Semiconductor Stocks Fall for 8 Days — An 18-Year Streak With a New

The longest losing streak since the financial crisis — and it feels different

The Philadelphia Semiconductor Index fell 3.8 percent on Oct. 16, extending its slide to eight consecutive trading days. The last time the SOX lost eight days in a row was October 2008, during the darkest weeks of the global financial crisis. Since Oct. 6, the index has shed roughly 15.7 percent. The iShares Semiconductor ETF has dropped for nine straight days.

This is not a routine correction. An 18-year losing streak places the current pullback in historically lonely territory. What makes this episode distinctly different from 2008, however, is the underlying cause. In 2008, demand collapsed across the board. Today, semiconductor demand has not vanished. Instead, investors are recalibrating expectations around artificial intelligence — specifically, whether the staggering capital expenditures being deployed to build AI infrastructure will ever generate proportional returns.

Who is getting hit hardest

Micron Technology dropped 5.6 percent. Intel fell 5.8 percent. Both are among the most heavily weighted names in the SOX, and both carry outsized exposure to the memory and custom-chip buildout that AI data centers require. Memory makers have been early beneficiaries of the AI boom, as training and inference workloads demand enormous capacity. But the same companies now face scrutiny over whether customers can absorb that capacity fast enough to justify the spending.

The selling pressure has been broad-based, not concentrated in a single name. That breadth matters. When a pullback is isolated, it is usually a company-specific problem. When it spreads across the entire index for eight straight days, it signals a shift in how the market prices the sector’s future cash flows.

This is not a demand crisis — it is a profitability test

Analysts note that semiconductor demand has not experienced the kind of sudden collapse that characterizes cyclical downturns. The drop is not driven by a cliff in orders or a sudden halt in shipping. It is driven by a reassessment of valuations that were built on multi-year growth assumptions for AI-related chip demand.

The question on every desk this week is simpler and more painful than the 2008 question. In 2008, the worry was whether chips would sell at all. Today, the worry is whether the companies building AI data centers will make their money back. Power constraints, rising energy costs, and the sheer scale of upfront investment are weighing on sentiment. The longer AI training runs and the larger the models get, the more electricity and cooling each data center consumes — and the harder it becomes to reconcile those costs with revenue per rack.

The market is not punishing the technology. It is punishing the math.

The Nasdaq and the Dow tell two different stories

On Oct. 16, the Nasdaq Composite fell 2.1 percent to close at 26,140.03. The S&P 500 declined 1.1 percent to 7,659.44. The NY Dow slipped 175.28 points to 52,002.41. Over the week, the Nasdaq dropped 4.5 percent while the Dow gained 0.7 percent. The divergence between the two indexes highlights how concentrated the pain is within AI-linked technology names.

The Dow’s resilience is a reminder that the broader U.S. economy is not unraveling. Large-cap industrial, financial, and energy companies did not participate in the AI capex rush, and they are not feeling the same margin pressure from energy costs and equipment financing. The Nasdaq’s decline is a sector-specific event, not a systemic one — at least so far.

Energy costs add another layer of risk

West Texas Intermediate crude settled at $101.41 a barrel. Middle East tensions and disruptions to Russian energy infrastructure are keeping prices elevated. For semiconductor manufacturers and AI operators alike, energy is not just an operating cost — it is a structural one. Data centers are among the largest single consumers of electricity in many regions. When oil holds above $100, electricity prices tend to follow, and the economics of AI infrastructure become tougher to model.

This introduces a compounding uncertainty. AI spending is already under scrutiny. Rising energy costs make that spending look even less efficient. If inflation holds firm due to energy, the Federal Reserve’s path toward rate cuts becomes more complicated. Higher rates for longer would further pressure growth-oriented valuations — exactly where semiconductor stocks have been priced.

What to watch next

Tokyo’s semiconductor apparatus makers — Tokyo Electron, Advantest, and related equipment suppliers — will set the tone on Monday. These names are closely watched because they sit upstream of the entire chain. If orders from memory and logic fabricators slow, equipment makers feel it first. A bounce in U.S. semiconductors after such a sharp move down is possible; profit-taking and short-covering can create relief rallies. But without a credible signal that AI capex is sustainable, any bounce may be shallow.

The key variables to track over the coming weeks are three: the trajectory of U.S. long-term Treasury yields, the path of oil prices, and any commentary from major cloud providers or chip buyers about their capital allocation plans. If hyperscalers signal that AI spending is on track despite energy headwinds, the sector could stabilize quickly. If they begin to moderate guidance, the downside has further to run.

The 18-year losing streak in the SOX is a dramatic number. But numbers alone do not define a cycle. What matters is whether the market’s growing doubt about AI profitability is a short-term adjustment or the opening move of a longer repricing. The demand side has not broken. The economics side is still being written.