Micron Surges 12% as Chip Stocks Rebound; Chinese AI Model Emerges as New Catalyst?

According to MarketWatch, chip stocks are heating up again, as investors gradually accept a new narrative: factors once seen as threats from the artificial intelligence (AI) investment boom are now becoming catalysts that drive the market.

One recent focus of market discussion is Moonshot AI from China, which launched the open-source AI model Kimi K3. Despite facing financial and technical constraints, Kimi K3 has demonstrated performance in certain benchmark tests that rivals top-tier closed AI models in the U.S. This has reignited market questions about whether other AI developers have overinvested in hardware and computing power.

Nancy Tengler, CEO of Laffer Tengler Investments, said this market reaction reminds her of last year when the Chinese open-source model DeepSeek triggered a sharp sell-off in chip stocks—only to create a 'perfect opportunity to buy the dip.'

On Tuesday (21st), major memory chipmakers rebounded sharply. Micron Technology (MU-US) surged 12.2%, SanDisk (SNDK-US) jumped 14.4%, and SK Hynix ADR (SKHY-US) rose 13.8%.

Storage device manufacturers also performed strongly: Western Digital (WDC-US) climbed 12.5%, and Seagate Technology (STX-US) gained 11.1%.

This rebound followed a broad sell-off in chip stocks the previous week, during which the Philadelphia Semiconductor Index briefly entered bear market territory.

Brian Mulberry, market strategist at Zacks Investment Management, said chip valuations have returned to a level 'where investors can re-enter and participate in future price appreciation.' He noted that the market still holds high expectations for sustained earnings per share (EPS) growth, and strong earnings reports recently from non-tech manufacturing firms have further boosted confidence in chip stocks.

Vivek Arya, analyst at Bank of America, said China's recently launched open-source AI models actually support his bullish outlook on the memory industry and Micron stock.

In a report issued Monday, Arya pointed out that Chinese firms charge developers significantly less for using their open-source models compared to Western companies. He believes this reflects a business model choice, not lower hardware costs.

More importantly, Arya emphasized that even though Chinese AI models use different technologies to improve efficiency and reduce computational intensity, the required memory capacity remains the same or even increases as model weights and active parameters grow.

Model parameters are variables used by AI to learn and recognize patterns during training, essentially serving as the model's 'memory.' For example, Kimi K3 has a total of 2.8 trillion parameters, with approximately 50 billion being active.

Additionally, Arya noted that whenever a company downloads an open-source model, it must independently configure hardware and memory to run it—unlike closed models, which do not require such setup.

Arya stated that Chinese models also have a cost advantage over U.S. AI models. He attributed this primarily to more efficient model architectures and lower physical infrastructure costs in China, such as electricity, labor, and land. Other cost differences stem from government subsidies and the abundant free cash flow of Chinese cloud providers like Alibaba (BABA-US) and Baidu (BIDU-US).

On the other hand, Arya does not believe that ChangXin Memory Technologies (CXMT), China’s DRAM manufacturer, expanding its DRAM production capacity poses a threat to Micron.

He pointed out that unlike global DRAM leaders Micron, SK Hynix, and Samsung Electronics, CXMT does not focus on HBM (High Bandwidth Memory), which is the product most in demand by AI chipmakers like NVIDIA (NVDA-US). Arya added that it remains unclear whether CXMT can serve U.S. customers.

Hendi Susanto, portfolio manager at Gabelli Funds, said that despite memory stocks having risen sharply in recent months, 'the investment case remains highly attractive,' as the market expects demand to exceed supply at least through next year. This implies 'favorable pricing and strong profitability' can continue.

According to Dow Jones market data, the Philadelphia Semiconductor Index rose 5.2% on Tuesday, marking its best single-day performance in over a month.