The World in 6, 12, 18 and 24 Months
AI analysis: this forecast was researched, drafted and red-teamed by AI agents. A PESTLE forecast overlaid with AI, robotics, resource and manufacturing constraints, as of 1 November 2026.
Bottom line
The most likely world in November 2028 (about 50%) is one that absorbed a loud, war-driven energy and interest-rate shock, while AI capability kept compounding faster than companies, workers and governments could adopt it. The single biggest swing factor is the Strait of Hormuz over the next six months. After that it is whether AI revenue grows fast enough to carry the debt now funding data centres.
The thesis in one line: loud shocks, slow absorption, fast capability. Capability moves on a months-long clock; adoption, grids, fabs, mines, central banks and elections move on a years-long one. Most of the surprises in this window come from the gap between the two.
| # | Headline call | Probability | Resolves |
|---|---|---|---|
| 1 | US or Israel strikes Iranian territory or forces again | 80% | by May 2027 |
| 2 | Brent above $100 a barrel | 45% in May 2027; 15% in Nov 2028 | on the date |
| 3 | Fed hikes again on 9 Dec 2026, and makes no cut before May 2027 | 78%; 80% | Dec 2026; May 2027 |
| 4 | US recession begins (NBER peak or two negative GDP quarters) | 25% by Nov 2027; 40% by Nov 2028 | cumulative |
| 5 | S&P 500 closes 20%+ below its 7,817 peak | 25% by Nov 2027; 38% by Nov 2028 | cumulative |
| 6 | AI-infrastructure credit event (Oracle cut below investment grade, a $5bn+ AI-infra default, or a big-4 hyperscaler cutting capex 10%+) | 30% | by Nov 2027 |
| 7 | Democrats win the House on 3 Nov; a Democrat wins the White House | 90%; 55% | Nov 2026; Nov 2028 |
| 8 | A frontier model is benchmarked at a 40-hour (one work-week) task horizon | 70% | by Nov 2027 |
| 9 | Xi stays in charge at the 21st Party Congress and names no clear heir | 91% | autumn 2027 |
| 10 | China blockades or quarantines Taiwan | 4% | by Nov 2028 |
| 11 | 2027 is the warmest year on record | 70% | Jan 2028 |
| 12 | Commodity offshore content, reporting and research work reprices 30%+ lower at renewal | 60% | by Nov 2027 |
Read the 25–40% numbers as seriously as the 80–90% ones. A world where at least one of the shocks in the shock register fires (cyber, sovereign bonds, nuclear, pandemic, institutional crisis) is more likely than not.
Method
The forecast was built in three passes, each one designed to catch the errors of the last.
- Scout pass. 18 Haiku research agents each took one slice of the world (macro, markets, US politics, China, US–China tech war, conflicts, Europe/Japan, India and the Global South, frontier AI, AI and jobs, AI law, chips, energy, minerals, robotics, other emerging tech, climate and health, society). Each was told to verify the state of play from pages dated Aug–Oct 2026, list dated catalysts to Nov 2028, and give its own probabilities.
- Hypothesis. The scout reports were merged into a draft thesis and a set of dated calls at 6, 12, 18 and 24 months, framed as PESTLE (political, economic, social, technological, legal, environmental) with an overlay of AI, robotics, resource scarcity, constrained manufacturing and US–China rivalry.
- Red team. Sonnet agents attacked the draft from different angles: too dramatic, too timid, China-sceptic, markets-bust, resource-constraint, base-rate calibration and black swans. Their critiques moved specific probabilities, recorded in the red-team section.
Probabilities are subjective and meant to be checkable. A call at 70% should be wrong about three times in ten. Horizons are counted from 1 November 2026: 6 months = May 2027, 12 = Nov 2027, 18 = May 2028, 24 = Nov 2028.
Where the world stands on 1 November 2026
The world enters this window in a war-driven energy shock layered on top of a debt-financed AI boom. Inflation is rising again, central banks are hiking into weak job growth, and long-term US borrowing costs are the highest since 2002.
| Dimension | State of play | Key numbers |
|---|---|---|
| Geopolitics | US–Iran war since 28 Feb 2026; Ali Khamenei killed; Strait of Hormuz largely closed; "no war, no peace" with talks in Doha and US strikes threatened after the midterms (CBS) | Hormuz transits fell to ~3/day vs ~130 pre-war (Aug) |
| Energy | Brent near $113–118; Qatar's Ras Laffan LNG offline since March; IEA sees 2026 oil demand falling (Argus) | Brent ~+75% y/y; 2026 demand −2.6 mb/d |
| Economy | IMF world growth 3.1% (2026), 3.4% (2027); global inflation 4.7% (IMF WEO) | Only 32 economies produce 95% of growth |
| US macro | CPI 3.7% (core 2.6%); payrolls +29k in Sep; Fed under Kevin Warsh hiked to 3.75–4.00% (Fed) | 10-year Treasury peaked ~5.35% in Oct |
| Markets | Record AI capex funded increasingly by debt and leases; Alphabet's first negative free-cash-flow quarter; Oracle cut to BBB− (IO Fund) | Big-4 hyperscaler capex >$700bn in 2026; AI debt ~$470–490bn YTD |
| AI | Cyber-capable frontier models gated (Claude Mythos, GPT-6 Astra); Anthropic valued at $965bn; ChatGPT at 1bn weekly users (Anthropic) | METR task horizon ~17 hours (Apr 2026) |
| Manufacturing | Memory supercycle: HBM sold out to 2029; TSMC revenue +41% y/y; advanced-packaging gap closing in 2027 (TrendForce) | DRAM contract prices +13–18% q/q (Q3) |
| Resources | Copper near record; China's rare-earth licensing intact, wider controls paused to 10 Nov; DRC cobalt quotas (Tech Times) | Dysprosium outside China ~16× the Chinese price |
| US politics | Midterms on 3 Nov; Democrats favoured; tariffs rebuilt on Sections 301/232 after the Supreme Court struck IEEPA tariffs (Decision Desk HQ) | House D ~82–95%, Senate D ~62–73% |
| US–China | Managed truce extended to 10 Jan 2027; two state visits in 2026 but no grand bargain; China proposing export controls on AI tech (The National) | H200 sales licensed; Blackwell banned |
| China | Growth ~4.5% carried by exports; property and stocks falling; PLA purge ahead of the 21st Party Congress (NBS) | Sep exports +36% y/y; 2026 surplus heading past $1.2tn |
| Europe & Japan | Energy inflation forced ECB, BoJ hikes; France in fiscal and political crisis with the far right leading 2027 polls; Lagarde leaving Feb 2027 (ECB) | Euro-area HICP 3.8% (Sep) |
| Society & work | Junior hiring in AI-exposed jobs suppressed; AI the top stated reason for US layoffs; youth protest wave; under-16 social media bans spreading (Pew) | 22–25-year-olds in exposed jobs ~19% below trend |
| Environment | Strong El Niño, possibly record, peaking this winter; August tied the hottest month on record; India's monsoon −12.6% (NOAA CPC) | COP31 in Antalya, 9–20 Nov |
| Legal | EU delayed AI Act high-risk rules to Dec 2027; DMA fines triggered a US Section 301 threat; US AI preemption stalled (iubenda) | Anthropic's $1.5bn authors' settlement approved |
Two threads matter most for what follows: how fast the Hormuz shock unwinds, and whether AI revenue grows fast enough to service the debt now funding its infrastructure.
The structural forces
Six forces drive almost every call below. The first two set the near-term weather; the last four set the climate of the next decade.
- The war-energy-rates chain. Hormuz disruption raised oil and LNG prices, which lifted headline inflation, which pushed the Fed, ECB and BoJ to hike into weak growth. A 10-year Treasury near 5% with equities at a CAPE near 41 leaves almost no equity risk premium. This chain, not AI, decides the next 12 months of markets.
- Debt-financed AI build-out. Hyperscaler capex has outrun operating cash flow, and the gap is filled by bonds, private credit, leases and vendor guarantees. Lab revenue is compounding fast (Anthropic's run-rate went from ~$9bn to $47bn+ in about a year), so the risk is a funding or timing mismatch, not a lack of demand.
- Capability outruns diffusion. Frontier task horizons doubled roughly every 4–7 months through 2025–26, and the most capable models are already gated for cyber risk. Yet only ~20% of US firms report using AI in production. Effects show up first where work is codified and junior: entry-level hiring, offshore content and research, software and customer service.
- Physical bottlenecks. In order of how hard they bind in 2027–28:
- Firm grid power and equipment: PJM is 6.8 GW short, large gas-turbine slots are booked to ~2030, transformers take 2–4 years.
- Memory: HBM sold out to 2029, DRAM wafer supply lags demand by 20%+.
- Leading-edge wafers and EUV: ASML's 2027 capacity does not meet demand; High-NA is not in volume until 2028–30.
- Rare-earth licensing: China's controls bind drones, defence, EVs and magnets through policy and two-tier pricing, more than through geology.
- Gulf and Asian energy: Qatar's LNG plant needs 1–3 years to repair regardless of Hormuz, raising power costs for Asian fabs and South Asian fertiliser.
- US–China: managed, eroding rivalry. Both sides now run export-control regimes on technology, including AI models. The truce holds through time-boxed extensions and summitry while tit-for-tat accumulates. China's export surge (record surplus above $1.2tn) is triggering trade defences across many partners, not just the US and EU.
- Slow-moving pressure: climate, demography, legitimacy. A strong El Niño makes 2027 a likely record-hot year and stresses South Asian food. Births in China, Japan and Korea keep falling. Youth unrest, populism and low trust raise the odds that governments change abruptly.
What the evidence does not support: copper, cobalt, helium and water are price risks or local fights, not global caps on AI or robots before 2028. Humanoid robots need only a few kilograms of magnets each; their real bottleneck is China-made actuators and reducers, now hit by Chinese export curbs and a new US Section 232 tariff.
6 months: to May 2027
The next six months are dominated by the post-midterm Iran decision and the inflation peak. Expect more strikes, oil staying high more often than not, one more Fed hike, a Democratic House, and another US–China truce extension. AI capability keeps rising, but the economic story is still energy and rates.
Dated catalysts: US midterms 3 Nov · rare-earth and BIS suspensions expire 9–10 Nov · COP31 9–20 Nov · Abu Dhabi Russia–Ukraine round mid-Nov · APEC Shenzhen Nov · Taiwan local elections 28 Nov · FOMC 9 Dec · CR expires 11 Dec · PJM capacity auction Dec · BoJ 18 Dec · US–China truce expires 10 Jan · Lagarde leaves 7 Feb · hyperscaler 2027 capex guides late Jan–Feb · China NPC early Mar · French presidential election Apr–May.
| Dimension | Call | Probability |
|---|---|---|
| Political | US or Israel strikes Iranian territory or forces again (publicly confirmed) | 80% |
| Political | Signed US–Iran framework or final deal | 20% |
| Political | Democrats win the House / the Senate (3 Nov) | 90% / 62% |
| Political | Federal funding lapse of 1+ day when the 11 Dec stopgap expires | 25% |
| Political | US–China truce extended or replaced past 10 Jan with no reinstated measures | 80% |
| Political | China extends or replaces its rare-earth pause; April 2025 licensing stays regardless | 85%; 95% |
| Political | Trump and Xi meet in person | 85% |
| Political | RN or allied candidate wins the French presidency | 38% |
| Political | Front-wide Russia–Ukraine ceasefire in force 7+ days | 8% |
| Political | Named large-scale PLA exercise around Taiwan | 45% |
| Economic | Brent above $100 in May 2027 (median ~$95, bimodal) | 45% |
| Economic | Hormuz crude flows at 85%+ of Feb 2026 for a full month (IEA) | 25% |
| Economic | Fed hikes on 9 Dec; no Fed cut before May 2027 | 78%; 80% |
| Economic | US headline CPI peaks between the Oct 2026 and Mar 2027 prints | 65% |
| Economic | 10-year Treasury sets a new cycle high above 5.35% | 35% |
| Economic | Sri Lanka CCPI inflation at or above 5% in May 2027 | 72% |
| Social | At least one more country passes a national under-15/16 social media ban | 75% |
| Technological | A Mythos-class model becomes generally available with safeguards | 55% |
| Technological | OpenAI or Anthropic files a public S-1 | 25% |
| Technological | DRAM contract prices still higher year on year | 95% |
| Technological | Big-5 hyperscaler 2027 capex guidance totals $900bn+ | 80% |
| Legal | China finalises its AI/LLM technology export-control rules | 65% |
| Environmental | CPC declares ENSO-neutral by its May 2027 update | 30% |
| Environmental | COP31 text goes beyond COP28's fossil-fuel transition language | 25% |
| Environmental | India imposes a new export curb on rice, sugar, onions, wheat or pulses | 30% |
The call most likely to be wrong in a way that matters: the oil path. It is genuinely bimodal; a deal in Doha sends Brent toward $75, a widened strike campaign toward $130.
12 months: to November 2027
By late 2027 the energy shock is more likely than not unwinding, and the question shifts to who pays for it. This is the window of maximum financial fragility: high rates, peak AI capex, a French election, a new ECB president and China's Party Congress all land inside it. AI's labour effects become hard to deny in junior white-collar and offshore work.
| Dimension | Call | Probability |
|---|---|---|
| Political | 21st Party Congress opens by 1 Nov 2027; Xi stays; a clear heir is signalled | 85%; 98%; 7% |
| Political | A sitting Central Military Commission member is purged | 60% |
| Political | Named large-scale PLA exercise around Taiwan (cumulative) | 65% |
| Political | Front-wide Russia–Ukraine ceasefire in force 7+ days (cumulative) | 15% |
| Political | Gaza international stabilisation force troops operating inside Gaza | 35% |
| Political | Federal funding lapse of 1+ day (cumulative) | 50% |
| Economic | Hormuz crude flows at 85%+ of Feb 2026 for a full month (cumulative) | 45% |
| Economic | Brent above $100 in Nov 2027 (median ~$80) | 25% |
| Economic | US recession begins / euro-area recession begins (cumulative) | 25% / 35% |
| Economic | S&P 500 closes 20%+ below peak (cumulative) | 25% |
| Economic | AI-infrastructure credit event (definition in Bottom line) | 30% |
| Economic | France–Germany 10-year spread above 120bp at some point | 30% |
| Economic | Disorderly yen or JGB episode forcing BoJ or MoF emergency action | 20% |
| Economic | IMF emergency or augmented financing for 2+ of Sri Lanka, Pakistan, Egypt, Kenya, Ghana | 50% |
| Economic | China's official 2027 growth at or above 4.0% / 2027 trade surplus above $1tn | 60% / 85% |
| Economic | New US or EU measure covering $10bn+ of Chinese imports | 70% |
| Economic | FAO food price index above 140 in any month | 50% |
| Social | Next Stanford-style estimate shows the junior gap in AI-exposed jobs at 25%+ | 40% |
| Social | A top-15 IT/BPO firm announces 10,000+ cuts explicitly tied to AI | 60% |
| Social | US unemployment reaches 4.7% at some point | 35% |
| Technological | A frontier model benchmarked at a 40-hour task horizon (METR or equivalent) | 70% |
| Technological | OpenAI or Anthropic files a public S-1 (cumulative) | 50% |
| Technological | AI-attributed cyber incident causing $1bn+ losses or a multi-day critical-infrastructure outage | 30% |
| Technological | A government again forces a frontier model to be gated or disabled for foreign users | 30% |
| Technological | First quarter-on-quarter fall in DRAM contract prices | 30% |
| Technological | A hyperscaler cites power or memory for a 2027 capacity slip | 80% |
| Technological | Waymo publicly reports 1m+ paid rides a week; Tesla runs unsupervised paid rides in 10+ metros | 75%; 40% |
| Legal | At least one more US state passes an AI-layoff disclosure law | 55% |
| Legal | Three or more US states adopt large-load tariffs or data-centre moratoria | 70% |
| Environmental | US residential electricity CPI at 6%+ year on year in some 2027 month | 60% |
| Environmental | LME copper sets a new all-time high | 75% |
The key interaction: if Hormuz reopens and oil falls, the Fed can cut and the AI financing structure gets time. If oil stays high into mid-2027, the rates-led bust path (scenario C) becomes the base case.
18 months: to May 2028
By mid-2028 the hardware cycle starts to turn as new fabs and packaging lines arrive, open-weight models close in on the frontier, and Taiwan's election and inauguration become the main geopolitical risk window. In the base case, central banks are cutting and oil is trending toward $70–80; in the bust case, this is the recession trough.
| Dimension | Call | Probability |
|---|---|---|
| Political | KMT-led ticket wins Taiwan's January 2028 presidential election | 45% |
| Political | Named large-scale PLA exercise around Taiwan (cumulative) | 85% |
| Political | Formal US rule controlling foreign access to frontier models published in the Federal Register | 30% |
| Political | Iran gives formal notice of NPT withdrawal (cumulative) | 10% |
| Economic | Brent above $100 in May 2028 (median ~$75) | 15% |
| Economic | US recession begins (cumulative) | 33% |
| Economic | S&P 500 closes 20%+ below peak (cumulative) | 32% |
| Technological | First quarter-on-quarter fall in DRAM contract prices (cumulative) | 45% |
| Technological | Two consecutive quarters of falling DRAM contract prices | 35% |
| Technological | Best open-weight model within 6 months of the closed frontier on the Artificial Analysis index | 75% |
| Technological | A Mythos-class model released with open or near-open weights (cumulative) | 50% |
| Technological | TSMC Arizona Fab 2 in volume production | 65% |
| Environmental | Copernicus confirms 2027 as the warmest year on record | 70% |
| Environmental | A second consecutive below-normal Indian monsoon (2027) | 20% |
The 2028 Taiwan window matters more than 2027. The PLA purge reduces near-term capacity for a major operation, but a new Central Military Commission, a contested election and a May inauguration raise the stakes of signalling.
24 months: to November 2028
November 2028 most likely looks like 2026 with lower oil, higher power bills, cheaper and far more capable AI, and a labour market visibly reshaped at its junior end. The US presidential election on 7 November 2028 is the closing event, fought mainly on cost of living, with AI and jobs as a rising theme.
| Dimension | Call | Probability |
|---|---|---|
| Political | A Democrat wins the presidency (7 Nov 2028) | 55% |
| Political | Gallup's top "most important problem" is the economy, inflation or cost of living | 80% |
| Political | US institutional stress event: open defiance of a final court order, removal of a Fed governor, an impeachment vote, or a disputed certification | 45% |
| Political | China blockades or quarantines Taiwan / coast-guard "quarantine-lite" at Kinmen or Pratas | 4% / 8% |
| Political | Front-wide Russia–Ukraine ceasefire in force 7+ days (cumulative) | 22% |
| Political | Signed US–Iran framework or final deal (cumulative) | 40% |
| Political | Two or more further governments fall in protest-linked episodes | 50% |
| Economic | Brent median ~$75, 80% band $45–115; above $100 | 15% |
| Economic | US recession begins (cumulative) | 40% |
| Economic | S&P 500 closes 20%+ below peak (cumulative) | 38% |
| Economic | US nonfarm productivity growth at or above 2% year on year (Q3 2028) | 50% |
| Economic | China reinstates its full October 2025 rare-earth controls | 25% |
| Economic | Dysprosium outside China still 5x+ the Chinese price | 80% |
| Social | Three or more further national under-16 social media bans in force, including a G7 country | 75% |
| Social | China's population falls for a sixth straight year (2027 data) | 97% |
| Technological | First quarter-on-quarter fall in DRAM contract prices (cumulative) | 55% |
| Technological | AI-attributed cyber incident causing $1bn+ losses or a multi-day critical-infrastructure outage (cumulative) | 45% |
| Technological | Tesla Optimus delivered to outside customers | 40% |
| Legal | A fair-use ruling in NYT v OpenAI | 75% |
| Legal | US tariffs explicitly tied to EU DMA/DSA enforcement in force | 25% |
| Legal | Federal statute pre-empting state AI laws | 10% |
| Environmental | New WHO public health emergency of international concern | 25% |
| Environmental | ENSO neutral or La Niña by late 2028 | 75% |
The election call is soft. A recession under the incumbent party lifts the Democratic odds toward 65%; a clean disinflation with oil at $70 pulls them back toward 50%.
Scenarios
The base case is that the shocks are absorbed (50%); a rates-and-oil-led bust is the main alternative (30%); a new phase of geopolitical rupture is the tail (20%). Separately, an AI-acceleration path (about 20%) can overlay any of the three.
| Scenario | Probability | What it looks like by Nov 2028 | First signs |
|---|---|---|---|
| A. Loud shocks, absorbed | 50% | Hormuz reopens through 2027, Brent $65–80, inflation back near 2.5–3%, rate cuts from late 2027; AI capex keeps growing more slowly; labour effects concentrated in junior white-collar and offshore work | A Doha deal or de facto reopening; Brent curve flattening below $85 |
| C. Stagflation and repricing | 30% | Oil above $100 into mid-2027 and a rates retest push the US and Europe into recession; S&P down 20%+; an AI-infrastructure borrower or Oracle loses investment grade; emerging-market financing requests multiply | 10-year above 5.35% on days equities fall; Oracle CDS above 300bp; negative payroll prints |
| D. Geopolitical new phase | 20% | Any of: China fully reinstates rare-earth controls with US retaliation; Iranian or Houthi strikes on Gulf processing or data-centre sites; a PRC quarantine of Taiwan | Strikes beyond Iran's borders; a rare-earth licence freeze; coast-guard inspections at Kinmen |
| B. AI acceleration (overlay) | 20% | Task horizons of 300+ hours, open-weight Mythos-class models, visible productivity gains in 2028 data, sharper job losses and a louder political backlash | A 40-hour benchmark result by mid-2027; lab revenue still compounding 1.4x a quarter |
Brent forecast (spot ~$113–118 on 30 Oct 2026)
| Horizon | Median | Chance above $100 |
|---|---|---|
| May 2027 | ~$95 | 45% |
| Nov 2027 | ~$80 | 25% |
| May 2028 | ~$75 | 15% |
| Nov 2028 | ~$75 (80% band $45–115) | 15% |
The oil forecast is a mixture of two worlds, which is why the median falls while the band stays wide: reopening points to $65–72, continued closure to $105–130.
Shock register. These are not exclusive and sit across all scenarios; the chance that at least one fires by Nov 2028 is roughly 70–80%.
| Shock (by Nov 2028) | Probability |
|---|---|
| US institutional stress event (court defiance, Fed governor removal, impeachment vote, disputed certification) | 45% |
| AI-attributed cyber incident with $1bn+ losses or a multi-day critical-infrastructure outage | 45% |
| Disorderly advanced-economy bond episode (France, Japan or a gilt-style event) | 30% |
| New WHO public health emergency of international concern | 25% |
| Iran NPT withdrawal notice or declared weapon | 15% |
| India–Pakistan cross-border strikes | 15% |
| North Korean nuclear test | 15% |
| Earthquake disrupting Taiwan or Japan chip fabs | 4% |
What the red team changed
Eight critics attacked the draft; the net effect was to raise the bust and rupture odds, cut the odds on company and policy promises, and reframe AI calls around what can actually be measured. The draft's slogan, "fast software, slow atoms", was replaced because the critics showed both halves were wrong: AI is itself gated by atoms, and the "slow" clocks (oil, tariffs, rates) moved violently in 2026.
| Call | Draft | Final | Why it moved |
|---|---|---|---|
| Brent above $100 in May 2027 | 35% | 45% | Geopolitics and macro critics: with Hormuz still impaired 70% of the time and Brent at ~$115, a $90 median was inconsistent |
| Hormuz "mostly restored" by Nov 2027 | 50% | 45% | Pinned to an IEA metric (85% of Feb crude flows); sources disagree on current flows (IEA 65%, US DOE ~7 mb/d) |
| US recession by Nov 2027 / Nov 2028 | 20% / 35% | 25% / 40% | Calibration critic: the draft's own scenario weights implied ~30% |
| S&P 500 20%+ drawdown by Nov 2027 | 20% | 25% | Equity risk premium near zero at a 5% 10-year; earnings concentrated in peak-cycle chips and memory |
| AI credit event by Nov 2027 | 25% | 30%, redefined | CoreWeave is already junk; split into Oracle, $5bn+ AI-infra default, or hyperscaler capex cut |
| Big-5 2027 capex $900bn+ | 65% | 80% | The draft's own starting facts pointed to ~$1tn |
| Fed hike on 9 Dec | 65% | 78% | Markets price ~85%; two hawkish dissents on 28 Oct |
| Frontier task horizon | "METR 100h, 50%" | "40h benchmark, 70%" | Fast-AI critic: the trend implies 60–160h; sceptics: the METR suite tops out at ~16h, so measurement risk dominates |
| Formal US frontier-model access rule | 45% | 30% | June 2026 controls were lifted in 18 days; the 2025 diffusion rule was rescinded |
| Tesla unsupervised rides in 10+ metros (12m) / Optimus external deliveries (24m) | 60% / 55% | 40% / 40% | Customer FSD slipped to Q2 2027; Q3 net income fell 43%; Optimus has no buyers |
| Copper new all-time high (12m) | 65% | 75% | Price sits ~3% below the record; a driftless walk alone gives ~80% |
| 2027 warmest year on record | 60% | 70% | Every strong El Niño was followed by a record year (1998, 2016, 2024) |
| Clear Xi successor signalled | 15% | 7% | No heir signalled at two Congresses |
| Scenario weights | 45 / 20 / 20 / 15 | 50 / 30 / 20 + 20% overlay | AI acceleration made an overlay; a rates- and sovereign-led bust folded into C; D redefined as a "new phase" |
The critics also added whole categories the draft missed: cyber risk from proliferating cyber-capable models, sovereign bond stress in France and Japan, US institutional stress, nuclear proliferation, health emergencies, food and fertiliser inflation in South Asia, and data-centre siting. Each is now a call or a line in the shock register.
Unresolved disagreements: the fast-AI critic would put scenario B at 30% and the sceptic at 12%; the macro critic would put C at 30% and A at 40%. The final weights split those differences.
Signposts to watch
Five indicators will tell you which scenario you are in, most of them within the first six months.
| Signpost | Points to A (absorbed) | Points to C or D (bust or rupture) | When to check |
|---|---|---|---|
| Iran after the midterms | Doha framework, or de facto Hormuz reopening; Brent curve flattens below $85 | Sustained US strike campaign; attacks beyond Iran's borders; Brent above $130 | Nov 2026–Feb 2027 |
| Rates and equities together | 10-year falls while payrolls weaken: a growth scare that lets the Fed cut | 10-year closes above 5.35% on days equities fall: a rates-led bust | Weekly, from Dec 2026 FOMC |
| AI financing | 2027 capex guides covered by revenue; IPO filings proceed | Oracle CDS above 300bp or a rating cut; AI-lease spreads widen; IPO window shuts | Hyperscaler results late Jan–Feb 2027; Oracle 14 Dec |
| US–China expiry cliff | Bundled extension of the 10 Nov, 27 Nov and 10 Jan deadlines | Rare-earth licence freeze; 50+ entity-list additions; China finalises tight LLM export rules | Nov 2026–Jan 2027 |
| Hardware cycle | First quarterly fall in DRAM contract prices driven by new supply | A fall driven by cancelled AI orders or collapsing device demand | TrendForce quarterly |
For the AI-acceleration overlay, watch for a published 40-hour task-horizon result, lab run-rates still rising about 1.4x a quarter, and a Mythos-class model with open weights. For the labour story, watch the next Stanford entry-level update, Indian IT and BPO headcount each quarter, and how many outsourcing renewals carry AI-productivity price cuts.
Sri Lanka and offshore services
Sri Lanka faces a stagflation grind, not a default: inflation stays elevated into mid-2027 on fuel and power pass-through, but IMF cover and surprisingly strong Gulf remittances (+29% y/y to August) cushion the shock. The bigger exposure for offshore knowledge services is the client side, where AI substitution and any US market drawdown hit discretionary client budgets at the same time.
Starting point (CBSL, early Oct 2026): CCPI inflation 8.0% in Aug and Sep; rupee ~372 per US dollar (−6.2% YTD); reserves $6.93bn; policy rate 8.75%; tourist arrivals Jan–Sep −1.8% y/y. The IMF completed the combined 5th/6th programme reviews in May 2026 but some criteria were missed (CBSL).
| Call | Probability |
|---|---|
| Sri Lanka CCPI inflation at or above 5% in May 2027 / in Nov 2027 | 72% / 30% |
| CBSL raises the policy rate before May 2027 | 40% |
| Rupee weaker than 400 per US dollar at some point by Nov 2027 | 30% |
| Next IMF review completed by Nov 2027 | 70% |
| New Sri Lankan sovereign default or restructuring by Nov 2028 | 5% |
| 2026 tourist arrivals below 2025 | 75% |
| Scheduled power cuts in Sri Lanka within 12 months | 20% |
| Commodity content, reporting and research-summary work reprices 30%+ lower at renewal (by Nov 2027) | 60% |
| Outcome- or gainshare-based pricing in most new large offshore deals (by Nov 2027) | 55% |
| Offshore entry-level analyst hiring 30%+ below 2025 levels at major KPO/BPO centres (by Nov 2027) | 65% |
| US federal tax on offshored services payments enacted (by Nov 2028) | 5% |
What gets automated first in offshore knowledge work: content drafting and localisation, research and news summaries, recurring reports, tagging and QA. What grows: AI operations and agent supervision, data engineering, compliance and regulatory work, and domain-heavy analytics. Evidence from the listed firms points the same way: EXL grew ~20% with data and AI at about two-thirds of revenue, while Wipro shed 6,700 staff in a quarter and HCLTech estimates a 2–3% revenue drag from AI price deflation (EXL, Zerodha). Revenue per head, not headcount, is the metric that matters.
Three signposts for offshore services:
- Renewals and RFPs: how many carry AI-productivity price-downs, and what large clients say about operating costs in Q4 results.
- The CBSL dashboard plus Brent: Brent above $100 beyond Q1 2027, monthly remittances, the rupee against 400 and the next IMF review.
- Client-market stress: an S&P drawdown of 15%+, a 10-year above 5.35%, or an AI credit event, any of which cuts client budgets.
Two regional risks sit outside Sri Lanka's control: El Niño and a 12.6%-weak Indian monsoon push South Asian food prices up through 2027, and a strike on Gulf infrastructure (scenario D) would hit remittances by an estimated 15–25%.
Sources
Pages opened by the research agents, Aug–Oct 2026. Several facts rest on secondary trackers or single sources; where sources conflicted (Hormuz flows, Brent level, lab revenue run-rates), the doc uses ranges. Probabilities are this forecast's own judgement.
- Macro and markets: IMF World Economic Outlook, Oct 2026 · Federal Reserve FOMC statement, 28 Oct 2026 · BLS employment situation, Sep 2026 · ECB monetary policy statement, 29 Oct 2026 · IO Fund on AI debt · First Trust market minute, Oct 2026
- Energy and conflict: IEA demand outlook via Argus · 2026 world oil market chronology · 2026 Iran war ceasefire · CBS live updates on Iran · PJM 2028/29 capacity auction (Utility Dive) · Al Jazeera on Ukraine energy ceasefire · Lloyd's List Red Sea brief · AEI China–Taiwan update
- US politics and law: Decision Desk HQ forecast · Jones Day on the IEEPA ruling · White House presidential actions · EU AI Act omnibus (iubenda) · Section 301 probe over Google DMA fine
- US–China and China: Mayer Brown on Commerce AI-model controls · Freight Figures on the truce extension · MercoPress on Xi's Washington visit · The National on China's AI export controls · NBS Q3 2026 release · AP on China's export surge
- AI, chips and robotics: Anthropic Project Glasswing · Claude Opus 5.5 announcement · OpenAI GPT-5.5 · METR time horizons · TSMC Q3 2026 results (SEC) · SK hynix Q3 (Korea Times) · TrendForce on CoWoS and 2nm · Electrek on Tesla Robotaxi · Tech Times on Optimus
- Work and society: Stanford "canaries" update (summary) · Challenger report, June 2026 · Pew on young adults and AI · Philippine BPO forecast cut · Al Jazeera on Australia's under-16 ban
- Resources and climate: Tech Times on rare-earth pricing · Kitco on copper · FAO Food Price Index · NOAA CPC ENSO discussion · Business Standard on India's monsoon · Climate Home on COP31
- South Asia: CBSL on the IMF reviews · EconomyNext on CBSL policy · India Q1 FY27 GDP