For most of the summer, investors worried that the U.S. labour market was weakening too quickly. Then August payrolls arrived. The economy added 162,000 jobs, almost three times the 56,000 economists expected, while unemployment remained at 4.1%. The labour force also expanded strongly. In almost any normal environment, that would be good news.
But markets did not celebrate. Treasury yields rose, stocks slipped and traders increased the probability of a September Fed hike to roughly 60%. The reason is simple: the market does not currently need more growth. It needs evidence that inflation can return to target without another round of tightening.
August services activity strengthened, new orders reached their highest level in three and a half years, and input-price pressures remained elevated. Manufacturing was weaker, but the broader economy continues to look considerably more resilient than the July payroll figures suggested.
This creates an awkward inversion. Weak employment data increase the chance of lower rates, which can support asset prices. Strong employment reduces recession risk, but also gives the Fed more freedom to tighten. Good economic news becomes bad market news.
That does not mean investors should root for unemployment. It means valuations have become unusually sensitive to the price of money. When long-duration assets already assume years of future growth, a small change in discount rates can matter more than a healthy employment report.
If next week's CPI confirms that price pressures are easing, strong employment may look like the soft landing everyone wanted. If inflation remains sticky, August's jobs report gives the Fed considerably more room to act.
Investors spent months asking whether the U.S. economy was slowing too quickly. It may turn out the more uncomfortable question is whether it has slowed enough.
Macro
Public markets
M&A
Staying Diligent
Things we are watching this week: 7–11 September:
Unhedged Commentary
AI is raising interest rates before it lowers costs.
The long-term economic promise of AI is straightforward. More output. More productivity. Lower marginal costs. But getting there is proving extremely expensive.
Hyperscalers are building data centres, buying GPUs, securing electricity and issuing debt at extraordinary scale. This week alone, ByteDance secured a $29.6bn loan, Anthropic was finalising a $15.0bn revolving credit facility, and AI infrastructure company Nscale was seeking another $3.5bn ahead of an IPO.
That money has to come from somewhere. And increasingly, economists think the AI investment boom may be contributing to a higher neutral interest rate, or R-star: the theoretical rate at which monetary policy neither stimulates nor restricts the economy.
The New York Fed's model puts U.S. R-star at around 1.65%, up from 1.36% in early 2025. Market participants argue the true rate may be higher still because AI investment and government borrowing are simultaneously competing for huge pools of capital.
That produces an interesting contradiction. AI may ultimately be deflationary. But before it makes goods and services cheaper, the infrastructure required to build it may make capital more expensive.
If data centres, hyperscalers and governments are all issuing debt at the same time, investors can demand higher returns. Mortgage rates rise. Infrastructure projects become harder to finance. Companies outside AI face a higher hurdle rate even though they receive none of the immediate productivity benefit.
The eventual payoff could reverse the effect. If AI materially raises productivity and pushes operating costs lower, inflation could fall and equilibrium interest rates could eventually follow.
But that is the second act. We are currently paying for the first. This matters because investors often treat AI's economic benefits as though they arrive simultaneously with the spending. They do not. Capital expenditure happens today. Productivity gains arrive later, and nobody yet knows how large they will be or who will capture them.
AI may eventually make intelligence abundant and economic activity cheaper. But building the infrastructure underneath it requires a historic mobilisation of capital. Before AI lowers the cost of everything else, it may first raise the cost of money.
The Thinking Corner
If a technology ultimately promises lower costs and higher productivity, but reaching that future requires trillions of dollars of capital today, how should investors decide whether higher interest rates are simply the price of getting there or a threat to the investment case itself?
Latest from the team
Market commentary and field notes from the team, on The Felix View.