Despite record highs in the Nasdaq and a robust semiconductor rebound, surging long-term borrowing costs and narrowing market participation highlight mounting risks as investors navigate a complex AI-driven landscape amid persistent bond market pressures.
Long-term borrowing costs and the narrowness of the market’s leadership took centre stage this week, even as the Nasdaq set fresh records and semiconductor shares rebounded. The 10-year Treasury yield, which had been near 4.92% early in the week, climbed to 5.11% on Wednesday and was trading around 5.18% by Friday, while only about 32% of S&P 500 constituents were above their 50-day moving average and roughly half were above their 200-day average. That split tells a familiar but increasingly uncomfortable story: artificial intelligence remains the engine of equity gains, but financing costs are rising and participation underneath the headline indices is still fragile.
Reuters reported that yields jumped on Tuesday after stronger-than-expected economic readings showed the services PMI at 58.7, up from 56.5, and manufacturing PMI at 57, up from 53.9. Those numbers suggested the economy is still expanding at a solid pace, and traders responded by raising the odds of another quarter-point Federal Reserve increase at the October meeting to about 66%. The 2-year Treasury yield rose 14 basis points to 4.89%, while the 10-year climbed by the same amount to 5.11%. Stocks reacted quickly: the S&P 500 fell 0.8%, the Nasdaq dropped 1.1% and the Russell 2000 lost 1.8%, with homebuilders, utilities and real estate among the hardest-hit groups.
The move in bonds was notable because it continued even as oil eased. West Texas Intermediate slipped towards $93 a barrel on hopes that negotiations could lead to a reopening of the Strait of Hormuz, yet the 10-year yield remained close to 5.2%. That suggests the pressure in the Treasury market is coming from more than energy alone. Strong growth, the prospect of further Fed tightening, heavy government borrowing and soft Treasury auction demand have all fed into the rise. As bond investors can now earn meaningful returns again, equities must compete with a much stronger risk-free alternative than they did a year ago.
Against that backdrop, the artificial intelligence trade remains powerful but more complicated. The Philadelphia Semiconductor Index jumped 4.3% on Monday, according to market reports, with Advanced Micro Devices rising almost 10%, Intel gaining more than 12% and Nvidia adding about 2.3%. The Nasdaq rose 2.3% that session and semiconductor stocks added another 2.1% the next day as the index moved to new highs. Yet the key question is shifting from whether AI demand exists to whether the spending boom can deliver durable returns. Anthropic’s seven-year commitment with Akamai, worth $11.6bn and potentially rising to $20bn, underscores that demand for computing power is not limited to frontier model training; inference, agents, memory, networking and data-centre capacity may keep expanding as AI is deployed more widely.
Even so, the financing of the AI build-out is becoming harder to ignore. Industry data cited in the market wrap puts AI-related debt issuance at more than $220bn this year, with some of the investment supported by vendor financing, guarantees, equity stakes and long-term purchase commitments. That does not mean demand is fake, but it does make it more important to distinguish between genuine customer adoption and spending that is being recycled inside the AI ecosystem. CNBC also highlighted comments from Steve Eisman, who has grown more cautious about parts of the cycle while arguing that regulation could help the largest model developers build moats and pricing power. Meanwhile, competition from Chinese and open-weight models is pushing down costs, which may be good for adoption but less helpful for margins at companies spending heavily to stay ahead.
For investors, the more practical issue is how to respond while rates stay high and breadth remains weak. The market is still behaving as though AI spending will continue to outrun efficiency gains, but that assumption is now being tested by a Treasury yield above 5%. A more balanced reading would allow for several outcomes: the AI build-out keeps generating revenue and cash flow, spending pauses while usage catches up, or the industry overbuilds capacity just as cheaper models and tougher competition compress returns. For now, the semiconductor rally suggests investors are still leaning bullish, but the burden of proof has risen.
The next important tests are straightforward. Micron’s earnings next week should offer a clearer reading on pricing, high-bandwidth memory demand, supply agreements and customer appetite. In the bond market, the key question is whether the 10-year yield can stabilise rather than keep making new highs. And in equities, breadth will matter more than a few record closes in the Nasdaq. More stocks need to recover their 50-day and 200-day moving averages before the rally can look healthier. Until then, patience may remain the more sensible strategy: stay selective in equities, keep duration short in fixed income and wait for the rate cycle to show signs that it is finally maturing.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
