As per Fundstrat’s head of research, Tom Lee, capital associated with the artificial intelligence sector is now flowing into Ethereum instead of chip makers. In a post he made on X on July 21, Lee implied that investors are moving away from investing in AI hardware companies toward investing in digital infrastructure, a development that might benefit Ether from institutional investment perspective.
The most recent statements come on the heels of Lee’s consistent optimistic vision on Ethereum. In the past, Cryptopolitan has disclosed that Lee believes that the network stands to gain from the increasing requirements of AI solutions for decentralized settlement, tokenization and on-chain infrastructure.
Lee pointed to the increasing performance gap between Ether and a fund tracking memory-chip manufacturers, deeming it an indication that the “AI downstream” trade is picking up steam.
The AI downstream trade continues to strengthen. Over the past month, $ETH has outperformed the Roundhill Memory ETF (DRAM) by 7,200 basis points.
The 72-point gap Lee is pointing to
As per Lee, Ether climbed by 24% over the last month whereas the Roundhill Memory ETF (ticker: DRAM) fell by 38%, making $ETH outperform the latter by 7,200 basis points or 72%.
Fundstrat provided a graph displaying the two assets going in separate directions. It did not, however, supply fund-flow statistics to back up the notion that money is leaving semiconductor stocks in favor of Ethereum. A theory about “rotation” is therefore based more on market performance than actual flows of capital.
This comparison is intentional. DRAM was released in April 2026, and it is the first ETF focused entirely on memory chip makers, including those that produce High Bandwidth Memory (HBM) chips, DRAM chips and NAND Flash chips. The chips have assumed great importance in the training and operation of large AI models, which makes the ETF a good marker for investments in AI infrastructure.
Why memory chips are the comparison
One of the largest beneficiaries of the AI boom has been memory chips. According to IDC, worldwide spending on AI is expected to reach $758 billion by 2029 and its enterprise infrastructure tracking revealed that storage systems designed specifically to support AI applications grew by 20.5% in Q2 of 2025, as businesses expanded their AI capabilities.
The prospects for the industry remain positive. TrendForce’s July report estimates an increase in the contract prices for conventional DRAM, expected to rise 13 – 18% in the third quarter of 2026, while NAND Flash prices will grow by 10 – 15% due largely to demand from AI servers.
That makes the recent downturn in the DRAM ETF much more pronounced. The decrease seems to stem from investors either cashing in their profits or making adjustments to their semiconductor stock holdings and not from any downturn in the market for memory chips.
Lee’s post did not clarify this issue, and it remained unclear whether Ethereum is benefiting from the funds being withdrawn from AI hardware or is simply profiting at a time when the chip stocks are consolidating.
What it means for the crypto market
For crypto investors, Lee’s argument offers a new way to look at Ethereum. Rather than viewing $ETH solely as the backbone of decentralized finance, the thesis casts the network as infrastructure that could benefit from the next phase of AI investment.
The concept still needs to be tested. A month’s worth of outperformance is not enough information to conclude that the capital exiting memory chip equities is going straight into Ether. Whether or not the trend is indicative of a larger institutional shift will depend on Ethereum investment products’ performance in the future and continuous flows into $ETH during the recovery phase of semiconductor stocks.
In any case, the “AI downstream” idea by Lee appears to be an interesting market event, although it still requires more proof before it can be called a lasting shift.
