The compute, networking, memory, storage, and power layer of the AI buildout. Owners of the bottlenecks: accelerators, custom silicon, switching, optics, HBM, fab equipment, test, and the server/power chain that turns capex into installed compute.
Follow the theme against the wider market.
Series before each basket's creation date are a backtest of the membership as of creation; live tracking starts at creation.
| Return | 1 day | 5 days | 20 days | 60 days | MTD | YTD | 1 year |
|---|---|---|---|---|---|---|---|
| Basket | -5.1% | -16.2% | -11.7% | -13.0% | -14.6% | +71.0% | +81.7% |
| vs S&P 500 | -4.0% | -14.6% | -11.8% | -16.2% | -14.8% | +52.2% | +57.5% |
1y · 2025-10-16 – 2026-10-16 · 252 / 252 sessions
Basket total returns · Annualized using 252 trading sessions · 0% risk-free rate and downside target. As of 2026-10-16 · close.
Sharpe: average daily simple return divided by its sample standard deviation, multiplied by √252. The risk-free rate is assumed to be 0%; this is not a cash-adjusted ratio.
Sortino: average daily simple return divided by downside deviation, multiplied by √252. Downside deviation is the square root of the average squared shortfall below 0%, counting all sessions, including zeros for gains.
Historical volatility: √(252 × average squared daily log return), matching the Vol page’s realized-volatility convention. This differs from the sample standard deviation used in Sharpe.
Each period requires complete history and at least 5 sessions. MTD and YTD follow the calendar; 1y is 252 sessions. Missing data or an undefined ratio displays — with a reason. Short periods can produce unstable ratios; intraday snapshots include a partial session. History before this basket’s creation is backtested. If a new basket has no stored daily return series yet, estimates from its rounded published chart levels are marked ≈.
Current members and the reasoning behind each inclusion. Returns are for individual holdings.
| Why it’s included | Added | |||||
|---|---|---|---|---|---|---|
| MPWRMonolithic Power Systems | Power management silicon for accelerators | -3.1% | -12.7% | -0.4% | +34.3% | 2026-06-11 |
| NVDANvidia | Accelerator monopoly economics; CUDA moat | -1.4% | -6.6% | -3.7% | +15.1% | 2026-06-11 |
| QQnity Electronics, Inc. | Advanced-node semiconductor materials (DuPont electronics spin-off) | -3.5% | -8.9% | -3.9% | +41.6% | 2026-06-11 |
| CIENCiena Corporation | Optical networking systems | -6.7% | -26.0% | -4.6% | +42.3% | 2026-06-11 |
| AMDAdvanced Micro Devices | Credible #2 in merchant accelerators | -6.0% | -13.0% | -5.5% | +146.9% | 2026-06-11 |
| KLACKLA Corporation | Process control intensity rises with node complexity | -5.0% | -15.3% | -6.4% | +36.9% | 2026-06-11 |
| MRVLMarvell Technology, Inc. | Custom AI silicon and electro-optics — re-enters untagged after joining the S&P 500 | -5.6% | -17.2% | -6.7% | +168.5% | 2026-07-21 |
| JBLJabil | AI hardware manufacturing services | -3.1% | -9.2% | -7.1% | +22.1% | 2026-06-11 |
| AVGOBroadcom | Custom AI ASICs (hyperscaler co-design) + networking | -4.5% | -9.1% | -7.9% | -4.5% | 2026-06-11 |
| AMATApplied Materials | Fab equipment breadth for leading-edge + memory | -4.8% | -19.3% | -7.9% | +59.8% | 2026-06-11 |
| LRCXLam Research | Etch/deposition leverage to HBM and advanced packaging | -4.9% | -17.2% | -8.3% | +54.6% | 2026-06-11 |
| ANETArista Networks | AI cluster networking share gainer | -4.6% | -18.3% | -11.2% | +35.1% | 2026-06-11 |
| GLWCorning Inc. | Optical fiber and interconnect | -4.6% | -15.8% | -12.1% | +51.6% | 2026-06-11 |
| TERTeradyne | Semiconductor test equipment | -5.3% | -19.3% | -12.5% | +68.1% | 2026-06-11 |
| ORCLOracle Corporation | GPU cloud capacity buildout (OCI) | -4.6% | -9.3% | -12.8% | -33.3% | 2026-06-11 |
| MUMicron Technology | HBM supply oligopoly member | -5.6% | -14.6% | -13.5% | +208.1% | 2026-06-11 |
| VRTVertiv Holdings Co | Datacenter power and liquid cooling | -3.9% | -15.0% | -17.2% | +27.5% | 2026-06-11 |
| SMCISupermicro | AI server/rack assembly | -9.9% | -24.4% | -19.1% | +8.1% | 2026-06-11 |
| DELLDell Technologies | AI server integration at scale | -4.6% | -23.0% | -20.5% | +262.0% | 2026-06-11 |
| WDCWestern Digital | Data-center hard-drive exabytes | -6.9% | -16.4% | -24.8% | +92.9% | 2026-06-11 |
| STXSeagate Technology | Data-center storage | -7.5% | -18.4% | -25.5% | +132.8% | 2026-06-11 |
| COHRCoherent, Inc. | Optical transceivers for AI networking | -7.1% | -25.5% | -26.6% | +26.2% | 2026-06-11 |
Partial price histories: Q from 2025-10-27.
Equal-weighted; monthly rebalance, buy-and-hold between rebalances; dated membership changes reset the target weights.
Tracked since 2026-06-11. Membership decisions are documented below.
Read the methodology →Basket returns include dividends. “vs S&P 500” compares the basket’s compounded growth with the benchmark’s. Green means positive; red means negative.
Cross-checked against SMH. Daily return correlation: 0.95 across 1707 sessions. semiconductor-only proxy; the basket’s networking, storage, and power names sit outside it
MRVL joined the S&P 500 and moves from broad-only to both universes (reverse of the ZION/MOH/MHK pattern): broad entry closed, untagged entry added same-day. The S&P version gains a member; the FW 3000 version is unchanged.
FN and CRDO removed — never FW 3000 constituents (offshore-domiciled, screened out of the universe), so the 2026-06-12 broad pass added them in error; the pipeline has flagged both as missing/ex-universe since. Entries deleted outright rather than closed with a removed date: neither ever had tape, so no computed history is restated. Optics coverage continues via COHR, CIEN, GLW, MTSI and the new optics-memory basket.
FW 3000 pass: added MRVL, CRWV, ALAB, CRDO, MOD, ONTO, MTSI, FN — names outside the S&P 500 but inside our broad universe; they count only in the FW 3000 version of this basket. Effective-dated to creation so the broad backtest covers the full set. The S&P 500 basket is unchanged.
Widening pass: added the optics/networking leg (COHR, CIEN, GLW), storage (WDC, STX), test and materials (TER, Q), power silicon (MPWR), and server manufacturing (SMCI, JBL). Reference pinned to SMH — semis-only, so expect high but not perfect tracking. New members are effective-dated to basket creation so the backtest and the reference correlation cover the widened set (the basket was one day old, with no meaningful live history to restate).
MRVL removed: not an S&P 500 constituent (basket rule violation at creation); custom-silicon exposure remains via AVGO.
Initial composition: 12 names across compute, networking, memory, equipment, power.