Why strategist Tom Lee believes the AI bloodbath and crypto selloff is almost over
Two of the decade’s most powerful themes—artificial intelligence and digital assets—just endured their ugliest stretch in years. Mega-cap AI leaders and semiconductor suppliers were hit by a classic de-risking: rates popped, positioning was crowded, and lofty expectations met a more complicated reality. Crypto, meanwhile, saw leverage unwind, miners sell into weakness, and ETF flows turn choppy. To many, it looks like the end of the story.
Tom Lee, the cofounder and head of research at Fundstrat and a long-time student of market cycles, sees it differently. His framework points to a late-stage washout rather than the start of a structural decline. Here are the pillars behind that view—and the signals he’s watching to judge whether the worst is indeed behind us.
1) Capitulation and positioning are doing the heavy lifting
– Sentiment snapped. Surveys of individual and professional investors swung sharply bearish, while the equity put–call ratio and volatility surged—typical markers of fear rather than complacency.
– Crowding unwound. AI-linked funds and semiconductor ETFs saw material outflows; short interest rose; factor models de-grossed. In crypto, perpetual funding flipped negative, open interest bled, and forced liquidations spiked—signs leverage has been meaningfully purged.
– Price damage reached “rubber band” levels. High-quality AI leaders retraced a large portion of their 12–18 month gains, compressing forward multiples toward (or below) their pre-hype averages. In prior cycles, that kind of rapid multiple mean reversion has tended to precede stabilization.
Lee’s playbook relies on the idea that major legs down often end with forced selling, not sober re-pricing. When you see cross-asset capitulation and a collapse in speculative positioning, the marginal seller gets exhausted.
2) Rates and liquidity are shifting from headwind to neutral (or tailwind)
– Real yields plateaued. AI and crypto both behave like long-duration assets; the worst drawdowns have clustered around spikes in real rates. Once real yields stop rising—and especially if they drift lower—valuation pressure eases quickly.
– Liquidity is less hostile. Measures like M2 growth stabilizing, a calmer Treasury General Account path, and a slower drain of excess reserves tend to support risk appetite at the margin. Historically, when the rate-shock fades, beaten-up growth leadership bounces first.
Lee’s bottom-calls often hinge on “macro stops getting worse.” He doesn’t need an aggressive easing regime—just the absence of new tightening impulses.
3) AI earnings power looks dented, not broken
– Demand is diffusing, not disappearing. The first AI leg was training-heavy and GPU-centric; the next is about inference, enterprise adoption, and software monetization. That transition is noisy quarter to quarter—but it broadens the revenue base beyond a handful of hyperscalers.
– Capex remains high where it matters. Leading fabs, network providers, and select accelerators still face multi-quarter backlogs for the bleeding-edge. Even modest delivery slippage doesn’t negate multi-year capacity plans.
– Unit economics are improving. Vendors are compressing costs per token, inference throughput is rising, and AI copilots and assistants are nudging attach rates higher. As monetization gets stickier, margin narratives become less speculative.
– Valuations reset. After the selloff, several AI platform and picks-and-shovels names trade at growth-adjusted multiples closer to pre-mania ranges. If revenue durability is intact, the multiple floor is likely higher than in past tech busts.
Lee’s central claim: the “AI bubble” frame misses the earnings math. If forward revenue and margins don’t crack, price damage outpaces fundamental damage—setting up rebounds.
4) From over-owned to under-owned in a hurry
– Active managers slashed exposure into weakness; quants and risk-parity de-risked as realized vol spiked. That flips the near-term “pain trade” higher: even a mild rebound can force performance chasers back into the winners they just sold.
5) Seasonality and the political calendar help, not hurt
– Equities often find footing after midsummer drawdowns, and presidential-cycle dynamics tend to reduce policy shocks into year-end. Lee frequently leans on these historical tendencies when they align with technical exhaustion.
6) Crypto shows classic post-washout cycle markers
– Halving lag and miner stress fade. Historically, Bitcoin’s halving is followed by a digestion phase as miners sell inventory and hash economics reset; once that pressure abates, supply overhang improves.
– Leverage reset. Negative funding, flat-to-negative basis, and a drop in perpetual open interest indicate a cleaner slate. That typically precedes more durable up-moves than reflex bounces off leverage.
– Spot demand stabilizes. ETF outflows slow or reverse; stablecoin float resumes an uptrend—a simple proxy for fresh crypto liquidity. On-chain, long-term holder supply remains tight, which historically limits downside once forced selling clears.
– Valuation bands reset. Metrics like realized price, MVRV, and cost-basis cohorts compress toward levels that have attracted longer-horizon buyers in prior cycles.
The throughline is familiar in Lee’s crypto work: once weak hands are out and structural spot demand is intact, the path of least resistance shifts back to the upside.
7) What would confirm the bottom is in
– Breadth thrusts in semis and AI software: a surge in advancing issues and strong up-volume days.
– Volatility cools while prices rise: VIX and credit spreads ease without equity givebacks.
– Reclaims of key trend levels: weekly closes back above intermediate moving averages for semis and AI bellwethers; higher lows on pullbacks.
– Crypto microstructure heals: funding normalizes slightly positive, basis turns modestly premium, ETF net flows stay green for consecutive weeks, and stablecoin supply expands.
Key risks to the call
– A renewed inflation shock or a policy surprise that reprices terminal rates higher.
– A genuine AI demand air pocket: delayed deployments, slower enterprise adoption, or regulatory frictions that defer monetization.
– Geopolitical or supply-chain hits that derail leading-edge semis or cloud buildouts.
– Adverse, unexpected crypto regulation or a major centralized venue failure.
Bottom line
Lee’s stance is not that pain can’t persist day to day; it’s that the mechanics of this selloff look far closer to a finishing move than an opening act. Positioning has been wrung out, policy headwinds have softened, AI earnings power appears cyclically noisy but structurally intact, and crypto’s leverage and supply overhangs have largely cleared. In that setup, even incremental good news can trigger outsized upside as underweight investors are forced back into the very themes they abandoned.
This article is for information only and not investment advice. Always do your own research and consider your risk tolerance.
