World War 3 Predictions: What a Calibrated Engine Actually Sees

Search interest in a third world war set records twice in two years. The measured state — record spend, 65 conflicts, 85 seconds — and the sealed numbers: 38%, 8%, 74%, each with a written way to be wrong.

A 38% is not a prophecy of war; an 8% is not a promise of peace. Both are numbers that can fail in public.

The fear, measured before the facts

The search phrase is blunt — world war 3 — and its volume is one of the most sensitive fear gauges the internet produces. During the twelve days in June 2025 when Israel, Iran and then the United States traded strikes, Google Trends coverage recorded searches for “World War III” rising roughly 2,000% and “is the US going to war” rising nearly 5,000%. This observatory runs its own instrument on the same nerve: the Attention Seismograph on the Psyche page reads live Wikipedia pageview traffic across six anxiety topics — World War III among them — and scores each one's current seven-day pulse against its own baseline. The seismograph exists because attention is data. It is just not the data most people think it is.

Then February 2026 removed the hypothetical. On 28 February, Israeli and American strikes on Iran killed its supreme leader and opened a war that has since produced the largest supply disruption in the history of the global oil market, a two-week ceasefire in April, a naval blockade, a signed memorandum in Islamabad on 17 June establishing a sixty-day window to negotiate final terms — and, on 8 July, the collapse of that ceasefire into renewed strikes. As this article is sealed in mid-July 2026, the war is neither concluded nor global. It is a live, contained great-power-adjacent conflict, and it is precisely the kind of event that converts a search phrase into a siren.

Here is what the siren measures, and what it does not. Attention spikes are real information about the people searching — their fear is a fact, and it deserves a serious answer rather than a monetized one. But attention carries no information about probability: the seismograph spikes after events, never before them, and its loudest readings in a decade (2017, 2020, 2022, 2025) each subsided without the war they named. The job of this article is to separate the measured state of the world from the imagined one — and then to quote, exactly, the sealed probabilistic simulations this platform carries on the components of the question, each with its uncertainty range and its written condition of failure.

What the doom content gets wrong

The WW3 content industry runs on the same three mechanisms this observatory has documented for prophecy generally: retrofitting, fabrication and translation drift. Nostradamus wrote no dated verse about a third world war — the phrase itself is a twentieth-century coinage, and the “three Antichrists” arc that listicles map onto Napoleon, Hitler and a conveniently unnamed third figure is a construction assembled by modern readers from undated quatrains. The genre's seer lists are recycled annually with the years quietly updated, which is the tell: a prediction whose date can be moved was never a prediction.

The structural failure is deeper than the sourcing. No viral war forecast states a probability. None states a window. None states what evidence would count against it — so none can ever be wrong, and a claim that cannot be wrong is decoration, not information. There is also a framing error specific to this topic: “will there be a World War 3?” is not a forecastable object. It bundles a hundred distinct events — an incident, an escalation, an alliance cascade, a nuclear use — into one cinematic composite, and composites cannot be scored. A calibrated engine must decompose the question into components that can.

Decomposition also disciplines the base rate. Direct war between great powers is rare — rare enough that the honest probability of the full composite in any given window is low, and anyone quoting high confidence in either direction is performing. What is not rare is the machinery that precedes such wars: rearmament, bloc formation, incidents at sea and in the air. Those are measurable, they are elevated, and they are where the sealed numbers below actually live. The engine prices the components; it refuses the composite. That refusal is not evasion — it is the difference between forecasting and screenwriting.

The measured state, July 2026

Spending first. SIPRI's April 2026 accounting puts world military expenditure at $2,887 billion for 2025 — a 2.9% real-terms rise, the eleventh consecutive annual increase, lifting the global military burden to 2.5% of GDP, its highest share since 2009. The composition matters as much as the total: European spending rose 14% in a single year to $864 billion, Asia and Oceania rose 8.1%, while US spending fell 7.5% to $954 billion — a world rearming faster than its largest military, which is historically the signature of a system redistributing power, not merely deterring.

Conflict count second. The Uppsala Conflict Data Program recorded 65 state-based armed conflicts in 2025 — the highest number since its statistics begin in 1946 — of which 13 met the definition of war, over a thousand battle deaths in the year. And the composite gauges agree: in January 2026 the Bulletin of the Atomic Scientists moved the Doomsday Clock to 85 seconds to midnight, four seconds inside the previous year's record and the closest setting in the Clock's history, citing nuclear risk, AI, biosecurity and climate together. Nine states hold roughly 12,200 nuclear warheads on the Federation of American Scientists' count.

Now the judicial reading of that table: every gauge is elevated, several are at records, and none of them is a war forecast. Military spending rose for a decade before 1914 — and also through the entire Cold War, which ended without great-power war. Conflict counts measure the periphery burning, not the center igniting. The gap between structural stress and outcome is precisely where calibration has to live, because it is where prophecy pretends the gap does not exist. The numbers below are this platform's attempt to price that gap honestly — sealed in July 2026, labeled probabilistic simulation, and built to be graded.

What the registry actually seals

GEO-02 assigns 38% (±15) to a direct US–China military incident in the western Pacific across 2026–2031 — not full war, but a collision, shoot-down or blockade probe that produces casualties and forces both capitals to the brink before de-escalating. The reasoning is on the record: Graham Allison's survey found 12 of 16 rising-vs-ruling power transitions since 1500 ended in war, and the modern tripwires are Taiwan — which fabricates roughly 90% of the world's leading-edge chips — and the Second Thomas Shoal. This is a probabilistic simulation sealed in July 2026, not an event on a calendar; read at face value it says the engine expects no such incident, three chances in five.

GEO-03 assigns 8% (±5) to battlefield use of a nuclear weapon by 2035. Eight percent is low, and it is not a rounding error: Russia formally lowered its declared use threshold in November 2024, Pakistan fields battlefield-range systems, and May 2025 produced the first direct missile-and-drone war between two nuclear powers. The entry is ruin-class — the expected cost dwarfs the odds — which is why the registry carries it despite the low number. The 80-year taboo is the single most valuable norm in the international system, and its survival through 2035 is the engine's base case at roughly eleven-to-one.

And the highest sealed number on this board is not military at all. GEO-01 assigns 74% (±10) to the world economy splitting into rival trading blocs by 2030 — average US tariffs at their highest level since 1934 on the Yale Budget Lab's measure, Chinese rare-earth export controls operational, parallel payment rails built, a hard two-bloc split priced by WTO simulation near 5% of global real income. The war most likely to arrive this decade is economic, and it is mostly already here. In the tail file sits the Taiwan semiconductor shock — carried as a black swan at 14% probability with an impact score of 90 out of 100 across 2026–2032, with Bloomberg Economics pricing a full conflict near $10 trillion in lost output.

What the 1910–1914 resonance means — and what it does not

The Historical Resonance Index scores July 2026 against nine past eras on five weighted components — debt load (.20), great-power rivalry (.25), internal polarization (.20), tech-disruption velocity (.20), monetary regime strain (.15) — with every input published and the arithmetic recomputed on each refresh. The current table: 1968–74 scores 83; 1910–14 and 1873–96 tie at 82; 1929–33 and 1848–49 score 79; 1937–39 scores 78; 1815–19 scores 77; 1999–2000 and 1987 trail. The pre-1914 world sits one point from the top of the table, and that fact deserves to be stated without either flinching or theater.

What the 82 means: the structural configuration of July 2026 — a rising challenger against an incumbent, rearmament, bloc formation, a general-purpose technology moving faster than institutions, monetary strain — resembles the 1910–14 configuration more closely than it resembles all but one other era measured. What it does not mean: a countdown. The index's own top two entries are the proof, because they resolved in opposite directions — 1968–74 produced inflation, the end of Bretton Woods and institutional reform; 1910–14 produced continental war within four years. Nearly identical structural pressure, opposite outcomes, decided by choices and accidents no index can see. Resonance identifies the fork, not the branch.

The 1914 lesson, stated properly, is about speed rather than fate: alliance tripwires and mobilization timetables converted a local assassination into systemic war in five weeks, faster than the diplomacy of the era could descend the ladder. The modern analogue is worse on exactly that axis — hypersonic flight times and algorithmic response cycles compress the five weeks toward five minutes. That is why the registry's monitoring layer watches tempo gauges (incursion counts, intercept behavior, exercise postures) rather than dates. The fork is real and measured. The branch is chosen, and it has not been chosen.

The de-escalation file

The Calibration Ledger's second entry is the one no doom headline will ever quote. CAL-02, sealed 3 July 2026 at 82%: no direct kinetic exchange between US and Chinese armed forces — no live fire, in any theater, at any scale — through the end of 2029. Read it carefully, because it is the engine's actual position: the base case, at better than four-to-one, is armed peace between the powers everyone is searching about. The number is sealed, it will be Brier-scored, and if live fire comes, it will be scored as the miss it was.

The bull case is not a mood; it is a file of precedents. The 2001 EP-3 collision — a US-China aircraft collision with a fatality — resolved diplomatically. The Cuban crisis resolved through secret concessions both sides denied. The May 2025 India-Pakistan exchange, the first direct missile-and-drone war between nuclear powers, held under the nuclear ceiling and stopped in four days. Even the live 2026 war has produced a signed memorandum and active mediation inside five months — wars still seek exits, even mid-fight. And note the shape of GEO-03's own falsifiability line: the engine can only be proven wrong about nuclear risk by a world that signs a major arms-control instrument. The upside scenario is written into the seal itself.

Deterrence arithmetic points the same direction. Public wargaming of a Taiwan invasion — most visibly the CSIS series published in 2023 — repeatedly finds no clean victory available to either side at costs both would recognize in advance; that is historically the configuration that produces probes and pressure rather than invasion. Economic entanglement is weaker insurance than the 1913 optimists believed, and this platform will not repeat their error — interdependence failed once security logic took over. But the honest summary of the evidence is this: the majority of the probability mass, on this engine's own sealed numbers, sits with no great-power war this decade. That is not complacency. It is arithmetic, and it is falsifiable.

The appointment: what this page will be graded on

Assemble the calibrated answer to the search phrase and it fits in one paragraph. There is no dated World War 3 prediction here, because none can honestly exist. There are sealed components: 38% (±15) that the US and China suffer a casualty-producing incident in the western Pacific by 2031; 8% (±5) that the nuclear taboo breaks by 2035; 74% (±10) that the world economy finishes splitting into rival blocs by 2030; 82% that no US–China live fire occurs through 2029. Every number is a probabilistic simulation sealed in July 2026 with sources attached, and some of them will be wrong — visibly, datably, in public.

What to watch, if you want signal instead of sirens: monthly PLA incursion counts into Taiwan's ADIZ, which Taiwan's defense ministry publishes daily — tempo is the tell; the behavior of intercepts around US freedom-of-navigation operations; each round of semiconductor export controls, which historically precedes signaling spikes; and, for the live war, whether the Islamabad process reconvenes. These are the gauges the registry's trigger lists are built on. None of them is improved by refreshing a feed at 2 a.m.; all of them are published on schedules.

The first public Brier scoring of the Calibration Ledger arrives in July 2027, and the geopolitical entries run to the end of 2029 — hits and misses published at the same size, arithmetic and unedited. Nostradamus has had 471 years without a single scored prediction about war or anything else; this ledger will have its first grades after one. If the fear behind the search phrase is yours, the honest offer is this: not reassurance, and not a countdown — a set of sealed, falsifiable numbers, the gauges they depend on, and a standing appointment to watch them be graded. Nothing here is advice. All of it is checkable.

Frequently asked

Will there be a World War 3?

No one can answer that honestly, and this registry does not try — the composite question cannot be scored. What it seals instead are components, as probabilistic simulations: 38% (±15) that a direct US–China military incident occurs in the western Pacific by 2031; 8% (±5) that a nuclear weapon is used in battle by 2035; 82% that US and Chinese forces exchange no live fire through 2029. The engine's base case is armed peace between the great powers this decade — sealed where it can fail.

What did Nostradamus predict about World War 3?

Nothing dated, and nothing named. The phrase is a twentieth-century coinage; the quatrains contain no verse assigning a war to any year of our era. The “three Antichrists” arc mapped onto Napoleon, Hitler and an unnamed third figure is an interpretive construction built by modern readers from undated, unlocated verse — and his one explicitly dated near-term prophecy, for the seventh month of 1999, passed without event. Every “Nostradamus predicted WW3 in 2026” headline is a retrofit wearing a byline.

What is the probability of US-China war?

This platform's sealed numbers, labeled probabilistic simulation: 38% (±15) for a direct military incident producing casualties in the western Pacific by 2031 — a collision, shoot-down or blockade probe, not full war — and 82% that no live-fire exchange of any kind occurs through 2029 (Calibration Ledger, CAL-02). Allison's survey of power transitions (12 of 16 ending in war since 1500) explains why the incident risk is priced seriously; the falsifiability conditions for both numbers are published beside them.

What are the warning signs of a world war to watch?

Tempo, not headlines. The registry's trigger lists watch: monthly PLA incursion counts into Taiwan's ADIZ (published daily by Taiwan's defense ministry); coast-guard and naval behavior at the Second Thomas Shoal; major exercises shifting to no-notice posture; each new round of semiconductor export controls; doctrine documents rather than rhetoric (Russia's 2024 nuclear revision was published, not leaked); and the Doomsday Clock and SIPRI series as slow composite gauges. Attention spikes — including this site's own seismograph — measure fear, and lag every one of these.

Sources

SIPRI, Trends in World Military Expenditure, 2025 (April 2026) — world total $2,887B, +2.9% in real terms, the 11th consecutive rise; burden 2.5% of GDP, highest since 2009; Europe +14% to $864B; US $954B, −7.5%
Uppsala Conflict Data Program / Uppsala University (June 2026) — 65 state-based conflicts recorded in 2025, the highest since records began in 1946; 13 at war intensity
Bulletin of the Atomic Scientists (27 January 2026) — Doomsday Clock set at 85 seconds to midnight, the closest in its history
Federation of American Scientists, Nuclear Notebook (2025) — ~12,200 warheads across nine states
Public record of the 2026 US–Iran war (Britannica; CNN, 9 July 2026) — war opened 28 February 2026; Islamabad Memorandum signed 17 June; ceasefire collapsed 8 July; the largest supply disruption in the history of the global oil market
Google Trends press coverage (June 2025) — “World War III” searches up ~2,000% and “is the US going to war” up ~5,000% during the Twelve-Day War
Graham Allison, Destined for War (2017) — 12 of 16 rising-vs-ruling power transitions since 1500 ended in war
CSIS, The First Battle of the Next War (January 2023) — wargamed Taiwan invasion scenarios ending without clean victory at acceptable cost to either side
Bloomberg Economics (2024) — a full Taiwan conflict priced near $10 trillion in lost output
Yale Budget Lab (2025) — US average effective tariff rate ~18%, highest since 1934; WTO two-bloc simulation — ~5% of global real income
Glenn W. Brier, "Verification of Forecasts Expressed in Terms of Probability", Monthly Weather Review, 1950