S&P 500 (Vanguard S&P 500 ETF / SPDR S&P 500)
Vanguard (VOO, the canonical low-cost tracker; State Street's SPY is the older liquidity-king alternative). Tracks the S&P 500, the 500-largest-US-companies cap-weighted benchmark maintained by S&P Dow Jones Indices.
The default 'you already own the AI trade' baseline. Cap-weighted, so the index has passively become a megacap-tech / AI fund: the Magnificent 7 (Apple, Microsoft, Nvidia, Alphabet, Amazon, Meta, Tesla) were ~32.7-33.8% of the index in mid-June 2026, and Information Technology alone is ~33% of the portfolio. An investor buying 'the market' here is putting roughly a third of every dollar into the same AI/compute names this page tracks actively.
Top holdings
Largest single weight; the AI-accelerator/compute core of the index (June 18, 2026)
On-device AI / consumer compute
Azure AI + OpenAI exposure
AWS compute / inference (Feb 28 2026 weight)
Combined A+C classes; Gemini, TPUs, DeepMind
Custom AI silicon / networking (Feb 28 2026)
AI capex + Llama (Feb 28 2026)
AI/robotics optionality (Feb 28 2026)
Recent moves
Index reconstitution is mechanical, not discretionary — weights drift with price. The structural 'move' is concentration: Mag-7 weight has held in a ~32-35% band over the trailing year, with Nvidia overtaking Apple as the top holding during 2026. June 2026 quarterly rebalance reflected continued AI-megacap dominance.
Our take
The honest baseline for every active AI bet on this page: matching the S&P 500 is the bar to beat, and the index is already ~third-weighted in Mag-7 / AI compute names, so 'I want AI exposure' is largely satisfied by owning VOO. Edge of an active manager must come from positions VOO under-weights (private/pre-IPO, non-US, picks-and-shovels like power/nuclear) or from sizing beyond cap weight. Limit: that same concentration is the risk — a Nvidia/AI drawdown hits the 'safe index' nearly as hard as a thematic fund, and cap-weighting offers no protection on the way down.
Two things the marketing won't say. (1) 'Passive' outsources active judgment rather than removing it: constituent selection sits with a human committee at S&P DJI, and cap-weighting is itself a momentum rule — the fund mechanically holds more of whatever has already appreciated and never rebalances toward what is cheap.
That is a factor bet, not the absence of one. (2) The diversification is worse than ~32% implies, because Mag-7 cash flows increasingly reference each other: hyperscaler capex is Nvidia's revenue, and Nvidia's largest customers are the same index constituents. Seven names with independent earnings would be a 32% concentration; seven names inside one capex loop is closer to a single correlated exposure wearing seven tickers. Headline weight is a lower bound on true factor exposure. The 0.03% fee is genuinely excellent — but fee is the solved problem here, and index construction is the unsolved one.
- 2026-07July 2026 split the loop rather than confirming it: MSFT +24.6% for holding capex flat while what that capex buys fell hard (MU -28.7%, SanDisk -46.6%, NBIS -31.1%). Opposite sides of one trade, not one exposure.
Thesis
VOO is the honest benchmark every active AI bet on this page must clear — and, less honestly marketed, it has become a concentrated AI vehicle in passive clothing. Top-10 holdings were 36.3% of the fund and the Mag-7 ~31.5% of it (summed from stockanalysis.com holdings, 2026-07-21; Forbes puts the Mag-7 at ~34% of the index, 2026-07-08), against a historical top-10 average of ~24% and a prior peak near 28%. At 0.03% the cost question is settled; the open question is whether someone who says 'I don't want single-stock AI risk' realises nearly a third of this portfolio already is that risk.
Full-replication tracking of the S&P 500: float-adjusted market-cap weighting, 520 holdings, quarterly rebalance, weights otherwise drifting mechanically with price. No manager discretion at the fund level — but the index itself is committee-selected, not purely rules-based: S&P DJI's index committee applies profitability, float, liquidity and sector-balance judgment to additions and deletions. Fee 0.03%; inception 2010-09-07.
Assessment
- 0.03% expense ratio removes cost as a variable — the drag an active manager must overcome is ~3bp, not 200bp+.
- Full replication across 520 holdings gives near-zero tracking error versus sampled or optimised alternatives.
- Total transparency: daily full holdings and published methodology — no lagged 13F guesswork, rare among vehicles on this page.
- Structurally the right bar. Any active AI thesis must explain what it owns that VOO's ~32% Mag-7 weight doesn't already supply.
- Top-10 at 36.3% vs a ~24% historical average is a materially different product than the diversification the S&P 500 brand connotes.
- Cap-weighting is an unhedged momentum rule — maximum exposure to a name exactly at its peak weight, with no valuation-based trim.
- 'Passive' overstates it: S&P DJI's committee exercises real discretion over constituents, an unaudited selection process.
- Mag-7 earnings are cross-linked through AI capex, so effective concentration exceeds the ~32% headline weight.
- AUM is widely misquoted — the ETF share class and Vanguard's whole S&P 500 fund complex get conflated routinely.
Record
+20.13% trailing 1-year total return and 14.79% annualised since the 2010 inception (stockanalysis.com, 2026-07-21). A 3-year annualised figure could not be verified at check time, so it is parked in 'unavailable' rather than asserted. Attribution matters more than the level: Forbes reports the S&P 500 gained 68.4% from 2023-2025 while the equal-weight version returned 34.2% over the same stretch — meaning roughly half the index's cumulative return came from cap-weighting into the megacaps rather than from broad US corporate earnings. Read plainly, much of this record is one factor (megacap AI) expressed through a weighting scheme, not 500 companies compounding. That cuts both ways: it is why VOO beat most active managers, and it is why the 'safe index' and a thematic AI fund would draw down together. Treat the recent record as a bull-regime number, not a through-cycle expectation.
- 2026-08A three-year annualised figure is now retrievable at +19.35% as of 1 Aug 2026 (financecharts), where previously none could be sourced and the rate was left unstated.
- 2026-07The first test went the other way: in July 2026 SOX fell -20.6% and SMH -17.6% while VOO was -0.02%. The index's AI weight sits in NVDA +0.3%, AVGO +3.1% and MSFT +24.6% — the haven side of the trade.
Risks & fit
- AI capex disappointment hits ~a third of the fund at once; cap-weighting offers no defensive rotation on the way down.
- Multiple compression: much of the recent return is re-rating, which reverses faster than earnings decay does.
- Index-methodology or committee changes (float rules, sector caps, dual-class treatment) reshape the book with no investor action.
- Reflexivity: passive inflows have themselves been a bid under the largest weights, and that mechanism runs in reverse on outflows.
- Single-country, single-currency; no non-US and no private-market representation at all.
The concentration critique fails if index returns prove separable from the Mag-7. Concretely: a Mag-7 drawdown greater than 30% during which the S&P 500 falls less than ~10%, or the equal-weight S&P (RSP) out-returning VOO across a rolling three-year window while the index still compounds at a normal rate. Either would show breadth genuinely carrying the index and that cap-weighting is not a disguised single-factor bet. Conversely, our read is confirmed if the next 20%+ index drawdown is attributable in majority to the top ten names.
Analytically this is a reference line, not a bet: the return series active managers on this page are implicitly measured against, and the composition that explains why beating it has been so hard. Readers using it as a comparison should note it is already an AI-weighted portfolio, so a separate thematic AI position overlaps exposure already held rather than adding a new one. Nothing here is a recommendation to hold, buy, or size any position.
0.03% expense ratio — about $3 per $10,000 per year. For contrast, SPDR's SPY charges 0.0945% for the same index (~$9.45), and equal-weight RSP ~0.20%. Fee gaps this small are immaterial next to the construction differences between cap- and equal-weight.