Breaking: July 22, 2025, 09:42 UTC — Tom Lee, Fundstrat’s head of research, just told Bloomberg that AI money is rotating into Ethereum. His evidence: ETH outperformed a DRAM ETF by 72% between June 25 and July 21.
The math checks out. The premise doesn’t.
Because Tom Lee isn’t just an analyst. He’s also chairman of BitMine, a publicly listed company that holds 5.77 million ETH — roughly 4.8% of the entire circulating supply.
Let me be clear: this isn’t analysis. It’s position management dressed up as macro research. And as someone who spent 2021 tracking whale wallets through the BAYC liquidity crash, I learned the hard way that when the largest holder speaks, you don’t buy the thesis — you audit the conflict.
Context: Why now?
The timing is deliberate. The DRAM ETF (ticker: CHPS) had a blistering run earlier this year, raising $6.5 billion rapidly and peaking at $81. Then supply overhang fears — Samsung’s legal tussle and memory glut — knocked it down 12% by June 25. That’s the starting date of Tom Lee’s “72% outperformance” clock.
Meanwhile, ETH had already bottomed at $1,500 (down 61% from its all-time high) and staged a modest recovery. The relative return looks impressive only because the denominator collapsed. Cherry-pick the right window, and any asset looks like a genius trade.
But the real context isn’t price action. It’s the narrative layer: “Institutional adoption.” BlackRock’s BUIDL fund, Robinhood Chain, and a few ETF filings are cited as proof that ETH is the settlement layer of choice. I’ve seen this movie before. In 2020, Yearn.finance had similar hype — until manual rebalancing lagged automated strategies by 15%, and the market realized the yield was mostly inflation.
Core: The 72% illusion and what it really means
Let’s break down the actual data, not the headline.
- The window is everything. From June 25 to July 21, DRAM ETF dropped another ~12%, while ETH rose ~11%. That’s a 23% divergence, not 72%. The 72% figure is the ratio of returns (11.9% ETH vs 6.9% DRAM, but misrepresented). A 72% relative outperformance implies ETH gained 72% more than the other asset, not a 72% difference in absolute return. In reality, ETH’s absolute gain over the period was about 10.9%, and DRAM’s was roughly -5%. The ratio is ~16x, not 1.72x. This is either sloppy arithmetic or deliberate framing.
- Supply concentration is the real story. BitMine’s 4.8% holding means that any positive news from its chairman can be used to juice the price for a discrete exit. I audited a similar pattern in 2021 when a prominent NFT collector pumped floor prices on Twitter before dumping. The mechanics are textbook: create a narrative, wait for FOMO, distribute.
- No evidence of rotation. The article cites zero data on actual capital flows from AI ETFs to ETH ETFs. The CoinShares weekly report shows ETH inflows were positive but modest over the same period — nowhere near the magnitude needed to support a “rotation” thesis. If AI money were truly rotating, we would see DRAM ETF outflows and ETH ETF inflows. We don’t.
- Ecosystem vs. token price. BUIDL and Robinhood Chain are real projects, but they don’t drive ETH demand the way people think. BUIDL is a tokenized fund with only $300M AUM on Ethereum — negligible relative to ETH’s $200B market cap. Robinhood Chain is an L2 that settles on Ethereum but uses its own token. The value accrual to ETH is indirect and diluted by L2 competition.
Contrarian: What everyone misses
Everyone is asking “Is AI money rotating into ETH?” The better question is: “Why would AI money leave one speculative asset for another?” The narrative assumes that capital must choose. In reality, institutional money can rotate out of both if risk appetite shifts. The 72% gap is not a moat; it’s a volatility trap. If DRAM ETF rebounds 15% (Jefferies predicts memory prices up 50% by Q3 2026), the relative performance collapses overnight.
Secondly, the article completely ignores Solana. Solana’s DeFi TVL has grown 200% YTD, and its AI-related projects like io.net are attracting real compute demand. If AI money is coming to crypto, it’s just as likely to land on Solana’s low-fee, high-throughput environment. Yet Tom Lee’s narrative conveniently omits competitors.
Lastly, the sheer volume of BitMine’s holdings (4.8% of supply) is a systemic risk. If the price rises 20%, selling just 10% of that stash could crash the market. This isn’t a bullish signal — it’s a Sword of Damocles.
Takeaway: Watch the data, not the mouthpiece
Over the next 14 days, the only signal that matters is the next memory-chip earnings report (Samsung, Hynix). If DRAM guidance is strong, the rotation narrative dies. If guidance is weak, ETH might squeeze higher — but only until BitMine decides to take profits.
I’ve been on the other side of these trades. In 2022, when Terra collapsed, I raced to audit stablecoin codebases and published a risk report that saved my readers millions. The lesson: speed without precision is just noise. Tom Lee’s 72% figure is noise.
17 reveals the true cost of trust. Yield farming isn’t innovation; it’s liquidity arbitrage until the music stops. The BAYC crash wasn’t a market correction; it was a liquidity extraction event.