AI compute proxies DRAM and SMH slide as bitcoin reclaims $61K
A reported Meta “Compute” push to sell excess GPU capacity added pressure to AI-infrastructure names as BTC rebounded from a sub-$58K dip.
AI-linked memory and semiconductor trades have started to lose momentum, with key ETFs pulling back from late-June highs. At the same time, bitcoin rebounded above $61,000 after dipping below $58,000 on July 1, reviving early talk of a cross-asset rotation back toward crypto risk.
Key Takeaways
- Roundhill Memory ETF (DRAM) fell roughly 25% from its June 22 record high, while VanEck Semiconductor ETF (SMH) dropped around 12%.
- Bitcoin dipped below $58,000 on July 1 and later traded back above $61,000, with BTC shown around $61,946.
- A reported Meta Platforms initiative to sell excess GPU computing capacity coincided with accelerated selling pressure across AI-infrastructure beneficiaries.
- Crypto-adjacent HPC and GPU-hosting equities including IREN, CIFR, and WULF were each described as down at least 20% from their all-time highs.
AI Compute Leaders Slip as Bitcoin Bounces Back Above $61K
The dominant 2026 equity trade in AI compute is showing a momentum break just as bitcoin is stabilizing off its lows. DRAM, a memory-focused ETF often treated as a proxy for AI compute demand, was described as down roughly 25% from its June 22 record high. SMH, a broad semiconductor benchmark, was described as down around 12%.
On the other side of the ledger, bitcoin dipped below $58,000 on July 1 and later traded back above $61,000, with BTC shown around $61,946. The pairing matters for cross-asset desks because the AI memory and semiconductor bid was framed as having sidelined much of crypto earlier in the year.
The setup is consistent with early-stage risk rebalancing, but the evidence in hand is still thin. A bounce in BTC alongside a pullback in AI proxies can be a rotation signal, or it can be a temporary de-risking in equities with a reflexive crypto bounce.
The Scoreboard: DRAM/SMH Pullback vs. BTC and IBIT Drawdowns
The magnitude of the prior run-up is part of why the current pullback is getting attention. DRAM more than doubled in the first half of 2026, while SMH climbed 60% over the same period, before rolling over from late-June highs.
Bitcoin’s proxy in the ETF complex has not been hiding the drawdown. BlackRock’s iShares Bitcoin Trust (IBIT), described as the largest bitcoin ETF, was down 30%, in line with bitcoin. For traders trying to map flows, that “in line” phrasing is the point: it suggests the ETF wrapper has tracked spot directionally during the period described, rather than showing a clear divergence that would imply unusual creation or redemption pressure.
Single-name leadership in the AI complex also underscores how crowded the compute narrative became. Sandisk (SNDK) was cited as up more than 530% in 2026, and Micron (MU) as up more than 230%, tied to demand for NAND flash, DRAM, and high-bandwidth memory (HBM) used in AI infrastructure.
Meta ‘Compute’ Report Adds Pressure to the GPU-Hosting Trade
Selling pressure in AI-related stocks was described as accelerating after Bloomberg reported Meta Platforms is creating a business unit called Meta Compute to sell excess GPU computing capacity to third parties. The packet includes no confirmation from Meta or details on scale or timing, but the direction of the catalyst is straightforward.
If a hyperscaler with deep GPU inventory starts commercializing “excess” capacity, it can compress the perceived scarcity premium that has supported parts of the AI-infrastructure stack. That is a plausible fundamental reason the market would lean harder on “neocloud” providers that lease GPU infrastructure to AI developers.
The same pressure can spill into a second cohort that matters for crypto traders: former bitcoin miners that pivoted into high-performance computing (HPC) and GPU hosting.
Rotation Checklist for Traders: Confirming or Fading the AI-to-Crypto Shift
The threshold that matters near-term is whether BTC can hold above $61,000 and avoid revisiting the July 1 sub-$58,000 low. Without that, the move risks reading as a dead-cat bounce rather than a rotation attempt.
On the equity side, traders will be watching whether DRAM’s drawdown from the June 22 high continues and whether SMH extends beyond the roughly 12% pullback. A stabilization in chips while BTC fades would undercut the rotation narrative.
The Meta Compute thread also needs more than a single headline. Any added confirmation, detail, or timeline beyond the Bloomberg report will matter, especially if AI-infrastructure selling pressure persists alongside it.
Finally, IBIT’s relative performance versus BTC is a clean read on ETF participation. Tight tracking would keep the focus on spot direction, while any widening gap would hint at wrapper-specific flows during any attempted risk rotation.
When AI’s Crowded Trade Unwinds, BTC Often Gets a Second Look
I treat this as a cross-asset positioning story first, not a crypto-specific fundamental shift. DRAM being roughly 25% off its June 22 high and SMH down around 12% at the same time BTC is back above $61,000 is consistent with early rebalancing away from the crowded AI compute trade, but it is not proof of durable rotation by itself.
The real test is whether BTC can defend the $61,000 area without revisiting sub-$58,000 while AI compute proxies keep bleeding and the Meta Compute narrative gains concrete details. If that holds, the setup starts to look structural rather than narrative-driven, with practical impact showing up as sustained relative strength in BTC and tighter, more persistent participation through IBIT.
