$TSLAB $GOOGLB $SPCXB #AI #谷歌 #X

An independent semiconductor industry researcher at @Evergreen Capital wrote this morning that now is a good time to buy memory stocks. He believes DRAM makers are among the stocks where the gap between share prices and fundamentals is currently widest, making them an attractive bet for investors looking to invest in AI. Meanwhile, in his view, Musk’s recent remarks about Terafab’s mass-production plans are also an “incremental tailwind” for the DRAM industry.

AI can’t exist without DRAM, and the industry is highly concentrated

Evergreen Capital first points out that DRAM offers high growth and high margins, and plays a critical role in AI workloads. He emphasizes that “there is no AI without DRAM.” He believes that if AI demand continues, memory earnings will remain significantly above past cyclical peaks for the long term.

On the supply side, he stressed that only four companies worldwide can manufacture DRAM at scale, and all are vertically integrated companies with proprietary IP. There is also currently no technology that can replace DRAM. In addition, increasingly strategic partnerships between DRAM makers and their customers, along with record capital returns, are among the reasons he is bullish on the industry.

Memory Valuations Already Price In a Steep Downturn in 2027–2028

He believes current DRAM valuations already price in a sharp cyclical downturn in 2027–2028. If AI demand continues as expected, the bearish scenario currently reflected in the market would amount to a mispricing—his main reason for believing investors should buy now.

He also noted reports that major cloud providers, including Google, are looking to secure memory supplies through 2035. If true, this would indicate a disconnect between customers’ long-term demand outlook and the cyclical downturn expectations reflected in share prices.

Musk’s Terafab to Start With SRAM, Delaying the Threat to DRAM

Citing Musk’s remarks yesterday, Evergreen Capital said Terafab will start with an SRAM chip, with mass production planned for the second half of 2028. Joking about “Elon-speak,” he suggested the actual timeline could be 2029–2030. In his view, Terafab is still several years away from mass production and will initially target logic processes rather than traditional memory or DRAM.

He speculated that this is because Musk understands how difficult it is to build memory capacity from scratch, particularly to achieve competitive yields and cost efficiency. As for Musk’s recent promotion of CXMT’s output, he believes the aim is to pressure the three major memory makers, including Micron (MU), into expanding capacity.

(Musk Meets With South Korean Giants to Push Terafab Forward! First Wafer Fab Aims for Completion Before 2030)

Could Micron Collaboration Be on the Horizon? Memory Makers’ Capacity Discipline Is Key

On the relationship between Terafab, SpaceX (SPCX), Tesla (TSLA), and the memory industry, he believes the best outcome would be collaboration among the parties. However, Terafab is still at an early stage, as are many of SpaceX’s and Tesla’s plans, so it is understandable that DRAM makers are taking a cautious approach.

As a potential partner, Evergreen Capital specifically mentioned Micron, suggesting that Micron executives could team up with Musk. The reasoning is that both sat near President Trump at the Superintelligence Summit last month, and both run pro-U.S., AI-related domestic companies of strategic importance.

In his view, the worst-case scenario would be memory makers blindly expanding capacity for the “Musk universe” before actual demand materializes, ultimately recreating a supply glut.

This article, “Starting With Musk’s Terafab Production Timeline: Three Reasons to Buy Memory Stocks Now,” first appeared on .

【Further Analysis】

【My Further Take】

From an industry supply perspective, DRAM and traditional logic chips have extremely high barriers to entry. If Musk’s Terafab initially chooses SRAM, it would essentially avoid the exceptionally complex capital spending and yield-control challenges involved in DRAM. This means the moat protecting the traditional memory duopoly (or the Big Three) will be difficult for new entrants to breach in the short term. Fears that new competitors will rapidly take market share are an overreaction, creating a margin of safety in valuations.

【Market Impact】

In the U.S. stock market, memory giants such as Micron (MU) may see the news as bullish, as the competitive threat has been pushed back and long-term supply contracts with cloud giants could support a valuation recovery. In Taiwan’s stock market, local companies such as Nanya Technology, Winbond, and memory module makers could benefit from improved sentiment among their U.S. peers, making this a potential medium- to long-term positive. As for the cryptocurrency market, although memory price fluctuations directly affect the cost of building high-performance mining rigs and AI node hardware, if overall semiconductor capital spending remains disciplined, the negative impact on broader macro liquidity and risk assets should be relatively limited.

【What to Watch Next】

Closely monitor whether cloud service providers (CSPs) actually sign large-scale, long-term memory supply contracts extending through 2035, as well as whether traditional memory giants maintain discipline in expanding capacity through capital spending. If the Big Three do not rush into blind capacity expansion in response to Musk’s remarks, the supply-demand balance should remain healthy.

【Risk Alert】

If high interest rates or geopolitical factors cause global macroeconomic conditions to weaken AI end-market demand, cloud giants may cut orders or delay purchases. In addition, if non-U.S. capacity, such as that of ChangXin Memory Technologies (CXMT), expands faster than expected, it could create potential price-war pressure on global memory prices.