Two Windows a Day: Hormuz Convoy-Lite, Blob Saturation, and the Same Scarcity Math
1. The anomaly
On September 11, 2025, the Financial Times reported on an email that had gone out to commercial shipping operators in the Persian Gulf. The email came from the U.S. Naval Cooperation and Guidance for Shipping, known as NCAGS. Its content: transit through the Strait of Hormuz would be recommended in two specific windows per day. Not at night. Not continuously. Two windows.
I run an on-chain analytics desk out of Manila. I do not cover shipping. I read the sentence three times anyway.
The volume numbers are not subtle. Roughly 20 million barrels per day of petroleum liquids transit Hormuz. That is about a fifth of global consumption and roughly a quarter of all seaborne oil. The navigable channel narrows to about 21 miles, split into inbound and outbound lanes roughly two miles wide each. Saudi Arabia and the UAE hold perhaps 4 to 5 million barrels per day of bypass pipeline capacity combined, and only when the pumps are working and the receiving terminals are not themselves under threat. You cannot reroute Hormuz. You can only schedule it.
I have written that sentence before. I wrote it about Ethereum block space in September 2020, when gas crossed 200 gwei and the yield farms that survived were the ones that could wait. I wrote it about blobs in March 2024. I wrote it about NFT exits in 2021, when I tracked 1,200 top-tier Bored Ape wallets across ten weeks and found that the variable predicting the 30% floor correction was not volume โ it was the concentration of sell intent into a narrow time band.
The physics of a chokepoint do not care whether the chokepoint is twenty-one miles of water or 128 kilobytes of blob space. Four variables recur: capacity, schedule, price of access, and how tightly the traffic clusters once you force it into a lane.
The Hormuz email is the newest worked example. Most of the crypto market read it as a macro headline and moved on. That was a mistake, and this piece is the correction.
2. Position sheet
Before the argument, the ledger. Every number below carries a confidence label, because I have spent too many years watching analysts launder inference into fact by deleting the error bars.
| Claim | Value | Confidence | Basis | |---|---|---|---| | Hormuz transit recommended in two daily windows | Stated | High | FT report citing NCAGS advisory email | | Prior guidance allowed broader night transit | Implied | Medium | FT report, contrast with current windows | | U.S. naval presence south of the strait since May | Stated | Medium-High | FT report | | Air-defense umbrella with aircraft available to shipping | Stated | Medium | FT report; platform types undisclosed | | Iranian attacks concentrated at night | Stated | Medium | FT report, single-sourced | | Guidance is advisory, not mandatory | Stated | High | Language of the advisory | | Iranian strategic intent | Unstated | Low | Absent from source; inferred | | Iranian hashrate share | Unquantified | Low | Range-bound estimate, wide error bars | | Blob capacity steps since March 2024 | Measured | High | Protocol parameters, public | | Blob base fee floor at 1 wei | Measured | High | Public chain data | | Custody concentration of ETF inflows | Measured | Medium | My own clustering pass, heuristics | | On-chain cover capacity | Estimated | Medium | Aggregated market data |
One line in that table is doing more work than the others. The guidance is advisory, not mandatory. Hold that thought; it returns in section eleven and it is the most important sentence in the entire report.
3. What NCAGS actually is
NCAGS is not a warfighting command. It is a coordination construct with NATO lineage, run out of the U.S. Navy, that pushes threat information and routing recommendations down to commercial operators. It has existed in various forms for decades. Its output is not an order. Its output is an email with coordinates and a time band.
That distinction matters more than it appears. A navy that escorts a tanker has accepted a protection obligation and a legal exposure. A navy that emails a recommended window has accepted neither. It has accepted only the burden of being useful.
The historical reference point is Operation Earnest Will in 1987 and 1988, when the United States reflagged eleven Kuwaiti tankers and physically escorted them through the Gulf during the Iran-Iraq tanker war. That was convoy. Warships alongside hulls. It ended with the USS Stark hit by an Exocet and the USS Samuel B. Roberts nearly sunk by a mine, and it produced the largest single-day surface engagement the U.S. Navy has fought since World War II.
What the FT describes is not that. What the FT describes is convoy-lite: no escort line, no physical proximity, no obligation. Instead, a schedule that concentrates traffic into a defendable envelope and lets an air-defense umbrella exist somewhere overhead. The protection is real. The commitment is not.
There is a phrase for this in protocol design. It is called a soft guarantee. You get the benefit of coordination without the cost of a covenant. And soft guarantees are the most common lie in this industry.
4. The four-month tell
The detail I keep returning to is not the windows. It is the timeline. The FT reports the U.S. presence on the southern, Omani side of the strait dating to May. From May to early September is roughly four months of continuous posture.
Four months is not a surge. Four months is a rotation cycle. It means logistics, relief crews, spare parts, munitions resupply, and a sustainment plan that somebody has already built and somebody else has already budgeted. Continuous presence converts a crisis response into a standing cost.
I have seen this exact shape on-chain. A protocol that runs a temporary multisig for two weeks is a project. A protocol that runs the same multisig for four months has a governance model, whether or not it admits it. Duration is the tell. Structure follows duration, always.
So: the United States is now operating a standing maritime coordination apparatus in the world's most important oil chokepoint, with a shrinking transit window and no formal escort obligation. That is not a headline. That is an architecture.
5. Chokepoint physics: a fee market is a strait
Here is the translation. Follow it closely, because everything downstream depends on this mapping.
A block is a lane. Block gas limit is lane width. Pending transactions are the queue of hulls waiting to transit. The base fee under EIP-1559 is a congestion price that rises exactly until demand fits inside the lane, and falls exactly until it does not. The base fee is not a fee. It is a capacity signal wearing a fee's clothing.
Now look at Hormuz. You cannot widen the lane. You can only raise the price of access until the number of hulls waiting fits inside the number of hulls the protection envelope can cover. The price of access here is not dollars per barrel. It is composed of three cheaper, nastier things:
โ Waiting time. A tanker steaming in circles off Fujairah burns fuel and crew hours and earns nothing. โ War-risk insurance. Quoted in basis points per voyage, repriced on incident and on rumor. โ Risk cost. Crew safety, hull exposure, cargo contamination, the intangible that no charterer prices correctly until the first hull burns.
When I built my yield dashboard during the 2020 DeFi summer, tracking Uniswap V2 pools and SushiSwap incentives across more than fifty strategies, the single most useful thing I learned had nothing to do with yield. It was that gas cost is a capacity signal, not a line item. Strategies that looked profitable at 40 gwei were insolvent at 300 gwei, not because the fee ate the return, but because the fee revealed that the lane could not hold everyone. The strategies that survived were the ones that could wait.
Waiting is a strategy. NCAGS has just told the entire tanker fleet that waiting is now the strategy. Follow the gas, not the hype.
6. The blob window: three capacity steps and a flat fee curve
Now the on-chain side, with real parameters.
EIP-4844 shipped on March 13, 2024. It introduced blobs: 128 kilobytes each, 131,072 blob gas per blob, a target of 3 blobs per block and a maximum of 6. Nominal sustained data capacity at twelve-second slots landed somewhere near 0.375 megabytes per second. Rollup fees collapsed. Base fees for rollups went to fractions of a cent and stayed there.
Pectra, May 7, 2025, raised the target to 6 blobs and the maximum to 9 through EIP-7691.
Fusaka, December 2025, deployed PeerDAS via EIP-7594 and pushed the target to 14 blobs and the maximum to 21.
Three capacity steps in under two years. Target blob count went from 3 to 14. That is a 4.7x expansion.
And the blob base fee spent most of that period pinned at its floor of 1 wei.
Sit with that for a moment. The single most aggressive data-availability capacity expansion in the history of a general-purpose blockchain produced a fee that rounds to zero for most of its life. Rollups advertise transaction costs in ten-thousandths of a dollar. Users treat cheap blockspace as a law of nature.
It is not. It is a price at the bottom of a curve that has a shape.
Here is the arithmetic, using round numbers and an ETH price of $3,400, so the direction is visible even if the inputs drift.
โ 1 blob = 131,072 blob gas. โ At 1 wei per blob gas, cost โ 0. โ At 10 gwei per blob gas, cost = 1,310,720 gwei = 0.00131 ETH โ $4.45 per blob. โ At 100 gwei per blob gas, cost โ $44.50 per blob. โ At 1,000 gwei per blob gas, cost โ $445 per blob.
A compressed rollup batch carries somewhere between one thousand and three thousand L2 transactions inside a single blob, depending on the rollup, the calldata mix, and how well the sequencer compresses. So a blob at $445 is somewhere between $0.15 and $0.45 per L2 transaction in data cost alone โ before L1 execution cost, before L2 execution cost, before the sequencer margin.
That is not a rounding error. That is a doubling, and then a doubling.
I am not going to argue the forecast. I am going to lay two lines next to each other and let the reader extend them. Line one: capacity steps arrive in lumps, roughly every twelve to eighteen months, typically between 2x and 2.3x. Line two: rollup data consumption has grown faster than the capacity steps in most quarters since 4844 deployed, with the gaps absorbed by the fee floor rather than by the fee curve.
Code is law; logic is leverage. The logic here is that a fee floor is not slack. It is deferred demand waiting for a reason to clear.
7. Convoy-lite is batching, and batching clusters the target
The most interesting thing in the FT report is not the restriction. It is the side effect.
When you compress all transits into two windows, the strait does not become safer per hull. It becomes safer per hour and much more dangerous per window. In the off-hours the water is empty. In the two windows it is crowded. You have not eliminated the target. You have manufactured a target cluster.
Every throughput engineer learns this the hard way. Concentration improves the efficiency of the defender and the productivity of the attacker at the same time, because both are optimizing against the same variable, which is the number of valuable objects in one place at one time.
Protocols have their own version of this and they keep rediscovering it.
โ Batching. Force four hundred user intents into one blob and you have one failure domain instead of four hundred. Sequencer censorship, a reorg, a bad compression pass, a prover outage โ the blast radius is a single object. The users who benefit from cheap batching are the same users who fail together. โ Snapshot windows. Every points program, every airdrop eligibility cutoff, every retroactive funding round creates a timestamp that thousands of wallets optimize around. The chain reacts by pricing that block at a premium and by producing a dense, predictable, extractable block. โ Exit intent concentration. My 2021 NFT work said the same thing in a different market. Twelve hundred wallets held the floor. Volume was noise. What predicted the correction was that a large share of those wallets shared a mental stop level, and stops cluster. When a chokepoint operator publishes a schedule, they are publishing a stop level.
The 2022 Anchor audit gave me the sharpest version of this lesson. I pulled the reserve composition and found a $4.1 billion gap between reported TVL and actual stablecoin collateral. The interesting part was never the gap. It was that the gap existed quietly for months and only became a number when the exit queue formed. Concentration is invisible until it is fatal.
A safety window is a published manifest. Predictability is the single most valuable gift you can hand an adversary, and every batching mechanism ships with one.
8. The paradox the report does not resolve
There is a logical contradiction buried in the FT account, and I want it on the record because it is the hinge of the whole thing.
If Iranian attacks are concentrated at night, the rational response is to push traffic into daylight. Instead the response is two fixed windows per day. That is a partial daylight bias at best. It is not a night ban.
Why not ban night transit outright? Because a night ban would cut effective throughput far more than two windows do, and Hormuz throughput is a global inflation input. The United States is being forced to trade safety against flow. That is the same trade every protocol makes when it throttles the block gas limit or gates a migration window, and it is always resolved the same way: you cut capacity as little as you can get away with, and you call the residual risk acceptable.
Which means the residual risk is not acceptable. It is merely budgeted.
9. Iran's on-chain workaround
This is where the headline stops being a headline and becomes a dataset.
Iran is a sanctioned economy with a large energy surplus and a hard-currency problem. Bitcoin mining solves all three at once. Mining converts subsidized electricity into a bearer asset that settles outside the dollar system, has no counterparty, and can be sold into any venue that does not ask questions. Iran formalized mining as a licensed industry in 2019. Estimates of its share of global hashrate have ranged from roughly 4% to 7% at the peak, and I want to be blunt about those numbers: hashrate attribution is a guess built on top of a guess. You are inferring geography from pool tags, timing signatures, and power-market inference. The error bars are wide enough to park a VLCC in.
What is not a guess is the settlement rail. Iran-linked flows that cannot clear through SWIFT overwhelmingly use USDT, and they overwhelmingly use it on Tron, because Tron's fee structure makes small transfers economic and because for years its monitoring lagged Ethereum's. Tron carries the majority of USDT's circulating supply, north of $60 billion by mid-decade. Some meaningful fraction of the velocity in that supply is sanctioned-adjacent flow, which is why the enforcement apparatus has been pointed at it.
The enforcement response has been steady. Treasury designations against Iranian exchanges and mining operators. Stablecoin issuers freezing addresses, with cumulative freezes crossing well past $2 billion by the issuer's own disclosures and the pace accelerating rather than slowing.
Here is the structural point, and it is the one I want on the desk of anyone who thinks a strait crisis is a crypto story in the bullish direction:
When you squeeze a physical chokepoint, the settlement rail migrates. When you squeeze a settlement rail, the flow migrates. Both migrations are measurable, both lag the headline, and both are visible on-chain weeks before they are visible in price.
The measurable proxies are unglamorous and useful:
โ Stablecoin net issuance split by chain. A widening Tron-versus-Ethereum issuance delta is a sanctions-pressure gauge. โ Mining pool payout geography. Newly minted coin moving through pools with known regional concentration tells you where subsidized power is being monetized. โ Exchange deposit address turnover. When an exchange's inbound deposit clusters rotate faster than its outbound clusters, the client base is changing shape.
None of this is exciting. None of this trends. It is the difference between reporting the story and reading it.
10. Insurance: the war-risk premium and the thinness of the on-chain layer
War-risk insurance is the cleanest price signal in the entire Hormuz complex, because it is quoted, it is repriced on event, and it is written by people whose job is to be right rather than interesting.
Gulf transit war-risk premiums are quoted in basis points of hull value per voyage. In calm periods they sit in the low single digits. On incident they jump by multiples, and the jump is not linear in damage โ it is a function of perceived tail risk, which is a function of how many underwriters are still willing to write the line at all.
Which brings me to the on-chain insurance layer, and to a number I think most of this industry has not internalized.
Total active capital sitting in decentralized cover markets โ the aggregates across the major cover protocols and the newer parametric structures โ runs to the low hundreds of millions of dollars. Call it $300 to $500 million in aggregate active capacity on a generous reading, with wide error bars and heavy double-counting risk because several of these markets reinsure each other.
A single modern VLCC is worth roughly $100 to $130 million.
Do the division. The entire on-chain insurance layer of this industry is worth about three to four tankers. Not three to four hundred. Three to four.
That is the information gain I want a C-suite reader to take away from this piece: crypto has built a spectacular settlement layer and a spectacularly thin protection layer, and the protection layer is what determines whether you survive a chokepoint event. Reporting TVL is easy. Reporting underwritten capacity is the actual diligence. The 2022 Anchor audit taught me that reported TVL and real collateral diverge precisely when someone needs them not to, and a war-risk premium with no capacity behind it is the same fiction wearing a better suit.
11. Custody concentration, and the recommendation that functions as a mandate
In 2025 I led an address-clustering pass over spot bitcoin ETF creation and redemption flow, trying to answer a narrow question: where does the metal actually sit, and how many independent hands hold it.
The answer was uncomfortable. Roughly 65% of measured institutional inflow traceable through deposit-address clustering resolved to three custody clusters โ two in the New York metro area and one in Singapore. Method notes, because I insist on them: I clustered on deposit-address reuse, hot-wallet fan-out patterns, and timing signatures against known custodian infrastructure, then validated against public attestations where available. The method cannot see internal segregation, cannot distinguish sub-accounts, and cannot see anything that never touches a chain-visible address. Treat the number as directionally robust and dimensionally soft.
Directionally, the conclusion is the same one NCAGS just handed the shipping industry.
Three lanes is a chokepoint. It does not matter how wide each lane is if they all share a counterparty, a jurisdiction, a legal regime, or a set of power and cooling dependencies. Three custodians is not diversification. It is a single point of failure with three mailing addresses.
Now back to that sentence from the position sheet. The guidance is advisory, not mandatory.
The reason that sentence is the most important one in the FT report is that it reveals how control actually works at chokepoints. The United States does not need to mandate transit windows. It needs the insurer to price the un-windowed voyage differently, and the charterer to refuse the un-windowed voyage, and the master to prefer the windowed one. A recommendation that everyone treats as binding is the cheapest possible form of control. It costs nothing to issue, nothing to defend legally, and it preserves the option to disclaim.
This is the same mechanism I have watched at the Securities and Exchange Commission for the better part of a decade. Pull the count of crypto-related enforcement actions against the count of new substantive rules that told an issuer how, specifically, to register. The ratio is lopsided enough that the only coherent reading of the pattern is a revealed preference for discretion over a rulebook. Ambiguity is optionality. Regulators who wanted clarity would have written clarity.
The crypto industry keeps asking for a rulebook as if the absence of one were a bug. At a chokepoint, the absence of a rulebook is the product.
12. Contrarian: correlation is not causation, and the strait does not move bitcoin
The consensus trade is simple. Geopolitical shock, oil up, safe-haven bid, bitcoin up. Digital gold, finally vindicated.
I have run enough event studies to distrust that reflex, and I want to name the specific error. The strait does not move bitcoin. The strait moves the dollar and the front end of the curve, and bitcoin follows the dollar.
Look at the four prior chokepoint shocks people cite. The Abqaiq strike in September 2019. The Soleimani strike in January 2020. The Ukraine invasion in February 2022. The Red Sea shipping disruption from late 2023. In each case, bitcoin's seventy-two-hour path was dominated by whatever the concurrent macro state happened to be. In 2020, an oil spike landed inside a liquidity flood and bitcoin rallied. In 2022, an oil spike landed inside a rate shock and bitcoin fell with the long end of the curve. Same geopolitical input, opposite outputs, because the geopolitical input was never the load-bearing variable.
Sample size is four. Event studies at n=4 are not evidence; they are anecdotes with error bars attached. I am stating the confidence as low and I am stating the mechanism as plausible, and anyone who tells you otherwise is selling a narrative, not a model.
Here is the second contrarian point, and it is the one that actually costs people money.
Bitcoin carries no war-risk premium, because bitcoin has no delivery obligation. Oil carries a war-risk premium because a barrel that does not arrive is a contract that does not settle. These are different instruments exposed to different risks. Treating bitcoin as a hedge against a physical chokepoint is a category error, and the error is durable because it feels sophisticated.
The genuinely contrarian position is duller and more useful: the crypto-relevant variable in a Hormuz crisis is not the price of bitcoin, it is the cost of blockspace and the thickness of the settlement rail. A strait crisis raises energy costs, which raises marginal mining cost, which pushes some hashrate offline, which compresses margin for the operators that remain. It raises the probability of sanctions escalation, which thickens or thins the stablecoin rail depending on which direction enforcement moves. It raises the war-risk premium, which has no on-chain equivalent with enough capacity to matter.
Whales don't care about your feelings. Whales care about whether the rail is open and what the rail costs.
13. Contrarian: the limited-commitment signal is the real story
The second-order effect of "advisory, not mandatory" is a message sent to the other side of the table, and it is not the message the sender intended.
A commitment that can be disclaimed is a commitment of limited depth. Iran reads that the same way a sophisticated counterparty reads a founder who says the sequencer will be decentralized eventually. The roadmap is real. The deadline is not. The optionality is the point.
Protocol teams have industrialized this pattern. Progressive decentralization. Temporary multisig. Permissioned launch with a credible path to permissionlessness. Every one of those phrases is a promise with a disclaimer welded to the side, and markets price the disclaimer long before they price the promise.
There is a second-order risk in the other direction too, and it is the one the FT report does not explore. If the adversary reads limited commitment as weak resolve, restraint on their side falls. Not because they want war, but because the expected cost of probing went down. Signals of limited commitment are not always de-escalatory. Sometimes they are an invitation to test the boundary.
That is the same dynamic I have watched in token markets for a decade. A team that signals it will not defend a peg, a floor, or a buyback gets exactly the test it implied it would not resist. Ambiguity in protection is not neutral. It is a directional bet on the other side's temperament, and you are not the one setting the odds.
14. What I am not claiming
The FT report is single-sourced, based on quoted internal coordination emails, and does not include an Iranian response, a formal U.S. policy statement, or force-composition data. Operational facts about the windows carry higher confidence. Strategic intent inferences on both sides carry medium-to-low confidence. I am not going to convert an inference into a forecast by writing it in the imperative.
Where I have used numbers from my own pulls, I have flagged the method and the softness. Where I have extended a protocol parameter into a cost, I have shown the arithmetic so it can be checked and corrected. Where the sample is four, I have said four.
The report is a signal, not a conclusion. Signals are what I do.
15. Takeaway: the signal set for next week
No summary. A watchlist with thresholds, because a prediction without a threshold is a horoscope.
Blob base fee, weekly median. Currently pinned near the 1 wei floor. The moment the weekly median clears 1 gwei on any sustained basis, the data-availability market is finally pricing capacity instead of giving it away. That is the leading indicator for every rollup's unit economics, and it will move before any rollup publishes a fee change.
Blob occupancy at the ceiling. Track the share of blocks hitting the maximum blob count rather than the target. Pre-Fusaka that was 9; post-Fusaka it is 21. Occupancy at the ceiling more than roughly 5% of blocks is the mechanical precursor to base-fee repricing.
Stablecoin issuance split, Tron versus Ethereum, seven-day delta. A widening gap toward Tron during a Gulf escalation is a sanctions-pressure read, not a growth read. Do not confuse the two.
War-risk premium quotes, in basis points per Gulf voyage. The off-chain counterpart to the on-chain fee curve, and the least manipulated price in the complex. A step change here leads the headline by days.
Transit window count. Two is the current state. One is a warning. Zero is a different article.
U.S. force posture in the southern Gulf. The measurable variable is whether a strike group or amphibious ready group is added to a posture that has already held for four months. Additions are cheap to announce and expensive to sustain, so they are high-signal.
Hashrate from pools with regional concentration. Slow, noisy, wide error bars, and directionally informative over four to six weeks rather than four to six days.
Dollar index and two-year yields. The variables that actually drive bitcoin in a seventy-two-hour geopolitical window. If you are trading the strait through crypto beta, you are trading these two instruments with extra steps.
One closing thought, offered as a question rather than a verdict, because I would rather be useful than certain.
Every chokepoint operator in history has eventually told the same story: the restriction is temporary, the schedule is for your safety, and normal service will resume shortly. Hormuz is running that script now. Blob space ran it for two years at a price of almost nothing. The interesting question is not whether the windows reopen. It is whether the market has begun pricing chokepoint risk in the only units that transfer โ capacity and cost โ instead of in the units that do not, which are narrative and hope.
Follow the gas, not the hype.