Wall Street is undergoing a profound restructuring of its AI investment logic. Nvidia, once the undisputed leader, is being marginalized by investors, with massive amounts of capital flowing into previously overlooked chip and hardware companies such as Intel, AMD, and Micron, reflecting a broader market bet on the long-term development of AI infrastructure.
This week's market movements provide the most vivid illustration. AMD and Intel both rose by about 25% in a single week, memory chip maker Micron surged over 37%, and fiber optic cable manufacturer Corning rose by about 18%. All four companies have more than doubled their stock prices this year, with Intel leading the pack with a gain of over 200%. In contrast, Nvidia has only risen by about 15% this year, roughly in line with the Nasdaq index.
Mizuho analyst Jordan Klein characterizes this phenomenon as a "power transition" in the AI field. Behind the shift in market logic are the new hardware demands spurred by the evolution of AI applications from chatbots to AI agents, as well as the supply-demand mismatch caused by a global shortage of memory chips.
Memory shortage drives Micron to record high
The structural shortage in the memory market is one of the core drivers of this round of price increases. Micron's market capitalization surpassed $800 billion for the first time this week, and its stock price has increased by more than 750% over the past year.
In a March interview with CNBC, Micron CEO Sanjay Mehrotra stated that due to supply shortages, major customers were currently only receiving "50% to two-thirds" of their demand. Jordan Klein, in a report to clients, noted that investors who overweight memory modules during cyclical upturns will reap substantial rewards when the market quickly enters a state of substantial shortage, prices surge while costs rise only moderately. "New capacity can't keep up quickly enough; the logic is that simple."
The global memory market is mainly dominated by three companies: Micron, Samsung of South Korea, and SK Hynix, with the latter two also experiencing historic growth.
The AI intelligent agent wave reignites CPU demand.
As the focus of AI applications shifts from large model training to agent deployment, the strategic value of central processing units (CPUs) is returning. Bank of America estimates that the data center CPU market is expected to more than double in size, growing from $27 billion in 2025 to $60 billion in 2030.
AMD's earnings this week exceeded expectations across the board, with strong growth in its data center business. CEO Lisa Su stated on the earnings call that AMD now expects the server CPU market to grow by 35% over the next three to five years, significantly higher than the 18% forecast given last November. In an interview with CNBC, she said, "Agents are driving huge demand throughout the AI adoption cycle, and we are very excited to be a part of it."
Goldman Sachs and Bernstein subsequently upgraded AMD's rating to "buy," both citing a structural tailwind in CPU demand. JPMorgan analysts stated that AMD's earnings report "clearly indicates a structural inflection point in the growth trajectory of server CPUs and data center accelerators."
Intel's strong recovery is attributed to news of Apple orders.
Intel's rebound is particularly noteworthy. The chip giant, which had fallen far behind in the AI wave, is experiencing a recovery driven by massive investments from the US government. Intel's stock price more than doubled in April, marking its biggest monthly gain in history, and continued to rise by 33% in early May.
According to Bloomberg, Apple is in talks with Intel and Samsung to manufacture main processors for its U.S. devices, a report that sent Intel's stock surging 13% in a single day. Subsequently, the Wall Street Journal reported that Intel and Apple had reached an agreement to manufacture some processors for Apple devices, causing Intel's stock to rise another 14%. Representatives from both Intel and Apple declined to comment.
Corning and Nvidia sign major fiber optic contract
In the landscape of new AI infrastructure, fiber optic technology is also a significant beneficiary. This week, glass and fiber optic manufacturer Corning signed a major collaboration agreement with Nvidia, involving the construction of three new factories in the United States dedicated to serving Nvidia's optical technology needs. Nvidia will receive investment rights of up to $3.2 billion.
This collaboration is seen as a key step in Nvidia's transition from copper cabling to fiber optic cables to support the expansion of its rack-mount systems. Nvidia CEO Jensen Huang told CNBC, "We will be scaling up fiber optic capacity on a scale never before seen in the optics industry," adding that the current economy is experiencing "the largest single infrastructure build-up in human history." Corning also signed a $6 billion fiber optic supply agreement with Meta earlier this year, covering the period until 2030.
Bubble warnings emerge, analysts warn of potential pullback risk.
However, some analysts have begun to compare this round of semiconductor market gains to the dot-com bubble era. BTIG analyst Jonathan Krinksy pointed out in a recent report that the magnitude of the semiconductor sector's gains is strikingly similar to that of 1999, and warned of a 25% to 30% correction risk for the Philadelphia Semiconductor Index—which has risen 66% year-to-date.
"We've written repeatedly about how extreme the semiconductor rally has been—in many ways, it's the worst since the dot-com bubble," he wrote. "In some ways, this rally is actually more extreme than that one."
This warning reminds investors that while the long-term logic of AI infrastructure development remains valid, short-term valuation pressures and risks revealed by historical precedents should not be ignored.
