The framework

Methodology

01

Map the opportunity

Evaluate demand, bottlenecks, pricing power, and replacement risk across the AI value chain. An AI label alone is not an investment thesis.

02

Size with intent

Set targets from fundamentals, relative strength, and entry conditions. Review single-name concentration, high-beta exposure, and speculative risk separately.

03

Keep a reserve

Adjust equity and reserve targets through a sector-trend framework. Capital without a qualified opportunity remains in the reserve.

Defense

60% equities 40% reserve

Neutral

75% equities 25% reserve

Attack

85% equities 15% reserve

Mode changes require confirmed signals. Reserve allocations do not guarantee a maximum drawdown or prevent losses.

Execution policy

Updated September 30, 2026

Routine weight adjustments use ongoing contributions. When the adopted mode and basket gates permit additions, Fidelity Smart Buys directs an authorized amount toward underweight holdings. Overweight positions are not sold to fund a routine rebalance; without eligible funding, weights may remain away from their targets.

Downgrades require three consecutive completed trading sessions with candidates below the adopted mode; the highest candidate sets the proposed mode. Additions pause on the first downgrade signal. Defense-to-Neutral recovery still requires three sessions, and Neutral-to-Attack requires five.

The larger of 25% of target weight or 2 percentage points remains the drift review threshold, not a trade trigger. Ordinary target reductions and monthly momentum changes guide future buys; they do not create matching sale orders. Smart Buys may also fund underweights within this review threshold.

Separately reviewed weak-stock trims, verified stops, thesis failures, hard risk limits and confirmed mode risk reductions still apply. A monthly review may conclude that no trim is needed.

Before a loss sale, review tax lots and purchases of substantially identical securities within 30 calendar days before and after the sale, including relevant accounts, Smart Buys, recurring investments and dividend reinvestment. A resulting underweight must not trigger an automatic buyback. Buying alone does not rule out a wash sale involving another sale; missing trade history leaves tax review pending. See IRS Publication 550: Wash Sales.

Additions require no pending downgrade and QQQ and SOXX closes above their 20-day averages. Smart Buys follows approved targets; each contribution need not buy every holding.

Individual-stock moving-average breaks prompt review; they do not automatically block a purchase or trigger a sale. Verified thesis failures, explicit stops and hard risk-limit breaches must be resolved in the target list and execution plan before the next basket purchase.

Smart Buys distributes the approved amount; BOXX reserves, unused funding tranches and account limits still govern how much can be invested. Authorized reserve deployment continues under those gates. Routine BOXX replenishment uses explicitly added funds, not sales of winners; the reference-high, funded-reserve and ledger requirements for a new cycle remain in place. This policy does not establish that a purchase is currently permitted or has occurred.

Review every sector when the mode changes.

Each mode-change proposal reviews all model equity sectors. The default comparison uses 1- and 3-month adjusted-price momentum (21 and 63 trading sessions), weighted by each sector’s current model targets at a common completed close. Both returns and their average are reported alongside business growth, risks and proposed weight changes. Candidates and BOXX are excluded; incomplete sector data cannot support a full rotation ranking.

Upgrades direct more weight toward stronger sectors with positive momentum in both windows and an intact investment thesis. Downgrades trim weaker sectors first to meet the lower equity budget; even strong sectors may lose absolute weight. Exact targets require approval and remain subject to position and risk limits. Monthly momentum adjustments within a sector retain their separate 6- and 12-month method; new base weights reset affected offsets.

Sector reviews are manual. This policy does not establish a current ranking, completed trade or performance result. Smart Buys does not set sector weights.

How Fidelity Smart Buys works

Questions

A closer look.

Is AI Stock Basket an ETF?

No. AI Stock Basket is an independent model portfolio and research project. It is not a registered fund, exchange-listed security, or product you can purchase.

Are these actual holdings?

The Allocations section shows dated model targets. Actual holdings shows a separate broker snapshot as percentages of invested stocks and BOXX, excluding uninvested cash and pending activity. Account details, amounts and share counts are private. Neither section is a live feed.

Why include a reserve?

The reserve leaves capacity for qualified opportunities and changes with the allocation mode. The model uses CASH as its reserve symbol and BOXX as the designated vehicle.

How often does this page update?

We aim to refresh model market data and the hypothetical YTD comparison each weekend, after the week’s final trading session. Updates are manual; each quote and performance result shows its own data date. We aim to update actual holdings separately each Friday after the US close, from that day’s broker CSV. Until then, the last verified holdings and unrealized gains retain their original snapshot date. Model updates do not require a new holdings file. New model decisions require a new frozen snapshot.

What are the main risks?

The basket concentrates on an investment theme. Company risk, valuation changes, technology shifts, and correlated declines can cause substantial losses. Diversification across AI layers does not eliminate those risks.