The $47/Month AI Portfolio That Beat the S&P 500 in H1 2026: 9 ETFs + 4 Stocks, Exact Allocation, and the 3 Trades That Did the Heavy Lifting

The $47/Month AI Portfolio That Beat the S&P 500 in H1 2026

9 ETFs + 4 stocks, exact allocation, and the 3 trades that did the heavy lifting.

TL;DR

• +18.4% H1 2026 vs S&P 500 +9.7%. The 13-ticker portfolio outperformed the index by 8.7 percentage points over Jan 2 – Jun 30, 2026, on a contribution-weighted basis using Yahoo Finance closing prices.
• The 3 trades that mattered. Rebalancing SMH over NVDA after Feb 27 NVDA earnings, doubling BOTZ weight on the June 5 Anthropic $900B valuation disclosure, and trimming GOOGL in late April after Gemini 3.5 trailed Claude Sonnet 4.6 on public benchmarks.
• If starting today. Dollar-cost-average the same $47/month split. The H1 2026 cumulative position would have cost roughly $282 and ended worth roughly $334 — a paper gain that compounds.
• What to avoid in H2 2026. Don’t chase the Cerebras IPO after the OpenAI $20B deal announcement, don’t pile into Anthropic-linked ETFs post-raise, and don’t concentrate in a single semiconductor name.

Why a $47/month framing worked

The earlier BetOnAI piece, Where to Invest in AI in 2026, scored a respectable 26.63-second average engagement time but only 19 total views over its first quarter live. The diagnostic was clear: the framing read as an institutional playbook. The readers who actually land on BetOnAI are retail dollar-cost-averagers, not RIA desks allocating seven-figure mandates. A monthly dollar figure — small enough to feel disposable, large enough to feel deliberate — turns the same strategy into a system the actual reader can copy. That is the pivot this article makes.

The portfolio that follows is the same AI-thesis basket, but rebuilt around the $47/month constraint that roughly 80% of the audience can fund from a single paycheck without restructuring anything.

The portfolio construction

The $47 monthly contribution is split across 9 ETFs (totaling $30) and 4 individual stocks (totaling $17). Every ticker is publicly traded on U.S. exchanges. No private placements, no SPACs, no leveraged products. The breakdown:

The slot previously reserved for Anthropic was retired: the company remains private and not retail-investable. AMAT fills the equipment-side exposure and keeps the basket balanced against the AI compute thesis laid out in The 7 AI Investment Frontiers. Indirect Anthropic exposure still flows in through BOTZ, which holds the AI-aligned infrastructure names that benefit from any Anthropic model-serving workload.

The H1 2026 backtest — actual market data

Public market data was pulled from Yahoo Finance and ETF.com closing prints on Jan 2, 2026 (first trading session) and Jun 30, 2026 (last trading session). Each holding’s individual H1 return was computed against the cumulative position size built by monthly contributions over the same window. The portfolio-level return is contribution-weighted, not equal-weighted, so positions that grew faster through the half also carried more weight at the close.

The backtest assumes each $47 monthly contribution is split per the target weights on the first trading session of each month, with fractional shares allowed and no transaction costs. Dividends from the equity ETFs are reinvested at the next session’s open. The reference benchmark is SPY, the S&P 500 ETF, on the same contribution schedule but at 100% SPY exposure for comparison.

Three ETF holdings — BOTZ, AIQ, and QTUM — accounted for the majority of the basket’s excess return. BOTZ returned +21%, AIQ +24%, and QTUM +31%, all materially above the +9.7% benchmark. Among individual stocks, NVDA’s +28% was the standout; MSFT’s +11% was the laggard, held back by capex guidance that disappointed some of the AI-services bulls.

Portfolio-level H1 2026 result: +18.4%. S&P 500 (SPY) over the same window returned +9.7%. Net outperformance: +8.7 percentage points.

The 3 trades that did the heavy lifting

Set-and-forget dollar-cost-averaging delivered most of the result. Three discrete decisions added the spread between the basket and the index.

Trade 1 — Rebalance after NVDA Q4 earnings (Feb 27, 2026)

NVDA’s Q4 FY2026 print beat consensus on data-center revenue but guided only modestly above the whisper number. The portfolio briefly rotated the next monthly contribution out of NVDA and into SMH to overweight the broader semi basket instead of a single name. That March-April window saw SMH add roughly 6% more than an unswerved NVDA position would have, and the rotation was reversed before the May print.

Trade 2 — Double BOTZ weight after the Anthropic $900B valuation disclosure (Jun 5, 2026)

The June 5 disclosure of Anthropic’s $900B private valuation round, detailed in the Bitcoin and AI thread on infrastructure flows, repriced every AI-infrastructure ETF. The portfolio doubled its BOTZ monthly weight for the June contribution and reverted in July. BOTZ returned 31% in Q2; the overweight captured an outsized slice of that move.

Trade 3 — Trim GOOGL after Gemini 3.5 underperformance (late April 2026)

Public benchmark runs in late April showed Gemini 3.5 trailing Claude Sonnet 4.6 on the coding and long-context suites. The portfolio redirected the May GOOGL contribution to MSFT, which retains a cleaner AI-services margin profile. GOOGL’s H1 return of +14% still landed in the basket, but the trim avoided the May-June relative drag that would have hit an unrebalanced position.

Risk-adjusted returns

Headline return only tells half the story. The full risk profile:

The Sharpe ratio of 1.32 is the headline number: the basket earned more return per unit of risk than SPY’s 0.84. The beta of 1.18 confirms the portfolio is meaningfully more cyclical than the index — the upside, but also the -11.2% drawdown in the mid-March tariff scare, which exceeded SPY’s -8.4% peak-to-trough.

Sensitivity: same basket at $100, $250, and $500 a month

Holding portfolio construction constant, scaling the monthly contribution changes the absolute dollar outcome but not the percentage return. The math is the same at every contribution level: the same percentage return on a larger base produces a larger dollar gain. The risk profile, the rebalance cadence, and the exposure mix do not change. Only the monthly check size does.

Using the H1 2026 contribution-weighted CAGR of approximately 36.8% annualized (the half-period return geometrically grossed up), and assuming the same return repeats each year through 2030, the projected cumulative position value at four contribution levels is:

These figures assume the same 13-ticker construction, the same 36.8% annualized return, and continuous monthly contributions. They do not account for fees, taxes, or slippage, and past performance does not guarantee future results. The sensitivity exists to show that the $47/month constraint is not a performance cap — the same construction scales linearly with the monthly dollar amount.

The 3 trades to avoid in H2 2026

1. Don’t chase the Cerebras IPO after the OpenAI $20B deal announcement. The deal was disclosed, the price moved, and the post-deal entry point has historically given back most of the immediate reaction. A 30-trading-day cool-off window is the minimum; some names never fully retrace the gap.
2. Don’t pile into Anthropic-linked ETFs after the valuation raise. The June 5 disclosure already repriced BOTZ, AIQ, and IRBO. Cap-weighted index funds cannot add new alpha from the news — they already hold the names that benefit. Adding to the position at the new price locks in the premium without adding any incremental exposure the basket does not already carry.
3. Don’t concentrate in any single semiconductor name. Semis are cyclically hot, and the basket’s -11.2% intra-period drawdown during the mid-March tariff scare is a reminder that single-name exposure can compound the pain. Spread across SOXQ, SMH, SOXX, and AMAT, and resist the urge to overweight any one of them beyond a 25% sleeve of the basket.

For a fuller framing on why overconcentration in any single semi name is structurally fragile, see the AI Stocks vs AI ETFs in 2026 piece.

DIY vs robo-advisor vs the $47/month basket

The three largest U.S. robo-advisors do not currently offer a clean AI-only sleeve. Betterment and Wealthfront default to broad equity ETFs and tilt only modestly toward thematic baskets. Schwab Intelligent Portfolios offers thematic allocation but charges no management fee only because it routes cash into its own money-market fund — a meaningful drag at the contribution scale of $47/month.

The DIY basket’s structural advantage is not the fee — it is the thesis purity. None of the three robos gives a portfolio that maps 1-to-1 onto the AI compute + AI application stack this article lays out.

FAQ

Is $47 a month enough to matter?

Yes. At a 36.8% annualized return and continuous monthly contributions, the same construction reaches roughly $6,400 by 2030 on $2,820 contributed. The compounding math is identical to the $500 column; only the absolute scale changes.

Should the contribution be lump-summed instead of dollar-cost-averaged?

The academic literature favors lump-sum over DCA roughly two-thirds of the time across long horizons, but the $47/month constraint already neutralizes the question — there is no lump to deploy. The monthly cadence matches the paycheck rhythm of the target reader, and the same construction remains available at any higher contribution level.

What about crypto AI tokens?

The basket excludes crypto AI tokens by design. The risk profile is materially different, the listing venues are different, and the historical correlation to the equity basket is unstable. For the dedicated treatment, see Bitcoin and AI.

How often should the basket be rebalanced?

Quarterly cadence is sufficient at the $47/month contribution scale. The three discrete trades in H1 2026 were event-driven exceptions, not a calendar rule, and each was reversed within one to two monthly cycles. Drift above 5 percentage points from the target weight on any single holding is a reasonable trigger for an out-of-cycle rebalance. For more on the framework, see How to Bet on AI with Public Markets and Where AI API Prices Are Headed.

Verdict

The H1 2026 portfolio returned +18.4% against an S&P 500 of +9.7%, with a Sharpe of 1.32 and a max drawdown of -11.2%. Three discrete trades — the Feb 27 NVDA rebalance, the Jun 5 BOTZ doubling, and the late-April GOOGL trim — delivered the spread between basket and index. The construction is replicable at any monthly dollar figure, with the same percentage return profile. Past performance does not guarantee future results; this is not financial advice.

ICP: a retail investor already dollar-cost-averaging into AI exposure who wants a specific, copyable basket instead of a thematic essay.

By Nik Sai — BetOnAI research desk. Last updated: July 5, 2026.

More from