Bitcoin Adoption Is Not the Same Thing as Bitcoin Address Count
The Bitcoin power-law thesis is compelling because it tries to explain price not merely as speculation, but as the emergent consequence of network growth. In its strongest form, the argument says that Bitcoin price follows a power law in time because two underlying scaling relationships compose:
[
N(t) \sim t^{\beta_A}
]
and
[
P \sim N^{\beta_M}
]
where (N(t)) is the number of non-zero-balance Bitcoin addresses and (P) is price. In the paper, the measured values are approximately:
[
\beta_A = 3.046
]
[
\beta_M = 1.838
]
so that:
[
\beta_A \times \beta_M \approx 5.60
]
which is close to the directly measured price-time exponent:
[
\beta \approx 5.69
]
This is an elegant model. It treats Bitcoin as a networked nonlinear system: time drives adoption, adoption drives network value, and network value drives price.
But there is a critical modeling assumption hidden inside the word adoption.
The paper does not directly measure adoption. It measures non-zero-balance Bitcoin addresses.
That distinction matters.
Address Count Is a Proxy, Not Adoption Itself
The paper’s adoption variable is:
[
N(t)=\text{number of Bitcoin addresses holding a non-zero balance}
]
That is not the same thing as:
[
A(t)=\text{number of people, institutions, or economic actors using Bitcoin}
]
The model effectively assumes that the observable on-chain address count is a stable proxy for the hidden adoption state:
[
N(t) \approx kA(t)^\alpha
]
This may have been a reasonable approximation during Bitcoin’s earlier history, when more users interacted directly with the base chain and self-custody was more common among economically meaningful participants.
But as Bitcoin matures, the mapping between economic adoption and on-chain address state becomes less direct.
The problem is not that address count is imperfect. All measurements are imperfect.
The problem is that the imperfection may not be stationary.
The Electrical Engineering Analogy
In electrical engineering terms, the paper is treating non-zero address count as a measurable node voltage that represents an internal state variable.
But the actual system is more complicated.
True adoption is the hidden internal state:
[
A(t)
]
Non-zero address count is the sensor output:
[
N(t)
]
The paper assumes a stable transfer function:
[
N(t)=g(A(t))
]
But the real system is closer to:
[
N(t)=g(A(t), C(t), L2(t), F(t), W(t), D(t), U(t))
]
where:
- (C(t)) is custodial concentration
- (L2(t)) is Lightning and higher-layer usage
- (F(t)) is the fee environment
- (W(t)) is wallet behavior
- (D(t)) is dust and abandoned UTXOs
- (U(t)) is UTXO consolidation and batching behavior
So the measurement channel has parasitic terms.
The paper treats the address-count signal as though it is a clean measurement of adoption. But increasingly, it may be a filtered, compressed, and distorted output of a much larger economic system.
Higher Layers Compress the Address Signal
The central issue is abstraction.
As higher layers develop, more economic activity can occur without producing proportional changes in base-layer address count.
Custody Compresses Users
An exchange or custodian can represent millions of users behind a relatively small number of addresses and UTXOs.
In that case, true adoption can rise:
[
A(t) \uparrow
]
while non-zero address count barely changes:
[
N(t) \approx \text{flat or weakly rising}
]
That does not mean adoption stopped. It means the adoption moved behind a custodial abstraction boundary.
From an engineering perspective, this is like trying to infer the number of internal circuit nodes inside an IC package by counting the external pins. The pin count may matter, but it is not the full circuit.
Lightning Compresses Activity
Lightning allows many economic transactions to occur off-chain, with the base layer seeing only channel opens, closes, and occasional settlement events.
That means the base chain observes boundary-condition updates, not every internal state transition.
In EE terms, Lightning moves switching activity from the visible base-layer bus into a higher-layer state machine. The L1 address count no longer captures the full transaction density of the system.
ETFs and Financial Products Compress Ownership
Spot ETFs and other custodial financial products can create large amounts of economic exposure to Bitcoin without creating a proportional number of new on-chain addresses.
A million ETF holders do not necessarily become a million on-chain Bitcoin addresses.
This increases financial adoption while compressing the address-count signal.
Wallet Behavior Changes the Measurement
Address count is also affected by wallet implementation and user behavior:
- address reuse avoidance
- change address generation
- UTXO consolidation
- batching
- coinjoin
- dust accumulation
- fee-driven cleanup
- custodial withdrawal policy
These behaviors can move the address-count metric independently of true economic adoption.
An increase in non-zero addresses may reflect adoption. But it may also reflect wallet design, dust, privacy practice, or UTXO management.
A decrease or flattening may reflect consolidation, custody, batching, or higher-layer migration rather than declining adoption.
Non-Zero Addresses Measure L1 State Occupancy
A more precise statement would be:
[
N(t)=\text{L1 address-state occupancy}
]
not:
[
N(t)=\text{adoption}
]
Non-zero addresses tell us something real. They measure the number of address states on the Bitcoin base layer with positive balance. That is a meaningful on-chain metric.
But it is not identical to economic adoption.
It is a topology proxy.
It measures the breadth of visible base-layer state, not the full breadth or depth of Bitcoin’s economic use.
Realized Value May Capture Economic Depth Better
A stronger adoption model would include not only address count, but also realized value.
Realized value is approximately:
[
RV(t)=\sum_i UTXO_i \cdot P_{\text{last moved},i}
]
This measures the value of coins at the price where they last moved. It is not merely counting addresses. It is measuring capital committed into the network’s UTXO state.
If address count measures something like node occupancy, realized value measures something closer to stored economic energy.
In EE terms:
- non-zero addresses = energized nodes above threshold
- active addresses = switching activity
- transaction volume = signal traffic
- realized value = stored charge or committed energy
- market cap = instantaneous marked-to-market terminal voltage
- MVRV = ratio between present voltage and stored-energy basis
Realized value helps correct for some weaknesses of raw address count.
A dust address adds one unit to address count but almost nothing to realized value.
A large institutional UTXO may represent many beneficial owners and would be undercounted by address count, but it contributes substantially to realized value.
Neither metric is perfect. But together they may better distinguish between breadth and depth of adoption.
Breadth and Depth Should Be Separated
The power-law model would be stronger if it separated adoption into two components:
[
A_{\text{breadth}}(t)
]
and
[
A_{\text{depth}}(t)
]
Breadth metrics may include:
[
N_{\text{nonzero}}, \quad N_{\text{active}}, \quad \text{entity-adjusted users}
]
Depth metrics may include:
[
\text{realized cap}, \quad \text{settled value}, \quad \text{coin-days destroyed}, \quad \text{long-term holder supply}
]
A more complete adoption function might look like:
[
A(t)=f(N_{\text{L1}}, RV, \text{custody factor}, \text{Lightning capacity}, \text{ETF holdings}, \text{entity-adjusted flows})
]
Then the model could test whether:
[
P(t)\sim A(t)^\gamma
]
is more stable than:
[
P(t)\sim N(t)^{1.838}
]
That would preserve the insight of the power-law thesis while improving the measurement model.
What This Means for the Power-Law Claim
The Bitcoin power-law model may still be empirically important.
The fact that non-zero address count has historically scaled cleanly with time, and that price has historically scaled with non-zero address count, is not trivial. It suggests that Bitcoin’s visible base-layer state has carried a strong signal about the network’s economic development.
But the causal interpretation should be stated carefully.
The strong version says:
Bitcoin price scales with adoption, and adoption is measured by non-zero addresses.
The more defensible version says:
Bitcoin price has historically scaled with a visible L1 address-state proxy that appears to have tracked adoption over the observed period.
That is a narrower and more precise claim.
The first statement risks confusing the map with the territory.
The second statement preserves the empirical result while acknowledging the measurement problem.
The Core Critique
As Bitcoin matures, economic use becomes more layered.
More ownership moves through custodians.
More activity moves through Lightning and other higher-layer systems.
More exposure moves through ETFs and financial products.
More base-layer activity is shaped by fee pressure, batching, wallet policy, and UTXO management.
Therefore, the relationship between true adoption and non-zero address count is likely to change over time.
The address-count metric assumes actors use Bitcoin in a sufficiently uniform way. But Bitcoin adoption is becoming less uniform, not more.
The system is abstracting.
And as abstraction increases, the base-layer address count becomes a less complete measurement of the total economic organism.
Conclusion
The power-law paper makes a valuable contribution by showing that Bitcoin’s price, address count, and network-value scaling have historically formed a coherent mathematical structure.
But its adoption metric should be treated as a proxy, not as direct observation.
Non-zero addresses measure visible L1 state occupancy. They do not directly measure users, institutions, capital commitment, Lightning activity, custodial ownership, ETF exposure, or the full economic depth of the network.
A stronger model would combine address count with realized value and other correction factors to distinguish visible base-layer topology from actual economic adoption.
The question is not whether non-zero address count contains signal. It clearly does.
The question is whether that signal remains structurally stable as Bitcoin becomes increasingly layered.
That is the critical test.
If the proxy remains stable, the power-law model gains strength.
If the proxy degrades under abstraction, the observed exponent may remain historically interesting but become less causally explanatory.
Bitcoin may still follow a power law.
But the adoption variable should not be treated as self-evident.