Tom Lee, head of research at Fundstrat, suggests that AI capital is flowing to Ethereum rather than chipmakers. In a July 21 X article, he noted that investors are shifting away from AI hardware companies and toward digital infrastructure, which could potentially boost Ether from an institutional investment perspective.
Lee’s comments follow his continued optimistic view of Ethereum. Cryptopolitan has previously indicated its belief that the network benefits from the growing demand for AI in decentralized settlement, tokenization and on-chain infrastructure.
The performance gap is widening
Lee highlighted the growing performance gap between Ether and a memory chip ETF. He sees this as evidence that “downstream AI” business is gaining momentum. Over the past month, he noted, Ether has soared 24% while the Roundhill Memory ETF (ticker: DRAM) has fallen 38%. This represents an outperformance of 7,200 basis points.
Fundstrat shared a chart showing the two assets moving in opposite directions. However, he did not provide flow data to confirm that money is leaving semiconductor stocks for Ethereum. The turnover theory relies more on market performance than on actual capital flows.
Memory chip background
DRAM was launched in April 2026 as the first ETF focused solely on memory chip makers, including producers of high-bandwidth memory, DRAM, and NAND Flash chips. These chips have become crucial for training and running large AI models, making the ETF a proxy for investments in AI infrastructure.
Memory chips have been a major beneficiary of AI. IDC predicts that global AI spending will reach $758 billion by 2029. Enterprise infrastructure tracking shows that storage for AI applications increased 20.5% in the second quarter of 2025 as companies expanded their capabilities. TrendForce’s July report projects conventional DRAM contract prices rising 13-18% in Q3 2026 and NAND Flash prices rising 10-15%, largely driven by demand for AI servers.
The cause of the slowdown is unclear
This makes the recent decline in the DRAM ETF more notable. This likely stems from profit-taking or stock sector adjustments, not a slowdown in the memory chip market. Lee’s post did not clarify whether Ethereum is benefiting from the withdrawal of funds tied to AI hardware or whether it is simply rising while chip stocks consolidate.
Lee offers crypto investors a new perspective on Ethereum. Instead of simply being a backbone of decentralized finance, the thesis presents the network as an infrastructure that could capture the next phase of AI investment. The idea still needs to be validated. One month of outperformance is not enough to conclude that capital flowing out of memory chip stocks is flowing directly into Ether. Whether this signals greater institutional change depends on the future performance of Ethereum investment products and sustained Ether inflows during the semiconductor stock rally.
For now, Lee’s talk of “downstream AI” is an interesting market observation, but it requires more evidence before it can be called a lasting trend.
