potentially mobilizes previously inefficient financial claims |
| Blockchains | Settlement, verification, and coordination | Potentially supplies neutral transaction infrastructure for software agents |
| Bitcoin | Scarce digital asset and collateral | Potential monetary, reserve, collateral, and institutional role |
The Productivity Boom Could Be Disinflationary and Inflationary at the Same Time
One of the most interesting macro consequences is that AI can push inflation in opposite directions. On one side, automation reduces the marginal cost of producing software, services, analysis, customer support, research, and potentially physical goods. That is disinflationary. On the other side, the construction of AI infrastructure requires enormous quantities of semiconductors, electricity, data-center capacity, networking equipment, land, and capital. Those bottlenecks can be inflationary.
This creates a potentially unusual investment environment in which the price of AI-enabled services falls while the price of scarce inputs rises. The Federal Reserve's current discussion of AI explicitly highlights this tension between capital intensity, labor substitution, productivity, and the distribution of returns. r9
The result could resemble previous technology waves, but with a crucial difference in speed. The internet required years of infrastructure construction before smartphones transformed it into a ubiquitous consumer platform. AI agents can potentially accelerate the transition from infrastructure to application because software itself can participate in building, testing, deploying, and improving software.
That is the mechanism behind recursive technological acceleration. It should not be confused with unlimited growth. Physical constraints remain: electricity, chips, cooling, transmission capacity, manufacturing, capital, regulation, and human institutions can all become bottlenecks. But the marginal cost of creating additional digital economic capacity can fall sharply once the infrastructure exists.
The Labor Market Is the Critical Transmission Variable
The greatest macro uncertainty is not whether AI can automate tasks. It is how quickly the resulting productivity gains translate into broad purchasing power. If AI primarily increases profits while reducing labor income, aggregate demand can become a constraint. If lower prices, new industries, capital income, transfers, entrepreneurship, and complementary employment redistribute purchasing power, aggregate demand can remain strong.
This is precisely why employment statistics may become harder to interpret. A company that replaces ten administrative workers with one agent may become more productive, but the immediate labor-market data show job destruction. If the company uses the savings to expand output, lower prices, build new products, or create entirely new business lines, the eventual macroeconomic effect can be positive while the transition is disruptive.
The IMF's global AI analysis emphasizes that the macroeconomic outcome depends heavily on diffusion and institutional readiness, not simply on frontier model capability. That is an important constraint on the more aggressive technological forecasts. r10
Our investment framework should therefore monitor productivity and corporate margins alongside employment. If earnings rise while labor demand stagnates, that is not automatically evidence of a weak economy. It may instead indicate a shift in the capital-labor composition of production. Conversely, if earnings rise only because of temporary cost cutting while final demand deteriorates, the same pattern would mean something entirely different.
Why Traditional Valuation Can Become Temporarily Misleading
Technology transitions create valuation problems because investors must price an uncertain terminal state. A conventional discounted-cash-flow model assumes that the future resembles an extrapolation of the present with changing growth rates and margins. But if AI materially changes the cost structure of entire industries, historical margins and labor ratios become poor anchors.
That does not mean valuation ceases to matter. It means the variables entering valuation must change. We need to understand compute intensity, inference demand, energy consumption, agent utilization, software substitution, distribution advantages, proprietary data, switching costs, and the durability of network effects.
The same applies to crypto. A token with increasing transaction volume is not necessarily capturing economic value. The relevant questions are whether the network is necessary, whether users are economically dependent on it, whether fees accrue to holders or merely subsidize activity, whether competing networks can replicate the function, and whether the underlying asset has a defensible monetary role.
That is a much more rigorous framework than assuming rising adoption automatically translates into rising token prices.
The Security and Governance Constraint
The bullish technological thesis also contains its own limiting factor: autonomous agents create new attack surfaces. An agent capable of executing transactions is also capable of executing unauthorized transactions if its permissions, context, credentials, or software dependencies are compromised.
The IMF's work on agentic payments explicitly identifies cybersecurity, traceability, opacity, legal uncertainty, authorization, liquidity, and systemic effects as central issues. The institution emphasizes that probabilistic AI behavior has to coexist with the deterministic requirements of payment systems. r11
This makes programmable financial controls economically important. Spending limits, identity layers, permission systems, audit trails, transaction simulation, collateral constraints, and reversible authorization mechanisms could become as important to agentic commerce as the underlying AI model.
In that environment, privacy technologies may also become economically relevant, but their adoption cannot simply be assumed. Businesses may want agents to negotiate anonymously to prevent counterparties from extracting information about their purchasing behavior, while regulators will require sufficient transparency to enforce financial laws. The resulting architecture will likely involve selective disclosure rather than absolute anonymity.
What the Market Should Actually Monitor
I would organize the macro dashboard around five observable transitions rather than relying on a single Bitcoin liquidity chart. The first is AI inference growth: rising real-world model usage indicates that AI is becoming an input to production rather than merely an experimental technology. The second is agent autonomy: the percentage of tasks that software can execute without human intervention. The third is tokenized settlement: evidence that assets, collateral, and payments are moving onto programmable infrastructure. The fourth is genuine stablecoin payment activity rather than speculative crypto turnover. The fifth is the distribution of productivity gains through margins, prices, wages, capital expenditure, and employment.
| Indicator | What it measures | Why it matters |
|---|---|---|
| AI inference demand | Real-world model utilization | Tests whether AI is becoming productive infrastructure |
| Agent autonomy | Tasks completed without human intervention | Measures expansion of machine economic activity |
| Corporate margins | Output relative to operating costs | Shows whether productivity is reaching company economics |
| Labor demand | Employment, hours, vacancies, wages | Tests whether capital substitutes for or complements labor |
| Stablecoin payment flows | Genuine economic settlement | Measures whether digital dollars escape crypto-native speculation |
| Tokenized collateral | Assets used within programmable financial markets | Tests whether tokenization is changing capital velocity |
| Agent-to-agent transactions | Autonomous software commerce | Tests the core hypothesis of machine-native economic activity |
The Institutional Finance Connection
The most important development may ultimately occur inside traditional finance rather than outside it. The SEC's August 2026 work on institutional tokenization describes a market increasingly focused on production environments, legal enforceability, custody, collateral, settlement, and structured finance. The regulator's September 2026 material similarly emphasizes preserving the legal and economic identity of an asset as it moves through digital infrastructure. r12
This matters because successful financial innovation rarely requires the existing system to disappear. It often requires the existing system to absorb a more efficient technological layer. Banks can remain banks while deposits become tokenized. Securities can remain securities while settlement becomes programmable. Dollars can remain dollars while software agents use them as machine-readable payment instruments.
The likely outcome is therefore less “crypto replaces finance” and more “financial infrastructure becomes increasingly software-defined.” That is a substantially more plausible macroeconomic pathway.
Our Five-Year Macro Framework
Looking forward, I would frame the next phase as a contest between three forces: monetary policy, technological productivity, and financial infrastructure. Monetary policy remains powerful, particularly for leveraged assets and sectors dependent on refinancing. But its relative influence can decline when technological productivity increases earnings capacity faster than financing conditions deteriorate.
The second force is productivity. If AI produces sustained efficiency gains, nominal earnings can grow even in an environment where labor-force growth is weak. That would create a different relationship between economic growth and employment than investors have historically experienced.
The third force is financial infrastructure. If tokenization and stablecoins move from experimentation toward genuine commercial settlement, the velocity of capital could rise. The resulting economic activity would not necessarily require proportionate increases in M2. That would represent a genuine structural change in the liquidity mechanism.
None of this makes the bullish AI-and-crypto thesis inevitable. The BIS explicitly warns that stablecoins create monetary and financial-stability risks, while the IMF emphasizes governance, settlement risk, fragmentation, and the importance of trusted monetary infrastructure. r13
Our conclusion is therefore more precise: the important macroeconomic opportunity is not simply the continuation of a crypto liquidity cycle. It is the possibility that autonomous software becomes a new class of economic actor and that programmable financial infrastructure becomes the settlement layer through which those actors transact. If that transition occurs at scale, conventional measures of liquidity, productivity, employment, and capital utilization will become progressively less sufficient on their own.
Bitcoin could benefit from that architecture, particularly if it maintains its role as a scarce, highly liquid digital collateral asset. But the broader economic opportunity is much larger than Bitcoin. It encompasses compute, semiconductors, energy, cybersecurity, AI applications, stablecoins, tokenized securities, programmable collateral, payment infrastructure, and blockchain settlement networks.
The decisive evidence will not be another prediction about a particular price level. It will be observable economic behavior: agents making purchases, agents managing corporate workflows, tokenized collateral moving continuously, stablecoins settling real commercial transactions, AI-driven productivity appearing in margins, and capital increasingly flowing toward infrastructure that enables autonomous economic activity.
If those signals compound, we are not merely witnessing another technology trade. We are watching the early formation of a different economic operating system in which software participates directly in production and finance. That is the macroeconomic transition I would watch most closely.