USD Implodes 17% And June 8th Could Be Worse

ProShares Ultra Semiconductors (USD) fell about 17% on June 5 as Broadcom’s weaker-than-expected AI revenue outlook, a hot payrolls report, and a 12 bps jump in the 2-year Treasury yield combined to hit semiconductor valuations. Broadcom dropped 8% to $385.73 after guiding Q3 AI semiconductor revenue to $16.0B versus a $17.2B whisper, while NVIDIA fell 6.2% and the chip complex lost roughly $1.2T in market cap. The article frames the move as a sector-wide AI capex reset rather than a tracking error in the 2x ETF.

Analysis

The key second-order effect is that the selloff is less about semiconductor demand vanishing and more about the market repricing the duration embedded in AI capex. That matters because the most levered beneficiaries of the buildout are also the names with the highest multiple sensitivity to any sign that hyperscaler spending is becoming more conditional, which is why AVGO can break the whole complex even on decent absolute growth. In practice, this shifts leadership away from the most crowded AI infra winners toward the less emotionally owned parts of the stack that can still compound if spending merely normalizes rather than accelerates.

The broader setup is a classic volatility regime change: low realized vol helped leverage work for months, but one macro hot print plus one guide miss exposed how fragile the tape was underneath. That usually creates follow-through risk over the next 1-3 weeks, not just one day, because systematic de-risking, call unwind, and dealer hedging can keep pressure on the largest index weights even if fundamentals remain intact. The important tell is that the move was concentrated in the long-duration growth segment of semis, which means the market is implicitly tightening the discount rate on future AI cash flows, not questioning near-term unit demand.

The contrarian take is that this may be a better entry point for selective exposure than the tape suggests, but not via broad beta or daily-reset leverage. If MU and TSM confirm that AI memory and foundry demand are still tracking, Friday becomes a temporary multiple reset rather than a top; if they disappoint, the downside likely extends because the market will start cutting 2026-27 build assumptions, not just 2025 guidance. The asymmetry is now in dispersion: alpha will come from being long the parts of the chain with visible order books and short the most consensus-sensitive compounders.

Key Decisions for Investors

• Avoid fresh long exposure in USD for 1-2 weeks; if holding, reduce size and cap downside with tight stops because daily-reset leverage will underperform badly if semis stay choppy.
• Initiate a relative-value pair: long TSM / short AVGO into the June-July data window. TSM should be less exposed to one-quarter AI guide risk, while AVGO is the cleanest proxy for consensus disappointment and multiple compression.
• Buy MU into the June 24 print only on weakness, but structure it with defined risk via call spreads or small cash equity size. Upside is a validation of AI memory demand; downside is a sharper leg lower if hyperscaler ordering is soft.
• Use NVDA only as a hedgeable core, not a momentum add. Consider selling upside calls against existing longs over the next 2-4 weeks to monetize elevated post-event volatility while the market digests capex uncertainty.
• If you want to express the bearish macro-duration view, short a semiconductor basket against QQQ or long-duration growth equities for 1-2 weeks; the thesis is that semis remain the most rate-sensitive equity cohort if yields stay elevated.