Traders are no longer treating the AI boom as a one-company story. The immediate trigger this week was Marvell’s violent re-rating after Nvidia CEO Jensen Huang called it the next trillion-dollar company at Computex, but the more important signal is what happened around that headline: the market started rewarding the infrastructure around AI, not just the obvious GPU winner.
That matters because the new leadership bucket is more practical than flashy. Marvell sits in the interconnect and custom-chip layer that helps large AI systems move data efficiently. Reuters also noted that SK hynix and Micron have already exploded in value over the past year, while Kioxia briefly became Japan’s second-most valuable company. In other words, the market is paying up for memory, storage, connectivity, and the tools needed to keep giant AI clusters fed.
The Korean leg of the story is straightforward: SK hynix remains one of the clearest public proxies for AI memory scarcity and ecosystem bargaining power. The Japanese leg is more interesting because it hints at breadth. If Kioxia and other Japan-listed suppliers are being rerated, traders are starting to believe the AI capex cycle is spreading into the less celebrated parts of the semiconductor stack, not staying trapped inside a few U.S. megacaps.
Europe still looks like the discipline layer of this theme. ASML’s first-quarter update said AI-related infrastructure investment continues to support demand for advanced logic and memory chips and for the lithography tools needed to build them. That does not make Europe the hottest momentum market in the chain, but it does make ASML a useful reality check: if equipment demand stays firm, the broader AI trade has industrial backing rather than pure message-board heat.
My cautious read is that this is a healthy rotation, but also the kind of rotation that can get crowded fast. Once the market moves from first-order winners into second- and third-order beneficiaries, valuation discipline usually weakens. Traders are effectively betting that AI spending will remain so intense that networking, memory, storage, testing, and wafer tools all deserve scarcity multiples at the same time. That can keep working, but it becomes much more sensitive to any slowdown in cloud capex or any sign that revenue timelines are slipping.
The cross-market signal is still bullish for the AI complex, but more selective than the headlines suggest. If the trade keeps broadening, the winners may come from the boring plumbing names rather than the celebrity chip names. If it stalls, those same second-derivative names could also unwind the hardest because they have just started absorbing the full optimism premium.
Risk notice: This article is for market commentary only, not personalized investment advice. AI-linked equities, semiconductor suppliers, and related derivatives can be highly volatile, and fast narrative shifts can produce sharp losses as well as gains.
Sources:
Reuters via Investing.com: Marvell shares touch record high after Nvidia’s Huang calls it ‘next trillion-dollar company’ (June 2, 2026)
Reuters via Investing.com: Morning Bid: Marvell, a fitting name for the latest AI darling (June 3, 2026)
SK hynix Newsroom main page, including 2026 AI memory and storage releases
ASML Q1 2026 financial results
Reddit discussion: Marvell +25% premarket after Nvidia’s Huang call
原创文章，作者：financial transaction，如若转载，请注明出处：https://www.fanbi.net/archives/213
