Chip stocks are surging at an historic rate. But Nvidia , the company that's been at the heart of the AI boom, isn't participating. Over the past month, stocks like AMD and Micron are up 90% and 76%, respectively. Nvidia is only up 17%. In recent trading, that underperformance has gotten worse. The GPU-maker is flat since April 27, just before big chip companies started reporting earnings. Meanwhile, Intel and Micron have popped more than 30% since that time. AMD is also up around 20% over those days, with the bulk of that advance coming Wednesday . The divergence came as first-quarter earnings revealed a bottleneck in memory chips along with progress among the hyperscalers in developing their own in-house chip systems, such as Alphabet's TPUs and Amazon's Trainium chips. One factor is concern that capex spending can't get much higher and Nvidia shares were already pricing in a best case scenario, while other stocks were not. "Within the semiconductor space, the large-cap AI bellwethers, which include Nvidia, are unfairly pricing a peak in capex, which is not in sight, certainly for the next [two to] three years. If anything, capex numbers are going up," Venu Krishna, head of U.S. equity strategy at Barclays, told CNBC on Wednesday. "The Street was materially too low in calendar year 2027 and 2028," he added. More money, more capex While more hyperscaler buildout is good for Nvidia in the long term, since much of that infrastructure will rely on Nvidia systems, the changing goal posts of additional expenditures could be weighing on the stock in the short term. Hyperscalers boosted capex projections for 2026 across the board in the first quarter. Alphabet's forecast was up 4% to $185 billion; Amazon increased it by 1% to $200 billion; Meta increased it by 8% to $135 billion; and Microsoft ratcheted it by a full 24% to $190 billion, according to a recent Bank of America tally. Both Evercore and Bank of America placed 2027 capex in excess of a $1 trillion in recent analyses . Nvidia spent $6.1 billion on capex during fiscal 2026 and said it expects "to increase capital expenditures in fiscal year 2027 relative to fiscal year 2026 to support the future growth of our business." Some analysts say Nvidia trades more like a sector than a stock. The company's dominance in GPUs is increasingly characterizing the stock as a value play, rather than a growth name. Goldman Sachs has a buy rating on the stock and a 12-month price target of $250 based on a 30 times price-to-earnings multiple. "For semiconductor companies, we see a clearly positive impact from ongoing capex spending from hyperscalers and LLM providers," Goldman analyst James Schneider wrote in a Tuesday note, citing "significantly more headroom for increased capex." Nvidia has a forward enterprise value to EBITDA ratio of 18.23, according to FactSet, while many analysts are projecting higher growth rates on top of that — a mismatch that could be weighing on the stock. "You can't escape the fact that this is a company that is valued around 20 times, but the growth rates out there that people are projecting are like 60% to 70% this year and next, and those two different factors don't add up," Matt Bryson, enterprise hardware analyst at Wedbush, told CNBC. Competitive factors While the demand growth for AI services and the bigger buildout is good for Nvidia, it's also intensifying competition, and some of the company's biggest customers are reporting successes with their own in-house chip developments. Amazon's Trainium chips and Alphabet's tensor process units (TPUs), for example, have been catching Wall Street's eye. "[There's] the idea that there's more demand for alternative solutions. We've been seeing Amazon talking about their strength in Trainium. Obviously, Google is doing really well with TPUs … there's been some speculation about Marvell building AI-related chips for Google. So there are all these alternatives," Wedbush's Bryson said. Wall Street wants more visibility into in-house chip developments among the hyperscalers. "A lot remains unknown today," Mizuho analyst Lloyd Walmsley wrote. "We assume Google can generate about ~$61B in TPU sales across '26E and '27E combined at overall Cloud segment margins of 32.4% / 38.0%." Shortages in memory chips and storage have boosted companies such as AMD and Intel, which is also likely widening the recent gap with Nvidia. However, Nvidia's GPU positioning could ultimately turn these tables. "NVDA has built a formidable moat for new compute-intensive workloads in the data center and could ultimately leverage its GPU architecture to more broadly displace INTC. New client and server CPUs from AMD also present a threat that we could be underestimating," UBS analyst Timothy Arcuri wrote. — CNBC's Ananya Chetia contributed reporting.
