The current assessment

Now

Long-term borrowing costs are rising across rich countries faster than their economies are weakening, while the labour market is quiet on the surface but hollow underneath.

US ten-year yields reached about 5.35% and thirty-year yields about 5.71% in early October, the highest since 2002; Britain's thirty-year yield passed 6%, and France's ten-year spread over Germany reached 152 basis points on 2 October. Oil near $100-115 adds inflation, and central banks are raising rates rather than cutting. In jobs, US unemployment is 4.2% and claims are low, but over twelve months all net job growth came from health care and social assistance, labour's share of business output is the lowest since 1947, and employers from HSBC to DNB now name AI in cuts to office work. On this reading, the strain shows first in government finances and in the composition of work, not yet in unemployment.

Our view, point by point

How far the evidence so far bears out each point of our view, on this assessment.

• 1. It may have begun two to four years agoUnder way
• 2. It could speed up as AI companies change how they sellBuilding
• 3. Government debt could become the central worryUnder way
• 4. A stock market bust, possibly larger and faster than the dot-com crashBuilding
• 5. A recession, as spending falls with wagesEvidence against
• 6. Care may stop absorbing the jobsNot yet seen
• 7. Tariffs in the US, spending cuts in EuropeUnder way

The coming months

The most likely course over the next six to twelve months is higher rates for longer, with fiscal tightening in Europe and a growing risk that the AI investment boom stalls on its financing costs.

France's 2027 budget (votes through the autumn) and its spring 2027 presidential election, Britain's budget on 28 October and US Treasury auctions will test whether buyers keep absorbing government debt. Debt-financed AI spending is large (about $450bn of AI-related debt issued this year by early September) and some private-credit funds have limited withdrawals. Should equity prices or credit markets break while oil stays high and central banks are tightening, there is a world in which job losses come quickly and, because firms automate rather than rehire, do not come back; US unemployment rising half a point from its low, with hiring still weak, would be the first sign.

The longer run

To about 2030-32, the evidence is consistent with a slow but persistent shift of income from wages toward the owners of capital and sites, with public budgets under growing strain.

Labour's falling share and the concentration of hiring in care fit a change that began two to four years ago, but the pace measured so far is well below the fastest paths one could draw. The more likely run assumes a smaller collapse of AI investment than the dot-com bust and a slower spread. A plausible worst case combines a severe bust with policy that turns to austerity and trade barriers as unemployment rises; in that world unemployment could reach levels not seen since the 1930s, and what would show it early is job losses that persist through a recovery.

Out of work counts the unemployed and those who have left the workforce after losing their job. The runs are the model's; they show what follows from their settings, not a forecast.