The AI investment frenzy is far from over, but the market spotlight is gradually shifting beyond GPU titan Nvidia. Two names now commanding attention are custom-chip specialist Marvell Technology (MRVL) and memory powerhouse Micron Technology (MU) — up roughly 129% and 700% respectively over the past year. Faced with already steep charts, investors are asking one unavoidable question: is it too late to buy?
Marvell’s explosive run is built on two high-growth pillars inside AI infrastructure.
The first is custom ASICs. Unlike general-purpose GPUs, Marvell designs application-specific chips tailored for large clients, delivering superior efficiency and lower power consumption in AI inference workloads. Goldman Sachs estimates that as soon as next year, custom ASIC shipments could equal GPU sales. Marvell signaled in May that its custom chip revenue could more than double in the coming fiscal year, dramatically accelerating from just 20% growth in the current period, with its customer base expanding rapidly.
The second pillar is optical networking. As AI data centers struggle to move massive datasets fast enough, optical connectivity has become a critical bottleneck and a booming market in its own right. Goldman Sachs projects a ninefold surge in optical component sales in just two years. Marvell expects its data-center interconnect business to grow 70% this year, while its switching unit is targeting $1 billion in revenue by fiscal 2028. Management sees the total addressable data-center market reaching $94 billion in 2028, and the company is aiming for a 20% share. That alone would translate into roughly $19 billion in revenue — more than triple its fiscal 2026 data-center top line.
It is this visible growth that prompted Nvidia CEO Jensen Huang to say publicly that Marvell could be the next chip designer to join the trillion-dollar market-cap club. With a current market cap of around $215 billion, that implies a potential near-fivefold upside. Yet the price tag is no longer modest. Marvell trades at a trailing price-to-earnings ratio of 94, a forward P/E of 67, and a price-to-sales ratio of 27. That premium embeds high expectations but also leaves little room for error.
If Marvell is a story of market-share expansion, Micron is undergoing nothing short of a historic reshaping of a cyclical industry.
As the only major U.S.-based manufacturer of DRAM, NAND flash, and high-bandwidth memory, Micron is deeply embedded in the physical backbone of AI data centers. HBM is stacked tightly alongside GPUs and is essential for training large language models and enabling split-second inference. With demand far outpacing supply, Micron recently disclosed it has secured roughly $100 billion in strategic long-term agreements, enjoying pricing power seldom seen in the memory industry’s history.
That momentum has set Wall Street ablaze. Melius Research analyst Ben Reitzes raised his Micron price target three times in two months — from $700 to $1,100, and ultimately to $2,200. That implies a nearly $2.5 trillion market capitalization and still more than 100% upside from current levels. A gain of nearly 700% over the past year looks increasingly like the opening act of a much longer narrative.
Yet alongside the euphoria come sobering warnings. Three decades of market history show that every disruptive technology in its early stages breeds bubbles that eventually burst. By some estimates, AI-related stocks now account for roughly 40% of total U.S. market value. Time and again, investors have overestimated how quickly new technologies are adopted, while underestimating how long it takes for businesses to truly optimize them and convert hype into profits. The internet revolution, for instance, took more than five years after the dot-com bust before its full earnings potential materialized. Micron’s forward earnings multiples may still look reasonable, but as a memory supplier sitting upstream in the AI supply chain, any jolt to AI capital spending or sentiment could hit it disproportionately hard.
Marvell and Micron — one betting on the customization wave in compute architecture, the other cornering the physical memory layer essential for data throughput — are both indispensable picks-and-shovels plays in the AI buildout. Massive addressable markets, triple-digit growth targets, and hard contractual backlogs provide powerful long-term support for the investment thesis.
But for those tempted to chase them today, the question isn’t whether the story is compelling. It’s whether the price already reflects more than perfect execution. Marvell’s near-100-times earnings multiple must be digested by consecutive quarters of above-consensus growth; behind Micron’s eye-popping price targets lurks the structural risk of a broader tech valuation compression. Investors with a strong stomach for volatility and a multi-year horizon may treat these names as a window into the next phase of AI. Those who demand a margin of safety might be better served waiting for the inevitable moment when euphoria cools and a better entry point emerges. On AI’s long and promising road, great companies are not the same thing as great stocks at any price.
