
Summary
Data released on 8 October 2026 show a US economy with few lay-offs and, in the Atlanta Fed’s nowcast, private domestic demand growing at an annualised 4.5%. Minutes published the previous day showed most Fed policymakers expecting a further rate increase by year end, and the 10-year real yield has risen to 2.87% from 2.44% at the end of August. For highly valued technology shares, we think the rise in real yields now matters more than the strength of demand. The 3.4% fall in semiconductor stocks on 8 October added a separate concern. It followed a report that OpenAI’s annualised revenue was well below an earlier, differently measured figure, and it reopened the question of whether the buyers of AI capacity can pay for it.
While the Fed is still raising rates, firm data lift expected profits and the expected policy path at the same time. For companies valued mainly on distant cash flows, the second effect can dominate. The 8 October releases also show that the headline GDP nowcast is a poor gauge of underlying demand.
Lay-offs remain scarce
Initial jobless claims fell to 197,000 in the week ending 3 October, below the 200,000 economists had expected, and the four-week average eased to 198,000. Continuing claims, which run a week behind, rose by 17,000 to 1,716,000, but their four-week average fell by 12,250 to 1,711,000. The weekly rise sits inside a declining trend.
Claims track lay-offs, not hiring, and hiring is slow: September payrolls rose by only 29,000. Firms are taking on few workers but letting few go. Household income is holding up, and the labour market gives the Fed no reason to step back from tightening.
The headline nowcast hides two offsetting flows
The Census Bureau’s August wholesale report, also published on 8 October, showed merchant wholesalers’ sales up 1.8% on the month to $817.5 billion and 15.6% on the year. The figures are not adjusted for prices. With consumer prices up 3.4% on the year and energy prices up 16.3%, part of that rise is price, though the report cannot say how much.
Inventories rose 0.5% in August to $964.2 billion, less than the 0.7% in the advance estimate, and 6.4% on the year. With sales growing much faster than stocks over twelve months, the inventory-to-sales ratio stood at 1.18, against 1.28 a year earlier. The downward revision to inventories was enough to move the Atlanta Fed’s GDPNow model: its estimate of annualised third-quarter real GDP growth slipped to 3.6% from 3.7% on 6 October, as the estimated contribution of inventory investment fell to 1.98 percentage points from 2.07.
The size of that contribution matters more than its revision. Inventories account for 1.98 of the 3.6 points; real final sales of domestic product, which exclude them, are growing at 1.6%. Real final sales to private domestic purchasers—consumer spending and private fixed investment—are growing at 4.5%. Neither measure moved on 8 October. The gap between the two can only come from government spending and net exports, and the model’s 6 October estimate put the contribution of net exports alone at −2.69 percentage points.
An inventory build of about two points alongside a net-trade drag of nearly three is consistent with imported goods going into stock rather than to final buyers, although import data would be needed to confirm it. If that reading is right, the two flows largely offset, and neither the 3.6% headline nor the 1.6% final-sales figure describes demand well. Private domestic final purchases are the cleaner measure, and at 4.5% they are strong. All of these are GDPNow model estimates, not BEA data or a Fed forecast.
The Fed is tightening against energy-led inflation
On 16 September the FOMC voted 12–0 to raise the federal funds target range by 25 basis points to 3.75–4.00%, describing domestic spending as resilient and inflation as elevated. Minutes released on 7 October showed most participants judging that a further increase would likely be appropriate by year end; several regarded the current rate as not restrictive, or only mildly so. Futures traders are less sure about timing. On 7 October they put the probability of an increase at the 28 October meeting at around 20%.
The inflation the Fed is fighting is concentrated in energy. Consumer prices rose 3.4% in the year to August, but energy prices rose 16.3% and the core index only 2.4%. That leaves the rate path unusually exposed to oil, and oil jumped on 8 October: West Texas Intermediate settled 3.6% higher at $91.49 a barrel and Brent at $104.28. With private demand this firm, we see little near-term case for easing. Whether the Fed tightens again, and when, is likely to turn on September’s CPI, due on 14 October, and on the oil price.
Technology: real yields now matter more than demand
The 10-year real yield closed at 2.87% on 8 October, 43 basis points above its end-August level, and the nominal 10-year yield reached its highest level in 24 years earlier in the week. The real yield is the closest market proxy for the real discount rate on long-dated earnings, and the further out a company’s cash flows lie, the more of its present value a given rise removes. Much of the Nasdaq Composite’s value rests on earnings expected well beyond the next few years. In our judgement, a move of this size outweighs what the 8 October activity data add to near-term revenue. Until real yields stop rising, we expect firmer US demand to support earnings estimates rather than multiples, and the activity data say nothing about which technology companies will capture that revenue.
8 October: the trigger was AI revenue
The day’s equity losses had a more specific trigger. In the afternoon the Financial Times reported, citing investor documents, that OpenAI had told investors its annualised revenue was approaching $50 billion at the end of September, about $20 billion below the figure of close to $70 billion reported late in the month. The two numbers are not on the same basis. According to CNBC, the higher figure included gross revenue from sales through partners, an adjustment investors had made to compare OpenAI with Anthropic, which counts cloud-partner sales that way; a person familiar with the presentation put OpenAI’s third-quarter run-rate growth at 77%. The report points to a lower revenue base than the market had assumed, not to falling sales. The Nasdaq Composite nonetheless closed 1.25% lower at 27,193.34 and the SOX semiconductor index 3.39% lower at 12,623.71; memory stocks were already weak after Samsung fell in Seoul. Rates do not look like the immediate trigger: the 10-year Treasury yield fell about five basis points, and the real yield by a similar amount. Their level, though, still governs how much room valuations have to absorb news of this kind.
The same day TSMC reported September revenue 54.6% higher than a year earlier. That is clear evidence of strong current demand for advanced chips, though one supplier’s monthly sales cannot rule out risks further along the chain. What the market questioned is whether AI developers’ revenue can grow fast enough to fund the computing capacity they have committed to buy. Those payments underpin cloud providers’ capital spending, and that spending drives semiconductor orders. Where revenue falls short, the gap has to come from cash reserves, operating cash flow from other businesses or new equity and debt. The more it relies on outside capital, the more the build-out depends on funding costs, and so on the same real yields weighing on valuations. One day’s trading does not show that the market has re-rated AI demand, and the macro data cannot settle the question. Third-quarter results, and cloud providers’ capital-spending guidance in particular, will be a better test.
Gold rose, with the GLD ETF up 0.73% to $378.62, in line with the day’s dip in real yields. The wider backdrop is less supportive: real yields are up more than 40 basis points since the end of August, and the dollar index stood near 102, close to the 18-month high of 102.54 it reached on 5 October. We would want both to reverse before turning constructive on gold.
What would change our view?
Four signals would change this assessment. A rise in continuing claims that carries through to the four-week average, together with a clear slowdown in private domestic final purchases, would weaken the case for another increase. A sustained fall in oil would relieve headline inflation and, with it, the pressure on real yields. For semiconductors, the test is third-quarter capital-spending guidance from cloud providers: cuts would confirm the doubts raised on 8 October, while plans held or raised would suggest the sell-off overstated them. For gold, real yields and the dollar would need to fall together. The first test comes on 14 October, when September’s CPI will show whether energy is still pushing headline inflation higher.