Article published May 23, 2026. Prices below use latest available snapshots.
Summary
Money is concentrated in tech/semis (XLK +30% 3m, SMH +40% 3m) with leveraged-ETF exhaustion signals. Defensives and cyclicals are flat-to-down — no rotation INTO them, but no panic OUT. The "weird ETF" canaries say: MSTR leverage is broken (MSTU/MSTX/MSTY all rolled), NVDA leverage is fine, miners decoupled from BTC, and the memory thesis has spread to a single basket-level instrument (DRAM ETF +90% 3m). The cleanest non-tech rotations are XLV (healthcare bounce), XLE (energy recovery, USO +74% 3m is the oil extreme), and Korea/Taiwan (EWY/EWT). The market is not rotating sectorally — it's narrowing into AI infrastructure stack (memory→networking→optical→power→space), with everything else passive.
What changed in weird-ETF land in 13 days (2026-05-10 → 2026-05-23)
The MSTR-leverage corner broke. Three wrappers — MSTU 2x, MSTX 2x, MSTY synthetic-income — all rolled from 2026-05-10 Tier 1 status to broken. MSTU sits at -94.2% from 52wk high, RSI 42.8, -18.75% 30d (was "+49.3% adj 30d" on 5/10). Crypto-spot (IBIT/FBTC/GBTC) all at RSI ~45 and -40% from 52wk high despite BTC holding up nominally. Meanwhile NVDA-leverage (NVDL) and TSLA-leverage (TSLL) held — that's a clean bifurcation, not a "leveraged ETFs broke" story.
What's still running and interesting from the weird-ETF set (data as of 2026-05-22, from the day's mechanism-aware ETF scan, including names that no longer made the printed tier table):
| Ticker | Mechanism | Reading |
|---|---|---|
| WGMI (BTC-miner basket) | thematic-rules 1x | +13.2% 7d / +23.4% 30d / +55.8% 3m, RSI 66 — miners decoupling from spot-BTC weakness (alpha is HPC pivot, not BTC price) |
| NVDL (2x NVDA) | daily-reset-leveraged | -6.95% 7d but +13.3% 30d / +18.5% 3m — NVDA leverage held while MSTR leverage didn't |
| TSLL (2x TSLA) | daily-reset-leveraged | +26.96% 30d, +5.53% 7d — quietly running on TSLA recovery |
| DRAM (Roundhill Memory ETF, ~brand-new) | thematic | +45.27% 30d / +90.27% 3m, RSI 65.9 — still building the parabola; SNDK/MU/STX trade in ETF wrapper form |
| CHAT, HUMN, BSTZ, LIT, MAGS | thematic 1x | Were 5/10 Tier 2 (RSI 72-76, at 52wk highs) — most cooled mildly but cluster held |
| NUKZ, BIGB (in universe) | thematic-rules | Listed in universe but didn't meet Tier criteria on 5/10 or 5/23 — coverage carriers, no Tier signal yet |
| BUFR | buffered S&P | Was the 5/10 melt-up tell at RSI 74; calibration still single-day, needs 3-6mo |
Three structural reads:
MSTR-leverage decay as the cleanest "wrapper breaks before underlying" signal. Three different MSTR wrappers (MSTU, MSTX, MSTY) all rolled while NVDA leverage held. This is volatility decay doing exactly what it's mathematically supposed to do — anybody long MSTU since April lost ~94% from highs. The wrappers themselves are the trade tell: when 2x products break first, the underlying tape is rotating, not crashing.
WGMI's decoupling from IBIT/GBTC. Miners +55.8% 3m while spot BTC ETFs +17% 3m. WGMI is now an HPC-pivot ETF in BTC-miner clothing. The wrapper hides the underlying thesis change. Classic thematic-ETF-gets-it-right-by- accident pattern.
DRAM ETF at +90% 3m — single-instrument expression of an old thesis. Pure-play HBM/memory ETF launched April 2026 has done what individual memory names (SNDK/MU/STX) took months to do. Cleanest single-instrument way to play the memory cluster.
Sector rotation read — where money is now (5/22 close, market-pulse summary)
From the market-pulse summary (SPDR sector ETFs):
| Sector ETF | 30d | 3m | RSI | Read |
|---|---|---|---|---|
| XLK (Tech) | +15.75% | +30.39% | 73.0 | 🟠 Dominant + overbought. Carrying the tape. |
| SMH (Semis) | +19.61% | +39.59% | 68.5 | 🟠 Engine inside XLK. Not RSI-overbought yet but extended. |
| VUG (Growth) | +6.7% | +14.49% | 67.8 | Riding tech |
| MTUM (Momentum) | +10.05% | +21.67% | 64.8 | Momentum factor stacking on tech |
| XLV (Healthcare) | +2.5% | -5.07% | 62.1 | 🟢 Recent bounce off 3m weakness |
| XLE (Energy) | +4.41% | +8.58% | 55.4 | 🟢 Quietly recovering — note USO +74% 3m is single-instrument extreme |
| XLF (Financials) | +0.27% | +2.91% | 56.0 | 🟡 Flat |
| XLY (Cons Disc) | +1.22% | +3.85% | 57.0 | 🟡 Flat |
| XLRE (Real Estate) | +1.36% | +2.86% | 58.6 | 🟡 Flat |
| XLP (Cons Staples) | +1.58% | -4.15% | 54.1 | 🟡 Late-cycle defensive softness |
| XLU (Utilities) | -1.61% | -2.18% | 50.5 | 🔴 Underperforming — surprising given AI-power thesis (GEV is named-ticker, not XLU which is utility utilities) |
| XLI (Industrials) | -1.32% | -1.47% | 50.1 | 🔴 Quietly weak |
| XLC (Comms) | -1.64% | +0.41% | 47.0 | 🔴 Weak — GOOGL/META not pulling weight at index level |
| XLB (Materials) | -2.93% | -4.63% | 44.7 | 🔴 Weakest — copper/metals leadership eroding |
Macro/risk gauges:
- VIX collapsed: VIXY -10.8% 30d, RSI 35.8 — complacency. Confirms JEPQ RSI 74.5 (income strats squeezed by low vol).
- TLT/IEF (long bonds): RSI 46/42 — yields creeping up but not breaking out.
- GLD pullback: -3.99% 30d / -14.02% 3m, RSI 39.6 — safe-haven unwind, risk-on confirmed (and confirms why we retired the gold-crash perspective's silver leg + closed food-security-cascade in R31-A).
- KWEB (China): RSI 37.9, -8.17% 3m — still broken. FXI same.
- EWJ (Japan): +5.21% 30d, RSI 57.5 — quiet rotation toward Japan (EWY/EWT showed up in the monster scan too).
How this connects to active perspectives
ai-power-bottleneck: the XLU/XLB weakness alongside named-ticker strength (BE +127% vs SMA200, POWL +85%, CGEH +112%, GEV +40%) confirms the perspective's frame — the AI-power trade is in named tickers (turbine makers, fuel cells, grid switchgear), not utility utilities or basic materials. Cleaner separation than 30 days ago.optical-supercycle: still extended. AAOI 188% vs SMA200, AXTI 369%, LITE 114%. Cluster state already updated to "contested / late-stage" in R31-B; this digest doesn't change that.
This digest is consumer-shaped synthesis citing two upstream scans — it does NOT itself produce new market data. All numeric claims are from the day's mechanism-aware weird-ETF tier scan and its May 10 predecessor, plus the validated ETF-monster-universe and market-pulse summaries (including names not printed in the tier table), data through 2026-05-22.
Sources
- Weird-ETF mechanism-aware tier scans, May 10 and May 23, 2026 — the baseline and the post-MSTR-leverage-decay snapshot (T1=4, T2=6, T3=5, T4=35).
Price truth: validated daily summaries (summaries).