Independent semiconductor researcher Evergreen Capital said in a post early Thursday that now is a good time to buy memory stocks. In the researcher's view, for investors looking to bet on AI, DRAM makers are among the clearest examples of a gap between stock prices and underlying fundamentals. Evergreen Capital also said Elon Musk's recent discussion of Terafab's production roadmap should be seen as an incremental positive for the DRAM industry.
Why Evergreen Capital is positive on DRAM
The first part of the argument focused on demand and industry structure. Evergreen Capital said DRAM combines high growth with high gross margins and remains critical to AI workloads, adding that "there is no AI without DRAM." If AI demand continues, the researcher said, memory profitability could stay well above prior cycle peaks for a long period.
On the supply side, Evergreen Capital said only four companies in the world can produce DRAM at scale. All four are vertically integrated and hold proprietary intellectual property, while no technology currently exists that can replace DRAM. The researcher also cited increasingly strategic relationships between DRAM makers and customers, along with record capital returns, as reasons for the constructive view on the sector.
Valuations already reflect a sharp downturn scenario
Evergreen Capital argued that current valuations for DRAM makers already price in a steep cyclical decline in 2027 and 2028. If AI demand continues as expected, the pessimistic scenario embedded in today's prices would amount to a market mispricing, which the researcher described as the main reason to buy now.
The post also referred to reports that large cloud providers including Google may be looking to secure memory supply through 2035. If true, that would indicate a mismatch between how customers see long-term demand and the cyclical downturn implied by current share prices.
Terafab starts with SRAM, not DRAM
On Terafab, Evergreen Capital cited Musk's comments from the previous day and said the project will begin with an SRAM chip, with mass production targeted for the second half of 2028. The researcher joked in "Elon-speak" that the actual timeline may land in 2029 or 2030.
Based on that reading, Evergreen Capital said Terafab remains several years away from mass production and is likely to target logic processes first rather than traditional memory or DRAM. The researcher said Musk likely understands how difficult it is to build memory capacity from scratch, especially if the goal is to reach competitive yields and economics.
Evergreen Capital also said Musk's recent promotion of ChangXin Memory Technologies, or CXMT, output appears aimed at pressuring the three major memory makers, including Micron (MU), to expand capacity.
Potential cooperation and the main downside risk
Discussing the relationship between Terafab, SpaceX (SPCX), Tesla (TSLA), and the memory industry, Evergreen Capital said the best outcome would be cooperation among multiple parties. At the same time, the researcher noted that Terafab is still at an early stage, and many plans at SpaceX and Tesla are also still developing, making a cautious response from DRAM makers understandable.
As for possible partners, Evergreen Capital specifically mentioned Micron and said it may be possible to see Micron executives work with Musk. The reason given was that both were seated near President Trump at the Superintelligence Summit last month, and both run domestic companies that are pro-America, tied to AI, and strategically important.
The worst-case scenario, according to the post, would be memory makers blindly expanding capacity for the "Musk universe" before actual demand appears, eventually repeating an oversupply cycle.