Fundstrat: Semiconductor Sell-Off Has Precedent, Recommends Holding Through the Dip
Miles Bennett
Micron is down 22% in a month and Sandisk 44%, as the memory-chip chain faces a concentrated sell-off. Fundstrat's Tom Lee argues the pullback is event-driven, not a signal that AI demand has peaked, and advises holding through it.
Why are semiconductors getting hit so hard?
Surging oil prices pushed Treasury yields higher, compounded by rising Fed rate-hike expectations — prediction markets put the odds of a 25-bp hike at roughly 27%, while Fed futures imply about 38%.
This means → two pressure lines are tightening at once: inflation fears + rate uncertainty. Capital is rotating out of high-multiple semis first.
Two sector-specific triggers added fuel: reports that Chinese state-backed firms have begun mass-producing certain chipmaking equipment, seen as a threat to ASML's dominance; and Nvidia's $250 billion guarantee for OpenAI's data-center buildout, which raised concerns about "circular spending" — companies buying from each other to prop up revenue, vulnerable to a chain reaction if one link breaks.
Why does Tom Lee say "don't panic"?
He sees the actual probability of a hike as "very low" and frames it as a binary event — markets can't find equilibrium ahead of a binary outcome, so some money is simply hedging, not signaling a trend reversal.
On oil, the easing of U.S.–Iran tensions has already pulled prices back roughly $10 from the peak. The 10-year Treasury yield has retraced nearly half its rise, and the market-implied probability of a U.S. invasion of Iran has dropped to 25%.
In plain terms = the lines driving the sell-off are all loosening. The probability of the worst-case scenario is falling, not rising.
Is capital really fleeing risk assets across the board?
Tom Lee notes that the Mag Seven, software stocks, and even crypto are all recovering — money has not left risk assets wholesale.
This reflects a structural factor: investors positioning for August are buying software and shorting semis, a rotation trade — moving money between sectors, not exiting — that has been running for some time.
This means → the semiconductor drawdown is amplified by rotation. It should not be read as "the market has turned bearish on AI."
What does the Cisco precedent show?
Tom Lee cites Cisco's trajectory during the 1990s internet boom: the stock rose from $0.80 to $9, pulled back 40%; a year later it reached $18, dropped 41% back to $9; and ultimately hit $80 in 2000 — roughly a 100× gain.
In plain terms = two near-halving drawdowns along the way did not break the decade-long thesis. Tom Lee believes AI is tracing a similar path — "we're roughly at that point today."
He also invoked Charlie Munger: "The big money is not in the buying and the selling, but in the waiting."
What comes next?
Tom Lee expects the S&P 500's week-long chop to "resolve in a day or two," with July's closing price likely above current levels.
Two key validation windows this week: the Fed rate decision and hyperscaler earnings.
This means → if the Fed holds steady and cloud-capex guidance stays intact, Tom Lee's thesis — event-driven pullback, not a demand peak — picks up its first round of supporting data.
Content is for reference only, not financial advice.