DEV Community: InvisibleHill

Has Bitcoin Bottomed? 7 Signals Point to a Local Low, Not a Confirmed Cycle Bottom

InvisibleHill — Fri, 14 Aug 2026 07:59:11 +0000

By InvisibleHill. Research cut-off: August 14, 2026. Bitcoin, ETF-flow, interest-rate, and on-chain readings can change after publication.

Bitcoin is trading near $63,000 after losing roughly half its value from the October 2025 peak. The easy answer is that a 49 percent drawdown, a June low near $58,600, and deeply compressed valuation indicators must mean the bottom is in.

The better answer is narrower.

Bitcoin may have formed a durable local bottom, but the evidence does not yet confirm a completed cycle bottom. Valuation is attractive. Price has defended an important support shelf. The macro environment is no longer uniformly hostile. But institutional flows are unstable, leverage has rebuilt, and Bitcoin has not recovered its 200-day average.

This is a market where cheapness has arrived before confirmation.

1. AHR999 says Bitcoin is in a bottom zone

The clearest valuation signal comes from BigRoom. On its live AHR999 page, the August 13 close showed:

Bitcoin price: $63,490.86
200-day average: $69,657.75
AHR999: 0.340
BigRoom classification: Bottom zone

AHR999 compares Bitcoin's market price with two slow-moving references: its 200-day cost line and a fitted long-term growth valuation. BigRoom's methodology guide treats readings below 0.45 as historically depressed.

That is meaningful. It says the current price is not merely below a recent high; it is compressed against both medium-term trend and long-cycle valuation references.

It does not say that $58,600 must be the final low. AHR999 is a valuation regime indicator, not a reversal detector. Bitcoin can stay in its bottom zone for weeks or months, and a liquidity shock can push an already-cheap asset lower.

The most useful interpretation is therefore: the long-horizon reward-to-risk ratio has improved, but the short-horizon trend is still damaged.

2. Price defended the first serious support shelf

BigRoom's daily series puts the latest one-year low at $58,624.71 on June 30. That is almost exactly where Coinbase Institutional's July bottom framework placed its first high-strength support zone: $58,000 to $59,000.

Coinbase reached that zone by combining unrelated methods, including pivot levels, volume-by-price, Fibonacci retracements, long-term-holder cost basis, and historical drawdown projections. The value of the exercise is not that every method is equally predictive. It is that several different methods converged on the same area.

Bitcoin tested that shelf, recovered above $60,000, and has spent more than a month without making a lower low. That is the strongest argument that a local bottom may already exist.

But the rebound has not completed the other half of the signal. Coinbase's August technical map placed resistance around $67,000, $69,000, and $71,000. BigRoom's 200-day average sits near $69,700, inside that resistance cluster.

A weekly recovery above the 200-day line would turn a valuation bottom into a trend repair. Until then, Bitcoin is still trading below the level that separates a damaged market from a recovering one.

3. The drawdown is large enough, but not historically final

A fall of roughly 49 percent from the October 2025 peak is not a routine correction. It has removed much of the cycle's speculative excess and forced late buyers to reconsider their time horizon.

It is also smaller than the 75 to 85 percent collapses associated with older Bitcoin bear markets. That comparison cuts both ways.

The bearish reading is obvious: history allows another leg down. Coinbase's confluence map identified $48,000 to $50,000 as the next strong zone if $58,000 fails, followed by lower support near $42,000.

The bullish reading is structural. Spot ETFs, deeper derivatives markets, larger institutional ownership, and a higher realized capital base may reduce the amplitude of each cycle. A maturing Bitcoin market does not need to repeat an 80 percent drawdown to clear leverage and reset valuation.

This is why the size of the decline cannot settle the question on its own. The current drawdown is severe enough to support a bottom, but not so severe that history forces one.

4. ETF flows have stabilized, then weakened again

The ETF channel is no longer a permanent bid. It is a two-way valve.

Farside Investors' daily table shows that U.S. spot Bitcoin ETFs took in about $865 million from August 3 through August 7. The next four reported sessions, August 10 through August 13, reversed by approximately $329 million.

That sequence matters more than either number alone. Institutions are willing to buy this range, but they are not yet accumulating with the persistence seen during the 2025 advance.

Coinbase made the same broader point in its August 7 commentary. Bitcoin had absorbed record June ETF outflows and weak July demand without breaking its range. That resilience is constructive. Yet the report also noted that falling oil, lower yields, and record equities produced little Bitcoin upside because crypto-native demand remained weak.

In other words, sellers are no longer overwhelming the market, but buyers have not taken control.

5. Positioning has reset, but it is no longer clean

The June selloff removed leverage. July began putting it back.

Coinbase's August positioning report found that perpetual futures, term futures, and options open interest all rose while spot and perpetual volumes fell. BTC funding stayed positive through most of July. That is not the profile of a fully washed-out market. Risk returned to balance sheets faster than organic trading activity returned.

The good news is that Bitcoin's leverage ratio eased slightly because exchange balances grew faster than open interest. The bad news is that open interest still rebuilt into quieter markets, creating more sensitivity to liquidations if support fails.

A confirmed bottom normally becomes healthier as price rises: spot volume expands, ETF demand persists, and leverage grows more slowly than real buying. The current market has only part of that pattern.

6. The macro environment is mixed, not easy

The Federal Reserve held the policy-rate target at 3.50 to 3.75 percent on July 29. More important for risk assets, three voters dissented in favor of a quarter-point increase. Monetary policy is restrictive, and the committee is not signaling an urgent rescue.

July data explain the tension:

Headline CPI rose 3.4 percent year over year; core CPI rose 2.5 percent.
Producer prices were unchanged in July but still 4.7 percent higher than a year earlier.
Nonfarm payrolls fell by 23,000, while May and June were revised down by a combined 103,000.
Unemployment remained relatively low at 4.1 percent.
On August 13, the U.S. Treasury curve showed a 4.15 percent two-year yield and a 4.63 percent ten-year yield.

This is not a clean recessionary easing setup. Employment is cooling, but inflation and long-term yields remain high enough to constrain the Fed. Bitcoin therefore lacks the powerful liquidity tailwind that confirmed several earlier cycle recoveries.

There is a more positive interpretation. If labor continues weakening while core inflation trends lower, the policy asymmetry can eventually shift toward easing. That would help Bitcoin. The market has not received that confirmation yet.

7. Strong equities are not pulling Bitcoin higher

One of the more unusual signals is Bitcoin's weak response to strength elsewhere.

Coinbase observed that Bitcoin's sensitivity to the Nasdaq-100 fell close to zero after being strongly positive earlier in 2026. Global equities reached records, oil declined, and rate expectations softened, yet Bitcoin stayed range-bound.

That divergence can mean two different things.

It may be a sign of exhaustion: Bitcoin has stopped falling even though its traditional buyers have stepped back. A market that refuses to break under bad internal conditions may be building a base.

It may also show opportunity-cost pressure. Investors can obtain high-growth exposure through profitable AI and semiconductor companies without taking crypto custody, regulatory, or drawdown risk. If equities remain strong while Bitcoin cannot reclaim $69,000, the base may become a long sideways process rather than the start of a new bull leg.

What would confirm the bottom?

The bottom case becomes materially stronger if three things happen together:

Trend repair: Bitcoin reclaims and holds the $67,000 to $70,000 region, including the 200-day average.
Real demand: ETF flows remain positive across several weeks instead of reversing after a few sessions.
Healthier positioning: spot activity improves without another disproportionate expansion in leveraged open interest.

The bottom case weakens if Bitcoin closes decisively below $58,000 while ETF redemptions accelerate. In that scenario, the $48,000 to $50,000 confluence zone becomes the next serious test.

The conclusion

Bitcoin looks cheap enough to be near a bottom, and the June low has survived long enough to be treated as more than an intraday accident. AHR999 is in BigRoom's bottom zone. Price defended a multi-method support shelf. The market absorbed heavy institutional selling without cascading lower.

What is missing is confirmation. Bitcoin remains below its 200-day average. ETF demand is inconsistent. Leverage has rebuilt. The Fed is still restrictive, and record equities have not pulled crypto higher.

The most defensible answer is therefore:

Bitcoin has probably formed a local bottom around $58,000 to $59,000. It has not yet proved that the cycle bottom is complete.

That distinction matters. Valuation indicators help decide when an asset is worth studying or accumulating over a long horizon. Trend, flows, and market structure decide whether the low is already behind us.

Sources

This article is for research and education only. It is not investment advice.

The ETF Changed Bitcoin's Four-Year Cycle, but It Did Not Create a Supercycle

InvisibleHill — Sun, 09 Aug 2026 05:07:06 +0000

Originally published by InvisibleHill Research. This cross-post preserves the original research cut-off and source list.

Research cut-off: August 9, 2026. Market prices, ETF flows, product availability, and indicator readings can change after publication.

Bitcoin reached a record above $126,000 in October 2025, about 18 months after the April 2024 halving. By August 8, 2026, its daily close was near $64,963, roughly 48 percent below that peak. The calendar looks familiar: a halving, a new high the following year, then a large drawdown.

The route was different. Bitcoin first broke its 2021 record before the 2024 halving, helped by U.S. spot exchange-traded products. ETF and corporate treasury demand later absorbed far more Bitcoin than miners created. When those flows weakened, the same regulated channel transmitted redemptions back into the spot market. AI stocks offered a profitable, liquid alternative, while crypto exchanges began letting stablecoin holders buy tokenized U.S. equities without returning to a bank or broker.

The four-year cycle therefore still exists, but it no longer explains the market on its own. The halving remains a supply event and a coordination point for investor behavior. Marginal demand now comes through vehicles that can move faster than miner supply, in both directions. That makes a permanent supercycle less likely than a different kind of cycle: more institutional, more correlated with equities, and less generous to altcoins.

The four-year pattern survived its first ETF cycle

Bitcoin has only three completed post-halving price cycles before the current one. That is too small a sample to treat the pattern as a statistical law. It is still difficult to dismiss the sequence.

The protocol halves the block subsidy every 210,000 blocks, or approximately every four years. Previous market peaks arrived in 2013, 2017, and 2021. The 2024 halving cut the subsidy to 3.125 BTC, and the market reached its latest record in October 2025. The subsequent decline has been smaller than the 75 to 80 percent collapses associated with older cycles, but a drawdown near 50 percent is not what strict supercycle forecasts promised.

This does not prove that halvings caused each peak. The timing also reflects a market convention. Miners, funds, media, and retail traders know the schedule, so capital moves before the event in anticipation of other buyers. The cycle combines programmed supply with a shared clock.

The latest cycle preserved that clock while breaking one of its familiar details. Bitcoin crossed its prior record in March 2024, before the halving. Demand did not wait for the supply cut. That was the first clear sign that ETF access could pull a later-cycle move forward.

The halving matters less to price formation

At 3.125 BTC per block and roughly 144 blocks per day, miners now create about 450 BTC daily, or 3,150 BTC weekly. At a Bitcoin price near $65,000, that is approximately $29 million of new supply per day. The number is meaningful to miners. It is small beside modern capital flows.

NYDIG estimates that spot ETFs and corporate treasuries absorbed more than 10,000 BTC per week during much of Bitcoin's 2025 advance, with one week above 48,000 BTC. Several ETF inflow periods alone reached 15,000 to 30,000 BTC per week, several times weekly issuance.

A halving still reduces the recurring sell pressure needed to fund mining operations. It also reinforces the scarcity narrative when demand is already rising. It cannot create that demand, and each absolute reduction becomes smaller relative to the outstanding stock. ETF investors, corporate buyers, and long-term holders now absorb or release thousands of coins each week. The halving sets the background; flows increasingly set the pace.

ETFs became a two-way valve

The U.S. Securities and Exchange Commission approved the first group of spot Bitcoin ETPs in January 2024. That gave brokerage accounts, advisers, retirement portfolios, and institutional allocators a familiar route to Bitcoin exposure without direct custody.

The supercycle argument treated this access as a permanent bid. In 2025, it sometimes behaved that way. Bitcoin's climb above $126,000 coincided with the strongest period of ETF creations and corporate treasury purchases.

An ETF is not locked capital. It is designed to create and redeem shares as demand changes. NYDIG found that ETF activity became much less consistent in late 2025 and 2026, with several weeks of net redemptions approaching 15,000 to 25,000 BTC. Bitcoin fell from about $126,000 toward $60,000 as ETF and treasury demand deteriorated.

Regulated access made the market deeper, but it also connected Bitcoin more tightly to portfolio rebalancing, risk limits, tax decisions, and macro shocks. Part of the old exchange-driven cycle is now a faster institutional flow cycle.

AI stocks competed for the same risk budget

The AI equity boom gave investors another liquid way to express a high-growth, high-duration view. The Nasdaq-100 returned 21 percent in 2025. Technology averaged 61 percent of the index and produced 88 percent of its total return. The ten largest securities made up 52 percent of the index and returned an average of 35 percent.

Many leaders also had revenue, earnings, buybacks, and large balance sheets. They compete with Bitcoin and altcoins for the same marginal risk budget even though the assets are fundamentally different. Both respond to liquidity, discount rates, momentum, and willingness to accept duration risk.

CME research found that Bitcoin's correlation with the Nasdaq-100 moved from almost zero before 2020 to a generally positive range afterward. In 2025 and early 2026, correlations across crypto assets and the Nasdaq-100 often reached 0.35 to 0.60. Crypto tended to rise with U.S. technology stocks, but by less, then fall more when those stocks sold off.

AI enthusiasm can support Bitcoin by keeping risk appetite high, then absorb capital that might once have moved into crypto. The competition is harsher for altcoins. They usually lack Bitcoin's ETF access, monetary narrative, and depth while also lacking the cash flows of public companies.

Tokenized stocks changed what stablecoin liquidity means

Stablecoin balances on crypto exchanges were once treated as probable crypto buying power. A trader holding USDT or USDC could wait, earn yield, withdraw, or rotate into Bitcoin and altcoins. That inference is now weaker.

Binance now puts bStocks inside its spot market. Eligible users can buy the underlying stock with supported crypto, including USDT and USDC, convert it into a bStock, or buy the token directly. The instruments trade around the clock, start from as little as $5, and can be withdrawn on BNB Smart Chain. Binance's distribution makes this more consequential than a niche RWA listing: stock exposure now sits beside crypto in one of the industry's largest pools of users and liquidity.

Bitget has gone further in product breadth and account integration. Its Stocks 2.0 product uses USDT for tokenized equities and connects eligible rTokens to unified accounts, margin, grid strategies, copy trading, and selected yield products. Bitget reported that its tokenized-stock spot volume had passed $1 billion by January 2026 and represented about 89 percent of Ondo-issued tokenized-stock volume in December 2025. By June 23, it had listed 529 rTokens. Those are company disclosures rather than an audited estimate of the whole market, but they show that the diversion channel is already operating at meaningful scale.

This creates a real diversion channel. A trader can move from USDT into an AI stock or equity index inside the same interface and, in some products, outside U.S. market hours. The stablecoin never leaves the platform, yet the capital no longer bids for a crypto asset. Trading attention and market-maker inventory can move with it.

Exchanges do not disclose how much tokenized-stock volume would otherwise have entered crypto, so the aggregate diversion cannot yet be measured. The effect can still be severe at the margin, especially in thinner altcoin markets. When the next buyer has dozens of profitable U.S. companies one click away, stablecoin growth no longer implies an approaching altseason.

Altseason is becoming narrower

Earlier cycles had a familiar rotation. Bitcoin rose first, profits moved into Ether and large-cap tokens, then smaller assets rallied as traders reached for more beta. The pattern depended on a relatively closed crypto venue where most speculative choices were crypto assets.

An investor who buys Bitcoin through a brokerage account does not automatically enter a crypto exchange, open an on-chain wallet, or gain a mandate to buy altcoins. Stablecoin liquidity already inside an exchange can now leave the crypto risk stack without leaving the account.

The relative performance supports this interpretation. CME's early-2026 comparison found that only Bitcoin, XRP, and Stellar were above their early-2024 levels in its sample, while Ether and Chainlink had fallen as much as 40 to 50 percent. The exact ranking will change, but the broad lesson is durable. Bitcoin can receive institutional demand without creating a broad altcoin bid.

Future altcoin rallies are more likely to concentrate in networks with real usage, credible economics, regulatory access, or their own investment wrappers. Memecoins can still surge. A market in which every large Bitcoin move lifts almost everything for months is becoming harder to finance.

A supercycle needs more than permanent access

There are two meanings of supercycle, and they lead to different answers.

The strict version predicts that structural adoption will eliminate the familiar boom, peak, and deep drawdown. The 2025 peak and 2026 decline argue against it. ETF demand reversed, treasury buying lost breadth, and Bitcoin fell by almost half even in a relatively supportive U.S. policy environment.

The weaker version predicts a secular rise interrupted by shallower cycles. That remains plausible. Fidelity Digital Assets observed unusually low realized volatility soon after the October 2025 record and argued that Bitcoin's larger market capitalization and deeper liquidity could reduce the old pattern of blow-off tops and 80 percent crashes. Fidelity's maturation thesis and NYDIG's cyclical thesis can both fit a cycle whose amplitude is falling.

A true supercycle would require demand that keeps expanding through equity corrections, tighter financial conditions, ETF redemptions, and competition from other assets. Bitcoin has not passed that test. The more defensible base case is a long-term adoption trend expressed through recurring liquidity cycles, not one uninterrupted cycle.

How to read the next cycle

The next halving remains relevant, but it should sit beside flow, valuation, and cross-asset measures rather than above them.

ETF creations and redemptions show whether regulated capital is adding or removing marginal demand. Bitcoin's relationship with the Nasdaq-100 shows whether the market is trading it as an independent monetary asset or as high-beta technology exposure. Stablecoin use now needs to be separated by destination: crypto spot, lending, payments, tokenized Treasuries, and tokenized equities do not create the same demand for Bitcoin or altcoins.

Long-cycle valuation tools still help with context. On BigRoom, the AHR999 Indicator compares Bitcoin's daily close with its 200-day cost line and a fitted long-term growth valuation. At the research cut-off, the indicator read approximately 0.355, inside BigRoom's bottom zone, while Bitcoin traded below its 200-day average.

That reading describes valuation temperature within Bitcoin's own history. It does not measure ETF redemptions, AI equity opportunity cost, or tokenized-stock adoption. A low AHR999 reading can support a long-horizon accumulation case without proving that the cyclical low is complete. It is more useful beside flow and cross-asset evidence than as a calendar.

Bitcoin's supply anchor remains, and the latest peak and drawdown still fit the broad sequence. What changed is the source and destination of marginal capital. ETFs can pull demand forward and reverse it quickly. AI stocks compete for the same risk appetite. Tokenized equities give stablecoins a route around crypto, with the largest effect on altcoins.

The base case is a more institutional cycle with smaller issuance shocks, larger flow shocks, tighter links to equities, and fewer broad altcoin rallies. That will not look exactly like 2013, 2017, or 2021. It also falls short of a permanent supercycle.

Memory Is the Bottleneck, but Not Forever

InvisibleHill — Sat, 08 Aug 2026 04:08:10 +0000

Originally published by InvisibleHill Research. This cross-post preserves the original research cut-off and source list.

Research cut-off: May 29, 2026. Market conditions and company guidance may have changed after publication.

Memory used to be the semiconductor industry’s least romantic business. It was
capital-intensive, brutally cyclical, and difficult to differentiate. Producers
spent billions of dollars making bits that customers mostly bought on price.
When supply exceeded demand by a few percentage points, profits disappeared.

Artificial intelligence has rewritten that story so thoroughly that the old
description now sounds quaint. World Semiconductor Trade Statistics expects
global memory revenue to rise about 250 percent in 2026, to more than $800
billion. Micron’s fiscal third-quarter revenue reached $41.5 billion, more than
four times the year-earlier level. It also guided to an 86 percent gross margin
for the next quarter. Sandisk’s gross margin jumped from 22.5 percent to 78.4
percent in a year.

Those are not normal semiconductor numbers. They are scarcity numbers.

The scarcity is real. AI accelerators cannot work without high-bandwidth memory;
inference servers need ordinary DRAM as well; vector databases, model
checkpoints, and retrieval systems consume flash; and the resulting oceans of
data eventually settle onto hard drives. But “storage” is not one market, and
real demand does not make every price rational. The same extraordinary profits
that prove the shortage also finance its eventual cure.

The useful investment question is not whether AI needs more memory. It does. The
question is who will still earn excess returns after customers, competitors,
and new capacity respond.

Four businesses hiding inside one theme

Investors often group Micron, Samsung Electronics, SK hynix, Sandisk, Western
Digital, and Seagate into a single AI-storage trade. Their economics are quite
different.

High-bandwidth memory (HBM) is stacked DRAM placed beside an accelerator.
Its value lies in moving enormous volumes of data without starving the GPU or
custom AI chip. HBM requires leading-edge DRAM, advanced packaging, thermal
management, a logic base die, and lengthy qualification with a small number of
accelerator customers. A failed stack can waste expensive silicon around it.
Yield and reliability matter as much as raw bit output.

Conventional server DRAM holds the working state of CPUs and accelerators.
It is less exotic, but AI servers carry far more memory than traditional
machines. It also benefits indirectly from HBM: every wafer and clean-room bay
directed toward complex HBM is capacity not used for ordinary DRAM.

NAND flash and enterprise SSDs sit farther from the processor. They store
model weights, embeddings, retrieval indexes, training checkpoints, and the
rapidly growing key-value caches created during inference. Enterprise products
earn premiums for endurance, firmware, latency consistency, and power
efficiency. Yet the underlying NAND bits remain more interchangeable and supply
is relatively elastic once factories are running.

Nearline hard drives are the archive. They are too slow for an accelerator’s
working memory but economically difficult to replace when hyperscalers need to
retain exabytes. Seagate and Western Digital now operate in what is effectively
a two-vendor market for high-capacity cloud drives. Higher areal density lowers
cost and power per terabyte, giving hard drives a durable role beneath flash.

HBM is a bottleneck, not a permanent monopoly

HBM deserves the highest-quality multiple in the group today. Its demand is
tied to the number and memory content of accelerators, not merely to replacement
PCs. Each generation is harder to manufacture. HBM4 doubles the interface to
2,048 data pins, adds a sophisticated base die, and integrates more deeply with
the customer’s system design. Suppliers cannot redirect commodity inventory
and call it qualified HBM.

The market therefore rewards execution. Micron is shipping HBM4 in volume for
its lead customer and expects HBM4E production in 2027. Samsung began commercial
HBM4 shipments in February and expects its HBM revenue to more than triple this
year. SK hynix entered the cycle with the strongest HBM franchise and continues
to sell high-value memory into AI systems.

This is a genuine moat, but it is a moving one. Product leadership must be won
again at every generation. Qualification is concentrated among a few enormous
buyers with the engineering staff and purchasing power to encourage a second or
third source. Packaging partners can add capacity. Yields improve. Samsung’s
return as a credible HBM4 supplier is good for the ecosystem and potentially
bad for scarcity rents.

Most important, HBM consumes capital. Micron expects roughly $27 billion of
capital spending in fiscal 2026, with quarterly spending rising again in 2027 as
it pulls forward clean-room construction. That expansion will not produce
finished supply immediately; fabs take years and HBM packaging remains complex.
But an 86 percent gross margin is an invitation written in very large type.

The bear case does not require AI demand to collapse. Supply only needs to grow
a little faster than customers’ urgent requirements, or accelerator makers need
to pause between architectures. In a fixed-cost industry, the marginal few
points decide the price.

Ordinary DRAM may be the most misunderstood winner

HBM attracts the headlines, but conventional DRAM explains why the current
cycle has spread across phones, PCs, automobiles, and industrial products.

In Micron’s latest quarter, DRAM bit shipments rose only in the low single
digits sequentially while average prices increased in the low-60-percent range.
That distinction is crucial. Revenue growth came overwhelmingly from price and
mix, not a sudden flood of physical bits. Samsung likewise attributed record
memory earnings partly to higher average selling prices and limited supply.

AI creates direct demand through memory-heavy servers and indirect scarcity by
pulling advanced capacity into HBM. Meanwhile, the producers that survived the
last downturn have become more cautious about adding commodity output. Micron
has signed 16 strategic customer agreements, generally running through 2030,
that provide supply assurance and may make future revenue less exposed to the
spot market.

That could represent a healthier industry structure. It does not repeal the
cycle. Long-term agreements differ in pricing, volume commitments, remedies,
and flexibility; “contracted” is not synonymous with “fixed high margin.” Node
transitions also create more bits per wafer without a new building. Once fresh
clean-room capacity arrives, ordinary DRAM lacks HBM’s packaging and
qualification barriers.

Investors should separate bit growth from price growth every quarter. If unit
economics are improving while bits remain constrained, earnings can stay
exceptional. If bits accelerate just as average selling prices flatten, the
profit peak may already be visible in the income statement.

NAND has the best demand story and the weakest memory

Enterprise flash has a compelling AI use case. Training creates checkpoints;
inference creates retrieval indexes and caches; agentic systems repeatedly read
large datasets. Micron has begun shipping a 245-terabyte QLC SSD. Sandisk’s
data-center revenue rose 233 percent sequentially in its April quarter. These
are not consumer USB drives being relabeled as AI.

But NAND has repeatedly punished investors who confuse a growing market with a
good industry. Producers can stack more layers, improve bits per cell, and turn
idle output back on. Customers can defer purchases when inventories rise.
Qualification and firmware create a moat at the drive level, but a large part of
the value remains sensitive to the price of NAND wafers.

Sandisk offers the clearest evidence. Its fiscal third-quarter revenue nearly
doubled sequentially to $5.95 billion, helped by a richer data-center mix and
higher pricing. Its 78.4 percent gross margin was 56 percentage points above the
prior-year level, and management guided to 79–81 percent for the following
quarter. The company has also signed multi-year arrangements with firm financial
commitments, which may improve durability.

The bullish reading is that AI has reset the value of enterprise flash. The
cautious reading is that no mass-produced storage medium retains an 80 percent
gross margin unless supply is extraordinarily tight. Both can be true. A
structurally larger market can still deliver violently cyclical earnings.

Hard drives: the quietest and perhaps cleanest case

The hard-drive thesis is less spectacular. It may be more legible.

After separating its flash business, Western Digital is a focused HDD company.
Its fiscal third-quarter revenue rose 45 percent year over year to $3.34 billion
and GAAP gross margin reached 50.2 percent. Seagate reported $3.11 billion of
revenue and a 46.5 percent GAAP gross margin. Both are benefiting from
hyperscale demand, disciplined supply, and higher-capacity drives.

AI strengthens “data gravity”: generated data may be processed in fast memory,
but useful records, model artifacts, video, telemetry, and backups must live
somewhere. Nearline HDD remains the cheapest practical home for much of it.
Heat-assisted magnetic recording and other areal-density gains can raise
capacity without equivalent growth in heads, media, power, or floor space.

The risk is that industry discipline is partly a product of consolidation and
long qualification cycles, not immunity from substitution. Flash cost per bit
continues to fall. Hyperscalers are a concentrated customer group. A delayed
capacity transition can create shortages; a successful one can produce more
exabytes than expected. And when a mature hardware company earns a 50 percent
gross margin, the market may capitalize a scarcity period as if it were a new
permanent baseline.

Is there a bubble?

There is no convincing evidence that AI memory demand itself is fictitious.
Shipments, customer commitments, product qualifications, and cash flow are
visible. This is not a pre-revenue theme assembled from investor slides.

There is, however, a classic bubble risk in the extrapolation.

WSTS expects the memory market to grow roughly 250 percent in 2026 and then 32
percent in 2027. Even the forecast embeds a dramatic deceleration. Micron’s
latest DRAM price increase, Sandisk’s margin expansion, and the HDD makers’
record profitability all describe an industry operating far above its
through-cycle economics. Stocks can fall while earnings rise if the rate of
improvement merely becomes less extraordinary.

Trailing price-to-earnings ratios are particularly unhelpful near a cycle peak.
The denominator is changing faster than the numerator. A producer can appear
cheap on the quarter in which scarcity is greatest and expensive one year later
without its share price moving. The better exercise is to normalize price,
utilization, and margin across a full replacement cycle, then ask what portion
of today’s demand and customer structure is genuinely new.

A disciplined investor should watch six things:

bit shipments versus average selling prices: price-led growth is powerful but more reversible;
capital spending and clean-room timing: today’s capex is tomorrow’s supply;
inventory days at producers and customers: shortages often end before management language changes;
HBM yield and customer qualification: leadership is valuable only while it remains hard to replicate;
contract quality: duration matters less than enforceable volume, pricing, prepayments, and cancellation terms; and
gross margin by product: a blended number can hide commodity exposure behind a small, premium category.

The investment conclusion

The memory boom is not one trade. HBM has the strongest near-term scarcity and
the deepest technical barriers. Conventional DRAM has the largest indirect
benefit from capacity diversion. Enterprise NAND has enormous AI volume
potential but the most dangerous history of supply elasticity. Nearline HDD
offers slower growth, a concentrated market, and a credible cost advantage for
cold and warm data.

The highest-quality companies will use this windfall to improve technology,
secure customer commitments, and return cash without assuming that today’s
prices last forever. The weakest investment cases will require both perpetual
AI demand acceleration and permanent scarcity margins.

AI has elevated memory from a component to a system constraint. That change is
structural. Constraints, however, attract engineers and capital. The winners
will not simply be the firms selling the scarcest bit in 2026. They will be the
ones whose advantage survives the moment the shortage begins to work.

Bitcoin’s Quantum Problem Is a Governance Problem

InvisibleHill — Sat, 08 Aug 2026 04:01:02 +0000

Originally published by InvisibleHill Research. This cross-post preserves the original research cut-off and source list.

Quantum computing will not “decrypt Bitcoin.” That phrase is technically wrong
and politically distracting.

A sufficiently capable quantum computer could do something narrower and more
dangerous: derive a private key from a visible public key, then produce a valid
signature and spend coins it does not own. The blockchain would not be hacked.
Its rules would be followed with a stolen key.

No existing machine can do this. Today’s best public demonstrations are still
learning how to preserve and operate a small number of error-corrected logical
qubits. But the resource estimates are falling, post-quantum standards now
exist, and Bitcoin upgrades slowly by design. Waiting for an attack is not a
serious plan.

The cryptography is solvable. The harder problem is the migration: how to move
millions of coins, what to do about owners who never move, and whether the
network should eventually reject signatures that are mathematically valid
under today’s rules. Quantum computing is therefore not only a security threat.
It is a test of Bitcoin’s claim that no committee decides whose money is real.

Two algorithms, two very different threats

Public discussion often combines Shor’s algorithm and Grover’s algorithm into
one quantum menace. They do different things.

Shor’s algorithm attacks ownership. Bitcoin uses elliptic-curve
cryptography on the secp256k1 curve for ECDSA and Schnorr signatures. Given a
public key, a large fault-tolerant quantum computer running Shor’s algorithm
could solve the discrete-logarithm problem and recover the corresponding
private key. With that key, the attacker could sign a normal transaction.

Grover’s algorithm weakens search. Bitcoin’s proof of work and many internal
commitments depend on SHA-256 and other hash functions. In an idealized quantum
computer, Grover’s algorithm gives a quadratic speedup: a 256-bit preimage
problem has roughly 128-bit quantum security. That is a meaningful reduction,
but 2^128 remains an enormous search space. Quantum mining would also face
serial gate costs, hardware scarcity, competition from classical ASICs, and the
network’s difficulty adjustment.

The practical conclusion is clear. The first systemic quantum risk is not a
miner rewriting the chain. It is a key-recovery attacker spending coins from
outputs whose public keys are already visible.

Not every coin is equally exposed

Bitcoin has several output types, and they reveal different information.

Early pay-to-public-key (P2PK) outputs place a public key directly on-chain.
Modern Taproot outputs also contain a public key in the output. Coins in those
formats offer a future quantum attacker unlimited time to work: the public key
is visible as long as the coin remains unspent.

Pay-to-public-key-hash formats, including common SegWit addresses, work
differently. The output commits to a hash of the public key. The public key is
revealed only when the owner spends. If the address has never been reused, a
quantum attacker must recover the private key and replace the transaction in
the brief interval before confirmation.

That creates two clocks. A long-exposure attack can target an already
visible key for months or years. A short-exposure attack must observe a
transaction in the mempool, recover the key, and front-run the owner, perhaps in
minutes. The latter requires a much faster machine.

Address reuse defeats the concealment. Once a public key has appeared in a
previous spend, any remaining or newly received coins controlled by the same
key become long-exposure targets. Extended public keys and some wallet
descriptors can create exposure off-chain as well.

This nuance cuts both ways. It means the entire supply is not waiting naked for
the first quantum computer. It also means the oldest coins, reused addresses,
and Taproot outputs are natural first targets. A successful theft from a famous
dormant address could cause a market crisis before the attacker had the speed
to threaten ordinary transactions.

How far away is the machine?

Nobody can responsibly give a date for a cryptographically relevant quantum
computer. Qubit counts in headlines are poor forecasts because a noisy physical
qubit is not a reliable logical qubit, and logical qubits alone say little about
gate speed, connectivity, error correction, or how long a circuit can run.

Google’s 105-qubit Willow processor crossed an important threshold in 2024. A
larger error-correcting code suppressed errors better than a smaller one, and a
101-physical-qubit code preserved one logical memory longer than its best
component qubit.
That was a scientific milestone, not a cryptographic computer. Shor’s algorithm
against secp256k1 requires a large universal fault-tolerant machine executing a
deep sequence of reliable operations, not merely storing one encoded state.

The gap is still enormous, but it is no longer useful to describe it with a
number from five years ago. A 2026 resource study accepted by PRX Quantum
describes circuit designs using fewer than 1,500 logical qubits and estimates
that, under specific superconducting-hardware assumptions, fewer than 500,000
physical qubits could recover a secp256k1 key in
minutes.
That estimate is not a delivery forecast. It assumes error rates, connectivity,
control, and manufacturing at a scale nobody has demonstrated. It does show
why “millions of qubits” should not be treated as a permanent safety margin.

There is another reason to start early: migration time. NIST finalized its
first post-quantum standards in 2024, including the lattice-based ML-DSA and the
hash-based SLH-DSA for digital signatures. The U.S. government’s transition
planning contemplates phasing out vulnerable public-key cryptography in the
2030s. Bitcoin cannot assume that banks, browsers, and governments will spend a
decade migrating while a global decentralized asset will improvise later.

The sensible stance is neither “quantum winter forever” nor “Q-Day in 2029.” It
is to treat arrival time as a wide distribution and make migration progress
observable.

The first step exists, but it is not the finish

Bitcoin Improvement Proposal 360 is the most concrete protocol step so far. It
proposes Pay-to-Merkle-Root (P2MR), a new output type similar to Taproot but
without the elliptic-curve key-path spend. The output commits directly to a
script-tree root, so it can avoid leaving a long-lived public key exposed.

The proposal is deliberately modest. Its own specification says P2MR protects
against long-exposure attacks, not the short-exposure window after a public key
is revealed.
Full protection still requires a post-quantum signature scheme inside the
spending script.

That separation is good engineering. P2MR can reduce today’s attack surface and
create a clean place for a future signature opcode without forcing the network
to choose an immature algorithm immediately. It can be activated through a
soft fork, preserving compatibility for nodes that have not upgraded.

The difficult part is size. A Schnorr public key is 32 bytes and a signature is
64 bytes. Standardized post-quantum signatures are generally measured in
thousands or tens of thousands of bytes. Bitcoin has intentionally scarce block
space. Replacing every signature naively would reduce transaction throughput,
raise fees, enlarge the blockchain, and make node operation more demanding.

Developers are studying lattice signatures, hash-based signatures, aggregation,
commit-and-reveal constructions, and zero-knowledge proofs that verify a
post-quantum authorization more compactly. Each trades bandwidth, computation,
assumptions, implementation risk, and maturity. NIST’s selection is valuable
evidence; it is not an automatic answer for a consensus system that may need to
validate signatures for generations.

The lost-coin dilemma

Suppose Bitcoin adds a sound post-quantum output and gives holders years to
migrate. Most active users, exchanges, custodians, and institutions move. Some
coins do not.

They may belong to people who missed the deadline. The keys may be in cold
storage. The owner may be dead. The coins may be deliberately dormant. Or the
keys may be permanently lost. The blockchain cannot distinguish among these
stories.

Once a quantum attacker can recover old private keys, the network faces three
uncomfortable choices.

Honor the existing signature rules. A quantum-generated signature is
valid, so miners include it. This preserves protocol continuity but rewards
theft. Dormant coins could suddenly enter liquid supply, and no observer
could tell whether the spender was an owner or an attacker.
Disable vulnerable signatures before the attack. This protects the
network from mass theft but strands any owner who failed to migrate. A rule
intended to defend property becomes a collective decision to freeze it.
Create a recovery mechanism. Owners could prove knowledge of an old
secret using a post-quantum construction and reclaim funds after elliptic-
curve spends are disabled. This reduces confiscation risk but adds protocol
complexity, privacy questions, and new failure modes.

The argument becomes most intense around coins attributed to early miners,
including Satoshi Nakamoto. If those keys are lost, a quantum computer could
turn economically dormant supply into active supply. If the network freezes
them, critics will say developers changed the rules to protect the price. If it
does not, critics will say it knowingly legalized theft.

No cryptographic primitive can answer that question. The answer will come from
social consensus expressed through software, miners, economic nodes, wallets,
and users.

A credible migration plan

A robust plan should be staged long before an emergency.

First, measure and reduce exposure. Wallets and custodians should inventory
output types, address reuse, extended-key handling, and recovery paths. Users do
not need to panic-move coins today. They do need modern backups and software
capable of participating in a future migration. Wallet defaults should stop
creating avoidable long-exposure keys.

Second, activate an upgradeable quantum-safer output. P2MR or a comparable
design would let cautious holders move before the final signature choice is
settled. Exchanges and hardware wallets need test vectors, descriptors, fee
estimation, recovery procedures, and years of operational testing.

Third, select at least two authorization paths. A primary scheme and a
conservative backup based on different mathematical assumptions would reduce
the danger of discovering a flaw after billions of dollars have migrated.
Hybrid signatures can require both classical and post-quantum authorization
during the transition.

Fourth, publish triggers. The network needs evidence thresholds for
accelerating migration and, eventually, disabling vulnerable spends. A July
2026 developer discussion proposed a cryptographic “tripwire”: a signature by a
published nothing-up-my-sleeve key could prove that secp256k1 had been broken
and automatically disable elliptic-curve paths in new protected outputs. The
idea is not finished, but it correctly asks who gets to declare the emergency.

Finally, decide the legacy policy while the choice is still abstract. A
multi-year notice period, transparent activation sequence, and narrowly scoped
recovery design are more legitimate than an overnight patch after a theft.
The worst time to debate property rights is while a quantum attacker is moving
coins.

What this means for Bitcoin’s value

Quantum risk is unlikely to appear in the price as a smooth annual discount. It
will arrive through discrete credibility events: a new error-correction result,
a smaller resource estimate, a state program, a public key-recovery
demonstration on a weaker curve, or a successful Bitcoin testnet migration.

If protocol work stays ahead of hardware, each milestone can increase
confidence. Bitcoin would demonstrate that fixed monetary rules do not require
frozen technology. If governance stalls while the hardware gap closes, the
same milestone can trigger a rush to safer output types and a repricing of
dormant-coin risk.

An actual theft would be different. The immediate problem would not be a change
to the 21 million cap; quantum computers cannot mint arbitrary bitcoin under
current rules. It would be a sudden transfer from inactive or compromised
owners to an adversary with no acquisition cost and every incentive to sell or
destabilize the market. Confidence in custody and settlement would matter more
than the formal supply schedule.

Bitcoin has survived bugs, forks, exchange failures, and political bans because
the protocol and its users adapted without changing the core monetary bargain.
Quantum computing asks whether they can do that prospectively, before the threat
forces a choice between theft and confiscation.

That is why the work should begin while it still feels early.

BIP 110 and the Cost of Policing Bitcoin's Block Space

InvisibleHill — Sat, 08 Aug 2026 03:49:09 +0000

Originally published by InvisibleHill Research. This cross-post preserves the original research cut-off and source list.

Research cut-off: July 30, 2026. Miner signaling is a live metric and may have changed after publication.

BIP 110 begins with a problem that many Bitcoin users can recognize. A miner
can collect a one-time fee for including an image, token payload, or other
arbitrary data, while thousands of node operators absorb the cost of
downloading, validating, and sometimes storing it. The proposal's authors see
that mismatch as a subsidy for data storage and a threat to Bitcoin's use as
money.

Their answer is a temporary soft fork. For about one year, BIP 110 would make
several currently valid transaction structures invalid under consensus rules.
It would cap OP_RETURN outputs at 83 bytes, limit many data pushes and witness
items to 256 bytes, restrict large output scripts and Taproot control blocks,
and disable several Taproot upgrade paths and script features that can carry
data.

Adam Back agrees with the premise more than his opposition sometimes suggests.
He has said that Bitcoin is about money and that spam has no place in its
timechain. He also designed Hashcash to make spam costly. His objection is to
the remedy. In Back's view, an annoyance that fits inside Bitcoin's existing
block limit does not justify a contentious consensus change, especially one
that can be bypassed, can interfere with legitimate scripts, and has not earned
technical or economic agreement.

That distinction is the center of the debate. BIP 110 asks whether Bitcoin
should discourage an unwanted use through local policy and fees, or declare
some forms of that use invalid for everyone. Back's case is stronger on this
question. The proposal identifies a real externality, but it offers an
asymmetric bargain: incomplete filtering in exchange for a new consensus
precedent, a more complicated upgrade path, and a material risk of a minority
chain.

A policy dispute became a consensus proposal

The immediate background is Bitcoin Core 30.0. Released in October 2025, it
raised the default -datacarriersize setting to 100,000 bytes, effectively
removing the previous OP_RETURN relay limit because the standard transaction
size limit is reached first. It also allowed multiple data-carrier outputs
within one transaction. Operators can still set the limit back to 83 bytes.

This was a change in node policy, not Bitcoin consensus. Policy determines
which valid unconfirmed transactions a node will relay and which valid
transactions a miner may place in its candidate block. Different nodes and
miners may choose different policies. Consensus determines whether a confirmed
block is Bitcoin at all. A transaction filtered by one node can still reach a
miner through another route, and a block containing it remains valid to the
network.

BIP 110 crosses that boundary. Formally titled the Reduced Data Temporary
Softfork, it would add seven consensus restrictions for one year:

New output scripts could not exceed 34 bytes, except for an OP_RETURN output of up to 83 bytes.
Many pushed data fields and script-argument witness items would be limited to 256 bytes.
Spending undefined witness or Tapleaf versions would be invalid.
Taproot annexes would be invalid.
Taproot control blocks would be limited to 257 bytes.
Tapscripts containing an OP_SUCCESS opcode would be invalid.
Tapscripts executing OP_IF or OP_NOTIF would be invalid.

Outputs created before activation are grandfathered. After the temporary
period expires, the extra rules disappear. The proposal reached "Complete"
status in the BIP repository on June 25, 2026, which means its specification is
complete. It does not mean Bitcoin has accepted or activated it.

The deployment is unusually assertive. Early lock-in requires 1,109 of 2,016
blocks, or 55 percent, to signal with version bit 4. If that threshold is not
reached voluntarily, BIP 110 nodes enter a mandatory signaling period at block
961,632. During that period they reject non-signaling blocks. Lock-in follows
no later than block 963,648, with the new transaction rules scheduled to become
active at block 965,664.

At 15:35 UTC on July 30, the public BIP 110 monitor showed 14 signaling blocks
out of 655 in difficulty period 476, a rate of 2.14 percent against the
55 percent target. That snapshot does not determine the final outcome, but it
shows how far the proposal remained from coordinated miner activation shortly
before mandatory signaling.

The strongest case for BIP 110

Reducing the proposal to a fight over JPEGs misses its best argument.

Bitcoin block space has a hard supply. Data transactions compete with payments
for that space and can raise fees when demand is high. The miner receives the
fee, while every validating node must process the block. Archival nodes retain
the data, and pruned nodes must still download and validate it during initial
sync. Large spendable output scripts are more costly because they can remain
in the UTXO set, which nodes need to query quickly.

The cost is bounded by Bitcoin's block-weight limit, and pruning reduces
storage requirements. It is not imaginary. One party can buy permanent
publication while distributing part of the bill to infrastructure it does not
pay. BIP 110 supporters are also right that a fee market does not compensate
every node operator. It compensates miners for block production.

The proposal has several thoughtful safeguards. It is temporary rather than
permanent. It preserves pre-activation UTXOs. It provides test vectors and a
reference implementation. Its limits are simple enough to audit. It openly
admits that BitVM, Miniscript constructions, future upgrade hooks, and some
experimental Taproot uses may be constrained. That is better than presenting a
filter as costless.

BIP 110 also does not claim that steganography can be eliminated. A determined
user can split data into smaller pushes or disguise it inside structures with
another protocol meaning. The authors' narrower goal is to make contiguous,
obvious publication more cumbersome and to state that arbitrary data storage
is unsupported.

That may work as a social and economic deterrent. Raising the engineering cost
of an activity can reduce casual use even when it cannot stop a determined
actor. A temporary rule could also buy time for a cleaner design that separates
optional data from the material every node must validate.

The difficulty is that this modest technical effect is paired with a much less
modest activation mechanism.

Adam Back's boundary: dislike is not a consensus rule

Back's position is useful because it does not depend on liking inscriptions,
Runes, or other data-heavy applications. He dislikes spam. In September 2025,
he wrote that Bitcoin is owned by humanity, that protocol developers are
stewards, and that material changes require user consensus. By February 2026,
he was arguing that BIP 110 would damage Bitcoin's credibility as a store of
value and secure monetary network. In the same exchange he called spam "just
an annoyance."

His longer July argument starts from permissionless money. The decentralization
needed to create bearer money also prevents one faction from imposing its
preferences on everyone else. A user can change personal software, but not
another user's software. Back therefore describes BIP 110 as an attempt to
police other people through consensus, even if its supporters believe they are
defending Bitcoin.

He also emphasizes the engineering process. Bitcoin's change resistance
depends on much more than a high miner threshold. It includes review by developers, protocol
researchers, wallet authors, miners, exchanges, custodians, and users who can
refuse a change. Back compares this to an IETF-style technical consensus in
which unresolved objections block progress. In July he summarized the outcome
plainly: "Bitcoin respectfully says no to what you want."

That sentence is rhetoric, not a formal vote. Back does not own Bitcoin, and
neither do Bitcoin Core developers. His argument should stand on its mechanics,
not his biography. Here the mechanics support him. BIP 110 has weak miner
signaling, no demonstrated coalition of major exchanges or custodians, and
unresolved technical objections. The proposal's mandatory path does not create
consensus. It schedules a test of whether consensus exists.

Back's position also explains why opposing BIP 110 is compatible with running
stricter relay policy. A node operator can filter locally. A mining pool can
decline transactions. Bitcoin Knots can offer different defaults from Bitcoin
Core. Those choices compete without declaring the other side's blocks invalid.
This diversity is less satisfying to anyone seeking a universal ban, but it
preserves a valuable boundary between individual judgment and shared validity.

The technical exchange is asymmetric

BIP 110's filtering is intentionally incomplete. In October 2025, Peter Todd
demonstrated a transaction that embedded the text of the proposal while
remaining compliant with its rules. Later examples showed that larger images
could also be encoded through permitted structures. These demonstrations do
not prove that the proposal has zero effect. They do prove that arbitrary data
will survive activation.

The users most motivated to evade the filter can update their encoding
software. Ordinary wallet and contract users may have a harder problem. BIP
110 disables tools that were reserved for future upgrades and restricts
features that are not inherently data storage. It forbids Taproot annexes,
OP_SUCCESS opcodes, undefined witness versions, deep control blocks, and
executed conditionals in Tapscript. Some are unused today. Others are building
materials for future protocols.

The proposal itself acknowledges the tradeoffs. A 257-byte control-block limit
can complicate BitVM and other large script trees. Current Miniscript compilers
can produce Tapleaves containing OP_IF. No new soft fork can use the disabled
upgrade hooks during the one-year window. Wallets and libraries must understand
one rule set during the intervention and another after it expires.

Grandfathering protects outputs confirmed before activation, but it cannot
cover every pre-existing commitment. A user may hold a pre-signed chain of
transactions in which a descendant confirms after activation. To BIP 110
nodes, that descendant creates a new UTXO even though the owner committed to it
earlier. If its only spending path violates the temporary rules, the funds can
become unspendable until expiry. In the Bitcoin development mailing list,
Gregory Maxwell pointed to inheritance plans and other pre-signed arrangements
as plausible examples.

The BIP calls such cases unlikely and imposes several conditions before funds
can be frozen or lost. That may be accurate. It is still the wrong comparison
to ask whether the number is probably small. The comparison is between the
expected benefit of a bypassable filter and the expected cost of changing the
validity of unknown, privately constructed spending arrangements.

Back's critique is compelling because the asymmetry points in the wrong
direction. Sophisticated data publishers route around the restriction.
Legitimate users cannot route around a consensus rule after their spending
conditions have been fixed.

The activation design carries the largest immediate risk

A soft fork is backward compatible only when the economic majority follows the
more restrictive chain. Old nodes will accept blocks produced under the new
rules. Upgraded nodes will reject blocks that violate them. If most miners
enforce the restriction, both groups stay on one chain.

If only a small minority enforces it, the result is different. At the mandatory
signaling height, BIP 110 nodes reject blocks that the rest of the network
accepts. Unless enough hash power follows them, those nodes move onto a
low-hashrate minority chain. Transactions may be valid on both chains, deposits
become difficult for exchanges to credit safely, and users face replay and
double-spend risk. The one-year expiration does not make that coordination
problem temporary in any reassuring sense.

Supporters compare the plan with the 2017 user-activated soft fork campaign for
SegWit. The analogy is incomplete. SegWit had years of engineering review,
broad wallet and infrastructure integration, and overwhelming economic support
before the threatened UASF date. Miners ultimately activated it through the
existing signaling process, so the UASF did not have to split the network.

BIP 110's 55 percent threshold is much lower than the 95 percent used by
traditional BIP 9 deployments, and the mandatory period is meant to force the
issue if voluntary signaling fails. A threshold can coordinate actors who
already want the same change. It cannot manufacture demand from miners,
exchanges, wallets, and holders who have not agreed to bear the fork risk.

This is where node counts can mislead. Reachable nodes are not ballots, and
creating many nodes is cheap. What matters during a fork is the economic weight
that miners need to serve: exchanges that value coinbase rewards, merchants
that accept payments, custodians that control deposits and withdrawals, and
holders willing to price the two assets differently. BIP 110 has generated a
visible movement. It has not yet demonstrated that coalition.

What BIP 110 changes even if it fails

BIP 110 is unlikely to make arbitrary data disappear. Its more durable effect
will be a clearer map of Bitcoin's governance limits.

First, policy diversity will matter more. Bitcoin Core 30's OP_RETURN default
helped turn an implementation choice into an identity dispute. Bitcoin Knots,
mining-pool templates, and operator settings provide a way to express stricter
preferences without a consensus fork. Future client teams may treat policy
defaults as politically consequential even when they do not alter consensus.

Second, a failed activation would raise the bar for the next UASF. The lesson
would not be that users are powerless. It would be that user power becomes
effective only when it is connected to wallets, businesses, liquidity, and
hash rate. A social-media node count cannot substitute for economic
coordination.

Third, the technical questions will remain. Node operators still bear costs
that block-space buyers do not fully pay. Pruning, faster initial sync,
assumeUTXO, more efficient validation, optional data transport, and better
policy tools can reduce that burden without classifying transactions by
purpose at consensus. None is a perfect answer. They attack the cost directly
instead of treating cultural disapproval as a script rule.

The fee market will remain contested as the block subsidy declines. Data
applications can crowd out low-fee payments during bursts of demand. They can
also provide transaction fees to miners when monetary settlement demand is
weak. BIP 110 chooses monetary purity over that source of security revenue
without showing that payments alone will reliably replace it. The ecosystem
needs evidence about sustained fee demand, node costs, and user behavior more
than it needs a permanent label for spam.

Finally, the precedent matters. If Bitcoin changes consensus rules to suppress
one legal but disliked use, future campaigns can argue that other controversial
transactions deserve the same treatment. Governments and private pressure
groups would learn that transaction eligibility is negotiable. Back's
neutrality argument is strongest here. Censorship resistance is credible
partly because participants can say that they lack the practical consensus to
make selective exceptions.

Why policy is the better layer

BIP 110 deserves a more serious answer than mockery. It documents a genuine
cost externality, gives node operators a political voice, and forces the
ecosystem to confront what scarce block space is for. Its temporary design and
UTXO grandfathering show that the authors tried to contain collateral damage.

The proposal still fails its own risk test. It cannot eliminate arbitrary data.
It interferes with legitimate and future script uses. It creates edge cases for
pre-signed transactions. Its changing rule set burdens wallet and protocol
developers. Most of all, its mandatory activation path can split the network
without evidence that the economic majority wants the change.

Adam Back's opposition is persuasive because it keeps the layers separate. He
can dislike spam, support money as Bitcoin's primary purpose, and still reject
a consensus filter. Local policy lets an operator refuse a transaction without
requiring every other operator to share that judgment. Consensus decides which
blocks belong to Bitcoin's ledger.

Bitcoin may eventually adopt narrower limits or a better architecture for
optional data. Such a proposal would need to reduce measurable node costs,
preserve upgrade hooks, protect existing spending arrangements, resist trivial
bypass, and earn broad technical and economic support before activation.
BIP 110 does not meet that standard.

Its longer legacy may be useful anyway. Bitcoin's participants plainly do not
agree on acceptable behavior. The network protects neutrality by making any
faction's preferred restrictions unusually difficult to impose on everyone
else, including when that faction believes it is protecting Bitcoin.