Markets ‘love chasing bottlenecks’: Wall Street weighs epic run in AI stocks

May 10, 2026
markets-‘love-chasing-bottlenecks’:-wall-street-weighs-epic-run-in-ai-stocks

Semiconductor stocks leading the S&P 500 (^GSPC) and Nasdaq Composite (^IXIC) to record highs have one common thread: They sit at the bottlenecks of the artificial intelligence trade.

It used to be that Nvidia (NVDA) graphics processing units (GPUs) — the primary engines of the AI boom — were hard to get.

But advancements in agentic AI are now accelerating demand for CPUs, or central processing units, on which agents perform tasks and generate outputs. Unlike chatbots that respond to prompts, AI agents can work on tasks autonomously for hours.

“Over the last several quarters, we’ve seen a new demand driver, which is really agentic AI workloads, which is really optimized better for server CPUs instead of GPUs,” KeyBanc Capital Markets analyst John Vinh told Yahoo Finance last week.

That’s been a boon for CPU makers like Intel (INTC) and AMD (AMD) as those stocks hover at all-time highs. Such is the demand that Nvidia unveiled its Vera CPU in March, positioning it as a competitor in the data center market.

“Market investors — whether hedge funds, retail investors, what have you — they love chasing bottlenecks,” Angelo Zino of CFRA told Yahoo Finance last week.

Read more: How to protect your portfolio from an AI bubble

The other growing bottleneck is the silicon memory inside AI infrastructure.

Micron (MU) stock hit all-time highs last week as memory demand has gone through the roof. Samsung (005930.KS), the world’s largest memory manufacturer, recently joined the $1 trillion market cap club. SK Hynix (000660.KS) also hit all-time highs.

Big Tech players Meta (META), Microsoft (MSFT), and Apple (AAPL) all discussed the rising cost of memory.

“Memory makers have been able to structure these long-term agreements with hyperscalers that is going to support, we think, a valuation re-rating on these memory stocks going forward,” Vinh said.

Analysts say the rush to produce high-bandwidth memory for AI chips is reshaping the entire semiconductor supply chain, including storage chips like those made by Sandisk (SNDK), which has rallied more than 400% year to date.

Minister Delegate for artificial intelligence and digital technologies Clara Chappaz (R) and Managing Director of OVHCloud, Benjamin Revcolevschi take part in a governmental visit at the data centre of French company OVHcloud in Roubaix, northern France on April 3, 2025. (Photo by Sameer Al-DOUMY / AFP) (Photo by SAMEER AL-DOUMY/AFP via Getty Images)

Government officials visit the data center of French company OVHcloud in Roubaix, northern France, on April 3, 2025. (SAMEER AL-DOUMY/AFP via Getty Images) · SAMEER AL-DOUMY via Getty Images

The third bottleneck is optics, as the industry advances toward moving data within chip infrastructure using light, or photons, rather than electricity.

Last week, Nvidia announced a partnership with Corning (GLW). The chipmaker has also made investments in Coherent (COHR) and Lumentum (LITE). Shares for those companies are also at all-time highs.

Bulls point to an AI-driven cycle that could extend the recent rally far longer. The tech industry is just starting to scratch the surface of robotics and autonomous systems, which are expected to be the next major drivers of AI.

Leave a comment