#Bitmine持有578万枚ETH Tom Lee in-depth interpretation: This round of an industry bull market is still in its early stage!

In late July 2026, global technology markets saw violent fluctuations: South Korea’s Kospi index plunged sharply back, the AI chip and storage semiconductor segments all crashed in unison, US tech-heavy stocks weakened at the same time, and market panic spread rapidly. Headlines proclaimed that the “AI bubble has completely burst.” Against the backdrop of widespread bearish sentiment, Wall Street’s top bull and Fundstrat founder Tom Lee delivered a fundamentally different rational judgment during an on-site interview at the NYSE on July 27: this round of selloff is by no means the end of the AI industry story, but rather a passive deleveraging cycle triggered by a concentrated liquidation of highly leveraged positions—weak-hand cleansing in the middle of a bull market. The current level is gradually revealing long-term allocation value. Unlike the sentiment-driven view of the market, Tom Lee’s judgment is not a subjective “bullish” call; it is based on detailed market trading data, industry fundamentals, valuation frameworks, and historical cycle规律. As a seasoned institutional manager with stable annualized performance, he runs the GRNY ETF with a size of $5 billion. This year, his performance has outpaced 92% of its comparable thematic funds. He also serves as Chairman of the Board of BitMine Immersion Technologies (BMNR), the world’s largest Ethereum- holding institution, having worked deeply in the technology sector and the crypto asset space for many years. He has precise research experience in the cycle of growth tracks, leveraged volatility, and the pace of industry implementation.

1. The truth behind the selloff: it’s not a collapse in fundamentals, but a chain-reaction liquidation crisis of extreme leverage
The market commonly attributes this round of AI semiconductor selloff to “AI demand peaking” and “bubble exhaustion,” but Tom Lee, by breaking down global trading data, shows that the core trigger of this adjustment is structural forced deleveraging, with no reversal in the underlying industry logic. The epicenter of this turmoil is the Korean market. Over the past two years, South Korea’s capital markets have significantly loosened leverage trading restrictions. More than a dozen leveraged ETFs linked to key semiconductor targets such as Samsung Electronics and SK hynix were launched in bulk with 2x leverage, while high-multiple margin trading and short-term financing & securities lending permissions for individual stocks were also opened up. In the earlier uptrend driven by AI compute power and memory chip price increases, many retail investors and small-to-mid institutions used extreme leverage to follow the trade, rapidly pushing up sector valuations—accumulating a massive stack of fragile leveraged positions across the entire AI semiconductor track. This extreme leverage structure has very strong bidirectional volatility characteristics: when the market rises, leveraged capital accelerates and amplifies the upside, magnifying the “making money” effect; once the market experiences even a modest pullback, mark-to-market losses in leveraged accounts expand quickly, triggering brokers’ and asset management firms’ forced liquidation lines. The resulting passive selling further depresses stock prices, which causes more leveraged accounts to liquidate, forming a negative loop of “decline—liquidation—another decline,” ultimately evolving into indiscriminate flash crashes.
Key data from US investment banks and prime brokers further corroborate this view: the recent speed at which hedge funds reduced long positions in tech stocks hit a peak not seen in nearly a decade. This is not that institutions were actively bearish on the AI industry; rather, many leveraged strategy products hit their risk-control thresholds and were forced to de-lever and cut positions to stop losses. At the same time, US market financing balances year-over-year increased sharply by 54%, reaching the sixth-highest growth rate in nearly 60 years. Historical data indicates that once leverage expands at this level, the market will inevitably undergo a concentrated deleveraging shock and adjustment cycle.
Tom Lee also emphasized the divergence between price action and fundamentals: throughout this decline, there was no core bearish industry signal. Global cloud computing vendors’ AI compute procurement orders continued to land, the semiconductor and storage chip consumption linked to global GDP kept rising steadily, and demand for AI server iteration and upgrades remained rigid. Meanwhile, South Korea’s leading semiconductor companies’ capacity orders and revenue guidance were not downgraded. In short, what is falling is the sentiment and positioning of leveraged capital, not the real demand and growth potential of AI. After this round of concentrated clearing, “weak-hand capital” that chased rallies, cut losses, and relied on leveraged games has basically exited the market. Remaining positions are mainly long-term institutional spot capital. The market’s position structure has improved significantly, and the subsequent AI segment’s volatility resilience will be markedly enhanced—setting a foundation for the repair of the next leg of the行情.

2. Core logic: in a structural bull market, you earn time—not the gap in trading swings
(#－.－) During this market panic, many participants liquidated out of fear after short-term plunge, or repeatedly tried to bottom-catch and top-sell in an attempt to trade swings. In the end, they missed the rally or incurred losses. On this, Tom Lee reiterated the core rule of long-term profitability in growth tracks: AI is an epic-scale structural bull market. The key to making money is steadfastly holding, not frequently trading to time entries and exits.
He cited classic ideas from two investment masters to support his core view: Peter Lynch famously said, “Selling an appreciating quality stock is like cutting off a blooming flower—you go water the barren weeds instead.” And Charlie Munger pointed out sharply that “the big money in the capital markets is never made by high-frequency buying and selling; it is made by waiting patiently and holding quality assets until the right time.”
Based on the current state of the AI market, Tom Lee provided specific interpretation: all players attempting to precisely forecast the tops and bottoms of AI segments and chasing short-term swing trades will, with high likelihood, after a rebound, chase at the high levels and repeatedly grind down returns—then miss the main advance. The current AI industry is in the early stage of technical implementation, the mid-stage of capacity construction, and the early stage of demand explosion—there is no fundamental signal that the industry has topped out. As long as investors recognize the long-term logic that “AI is reshaping global industries, and compute power becomes the core productive asset,” the optimal strategy is to buy core assets and hold long term, ignoring short-term emotional volatility.

3. Valuation breakdown: valuation diverges across sub-segments; AI core assets are not overvalued
The market’s main reason for being bearish on AI is “valuation is too high and bubbles have been priced in,” but Tom Lee, through a precise breakdown of valuations and business models across sub-segments, shows that current AI core assets are fairly valued. In some sub-segments, assets are even severely undervalued. The market’s valuation anxiety stems from broad-brush judgments about whole sectors, not from a refined breakdown.
1、Nvidia: the core moat is solid; current valuation is significantly undervalued
There is a widely held rumor that Nvidia’s P/E is as high as more than 20x and that valuations are bubble-like, implying a major pullback risk. But Tom Lee corrected the market’s valuation misconception: Nvidia’s current forward dynamic P/E is only 16x, far below what the market subjectively believes. In terms of barriers and growth, Nvidia has a unique CUDA ecosystem moat. Global compute software adaptation, developer ecosystem, and server deployments are highly bound to the Nvidia system—there is no near-term substitute. At the same time, the company has a clear and continuous chip iteration roadmap. Demand for AI servers, edge computing, and industrial AI keeps expanding, fully supporting a “frontline growth stock” valuation of 25-30x. Compared with the valuation frameworks of global tech leaders, Nvidia at a 16x PE is at a historical low valuation level and offers ample room for valuation repair.
2、Memory chips and semiconductor equipment: strong cyclical attributes; low P/E does not mean low risk
Unlike Nvidia, the memory chip and semiconductor equipment tracks have very strong cyclical attributes, so you cannot simply judge value using static P/E. Currently, the P/E of South Korean memory-chip stocks has fallen to around 4.5x, which appears extremely cheap. But Tom Lee reminds investors not to blindly bottom-fish. A low static P/E might be a leading signal of a cycle downturn. The core criterion is not the current valuation, but whether future enterprise earnings can continue to be revised upward. Over the long term, the explosion of “machine-to-machine” compute interaction scenarios driven by AI will make the demand ceiling for semiconductors and storage chips far exceed traditional human consumption scenarios. The long-term room for earnings recovery is clear, but in the short term, the market still needs to absorb the risks of cyclical volatility.

4. Historical cycle comparison: comparable to Cisco’s “golden decade”; the AI bubble is far from its end stage
To help the market more clearly understand which phase the current AI industry is in, Tom Lee used Cisco, the core leading enterprise of the internet era, as a classic benchmark. 1993-2000 was a “golden seven years” of internet infrastructure explosion. Over seven years, Cisco’s stock price surged 100x, completely replicating the growth logic of this round of AI industry. But in this super bull market, Cisco was not a straight-line ascent; during the period it experienced at least four deep drawdowns at “halving”-level severity. Tom Lee clearly defined the characteristics of the true end-stage of an industry bubble: the industry’s overall valuation reached above 200x P/E, and the market was filled with unrealistic growth expectations. But the current AI industry fundamentals are exactly the opposite: the market is collectively panicking and疯狂ly liquidating. That is precisely the emotional bottom characteristic of the middle of a bull market.

5. Rationally view the impact of domestic open-source models: competition intensifies; downstream opportunities stand out
Tom Lee acknowledged the rapid iteration of domestic open-source models, which significantly lowers the application barrier for the AI industry. But he also pointed out that the widespread adoption of open-source models is not a negative for the industry—instead, it accelerates the commercialization and deployment of AI, creating downstream segment tailwinds such as AI software applications, industry solutions, data services, and encryption security. This is also the core logic behind the GRNY ETF he manages, focusing on AI downstream applications and network security.

6. Ethereum deep logic: the era of yield-bearing assets is arriving; institutions’ allocation window opens
As Chairman of the board of the world’s largest Ethereum-holding institution, Tom Lee clearly stated: Ethereum is a yield-bearing digital financial asset, and the institutional era has just begun. Currently, Ethereum staking annualized yield is stable at about 3%. BitMine holds 5.78 million ETH; its current weekly staking收益 is about $6 million, and its annual stable收益 can reach $300 million. If Ethereum’s price subsequently rebounds to $5,000, the company’s annual staking收益 will approach $1 billion.

7. Macros and the bigger trend: inflation turning point shows up; Fed dovish cycle supports growth tracks
Beyond industry and track logic, the macro environment provides solid support for the subsequent行情 of AI and crypto assets. Tom Lee expects that in the second half of 2026, global inflation will continue to stay below market expectations. The upward pressure on oil prices that previously hit the market has already peaked and fallen back. Two major core inflation stickiness indicators—real estate rents and labor wage growth—have continued to weaken, while global inflation pressure has eased steadily. Cooling inflation will directly push the Fed to end tightening policy and shift to a dovish easing cycle. Market liquidity conditions will keep improving. In a liquidity-easing cycle, high-growth AI technology and digital asset tracks often receive an excess valuation premium. Based on this, the GRNY ETF continues to stick to three core tracks: AI downstream applications, network security, and quality small-cap growth stocks. It ignores short-term market volatility and holds core positions long term.

8. Final conclusion and practical suggestions
Market characterization: This round of global AI and semiconductor selloff is an emotion- and trading-driven adjustment triggered by concentrated liquidations of highly leveraged funds. It is absolutely not a reversal of industry fundamentals. The story of the AI super-growth track is far from over. The current bubble is only in its early stage. Position structure: short-term high-volatility clearing has flushed out short-term speculative capital, while long-term institutional positions have lower cost and more stable holdings. Overall, the market’s collective panic comes from players confusing “leveraged-行情 volatility” with the “end of the industry cycle.” From a long-term industry perspective, the progress of AI compute infrastructure buildout, digital transformation, and intelligent ecosystem deployment has only just started. The short-term crash is just a deep shakeout during the bull market. The real industry dividends and the main upswing still lie ahead. $BTC ‌