Brent at $109, the 10-year at 4.94%: the discount rate is beating the AI trade

The market repriced the cost of money, not earnings

The problem with Sept. 10 was not the size of the decline. It was the persistence of the direction. The S&P 500 fell for a fourth consecutive session, and the Russell 2000 dropped the most of the four major indexes. When small and mid caps, the names most exposed to long yields and funding costs, sell off alongside megacap tech, this stops looking like a valuation adjustment confined to a handful of large stocks.

The immediate pressure came from prices and energy. The August producer price index rose 0.4% on the month and 5.4% from a year earlier. The monthly figure matched expectations, but the annual rate ticked back up from July, and energy prices jumped 4.2% over the month. The market priced the risk that expensive oil slows the disinflation path before it priced any expectation that prices resume falling.

Higher oil, higher inflation expectations, selling in long Treasuries, a higher 10-year yield, a lower present value of future earnings, compressed multiples across AI, semis and software.

The 10-year traded around 4.94% and touched roughly 4.95% intraday. At this level, the number itself matters less than whether the trend keeps grinding toward 5%. Another leg higher in long rates means the fastest-growing companies get the steepest discount rate. That is why fair value falls even when the earnings estimates do not move.

The whole AI supply chain fell together, and that is the clue

Nvidia and the rest of the semiconductor and AI infrastructure complex were weak. But reading that as a demand signal requires more evidence than one session provides. When design, foundry, memory and server names are all pressured on the same day, the cleaner explanation is not a broad drop in orders. It is that the multiple on the entire theme reset to the level of rates.

That means the recovery condition is more likely to show up in the bond market than in a company press release. If the 10-year drops back below 4.9% quickly, AI and software get room to bounce. If it settles above 4.95% and creeps toward 5%, the ceiling on valuations comes down for good businesses and bad ones alike.

Apple +3.56%: the market paid only for certainty

Apple closed at $326.57, up 3.56%. Alphabet gained 0.61% and Microsoft added 0.16%, while Meta fell 1.42% and Amazon slipped 0.20%. This was not an indiscriminate liquidation of tech.

Apple's strength shows that even in a rising-rate tape, money stays with names that have a visible product event and visible earnings. It is harder to call that a return of risk appetite. The real change comes when the Apple pattern spreads to the rest of megacap tech and to semiconductors.

Then, after the bell, Oracle posted very different numbers

Oracle reported fiscal 2027 first-quarter non-GAAP EPS of $1.92 on revenue of $19.3 billion. Total cloud revenue rose 62% to $11.6 billion, and OCI, the core of the story, grew 121% to $7.4 billion. The more important line is RPO, the remaining performance obligations that have been contracted but not yet recognized as revenue. RPO reached $664 billion, up $209 billion from a year earlier.

That is a stronger statement than a forecast that AI data center demand will hold up. It means customer demand for AI compute has already converted into contracts and backlog. Oracle guided fiscal 2027 revenue to at least $90 billion. For anyone arguing the AI infrastructure investment cycle has rolled over, this is an inconvenient print.

It supports the case for sustained GPU demand at Nvidia, leading-edge volume at TSMC, networking and custom accelerators at Broadcom and Marvell, and server and rack integration at Dell and Supermicro. It points the same direction for memory and enterprise SSDs.

Good news comes with a bill. Oracle is spending heavily to expand data center capacity. According to Reuters, capital expenditure in the quarter reached $28.5 billion. AI demand is being confirmed in contracts, and at the same time the capital required to turn that demand into revenue and cash flow is getting larger.

What to watch

The 10-year yield. A settled range above 4.95% and a move toward 5% signals further multiple compression in AI, semis and software. A quick return below 4.9% is the first condition for a tech rebound.

Crude. Above $110, energy-driven inflation and the pressure on the Fed both build. Long rates will not stabilize until oil does.

Breadth in the AI complex. Does Oracle rally alone, or do Nvidia, TSMC, Broadcom, Dell and Supermicro move with it? The second case would mean the demand signal is starting to beat the macro pressure.

CPI. With PPI and expensive oil already in hand, the consumer print is the next test. If long-rate volatility stays high into the meeting, high-multiple names keep facing position-cutting pressure.

Megacap breadth. Watch whether only Apple, with its product event, holds up, or whether buying widens across big tech as rates settle.

FAQ

Does Oracle's quarter mean the AI trade is back? No. Oracle confirms the strength of demand, but a discount rate is applied to the stock at the same time. If long rates keep rising, even good results will show up only partially in the share price.

Was the semiconductor selloff a demand warning? It is too early to say that from this data. The entire supply chain fell at once, and on the same day Oracle's OCI growth and RPO confirmed strong AI demand. Multiple compression explains the tape better than demand destruction.

Why does the Russell 2000 matter here? Small and mid caps are more sensitive to borrowing costs and domestic financial conditions. When the Russell underperforms large caps, the market is worrying about financial conditions broadly, not one sector.

What would change the read? If oil rolls over quickly and the 10-year drops below 4.9% but the AI supply chain stays weak, the issue is no longer the discount rate. At that point, look again at earnings estimates, the quality of the order book, and whether the capex is earning its return.

Insight Times Editorial Desk