The US war with Iran

Every IAEA report from February 2017 to May 2019 found the stock under the 2015 deal's limit and enrichment no higher than 3.67%. In August 2018 the IAEA said it had made complementary access visits "to all the sites and locations in Iran which it needed to visit".

In 2016 Iran slightly exceeded the deal's cap on heavy water, a material that is not enriched uranium, and was back within the limit by December.

The US left the deal in May 2018. A year later Iran said it would no longer be bound by the stock limits, and in July 2019 the IAEA confirmed the limit had been exceeded. A week later Iran began enriching above the deal's cap.

From there the stock rose in almost every quarter.

The total fell in four reports between 2021 and 2024. For August 2021 the IAEA gives the reason: uranium enriched to 2% was used "to produce uranium enriched up to 5%". The other reports record the falls without explaining them, so a dip in this line is not, on its own, a sign that Iran slowed down.

From February 2021 the IAEA lost access to its monitoring data, and from May 2021 it could no longer verify the total; each later figure is its estimate from Iran's declarations and partial checks. The final figure, for 13 June 2025, also rests on "the past operating records" of Iran's facilities.

Iran began enriching to 60% in April 2021 and built the stock slowly for three years. It then more than doubled between October 2024 and May 2025.

The IAEA's last estimate, for the day the strikes began, is 440.9 kg of 60% uranium.

It had verified 432.9 kg of that. As of September 2026 it "cannot provide any information on the current size, composition or whereabouts" of the stock.

Grossi said after the June 2025 strikes that most of it, "but less than 70 percent", was in the tunnels at Isfahan, as ISIS reports him. Iran's foreign minister said in March 2026 that it is "under the rubble" and Iran has "no plan" to recover it.

In May 2025 the IAEA called Iran "the only non-nuclear-weapon State in the world that is producing and accumulating uranium enriched to 60%", which it said "remains a matter of serious concern". It repeated the point in June 2026.

In June 2026 the Board recalled that Iran's failures "constitute non-compliance" and said the IAEA could not verify "that there has been no diversion" of material to weapons. In September it asked for its resolutions to go to the UN Security Council and General Assembly. Each 2026 resolution passed with three votes against and at least eight abstentions.

Complementary access ended in February 2021 and has not resumed. By February 2026 the IAEA had inspected about half the facilities not hit in June 2025, and none of those that were.

The IAEA stopped verification when the US and Israel struck on 28 February 2026. Its one visit since was to the Bushehr power plant in June 2026.

The IAEA's findings of 2011, 2015 and 2025 all place organised weapons-relevant work before the end of 2003.

The 2025 report adds that Iran used about 10 kg of undeclared uranium metal in 2002–03 and kept material from the old programme at Turquzabad until 2018. Iran also "sanitised locations".

Because of this, the IAEA said in May 2025 it "will not be in a position to provide assurance" that the programme is "exclusively peaceful".

The same May 2025 report says the IAEA "has no credible indications of an ongoing, undeclared structured nuclear programme". Its 2015 assessment found nothing credible after 2009.

In March 2026 Grossi said that in June 2025, when the first strikes came, the IAEA had no "information indicating that there was a structured systematic [Iranian] program to build or to construct a nuclear weapon".

The March 2025 threat assessment judged that Iran "is not building a nuclear weapon" and that Khamenei had not restarted the programme halted in 2003. It added that pressure on him to do so had "probably built".

The March 2026 assessment says Iran was "intending to try to recover" its nuclear infrastructure before the war. As CRS puts it, the assessment "does not address Iran's nuclear weapons activities". So the most recent public US judgment on a weapons decision is the 2025 one.

The IAEA's chief attended indirect US–Iran talks in Geneva on 17 and 26 February 2026 and advised on verification. The US and Israel struck two days after the second round.

The June 2026 US–Iran memorandum states that "Iran reaffirms that it shall not procure or develop nuclear weapons".

In May 2025 the Defense Intelligence Agency judged that Iran needed "probably less than one week" to make enough weapons-grade uranium for one weapon. A November 2024 intelligence report said the stock could fuel "more than a dozen nuclear weapons" if further enriched.

In June 2026 ISIS wrote that "for the first time in 20 years, Iran has no identifiable route to produce weapon-grade uranium" in its enrichment plants. It declined to give a time estimate, which "would require unsubstantiated speculation" about centrifuges that survived.

Time to weapons-grade uranium, time to a working weapon and the size of the stock are different measures, and the table keeps them apart. Most post-strike figures are reported second-hand, through CRS or FDD; CRS concludes that "the current assessed timeline is unclear".

The Pentagon said "nearly 80%" of Iran's nuclear industrial base was "hit", not destroyed. US and Iranian officials say enrichment has not resumed, but the IAEA cannot verify whether it has. No source read here publishes evidence that settles which claim is right; CRS notes that neither the US government nor the IAEA has described evidence of covert production.

The House rejected the first three resolutions by margins of seven votes or fewer, including a tie, then adopted the next three, from June to September (chart).

Every vote was on a concurrent resolution under section 5(c) of the War Powers Resolution. The House has held no roll call of any kind since mid-September, as of 29 October (see the chart note).

A concurrent resolution is not sent to the President to sign or veto. CRS calls its effect "uncertain" since the Supreme Court's 1983 Chadha ruling, and Lawfare says it was brought into "serious constitutional doubt".

A joint resolution would be binding but faces a veto, which two-thirds of each chamber would have to override.

Republicans voting to end the war rose from two in March to seven in September, out of more than 200 Republicans voting. Democrats voting against fell to none from June on (chart).

In the Senate, Republican yes votes rose from one to four (chart below). With the House this evenly divided, these few members and absences decided each result.

Twelve of the Senate's votes failed. A motion to discharge succeeded in May, but the resolution then failed a motion to proceed. In June the Senate adopted H.Con.Res.86, which the House had passed, so both chambers adopted one resolution (chart).

The Senate rejected the next House resolution, H.Con.Res.89, by one vote in September. GovTrack's list shows no Iran war-powers vote after that, as of late October.

One Republican voted yes on each of the first five votes, and up to four from May. One Democrat voted no on every vote with a member record (chart).

The same pattern, one Republican yes and one Democratic no, appeared in the failed June 2025 Senate vote on an Iran war-powers resolution.

In Pew's surveys about a third of adults approve of the President's handling of the war, down slightly since March; Pew's separate question on whether force was the right decision moved the same way (chart).

Gallup in June and Quinnipiac in March, asking about the military action itself, also found majorities against (charts below).

Pew finds the whole decline in approval came from Republicans; Democrats' views did not change. Republican support is still a clear majority in every poll here, highest in Quinnipiac's March poll and Gallup's June poll, which exclude leaners (charts below).

Gallup's June poll found independents opposed by more than two to one, and Quinnipiac's March poll by nearly two to one (charts). Pew does not report independents separately.

In Quinnipiac's poll a week after the war began, Republicans supported the action and Democrats opposed it by similar margins; independents sat closer to Democrats (chart).

In March, 59% of registered voters said the President should have received approval from Congress before acting, including most independents and about a fifth of Republicans (chart). This is one poll; no later reading was found.

Of the War Powers reports filed from 1975 to March 2017, CRS finds only one cited the section that starts the 60-day withdrawal requirement (chart).

Each 2026 letter that could be read cites the President's Article II authority and is filed "consistent with" the Act.

On election night 2024 the President said "I'm not going to start a war. I'm going to stop wars." The remarks did not mention Iran.

In his 2024 convention speech he said Iran was "very close to having a nuclear weapon" and blamed his predecessor. Both lines were said by the same candidate in the same year.

The 1 May letter said hostilities "have terminated" because there had been no exchange of fire since the ceasefire. The blockade that began in April was still running, and from June the administration filed at least five letters reporting new military action (timeline).

Goldsmith calls the "no exchange of fire" claim "highly misleading". The 1 May letter is known only through these quotations.

Each later letter reports military action "commenced" on a new date, in response to a named Iranian or militia attack. Lawfare notes the tension between this and treating the blockade as anything but a continuation of the March operation.

The Constitution gives Congress the power "To declare War" and makes the President "Commander in Chief".

CRS counts eleven declarations of war covering five wars, and "hundreds of instances" of US forces used abroad in conflict or potential conflict. The administration did not seek authorisation for this war before the 60-day mark.

CRS does not code the instances by authorisation, length or scale (see Data still needed).

Joint resolutions on Yemen in 2019 and Iran in 2020 passed both chambers and were vetoed; neither override reached two-thirds. In 2011 the House rejected both authorising the Libya operation and limiting its funding; the 2025 Iran resolution failed in the Senate (table).

Kriner's study, as described on its publisher's page, finds that Congress "materially shapes the initiation, scope, and duration of major military actions", and that the party balance in Congress helps explain how long conflicts last. The publisher's page gives no effect sizes.

The President's own letters record Iranian attacks and US strikes during the ceasefire, before and after the memorandum (timeline). In each letter the US strikes answer a named Iranian attack.

Every entry comes from the US government or ISW/CTP. No Iranian or neutral record was found to check the sequence.

Within nine days of the signing, the President reported Iranian attacks on shipping and US strikes in reply. Strikes resumed from early July, and ISW/CTP judged that they and the renewed blockade "mark the end of the ceasefire".

The IAEA records indirect US–Iran talks in Geneva on two dates in February 2026, the second two days before the strikes. In 2025 a sixth round was suspended after Israel's attack, according to Wikipedia's account.

The June memorandum states that "Iran reaffirms that it shall not procure or develop nuclear weapons". Its other terms, including any on inspections, have not been published in any source found, so it cannot be compared clause by clause with the 2015 deal.

The President's pledge spoke of "productive discussions", and in late October Qatar, Pakistan, Saudi Arabia and Turkey were pressing a deal. Regional sources told ISW/CTP the Tehran meeting "made no progress on any issue".

The British, French and German leaders said they "did not participate", while condemning Iran's attacks on its neighbours "in the strongest terms". This list is the statements found, not a complete record of allied positions.

Oman's foreign minister describes the war as "a catastrophe" launched "without United Nations authorisation". This is his statement; no UN legal text was checked, and resolution 2817 could not be fetched.

ISW/CTP assesses that Iran "continuously made efforts to reconstitute its ballistic missile and drone programs" from the April ceasefire. In October it read the new defence minister's appointment, and his plan to rebuild "in mountains and security structures", as prioritising missiles.

These are assessments by ISW and AEI's Critical Threats Project, not imagery or official findings.

The NYT reported plans for "a short and intensive operation" after 3 November, and AP reports allies were told to prepare for renewed action and Iranian retaliation. AP's diplomats say the President has made no final decision.

AP reports no US strike on Iran since early September (as of 28 October).

ISW/CTP reports that the Director of National Intelligence warned on 23 October that Iran is in the "advanced stages" of planning attacks on US targets in the region. The memo itself could not be fetched, so this rests on ISW/CTP's account.

Officials told Bloomberg of 23 attacks on vessels near Hormuz in the first two weeks of October, and the Houthis fired missiles and drones at Saudi Arabia from 24 October (ISW/CTP). The US blockade of Iranian ports was still in force in late October.

Wounded came in two surges, in March and in July, when US strikes resumed (chart). More than half of all the wounded were recorded after the July split into a second category.

Few were wounded in the quieter months of May, June and August, and a smaller rise followed in September.

Of the US deaths DCAS records by late October, half were non-hostile, mostly Air Force accidents in March (chart).

Reports in early October gave a total of 21; DCAS has since added one more non-hostile death, in October.

HRANA counted 1,701 civilians killed by the April ceasefire, including at least 254 children. It counts incident by incident, takes minimums, and warns of undercounting.

Hengaw counts 6,620 military personnel killed, mostly Revolutionary Guards and army. Its civilians are 13.5% of its total.

Its method is not published, and it also alleges that military units moved into civilian buildings.

All three counts put deaths in the thousands by spring. HRANA's civilians are slightly under half of its total; Hengaw's are about one in seven (chart).

The government's count has no split at all, so it cannot settle the question.

HRANA's and Hengaw's counts end at the April ceasefire. After that, only the authorities' totals and Airwars' incident tallies exist.

Airwars tracked 267 civilian-harm incidents before the ceasefire and 34 more after strikes resumed in June. WHO notes the authorities have yet to publish cumulative figures.

Airwars' one-month report of at least 1,443 civilians killed is HRANA's count, not a separate one.

That is about one in every 29 people in Iran (chart). WHO calls the figure an estimate and does not say how many have returned home.

CBO put the cost at about $38 billion by August, and the Pentagon at $43.6 billion by early September (chart).

AP reports a later Pentagon figure of $47.1 billion, without its as-of date. Most of the Pentagon's September figure is replacing munitions and equipment, not operating costs.

CBO estimates each further month at $2 billion at the low intensity of May and June, and $3 billion at July's level.

CBO notes that DoD did not answer its requests for information, so it worked from public reports.

The Iran war's US deaths are about one two-hundredth of Operation Iraqi Freedom's (chart).

CBO attributes the war's lower cost to "relatively few U.S. forces" being involved compared with Iraq and Afghanistan. The comparison is of totals: Iraq ran for years, this war for eight months.

The Pentagon's latest figure is about a seventeenth of the $815 billion appropriated for Iraq through FY2014 (chart). Iraq's appropriations built up over 13 years.

In 2002 the White House budget director put the war at $50–60 billion.

Congress appropriated $815 billion for Iraq through FY2014, more than thirteen times the top of that range.

Even the high estimate the budget director called "very, very high" was a quarter of the outcome (chart).

CSIS says the war "required extensive use of interceptors", and that lower stocks "may force" the US and its partners "to take more risks with interceptions".

These are CSIS estimates built from budget documents, with deductions for testing and training, not Pentagon figures.

The Washington Institute's tracker confirmed 223 attacks on US and coalition bases in Iraq and Syria between October 2023 and December 2024 (chart).

The State Department counted "more than 100" by the end of 2023 alone. These attacks predate the February 2026 war.

About four in five of the confirmed attacks fell in the first four months, to January 2024 (chart). From February to mid-July 2024 the tracker confirmed only a handful, during what it calls a "temporary cessation". Attacks resumed at a lower rate from August.

The State Department says the October 2023 attacks "came after an 18-month general pause in attacks in Iraq and a six-month pause in attacks in Syria" (marked on the chart).

The pauses show the militias' attacks start and stop. The sources do not say whether Iran ordered the pauses or the restarts.

It says "there is no evidence that Iran knew about Hamas's October 7 attacks on Israel in advance", while its support "enabled Hamas to execute the attack".

The UN Panel found strong similarities to Iranian materiel in more inspected items than to any other source (chart). It says the Houthis receive "technical assistance, training, weapons and financial support" from Iran, Iraqi armed groups and Hezbollah.

The finding is that items are "similar" to those Iran develops or operates, not that each was shipped from Iran. Similar items are also operated by armed groups in Iraq and Lebanon.

In April 2024 the Cassation Court ruled that both Buenos Aires bombings were carried out by Hezbollah "sponsored by the government of Iran", and blamed then-president Rafsanjani and other officials. Iran denies involvement; no one has been tried in person.

Three men have been sentenced for plots against Masih Alinejad that DOJ says were directed by the Iranian government and the IRGC (timeline). In Belgium an Iranian diplomat was convicted of plotting to bomb an opposition rally.

Every US case on the timeline is a plot or conspiracy; none records a completed killing.

The alleged tasking against Donald Trump appears only in a 2024 criminal complaint against a man who remains in Iran.

Ahmad Vahidi, former Quds Force chief, is among the officials an Argentine judge charged in September 2026. He was serving as interior minister in March 2024, when he announced that year's election turnout.

Iran is the longest-listed state on the list after Syria (timeline). Whether Syria has since been removed could not be checked, because state.gov refused automated requests.

Its 2023 report calls Iran "the leading state sponsor of terrorism". It says Iran gave Hezbollah "hundreds of millions of dollars" and "thousands of rockets, missiles, and small arms" since 2006. This is the US government's own assessment, with no published accounting behind the dollar figure.

Of the plots in Europe from 2021 to 2024 in Levitt's dataset, most targeted Iranian dissidents, including Farsi-language journalists (table). Across the whole 1979–2021 dataset, dissidents were the largest single target group, ahead of Jews or Israelis.

Levitt finds a gap in Iranian plotting of "about 23 months" after the September 2001 attacks, and very few plots in Europe in the following decade.

Plotting then rose: more than half of all his cases fall in 2011–2021 (table).

The dataset more than doubled between the 2022 and 2024 publications (table). The 2022 study covered assassination, surveillance and abduction plots; the 2024 map adds intimidation.

Levitt works at a think tank whose analysts generally favour a hard line on Iran. Neither publication says how many plots were foiled and how many carried out.

MI5 says it tracked more than twenty potentially lethal Iran-backed plots in the year to October 2025, and that Iran-linked operations led to more than fifty UK arrests in the following twelve months. Neither figure comes with cases that can be checked.

MI5 says a wave of arson against Jewish communities in the UK and Europe was claimed by a front group, and "behind this front sat members of the IRGC". It also reports "Iranian ambitions to come at harder targets, like military sites".

MI5 describes Iran using "criminal proxies to try to intimidate oppositionists, Iranian media organisations, and Israeli sites". In 2025 fourteen countries jointly condemned "Iran's efforts to kill, kidnap and harass in Europe and North America".

On 27 February 2026 the Secretary of State designated Iran "a state sponsor of wrongful detention", citing "its pattern of hostage-taking and arbitrary detention of Americans".

He said releasing detained Americans were "steps that could end this designation". No series of cases by year was found.

Turnout fell to 48.8 percent in 2021 and to 40 percent in the 2024 first round, each a record low since 1979 (chart).

Parliamentary turnout fell the same way, to the lowest since the revolution in 2024 (chart below). Earlier swings were large: turnout was highest in 1997 and 2009 (chart).

After Raisi's death, Masoud Pezeshkian, "the sole moderate in an original field of four", won the runoff with 53.7 percent of the vote.

Turnout rose to 49.8 percent in the runoff, about ten points above the first round. He was one of only six candidates the Guardian Council approved (next chart).

From 2001 to 2021, about one in a hundred or fewer of those who registered for president were allowed to run (chart). In 2021 none of the 40 women who registered was approved.

In 2024 the share was higher, six of "more than 80", but well-known figures such as Ali Larijani and Mahmoud Ahmadinejad were barred.

In January 2004 the Guardian Council rejected 3,600 parliamentary candidates, including dozens of sitting members (timeline below). About a thousand were reinstated after a third of the Majlis resigned.

In 2024 it barred a smaller share of sitting members who sought re-election.

IHRNGO recorded at least 1,639 executions in 2025, the highest number since 1989 (chart). Its count has risen every year since 2020.

Amnesty's separate count for 2025 is higher still (chart below).

The monthly rate under Pezeshkian is more than double the rate under Raisi, by IHRNGO's count (chart below).

IHRNGO describes "a consistent upward trend ... regardless of the president in power".

Executions fell sharply from 2016 to 2020 (chart). IHRNGO links the decline to the 2017 drug-law amendment, which narrowed the death penalty for drug offences.

It says the amendment "has been reversed in practice" since 2021: drug executions averaged 26 a year in 2018–2020 and reached at least 795 in 2025.

IHRNGO says "a new surge in executions began in the months after the 12-day war" and that 60% of 2025's executions came in the second half of the year.

It finds executions "tend to increase when authorities fear social unrest and when the attention of the international community is focused on other geopolitical issues". The UN mission called October 2025's total "the highest monthly total recorded in at least 25 years".

HRANA counts from October to October. Its latest year, to October 2026, was its highest in ten reporting years (chart).

Like IHRNGO, it shows a low around 2018–19 and a rise every year since.

HRANA recorded hundreds of executions a month from November 2025 to February 2026, then its lowest full month in March 2026, when the strikes began (chart).

The monthly total recovered after the ceasefire but stayed well below its pre-war level through September. HRANA does not say why it fell.

Executions on political and security charges ran the other way: HRANA's peak months were April and May 2026, with 29 between them (chart).

Of the political and security executions HRANA recorded in its latest year, 35 were "in connection with the January 2026 (Dey 1404) protests".

Growth was strongest in 2016, the year the nuclear deal took effect, and turned negative in 2018 and 2019 after the US withdrew (chart). Inflation was at its lowest in decades in 2016 and 2017 and then climbed (next chart).

The timing fits sanctions relief and their return, but these series do not separate sanctions from oil prices or other causes.

The March 2024 parliamentary election drew 41 percent, "the lowest participation since Iran's 1979 Islamic Revolution", just below 2020's pandemic-year 42 percent.

HRANA's 7,007 counts every confirmed death in its 50-day record, protesters and others (chart).

It is a floor: 11,744 more cases were still under review when it was published.

The government counted 3,117 dead by late January, including people it called "rioters".

The UN mission found the crackdown amounted to "crimes against humanity of murder, imprisonment, torture".

Amnesty documented 304 deaths in November 2019, most of them in two days.

The UN mission counted at least 551 protesters killed in 2022–23. The January 2026 toll is many times either (table).

No primary source for deaths in the 2009 post-election crackdown was found, so it is left blank rather than filled from secondary reports. See Data still needed.

HRANA's confirmed total rose by more than half in a month while thousands of cases stayed under review (table).

Its published list names 4,555 people, against 3,038 on the government's list.

In the three months after the June 2025 war, Amnesty and Human Rights Watch counted over 20,000 arrests, including "dissidents, human rights defenders, journalists" and foreign nationals; the police's own figure was about 21,000. At least nine men were executed on political or espionage charges.

HRANA counted 53,777 arrests during the crackdown, including hundreds of children and school students.

Amnesty recorded at least 2,159 executions in Iran, 80% of all it recorded worldwide (chart). China is not included: Amnesty believes it executed thousands but cannot count them.

Saudi Arabia recorded the next-highest total, and per resident it executed at about two-fifths of Iran's rate (both charts). Together the two account for 93% of recorded executions.

IHRNGO and Amnesty were close for 2023 and 2024, but about 500 apart for 2025 (chart). Both count only cases they can confirm, so each is a minimum; they are never added together or merged here.

By IHRNGO's count, the average monthly rate rose from Ahmadinejad to Rouhani to Raisi, then more than doubled under Pezeshkian (chart).

Freedom House scores Iran one point above Saudi Arabia, out of 100 (chart).

Both are rated "Not Free" and are among the 20 lowest-scoring countries it rates.

Iran scores higher on political rights than Saudi Arabia, which holds no national elections, and lower on civil liberties.

In GAMAAN's June 2026 online sample, about 13% chose the Islamic Republic, down about 7 points on the previous year (chart). Monarchy and republic each drew more, and about a quarter had no clear alternative.

42% in GAMAAN's sample called the US–Israeli action "military aggression against Iran", while about 40% called it "humanitarian intervention" (chart). Most opposed Iran building nuclear weapons.

About 60% "felt satisfaction at the weakening of the Islamic Republic during the war", and only about 14% approved of its response to the January protests.

In December 2024 the Supreme National Security Council put the new hijab law "on hold"; its decisions "require the Supreme Leader's signature". In May 2025 the Speaker said he had been told in writing not to enact it.

What was halted is the new law's enactment. The existing hijab rules stayed, and Freedom House records arrests in December 2025 over unveiled women at a marathon.

Under Khatami, hardliners closed sixteen reformist publications in a week in 2000, and in 2004 the Guardian Council purged reformist candidates for parliament (timeline). The November 2019 killings came under Rouhani.

On 8 March 2026 the Assembly of Experts named Mojtaba Khamenei "the third leader", by "the decisive vote of its members", with no tally published. Iran International reports he was chosen "under pressure from the Revolutionary Guards"; that is a single outlet's claim.

Brent went from $71.32 a barrel on the last trading day before the strikes to $138.21 in early April.

After falling in June and July, it climbed again to $135.51 in early October.

The latest daily price, $123.10 on 27 October, was still nearly three-quarters above the prewar level.

Brent's daily record is $143.95, set in July 2008. Its high in 2022 was $133.18.

Monthly averages in 2026 also stayed below the 2008 peak (chart).

WTI, the US benchmark, went from $66.96 before the strikes to $114.58 at its April peak. In proportion, its rise was smaller than Brent's.

In October 2026 dollars, the refiner price of imported crude peaked at $98.85 in May 2026 (chart).

That is well below January 1981, July 2008 and June 2022 in the same series, and above the 1974 and 1990 levels. EIA's estimates for September and October 2026 are close to May's, so the autumn may set a slightly higher 2026 peak.

The weekly US average for regular rose from $2.937 a gallon just before the war to $4.500 in mid-May.

In the latest week it was $4.269, below the record of $5.006 set in mid-2022.

BLS reports gasoline up 34.2 percent over the year to September, accounting for most of the energy index's rise. Prices excluding food and energy rose far less (chart).

CBO estimates headline PCE inflation in early 2027 will be 0.5 percentage points higher than it projected before the war, and core inflation 0.3 points higher.

BLS does not attribute the rise, and the Federal Reserve's September statement cites only "geopolitical developments".

PortWatch counted 57 ships on the day of the first strikes and none four days later.

Monthly averages have stayed far below any earlier month since PortWatch's records began in 2019 (chart).

The strait normally carries oil equal to about 20% of world consumption.

Traffic picked up after the June memorandum, peaking at 51 ships on 24 June, then fell back after the blockade resumed in July (chart).

CENTCOM says more than 3,000 vessels passed through with US support in six months, and that Gulf partners exported more than 1.5 billion barrels since summer. Kpler, via the Washington Post, reports Gulf exports approaching prewar daily averages.

EIA warns that ship-tracking signals in the strait have been "especially unreliable" since February. Recorded transits are therefore a floor, and the true level is unsettled.

EIA puts oil flows in the second quarter of 2026 at under a quarter of their 2025 level, and LNG flows near zero (chart).

EIA estimates production shut in across the region peaked at 10.9 million barrels a day in May and fell to 4.8 million in September.

In September, most of the 28 chokepoints PortWatch tracks were close to their 2025 traffic (chart). Only the Kerch Strait, near zero, ran lower than Hormuz.

Vortexa data put Iran's exports in May at about 209,000 barrels a day, the lowest since early 2020 (chart). The first blockade had begun in mid-April.

ISW/CTP judges the spring blockade had "a limited but notable effect" on Iran's willingness to keep fighting.

CENTCOM reports turning around 135 ships and disabling four by 10 October, and says enforcement will continue against "any vessels trying to run the blockade". These figures have no independent check.

CENTCOM's early-October releases say US forces destroyed 13 commercial vessels in one, and 10 IRGC-linked tankers in the other.

The two releases use different categories, and CENTCOM does not explain the difference. The 1 September War Powers letter reported strikes on Iranian state oil tankers.

Officials told Bloomberg of 23 recorded attacks on vessels in or near Hormuz in the first two weeks of October, at least five hitting engine rooms. The report, relayed by ISW/CTP, does not say who carried out each attack.

Of the 31 states Bleek codes as exploring a weapon, ten acquired one, and fourteen stopped before launching a programme. Only South Africa built weapons and then gave them up.

Bleek dates the end of Iraq's pursuit to 1991, when after the Gulf War it was forced to give up its weapons capabilities "and submit to an intensive verification regime". Air strikes alone did not end it: Iraq's pursuit is coded from 1981, after Israel bombed the Osirak reactor.

After Israel destroyed the Al-Kibar reactor in 2007, Syria "returned to exploring". Bleek finds no public evidence of a decision either to continue or to stop.

Libya "returned to no activity in 2003", announcing it was abandoning its weapons of mass destruction programmes "and allowing international inspections". Hufbauer and colleagues count Libya as a sanctions success, but "only after 20+ years".

Braut-Hegghammer finds the 1981 strike "triggered a covert nuclear weapons program that did not previously exist", and Bleek follows her in dating Iraq's pursuit from 1981.

Other datasets disagree: Jo and Gartzke date Iraq's pursuit from 1973, and Bleek grants that the earlier work could "reasonably" be coded as pursuit. Which coding is right decides whether Osirak counts as a strike that started a programme or one that hit a programme already under way.

Bleek codes Iran's pursuit as open-ended from 1989, but adds that one could "reasonably" code Iran as dropping back to exploration after 2003 or in 2015, when it accepted the nuclear deal's limits.

In the TIES data, the success rate roughly doubles between the strictest count over all cases and the relaxed count over cases with a known outcome. Counting negotiated settlements as success, as the relaxed definition does, means the sender also gave something up.

Hufbauer and colleagues count a success when sanctions "contributed significantly" to the goal, and find one in "about one-third" of 204 cases. Pape argued that sanctions must have caused the result, not merely contributed; his stricter count was not read here.

Hufbauer and colleagues find sanctions more likely to work when targets are "friendly or democratic countries, rather than enemies, autocrats". For Iran and North Korea they judged "destabilization of either existing regime unlikely", and found "carrots plus sticks" more effective than sticks alone.

The Global Sanctions Data Base, the largest set here, scores each imposed sanction by its objective, one of which is "regime destabilization". Its authors' 2021 announcement put the success rate at "30 to 50 percent", without saying which outcomes count.

Horowitz and Reiter count 19 successes in 53 attempts from 1917 to 1999. Success was likelier when the target's military was vulnerable; how exposed its civilians were made no difference.

"Success is less likely when the attacker demands the target change its leadership." The target's regime type made no difference.

In Downes's data a third of imposed regime changes were followed by civil war within a decade. Imposing a new government "more than doubles" the likelihood, and replacing only the leader makes civil war "four and a half times more likely".

Nearly half of installed leaders were later forcibly removed, and about two-thirds of those installed by leadership change. One reviewer notes the base rate of civil war is low, so even a doubled risk might look acceptable to an intervener.

Downes finds that regime changes restoring a previously ousted ruler "tend to have best outcomes".

Overall, regime change had no significant effect on later disputes with the intervener, and leadership change more than doubled them. Targets were "no more likely" to vote with the intervener at the UN. Downes and O'Rourke find it "damages relations … more often than it improves them".

Downes and Monten find that interveners "will meet with little success" unless the target has "high levels of economic development and societal homogeneity, and previous experience with representative governance". Their data include Iran in 1953 but exclude the Reagan-era covert operations.

O'Rourke finds Cold War covert operations "unsuccessful more than 60 percent of the time", with targets more likely to have a civil war within ten years. Most backed authoritarian forces, and the successes came mostly against "democratic governments, weak states, and American allies".

Kreps and Fuhrmann find that peacetime strikes delayed programmes, "although the size of this effect is rather modest", and wartime strikes did less well. Their successes came when the attacker struck "well before a nuclear threat was imminent".

Writing in 2011, Kreps and Fuhrmann concluded that "many of the conditions that were conducive to past success are not present in the contemporary Iran case".

In Miller's data, once a programme had started, US sanctions succeeded "in only three of 22 cases". His reviewer Fuhrmann adds that "we cannot infer much" from those cases alone: counting states deterred from starting a programme gives "a much more sanguine picture".

Threats worked when "credible" and when the state was "highly dependent on the United States".

Mehta finds reversal "most likely" when states face sanctions and are offered "face-saving rewards", especially with force "on the table". The summary read gives no figures.

Fortna finds that "strong agreements markedly lessen the risk of further war", naming demilitarised zones, dispute commissions, peacekeeping and outside guarantees.

Press expected backing down to hurt credibility and found that leaders who did were "still taken seriously in the future". Weisiger and Yarhi-Milo find that past action is "closely connected" to later challenges, and that judging credibility by current power and interests alone "is misleading".

In 2018 Mohammed bin Salman said Saudi Arabia did not want a bomb, "but without a doubt, if Iran developed a nuclear bomb, we will follow suit as soon as possible".

Potter and Mukhatzhanova, reviewing the research, describe "past predictions of rapid proliferation" that "failed to materialize". Bleek's record shows ten states with weapons after some 70 years of exploration by 31.