VGT: The Invisible Architecture Of American Technology

VGT: The Invisible Architecture of American Technology

What you actually own — and what you don't — when you buy Vanguard's most popular tech ETF

Key Takeaways

• VGT tracks the MSCI (MSCI) US Investable Market IT 25/50 Index — a rules-based GICS sector fund that owns semiconductors, systems software, and platform software, but deliberately excludes Alphabet (GOOGL), Meta (META), and Amazon (AMZN).

• The ETF's 322 holdings are misleading: the top 5 stocks account for over 50% of the fund, making VGT effectively a concentrated bet on a handful of mega-cap names with a long tail attached.

• Whether OpenAI ends up in VGT after its IPO depends entirely on one question: does ChatGPT become an advertising business or stay a software business — a question OpenAI itself has not yet resolved.

• VGT and XLK have a 0.99 correlation and nearly identical 10-year returns (24.9% vs. 24.6% annualized), making the popular debate between them largely a false choice.

• Passive ETFs like VGT do not fund the companies they hold — they buy existing shares on the secondary market — meaning VGT's small-cap tail provides diminishing optionality as the best tech companies stay private longer.

What VGT Actually Is

Most investors think of VGT as a technology fund. That description is accurate but incomplete. VGT is specifically a fund that owns companies classified as Information Technology under the Global Industry Classification Standard — a rulebook jointly maintained by MSCI and S&P Global (SPGI) that assigns every publicly traded company to exactly one economic sector based primarily on where it earns its revenue.

That distinction matters more than it first appears. The GICS Information Technology sector covers three specific economic activities: technology software and services, technology hardware and equipment, and semiconductors and semiconductor equipment. It does not cover companies that use technology to sell advertising, run e-commerce marketplaces, or stream entertainment — even if those companies are among the most technologically sophisticated enterprises on earth.

The result is a fund that looks like a broad technology bet but is actually something more precise: a bet on the companies that build and sell the infrastructure and operating layer of the digital economy, rather than the companies that monetize that infrastructure through consumer-facing applications.

The Companies You Don't Own — and Why

When you buy VGT, you do not own Alphabet, Meta, or Amazon. This surprises most investors who associate all three with technology. The reason sits in their revenue structures.

Alphabet generates approximately 77% of its revenue from advertising. Meta generates approximately 98% from advertising. GICS classifies both under Communication Services — the same sector that houses traditional broadcasters and telecom companies — because their principal business activity is selling access to audiences, not selling software or hardware.

Amazon presents a different case. Its AWS cloud division is pure IT infrastructure, but its retail and advertising businesses dominate revenue, placing it under Consumer Discretionary in the GICS framework.

This is not an oversight or a flaw in the index. It is the classification system working exactly as designed: follow the money, not the technology. A company whose revenue comes from advertising is an advertising company that happens to use sophisticated technology — not a technology company.

The practical consequence for VGT investors is significant. The three largest consumer-facing AI platforms — Google Search, Instagram, and Amazon — are absent from the fund. What VGT owns instead is the layer underneath: the chips that power those platforms, the software that runs them, and the hardware infrastructure that connects them.

The Invisible Rulebook: MSCI, GICS, and the 25/50 Cap

Vanguard does not decide what VGT holds. MSCI does — and understanding that distinction changes how you think about the fund.

MSCI is a publicly traded financial data company with $3 billion in annual revenue and a 56% operating margin. Its core business is defining, calculating, and licensing indexes to fund managers who build investment products around them. When Vanguard licenses the MSCI US Investable Market IT 25/50 Index, it agrees to hold whatever MSCI puts in that index, in whatever proportions MSCI specifies, updated on MSCI's quarterly review schedule. Vanguard's role is execution, not judgment.

The "25/50" in the index name describes a regulatory concentration constraint. No single stock can exceed 25% of the fund at rebalancing. All stocks individually exceeding 5% of the fund cannot collectively exceed 50%. This is why Nvidia (NVDA) sits at approximately 18.6% rather than the 25%-plus weight its raw market capitalization might otherwise justify. Without this cap, VGT would be even more concentrated in its largest holding than it already is.

MSCI reviews the index quarterly — in February, May, August, and November — adding companies that have grown large enough to qualify, removing those that have shrunk below minimum thresholds, and updating weights as market capitalizations shift. These changes are announced in advance, giving Vanguard time to trade at the new composition before it takes effect.

For large IPOs — those with free-float-adjusted market caps above approximately $13 billion — MSCI can move even faster. Under its early inclusion rules, a qualifying company can enter the index as soon as 10 trading days after its IPO. This means VGT could be a forced buyer of a major new listing within two weeks of its debut, before any fundamental analysis of the company has been widely published.

Fund at a Glance

Top 10 Holdings (March 31, 2026)

The OpenAI Question: Software Company or Advertising Company?

OpenAI officially launched advertising inside ChatGPT in February 2026, shifting from a pure subscription and API model toward a consumer advertising business. The company projects $2.5 billion in ad revenue for 2026, scaling to $100 billion by 2030, while simultaneously planning an IPO filing in the second half of 2026 at a potential valuation approaching $1 trillion.

For VGT investors, this creates a genuinely live and unresolved question: will OpenAI land in the Information Technology sector or the Communication Services sector when it goes public?

The answer depends entirely on revenue composition at the time of classification. If subscriptions and API enterprise contracts remain the dominant revenue source at IPO, GICS classifies OpenAI as Application Software under IT — and VGT automatically acquires it within 10 trading days under the fast-track inclusion rule. If advertising revenue has grown large enough to represent OpenAI's principal business activity by that time, GICS routes it to Communication Services alongside Alphabet and Meta — and VGT gets nothing.

The Alphabet and Meta precedent is instructive. Both companies began as technology companies and were reclassified out of IT in 2018 when GICS updated its methodology to reflect that advertising had become their dominant revenue model. OpenAI is actively building that same revenue architecture while simultaneously targeting public markets. It is currently unknowable which revenue stream will dominate at the moment GICS makes its classification decision.

Anthropic, by contrast, has explicitly rejected advertising. Revenue comes from API access and enterprise subscriptions — a profile that maps cleanly to IT Application Software regardless of eventual IPO timing. If Anthropic were to go public, its classification would be essentially unambiguous.

The broader point is this: the technology companies investors most associate with the AI revolution may or may not end up in VGT. The classification outcome is not a technology question — it is a business model question that several of the most important companies in the sector have not yet resolved.

Winner Takes All — True at the Company Level, Irrelevant at the ETF Level

The winner-takes-all argument for technology investing is well-grounded in economics. Digital platforms have high fixed costs and near-zero marginal costs, creating powerful economies of scale. Network effects — where a platform becomes more valuable as more people use it — amplify early leads into nearly insurmountable moats. The combination of scale economics and network effects has produced the most concentrated corporate value creation in market history.

Where the argument breaks down is in its translation to ETF selection. The question is not whether technology winners take all — they clearly do. The question is whether owning a more concentrated large-cap-only fund captures that dynamic better than owning a broader fund. The empirical answer over 10 years is: it does not.

VGT and XLK — the two dominant pure IT ETFs — have a 0.99 price correlation and produced nearly identical 10-year returns: 24.9% annualized for VGT vs. 24.6% for XLK. The difference is statistical noise. VGT holds 322 stocks across large, mid, and small cap; XLK holds 71 large-cap-only names. Despite the structural difference, the outcome is functionally the same.

The reason is market cap weighting mechanics. Both funds are dominated by the same 3-5 companies. VGT's additional 250 holdings collectively represent a small fraction of total fund weight — enough to occasionally add performance during periods of mid-cap semiconductor outperformance, but not enough to meaningfully differentiate the two funds over most market cycles.

VGT vs. XLK Annual Returns (2016–2025)

The Concentration Hiding Inside Diversification

VGT's 322 holdings create an impression of diversification that the fund's actual risk profile does not support. The top 5 holdings — Nvidia, Apple, Microsoft, Broadcom, and Micron — account for approximately 53% of total fund weight. The top 10 account for nearly 60%. The remaining 312 companies collectively represent the other 40%.

This creates a specific risk that passive investors rarely consider: single-theme concentration. Nvidia, Microsoft, and Apple are all deeply exposed to the AI infrastructure buildout — Nvidia through GPU sales, Microsoft through Azure cloud and AI services, Apple through the device layer. A significant cut in hyperscaler capital expenditure would simultaneously pressure all three of the fund's largest positions, leaving Apple as an insufficient hedge given its own AI exposure through on-device processing.

The 322-stock wrapper does not distribute this risk. It packages it alongside hundreds of smaller companies that are too small to provide meaningful protection when the top holdings decline in concert.

This is not a critique of VGT's design — it is an accurate description of what the IT sector actually looks like. Technology leadership in the current era is extraordinarily concentrated. The index reflects that reality faithfully. What investors need to understand is that buying a "diversified" tech ETF with 322 holdings is structurally closer to buying 5 stocks than it appears.

ETFs Are Followers, Not Funders

One of the most persistent misconceptions about passive investing is that ETF inflows fund the companies held in the fund. They do not.

When you buy VGT, Vanguard does not send money to Nvidia, Microsoft, or any small-cap IT company in the index. ETF shares trade on stock exchanges in the secondary market — between investors. The underlying securities change hands through institutional intermediaries called authorized participants, who assemble and disassemble baskets of stocks to create or redeem ETF shares. No cash flows from the ETF to the companies themselves.

The only time a company receives capital from equity investors is at its IPO or through a subsequent secondary offering — events driven by investment banks, venture capital, and private equity, not by passive index funds.

This has a direct implication for how to think about VGT's small and mid-cap holdings. The traditional argument for owning a broad index was that it captured future winners before they became well-known. But the companies most likely to become the next generation of technology leaders are increasingly spending their entire high-growth phase in private hands — funded by venture capital and sovereign wealth funds — and arriving on public markets only when growth has already begun to decelerate. The median age of companies going public has risen from 8 years in the mid-1990s to 14 years in 2024.

VGT's small-cap tail does not fund the next Nvidia. It buys the next Nvidia's shares only after Nvidia has already done most of its compounding. The optionality argument for broad index inclusion is structurally weaker than it was a decade ago.

The Sub-Industry Structure That Actually Drives Performance

VGT's modest outperformance over XLK in certain periods is not explained by its small-cap breadth. It is explained by sub-industry composition.

VGT's MSCI IMI index casts a wider net across the IT sector, capturing more mid-cap semiconductor and semiconductor equipment companies that do not yet qualify for S&P 500 inclusion — XLK's eligibility requirement. This difference is marginal in most years but decisive in periods when semiconductors outperform the rest of the IT sector.

The 2020-2024 semiconductor surge — driven by AI infrastructure demand, data center buildout, and the end of the chip shortage — favored VGT precisely because its slightly higher semiconductor sub-industry weight compounded over multiple years of extraordinary returns. Nvidia's trajectory from mid-cap GPU maker to the world's most valuable semiconductor company was captured more fully by VGT's index methodology than by XLK's large-cap filter.

VGT Sub-Industry Breakdown

Who Should Own VGT — and at What Weight

VGT is a high-conviction, high-volatility instrument. Its maximum historical drawdown is 54.6%, recorded in November 2008. Its 2022 drawdown was 29.7%. Investors who held through both events and maintained their positions captured a 10-year annualized return of approximately 24.9% — one of the strongest long-term records of any passive fund available to retail investors.

The fund is appropriate for investors who have made a deliberate, informed decision to concentrate in the GICS Information Technology sector — specifically its infrastructure and software layers — and who have the risk tolerance and time horizon to hold through drawdowns that may last 18 months or longer. It is not appropriate as a diversifying position within an equity portfolio, because its 0.99 correlation with XLK and its high correlation with broad growth indexes means it moves in lockstep with whatever tech exposure the investor already holds.

Investors who hold VGT alongside VGT-heavy broad market funds — VOO, QQQ, VONG — should audit their actual aggregate exposures. The same top 5 names appear in each fund. Diversification across tickers is not diversification across risks when the underlying holdings substantially overlap.

The most defensible use case for VGT is as a deliberate sector overweight — an investor who wants US equity exposure tilted toward the IT infrastructure layer beyond what a broad market fund already provides, with clear understanding of what the fund holds, what it deliberately excludes, and what external classification decisions could change its composition overnight. 

VGT Historical Performance Summary

The Bottom Line

VGT is not what it appears to be on the surface. It is not a broad technology fund — it is a GICS IT sector fund that owns the infrastructure and software layer of the digital economy while deliberately excluding the consumer-facing companies that dominate technology headlines. It is not a diversified fund — it is a concentrated bet on 5 companies dressed in a 322-stock wrapper. And it is not a vehicle for funding the next generation of technology innovation — it is a secondary market instrument that tracks the outcome of a capital formation process that happens primarily in private markets.

Understanding these distinctions does not make VGT a worse investment. Over 10 years, it has delivered approximately 24.9% annualized returns with a Sharpe ratio of 0.89 from 2015 to 2025 — the strongest risk-adjusted performance of any major passive fund in its comparison set. But those returns came specifically from the semiconductor and platform software sub-industries that happen to be VGT's core exposure, not from the breadth of its holdings.

The investor who understands VGT as a concentrated semiconductor and IT infrastructure bet — with an upward-drift lottery ticket on 300 additional names — is better positioned to hold it through its inevitable drawdowns than the investor who believes they own a broadly diversified technology fund. In investing as in science, precision of understanding is the foundation of sound decision-making.