US technology shares, led by semiconductor companies, came under heavy selling pressure on Tuesday as investors grappled with a combination of broad economic anxiety and growing skepticism about the enormous sums being spent on artificial intelligence infrastructure. The retreat was sharpest among chipmakers, the companies that supply the computing power behind the AI boom and that have been among the market’s biggest winners over the past year.
The Philadelphia Semiconductor Index, a benchmark that tracks major US chip firms, fell 5.6 percent. Nvidia, the dominant supplier of AI accelerators, declined 2.2 percent. The losses were not confined to chip designers. Memory and storage specialists Sandisk and Western Digital, whose shares had surged over the past year on overwhelming demand tied to AI data centres, dropped 9.8 percent and 7 percent respectively. The scale of those moves underlined how quickly investors have turned cautious toward companies whose valuations depend on continued explosive growth in AI-related hardware demand.
Broad market pressure
The semiconductor selloff rippled across the wider market. The S&P 500 index fell 0.6 percent, while the tech-heavy Nasdaq dropped 1.6 percent. The divergence reflected the concentration of AI-related risk in the technology sector and the sensitivity of high-growth names to changes in interest rates and investor sentiment. For much of the past two years, chip stocks have been propelled by expectations that AI models will require ever-larger data centres, more advanced processors and vast amounts of memory and storage. Tuesday’s session showed that the same trade can reverse quickly when the assumptions behind it are questioned.
One immediate source of pressure came from the bond market. The yield on 30-year US Treasuries rose to its highest level since 2007, driven by concerns about government borrowing linked to the Iran war. Higher long-term bond yields can depress the value of future cash flows at quickly growing companies, because investors discount those future profits at a higher rate. That math is especially relevant for technology firms whose valuations rest on earnings expected far into the future. When yields climb, the present value of those distant earnings falls, and investors become less willing to pay premium multiples for growth.
Global rise in long-term yields
The move was not limited to the United States. German, French and UK bond yields of the same maturity similarly rose to levels not seen in decades, while 30-year yields in Japan approached their highest levels ever. The synchronized increase in long-term borrowing costs points to a broader repricing of duration risk across major economies. Governments are facing heavier fiscal burdens, central banks are navigating uncertain inflation paths, and investors are demanding more compensation for holding long-dated debt. For equity markets, and particularly for long-duration growth sectors such as technology, the shift matters because it changes the discount rate applied to future earnings and raises the bar for any company that depends on cheap capital.
Technology companies themselves have contributed to the long-end selloff through their issuance of long-dated debt, according to analysts at a major investment bank. Large tech firms are issuing debt at record levels to finance unprecedented spending on data centres that support AI services. Investment-grade issuance this year is expected to reach a record $1.9 trillion, or about £1.4 trillion, compared with $1.44 trillion last year, the analysts said. That surge in corporate borrowing adds supply to the bond market and can push yields higher, especially at the long end of the curve. It also ties the AI investment cycle more tightly to credit conditions: the more companies borrow to build capacity, the more sensitive their prospects become to changes in interest rates and investor appetite for debt.
The debt-funded AI buildout
The scale of the AI buildout is difficult to overstate. Data centres require land, power, cooling systems, networking equipment and enormous quantities of advanced chips. The largest technology companies have been racing to secure computing capacity, often committing to multi-year projects that will not generate immediate revenue. Debt financing has become a central part of that strategy. By issuing long-dated bonds, companies can lock in funding for projects that may take years to mature. But the approach also introduces new risks. If AI revenue grows more slowly than expected, or if borrowing costs remain elevated, the burden of servicing that debt could weigh on earnings and limit flexibility.
Investors are also watching the demand side of the equation more closely. The initial phase of the AI boom was driven by enthusiasm about model capabilities and the potential for transformative productivity gains. The next phase is likely to be judged on revenue, margins and return on invested capital. That is why disclosures from leading AI developers can move markets far beyond their own private valuations. They provide clues about how quickly businesses and consumers are actually paying for AI services, and whether the infrastructure spending underway can be justified by demand.
Anthropic’s growth signals
One factor in this week’s market moves was Anthropic’s disclosure to investors over the weekend that its annualised run rate reached $65 billion at the end of July. The figure represents a rapid growth rate, but it may have fallen short of the expectations of some investors. Separate reporting indicated that Anthropic is providing guidance for 2028 revenue of $190 billion to $200 billion, a range that may have disappointed some investors hoping for even more stellar growth. The reaction illustrates the tension at the heart of the AI trade: expectations are so high that even strong growth can be judged insufficient if it does not exceed the most optimistic forecasts.
The debate over AI spending is not merely about one company or one quarter. It touches on the entire chain of suppliers, from chip designers and manufacturers to memory producers, networking specialists and data-centre operators. Companies such as Nvidia, Sandisk and Western Digital have benefited enormously from the buildout, but their shares are also vulnerable to any signal that demand could plateau or that customers might slow construction. When investors reassess the pace of AI adoption, the effects can cascade through the sector quickly, as Tuesday’s trading showed.
At the same time, the long-term case for AI infrastructure remains intact for many investors. The technology is being embedded in software, cloud services, advertising, customer support and industrial applications. Large language models and other AI systems require substantial computing resources, and demand for inference, the process of running trained models, is expected to grow as adoption spreads. The question is not whether AI will require significant infrastructure, but how much spending can be sustained, how quickly it will translate into revenue, and who will bear the financing risk if the timeline slips.
What the market is watching
For now, market participants are focused on the interplay between bond yields, corporate borrowing and earnings expectations. A further rise in long-term yields could intensify pressure on high-multiple technology stocks. A stabilization in yields, combined with evidence of strong AI revenue growth, could restore confidence. The coming weeks will bring more data on data-centre spending, cloud demand and corporate technology budgets, all of which will help investors judge whether the current pace of investment is justified. The semiconductor index’s sharp drop and the weakness in memory and storage shares have set a cautious tone, but they have not resolved the central debate over the durability of the AI spending cycle.
Tuesday’s moves also highlighted how interconnected the AI trade has become with the broader financial system. Record investment-grade issuance from technology companies links the sector to credit markets, while the global rise in long-term yields connects AI valuations to fiscal policy and inflation expectations. Investors who once treated chip stocks as a pure growth play are now factoring in duration risk, debt supply and the cost of capital. That more complex calculus helps explain why a sector that has delivered extraordinary returns can experience sudden, broad-based selling when the macroeconomic backdrop shifts.
The magnitude of the declines in Sandisk and Western Digital was particularly notable because memory and storage have been viewed as direct beneficiaries of AI data-centre demand. Their shares had surged over the past year as customers raced to secure components. A pullback of nearly 10 percent in Sandisk and 7 percent in Western Digital suggests that investors are taking profits and reassessing how much of the good news is already priced in. It also shows that even companies with strong fundamental momentum are not immune to a change in market sentiment when rates and growth expectations move together.
Nvidia’s 2.2 percent decline was smaller than the drops in memory names, but it carried symbolic weight because the company has become the most visible proxy for the AI trade. Its chips power many of the largest AI training and inference workloads, and its guidance has repeatedly shaped expectations for the entire sector. A modest pullback in Nvidia shares can still signal broader caution among investors who use the stock as a bellwether for AI demand. The Philadelphia Semiconductor Index’s 5.6 percent fall showed that caution was widespread across chipmakers rather than concentrated in a single segment.
The bond market’s message was equally important. Thirty-year US Treasury yields at their highest since 2007 reflect concerns about government borrowing linked to the Iran war, but they also reflect a global repricing of long-term risk. German, French and UK yields at multi-decade highs and Japanese 30-year yields near record levels indicate that the rise in term premiums is not a US-only phenomenon. For technology companies, that means the cost of financing the AI buildout could remain elevated, and the discount rate used to value future earnings could stay higher for longer. The combination of record debt issuance and rising yields creates a feedback loop that investors will continue to monitor closely.
Anthropic’s revenue disclosures added another layer to the debate. An annualised run rate of $65 billion at the end of July is a significant figure for a company still in a rapid growth phase, but markets are forward-looking. Guidance for 2028 revenue of $190 billion to $200 billion may be enormous in absolute terms, yet it can still underwhelm investors who have grown accustomed to exponential forecasts. The episode underscores how sensitive AI-linked stocks have become to any perceived shortfall in growth expectations, even when the underlying numbers remain strong by conventional standards.
As the week progresses, investors will weigh the selloff against the longer-term demand picture. Semiconductor companies remain central to the AI economy, and the need for advanced chips, high-bandwidth memory and storage is unlikely to disappear. But the market is increasingly demanding evidence that spending will generate returns within a reasonable timeframe. The sharp declines in chip and memory stocks, combined with rising long-term yields and record tech debt issuance, have created a more cautious environment. That caution is likely to shape trading until clearer data on AI revenue and infrastructure demand emerges.
Source: Silicon UK News