美光和 SK 海力士的万亿美元级上涨显示，AI 基础设施交易正在走出 GPU

Memory is Becoming the Next AI Bottleneck After GPUs

The latest round of stock market movements shows that investors are no longer pricing AI solely through GPU leaders. Large-scale AI models require not only GPUs but also high-bandwidth memory, advanced DRAM, storage, networking equipment, and high-energy-consumption data center capacity.

This is why the recent rallies of Micron and SK Hynix are noteworthy. Analysts have significantly raised their price targets, providing a new valuation anchor for Micron's memory chip narrative; the rise in SK Hynix further reinforces a key judgment: high-bandwidth memory is becoming one of the most clearly identifiable bottlenecks in AI hardware.

When a bottleneck becomes sufficiently clear, the market typically first re-prices the companies closest to that bottleneck.

This Rotation Trade is About Infrastructure Beta, Not Just AI Labels

The signal from this multi-asset linkage is not that all AI-related assets should rise together. A clearer interpretation is that capital is seeking exposure to the next layer of AI infrastructure.

In the stock market, this may manifest as attention on companies related to memory chips, semiconductor equipment, data center power, and networking equipment. In the crypto market, the more relevant assets are not simply those with an AI label, but networks related to compute, storage, data, oracle infrastructure, or AI agent tooling.

Therefore, projects like Bittensor, Render, Akash Network, Filecoin, Internet Computer, NEAR Protocol, Chainlink, Artificial Superintelligence Alliance, Virtuals Protocol, Worldcoin, and Grass may fall under observation, but they do not belong to the same asset class.

The connection between compute/storage-related projects and AI infrastructure is clearer, while AI agent-related projects are often more sentiment-driven, potentially rising faster during periods of heavy AI news flow, but also exhibiting greater volatility.

The Biggest Risk is Mistaking a Memory Cycle for a Permanent AI Supercycle

The greatest risk in this trade is that memory remains a highly cyclical industry. When supply is tight, pricing power and earnings expectations can rise rapidly; but when supply catches up with demand, inventories build, or demand expectations cool, the same trading logic can reverse just as quickly.

This is important across multi-asset markets. Semiconductor stocks have support from earnings, profit margins, supply agreements, and analyst models, whereas many AI-related crypto assets are still traded more heavily on narrative and future potential.

If the rally in memory chip stocks can sustain for multiple trading sessions, the AI infrastructure theme may continue to spread to high-beta assets. Conversely, if chip leaders begin to fail at breakout levels or memory price expectations weaken, AI and DePIN-related crypto assets might decline faster than their equity counterparts.

FAQ

Why are memory chip stocks rising?

Because investors are re-pricing the importance of high-bandwidth memory and advanced DRAM within AI infrastructure. GPUs remain critical, but AI systems equally require memory, storage, networking, and data center capacity.

Why is this important for the broader AI trade?

It indicates the market is no longer just rewarding the most obvious AI leaders. Capital is flowing deeper into the AI supply chain, specifically targeting segments that could become bottlenecks.

Which crypto assets are more closely related to this theme?

More direct mappings exist on infrastructure-related assets, including compute, storage, data, oracles, and DePIN networks. Related projects include Bittensor, Render, Akash Network, Filecoin, Internet Computer, NEAR Protocol, Chainlink, and Grass. AI agent or AI application-related projects like Artificial Superintelligence Alliance, Virtuals Protocol, and Worldcoin may also be influenced by sentiment, but their connection to the memory chip cycle is typically more indirect.

Will rising memory chip stocks directly improve the fundamentals of AI-related tokens?

Not necessarily. Micron and SK Hynix can directly benefit from stronger memory demand and price expectations. However, most AI-related crypto assets do not directly earn revenue from memory chips, so their price reaction is more likely driven by narrative beta and risk appetite.

What should be watched next?

Key factors include whether the semiconductor rally can continue, whether memory price expectations remain strong, whether AI infrastructure assets see broader participation, and whether the rally in AI and DePIN-related crypto assets is supported by genuine trading volume, rather than just short-term news-driven sentiment.