There is a number that connects the tower that collapsed at Iran's Chabahar port early Friday and the digits on a gas-station sign in a Pennsylvania exurb. Nobody prints it. Everybody trades on it.

Overnight into Friday, the United States expanded its air campaign against Iran for a sixth consecutive night, hitting bridges in the southern Hormozgan province and collapsing a tower at the Chabahar port on the Gulf of Oman. Iranian state media reported at least seven killed at Bandar-e Khamir, a city on the strait itself. It is part of President Donald Trump's stated plan to strike infrastructure until Tehran loosens its chokehold on the Strait of Hormuz. Iran answered with fresh missile fire at U.S.-allied states, including Qatar, the mediator it needs, and warned that its retaliation "will spread to new areas."

The interim ceasefire, the Islamabad Memorandum signed barely a month ago, is functionally dead. What has replaced it is not a war in the old sense of a thing that starts and stops. It is a standing condition with a price feed. Brent crude traded near $85 on Friday, up almost 12 percent on the week. War-risk insurance on a hull crossing Hormuz now runs 2 to 6 percent of a ship's value, repriced every 24 to 48 hours on seven-day terms, a number that used to sit at a quarter of one percent. On a $100 million tanker, a single transit can cost $6 million in cover alone. That is not a market anomaly. That is a toll, collected by underwriters on behalf of a war.

Here is where the geometry closes. The pump price is the toll passed downstream to a voter. Gasoline has climbed roughly ninety cents a gallon nationally since the war opened on February 28, brushing $4. Reuters/Ipsos polling has 77 percent of registered voters assigning Trump at least a fair share of the blame, and 58 percent saying they are less likely to back candidates who support his Iran approach. The president's own advisers, per the Wall Street Journal, are privately racing to end the war before the fuel-cost surge reaches November's ballots. Airfares are up because jet fuel roughly doubled. The IEA's Fatih Birol calls global oil security "fragile" out loud.

So the war has two clocks running at once, and they are not synchronized. The military clock is measured in nights, and it wants to keep striking until Iran folds. The political clock is measured in months, and it wants prices down by autumn. Trump says the Revolutionary Guard has lost up to 90 percent of its weapons capability. Iran says the Strait "will never return to how it was before." Both statements can be true, and if they are, the campaign that is winning tactically is the one bleeding the party that ordered it.

This is what makes the moment worth watching rather than merely dreading. A great power has discovered that it can win a war and lose the arithmetic. Every bridge dropped in Hormozgan tightens a supply line that ends at a swing-district gas station, and the analysts León at Rystad and others keep noting the same thing: the U.S. wants cheap oil before the midterms and Iran wants its frozen assets back, which means both sides have an incentive to stop that has nothing to do with who is right. The likeliest path to peace here is not diplomacy or victory. It is a spreadsheet.

We are choosing to lead with the meter instead of the missiles because the missiles have led the front page of every paper for a week, and the meter is the part almost nobody is naming plainly: that the pressure campaign against Tehran is, functionally, a pressure campaign against the administration's own electorate, and the White House is the party that will crack first. Watch the pump, not the podium. The pump is where this war is actually being fought, and it is the only front on which Iran is currently winning.

If you have a Hormuz story, a stranded relative on a tanker, a fuel bill that broke your month, write to us. This is the number we intend to keep.

U.S. Central Command opened its sixth straight night of strikes at 6 p.m. GMT Thursday, describing the wave as an effort to "further degrade Iranian military capabilities." By Friday morning the targets were physical geography: bridges near Bandar-e Khamir, roughly ten sites on Qeshm Island, targets at Bushehr and Ahvaz, and a collapsed tower at Chabahar port on the Gulf of Oman, a lifeline for landlocked neighbors. Iranian officials put the week's dead above 35 and the wounded above 300; Fars reported seven killed at the Bandar-e Khamir bridge alone. A strike near Ahvaz forced the evacuation of more than 200 patients from a children's cancer hospital, a detail Tehran has pushed hard and Washington has not addressed.

The strait fight is now spilling toward a second one. Three sources told Reuters that Iran's leadership has asked the Houthis to stand ready to close the Red Sea oil route, the Bab el-Mandeb, if the U.S. strikes Iranian power infrastructure. With Hormuz already throttled, a simultaneous Red Sea closure would pinch both of the Middle East's principal export arteries at once. The U.S. Navy-led Joint Maritime Information Center has already raised its Hormuz advisory to "severe," its top tier, for the first time since mid-June, citing the likelihood of "deliberate hostile action."

CENTCOM says it turned back three commercial ships in the latest window, disabled one that refused to comply, and boarded another in the Gulf of Oman. Lloyd's List Intelligence reports that no large vessel has transited the U.S.-coordinated Omani coastal route with its transponder on since July 7. The human residue of that statistic: the UN counts roughly 6,000 seafarers stranded aboard vessels in the channel, crews who signed onto a shipping contract and found themselves inside a blockade.

The diplomatic track is not dead so much as sidelined. China's Wang Yi and Pakistan's Ishaq Dar met in Shanghai and jointly called for a ceasefire and a return to the memorandum's terms. The White House says Trump will hold Iran "accountable" but is "always open to diplomacy." What is missing is any mechanism to make either side stop, because the memorandum that was supposed to be that mechanism is the thing that just broke.

Ignore the Brent number for a second. The commodity screen is the crowd's guess. The truest price of this war is set in a room in London by people who are legally obligated to be pessimists.

War-risk insurance on a vessel crossing the Strait of Hormuz is now quoted at 2 to 6 percent of the ship's value, up from a pre-war baseline near 0.25 percent, according to brokers at Marsh and underwriters cited by Insurance Business. Some cargo-value estimates run 3 to 8 percent, a 30- to 80-fold jump. The cover is sold in seven-day slices and re-underwritten every 24 to 48 hours, which means the market's view of the odds of a hull being hit is refreshed twice daily and denominated in cash. At the top of the range, a single transit for a $100 million tanker carries a $6 million premium. That is the war, expressed as a number that someone actually has to pay before the ship moves.

What the insurance market is telling you that the oil market is not: this is not being priced as a spike expected to pass. Underwriters have begun advising some owners to simply pause Hormuz voyages. The IMO's secretary-general, Arsenio Dominguez, is publicly asking governments to lean on insurers so premiums "reflect current realities, rather than continuing to reflect the peak of the crisis" — a polite way of saying the market has priced in permanence and he wishes it wouldn't. When the people whose job is to model catastrophe stop treating a war as temporary, the temporary war is over.

There is a grim elegance to how this cost cascades. The premium is a per-transit tax; the tax lifts delivered energy prices 3 to 8 percent for European utilities, per shipping analysts, which makes long-dated supply contracts economically incoherent, which pushes buyers toward diversification, which is the polite name for permanent structural change. A seven-day insurance policy is quietly rewriting the geography of global energy, one 48-hour repricing at a time.

The lesson for anyone reading past the crude ticker: in a slow-motion crisis, watch the instrument with the shortest duration. It has the fewest reasons to lie.

On June 16, SpaceX exercised an option to acquire Anysphere, the maker of the Cursor coding tool, in an all-stock deal valued at $60 billion, the largest acquisition of a venture-backed startup on record. The strategic story writes itself, and that is exactly the problem: SpaceX, having merged with Musk's xAI in February, gets a stack spanning model (Grok), coding surface (Cursor), and compute (the Colossus cluster). Vertical integration. Clean slide.

Now read the mechanics. The option was signed April 21, with a roughly $10 billion break-up fee if SpaceX walked. Anysphere shareholders receive SpaceX Class A shares priced on a seven-day volume-weighted average before close, expected in Q3. And the currency doing the buying is a stock that went public barely a week before the deal and ran from a $135 IPO price toward $210, adding close to $1 trillion in market value in days. In the founder's own arithmetic, that is "roughly 16 Cursors" of paper wealth created faster than the thing being bought is worth. When your stock adds sixteen targets in a week, spending one of them is nearly free.

The deal also quietly admits a failure. Grok, per multiple accounts, has not competed effectively against Claude Code, Codex, or Cursor in the coding market that was a centerpiece of SpaceX's IPO promises, and the AI division has been restructuring after a string of controversies. Buying the incumbent you couldn't beat is a legitimate strategy. It is not the same story as building the winner, and the vertical-integration framing exists to blur that distinction.

The part worth watching is the collision course inside the portfolio. Grok Build is a terminal-native agent; Cursor is IDE-native. SpaceX now owns both and will have to either divide the territory or consume one into the other. Cursor's users, roughly the entire professional developer class that adopted it, are about to learn whether they bought a tool or a hostage. The $60 billion headline is the least interesting number in the filing. The interesting one is how many of those users are still there a year after a rocket company decides what an editor is for.

Governor Kathy Hochul's executive order this week made New York the first state to pause new hyperscale data centers, freezing environmental permits for facilities over 50 megawatts for up to a year while the state writes rules. The headline is a moratorium. The event underneath it is a price signal.

"The headline is a ban," a PitchBook-cited investor, Rolfes, put it. "The actual event is that time-to-power just became a valuable commodity in the AI stack. A state legislature, not a chip shortage, is now setting the price." That reframing matters because the physical constraint was already biting: of the roughly 241 gigawatts of U.S. data-center capacity in the pipeline by end of 2025, only about a third is under active development, per Wood Mackenzie, the rest stalled on power. New York alone is sitting on nearly 12 gigawatts of load requests, more than 8 of them added in 2025.

During the pause, the state's Department of Public Service will build a Generic Environmental Impact Statement to weigh water, grid, and air-quality effects, and is weighing a fund operators might pay into to shield ratepayers from stranded grid costs. On a 60-day clock, a Community Investment Framework will set baseline expectations for what operators owe host towns. Hochul also wants to repeal sales-tax exemptions for large data centers. In other words: the state is not saying no. It is assembling an invoice.

And New York is not alone, which is the part that should worry Washington more than any single order. Twenty-seven states are advancing large-load legislation; California, Ohio, and Utah have already enacted laws going beyond the federal government's voluntary, unenforceable Ratepayer Protection Pledge. Maine nearly beat everyone to a moratorium before a gubernatorial veto. The federal posture is acceleration; the ground truth is a patchwork of states, red and blue alike, deciding that the grid is theirs to protect. The AI build-out spent two years treating electricity as a solved input. A dozen statehouses just turned it back into the binding one.

OpenAI moved its GPT-5.6 family to general availability on July 9 after a government-coordinated preview, in three durable tiers: Sol at the top ($5/$30 per million input/output tokens), Terra in the middle, Luna at the floor ($1/$6). Read the launch language closely and the interesting admission is what it doesn't claim. OpenAI describes Sol as "competitive with" Anthropic's restricted Mythos tier "using only ~1/3 of the output tokens" on a cybersecurity benchmark. Competitive with, at a third the tokens. Not better than. The pitch has shifted from intelligence to efficiency, from the smartest answer to the same answer for less money.

That is a meaningful turn. Anthropic's Claude Fable 5, out since June 9, remains the model rivals measure against, with an independently confirmed 95 percent on SWE-bench Verified. Rather than try to top that number, OpenAI is competing on performance-per-dollar and introducing an "ultra" setting that coordinates multiple agents across parallel workstreams. The whole framing, tiers you route between by task, concedes that not every job needs the frontier and that the real product is knowing which model to spend on.

Meanwhile each lab is placing a different bet, and they no longer rhyme. Meta is going cheap and embedded, dropping models straight into WhatsApp and Instagram. SpaceX's xAI is building a specialist for developers, which is precisely why it just spent $60 billion on Cursor. OpenAI is chasing the efficient generalist and shipping GPT-Live voice models that talk and listen at once. The era where everyone competed on a single leaderboard number is closing. The new competition is over who owns the surface where the model actually gets used, and how little that use has to cost. "Smartest" was a benchmark. "Cheapest good enough" is a business.

The critics called Antoine Fuqua's Michael "mixed," which is the polite register for a shrug. Audiences didn't care. Over the weekend of July 12–13 the film crossed $1 billion worldwide, $371.8 million domestic and $629.8 million abroad, becoming the first biographical film of any subject in cinema history to reach the mark, past Oppenheimer, past Bohemian Rhapsody, the biggest release in Lionsgate's 29-year existence. That is the story everyone is telling. It is the boring half.

The interesting half is what the movie is actually for. In the weeks after its April premiere, Michael Jackson's Spotify monthly listeners went from about 68 million to roughly 105 million. His top 60 tracks jumped from 467 million monthly plays to 1.21 billion, an extra 748 million streams in under four weeks. He added nearly 14 million permanent followers. The film, in other words, functioned as the most effective catalog-marketing campaign ever run, and it did it while charging admission. For the Jackson estate and Sony's publishing arm, the billion-dollar gross is almost a rounding detail against a permanent uplift to the value of the underlying rights, one that cost them nothing.

This is the model now, and it should unsettle anyone who still thinks a biopic is a story about a person. The same weekend Michael cleared a billion, Disney's live-action Moana opened 35 percent below projection and toward a nine-figure loss, because Disney owns the streaming service that already gives you the original for free and Lionsgate owns no such thing competing with a Jackson catalog. The film that wins is the one that drives you back to an asset the studio can meter forever.

So we have arrived somewhere strange: the year's largest cultural event is a two-hour advertisement for songs that are forty years old, made about a man whose reputation the film mostly declines to litigate, and it worked precisely because it treated the catalog as the product and the person as the packaging. The box office is booming, on track for its first $10 billion year since 2019. Some of that is Nolan's Odyssey and a healthy slate. But the tentpole of the moment is a liquidation event with a soundtrack, and audiences forwarded it to each other by the tens of millions. The ghost outsold the living. Ask yourself what that says about what we go to the movies to buy.

The interim deal is dead, and everyone is performing surprise at a funeral they scheduled. The Islamabad Memorandum was signed a month ago and collapsed inside twenty-six days, and if you go back and read what it actually promised, the wonder is that it lasted that long. It was not a settlement. It was a device for letting two governments each tell their own domestic audience that the bill had been deferred.

Consider what an honest agreement would have had to say out loud. It would have had to admit that Iran regards its ability to threaten the Strait of Hormuz as its single most valuable strategic asset and will not surrender it for a waiver package it doesn't trust Washington to honor past the next news cycle. It would have had to admit that the United States cannot both bomb Iranian infrastructure into compliance and keep gasoline under $4 for a midterm electorate, because those are the same barrel of oil pointed in opposite directions. It would have had to name the price of permanence: that a militarized Hormuz is now a standing feature of the world, priced by insurers at 2 to 6 percent of a hull and by voters at ninety cents a gallon, and that no memorandum makes that number disappear.

Instead the parties signed a placeholder, called it peace, and went back to the thing they were always going to do. This is a pattern, not an accident. The ceasefire, the sixty-day understanding, the sanctions waiver, the pledge to swap tolls for trade — all of them were IOUs written against a reconciliation nobody has any intention of funding. A placeholder is a promise to argue later. The trouble is that the placeholders are all expiring at once, and the argument they deferred is arriving with interest.

Here is the contrarian position, and I hold it firmly: the collapse is more honest than the deal. The memorandum was a sedative. Its death at least forces the real question onto the table, which is not "how do we restore the ceasefire" but "what does a permanent, priced, semi-militarized Hormuz actually cost, and who agrees to pay it on purpose." That is a hard conversation. It is also the only one worth having, and every interim agreement is a way of not having it for another twenty-six days.

The likeliest thing that ends this is not a better memorandum. It is a spreadsheet in the West Wing showing gas prices and a calendar showing November, and Tehran reading the same spreadsheet from the other side. That is a grubby basis for peace. But it is at least a real one, denominated in the currency both sides actually respect. Stop signing ceasefires you intend to break. Sign the invoice, or don't, but stop pretending there's a version of this without one.

The Valley's response to New York's data-center moratorium was as predictable as it was wrong: a regulator has thrown sand in the gears of progress, a legislature is now dumb enough to price a chip shortage, innovation flees to friendlier states. Every word of that framing assumes the thing it needs to prove, which is that the pre-moratorium arrangement was a functioning market. It wasn't. It was a subsidy dressed as a build-out.

Strip the rhetoric and look at the mechanics of what a hyperscale campus was actually doing. It arrived in a jurisdiction, drew tens or hundreds of megawatts off a shared grid, frequently collected a sales-tax exemption for the privilege, and left the cost of grid upgrades and the risk of stranded infrastructure to be socialized across ratepayers who never voted for a server farm. When a business consumes a scarce shared resource and pushes the cost onto third parties, an economist has a word for it, and the word is not innovation. It is an externality. New York's order is a machine for internalizing it: a fund to protect ratepayers from stranded costs, a repeal of the tax break, a requirement that operators contribute to the towns they burden. That is not the market failing. That is the market finally being made to clear.

Notice, too, who is doing this. Twenty-seven states, red and blue, are moving on large-load rules; California, Ohio, and Utah already have laws with teeth the federal Ratepayer Protection Pledge conspicuously lacks. When resistance to a policy spans the entire ideological spectrum, the honest read is not that a mob has formed. It is that a genuine cost was being hidden and constituents across the map independently found it on their electric bills. The federal pledge is voluntary and unenforceable precisely because the industry knew a real price would change the math, and the states are calling that bluff.

Here is the part the accelerationists should actually welcome if they believe their own religion about markets. The investor who called "time-to-power" the new scarce commodity is right, and pricing a scarce commodity is what markets are for. A moratorium that ends with a real cost of power attached to a permit does not kill AI infrastructure. It kills the speculative, half-funded, never-going-to-be-built third of the pipeline that was clogging the grid on the strength of a subsidy, and it directs capital to projects that can pay their own way. If your business model only works when someone else covers your electricity, you do not have a market objection to regulation. You have a confession.