VGT Gets Both AI Winners: What the OpenAI and Anthropic IPOs Mean for Passive Tech Investors
Updated: Jul 18
The two most consequential AI companies are racing to public markets in 2026. Here’s why the revenue model framework — not market hype — determines where they land.
Prepared by Richstorm.co
Key Takeaways
OpenAI filed a confidential S-1 on May 22, 2026, targeting a September Nasdaq debut above $1 trillion valuation. Anthropic filed its confidential S-1 on June 1, 2026, targeting an October listing at approximately $965 billion.
Both companies qualify for MSCI’s fast-track index inclusion rule, entering eligible indexes at the close of the 10th trading day after IPO — making VGT a forced buyer within two weeks of each listing if they receive IT classifications.
Anthropic’s GICS classification as Information Technology is almost certain — its revenue is 100% API access and enterprise subscriptions with no advertising component whatsoever.
OpenAI’s classification is genuinely uncertain. Despite $25 billion in subscription and API revenue, it is aggressively building an advertising business that could shift its revenue mix before or after its September listing.
Initial index weights for both companies will be constrained by free-float — if each floats only 5–10% of shares, VGT’s actual exposure at entry will be modest, growing over 12–24 months as lockup periods expire.
The Revenue Model Framework: Two Categories, Two Destinations
Understanding where OpenAI and Anthropic land in passive indexes requires one analytical tool: the revenue model classification framework that underpins everything GICS does.
Every company that goes public receives a GICS classification based primarily on where it earns its revenue. The framework creates two distinct destinations for technology companies. Companies that sell technology directly — software licenses, API access, enterprise subscriptions, hardware — are routed to Information Technology. VGT tracks this sector. Companies that use technology to sell audiences — advertising platforms where user attention is the product sold to advertisers — are routed to Communication Services. VOX tracks this sector.
This is not a subtle distinction. It is the same framework that kept Alphabet and Meta inside Information Technology until 2018, then moved them to Communication Services when advertising became their dominant revenue model. It is the same framework that keeps Microsoft in IT despite Teams, Bing, and Xbox. And it is the framework that will determine which fund — VGT or VOX — becomes a forced buyer of the two most anticipated AI IPOs of 2026.
The answer for each company is different — and the difference matters enormously for passive investors who hold VGT.
How Fast-Track Index Inclusion Works
For most IPOs, joining a passive index requires three months of trading history and a minimum liquidity screen — a delay of up to six months from listing. MSCI operates a fast-track rule that bypasses this entirely for sufficiently large companies.
To qualify, a company needs a full market cap above approximately $26 billion and a free-float-adjusted market cap above approximately $13 billion, assessed at the close of the first or second trading day. Companies that qualify enter MSCI’s Global Investable Market Indexes at the close of the 10th trading day. OpenAI at $852 billion to $1 trillion and Anthropic at $965 billion exceed the threshold by roughly 70 times. Both will qualify as a mathematical certainty.
One important constraint: initial index weights are based on free-float-adjusted market capitalization, not total valuation. If OpenAI floats 5% of shares in its IPO, VGT must buy shares representing 5% of the float-adjusted value — not the full trillion-dollar company. Initial weights will be modest — perhaps 1–2% of VGT each — with subsequent reweighting events as lockup periods expire and public float increases over 12 to 24 months.
The question is not whether they enter the MSCI universe. It is which sector index they enter.
Anthropic: The Clean Case for VGT
Anthropic’s revenue model is unambiguous. The company generates revenue through API consumption priced per token, Claude subscription tiers for individuals and teams, and enterprise contracts for higher-volume deployments. There is no advertising component, no ad-supported tier, and no stated intention to build one. Anthropic has explicitly and publicly rejected advertising as a monetization path.
The revenue trajectory has been extraordinary by any benchmark in software history. Annualized revenue went from $9 billion at end-2025 to approximately $47 billion by the time of its June 2026 IPO filing. Eight of the Fortune 10 are Claude customers.
Over 500 customers spend more than $1 million per year. Claude Code alone generates $2.5 billion in annualized revenue, with enterprise users representing more than half of that. Every dollar of Anthropic’s revenue comes from selling access to the technology — not from selling user attention to advertisers.
GICS will almost certainly classify Anthropic as Information Technology, Application Software sub-industry — the same sub-industry as Palantir, Adobe, and Salesforce. VGT would be a forced buyer within 10 trading days of Anthropic’s October 2026 listing.
One structural caveat applies. Amazon has invested $8 billion in Anthropic and provides primary cloud infrastructure through AWS. Google has committed up to $40 billion more. These large strategic stakes raise questions about customer concentration and the durability of Anthropic’s standalone revenue model if either partner eventually builds a competing frontier model. They do not affect GICS classification — Anthropic files its own financials as an independent entity — but they are legitimate investment risks that public market investors will scrutinize in the prospectus.
OpenAI: The Uncertain Case
OpenAI’s classification is genuinely uncertain — and that uncertainty has grown in the weeks leading up to its S-1 filing.
The base case at IPO remains Information Technology. OpenAI generates approximately $25 billion in annualized revenue, with subscriptions and API access representing approximately 99.6% of that total. Advertising represents approximately $100 million — roughly 0.4% of revenue. At that revenue mix, GICS classification as IT Application Software is the almost certain outcome for the September 2026 listing. No reasonable application of the principal business activity test routes a company earning 99.6% of its revenue from software subscriptions to Communication Services.
But the advertising pivot is real, aggressive, and accelerating in a way that matters for long-term classification risk. OpenAI launched cost-per-click advertising inside ChatGPT in May 2026, hired a 21-year Google search advertising veteran to lead the business, partnered with major programmatic advertising platforms, and publicly targets $100 billion in advertising revenue by 2030. The infrastructure for a large-scale advertising business is being constructed at remarkable speed.
The classification risk is not at September 2026 IPO — it is at the first GICS annual review after listing, expected in 2027. If advertising grows from $100 million to $2–3 billion while total revenue reaches $35–40 billion, advertising represents 7–10% of revenue — unlikely to trigger reclassification. But if OpenAI’s advertising buildout succeeds and advertising approaches 20–30% of total revenue within two to three years, a reclassification review becomes plausible — following the exact precedent set by Alphabet and Meta in 2018.
An investor buying VGT after OpenAI’s September IPO should understand they are getting OpenAI at its current revenue mix. Whether they keep it depends on a business model trajectory that OpenAI’s own management has not resolved.
GICS Classification: OpenAI vs. Anthropic
The Big Four AI Tools: Who Is Winning and What It Means for Investors
Understanding the OpenAI and Anthropic IPOs requires understanding the competitive landscape they are entering as public companies. The AI tool market in 2026 has four dominant players — ChatGPT, Gemini, Copilot, and Claude — each representing a different strategic approach and a different investment structure.
ChatGPT remains the consumer leader with approximately 900 million weekly active users, but its web traffic share has fallen from 87% to approximately 55% in 14 months as Gemini, Claude, and others have scaled. Despite losing relative share in a fast-growing market, ChatGPT’s absolute user base continues growing and its enterprise revenue now represents approximately 40% of OpenAI’s total — growing fast enough to potentially match consumer subscriptions by year-end 2026.
Google Gemini has been the most aggressive share gainer, growing from 5.7% to approximately 21.5% of AI chatbot web traffic in 12 months — the fastest market share expansion in the category’s history. Gemini’s investment exposure already exists inside VOX through Alphabet, which represents approximately 22% of the fund. But Gemini’s financial contribution flows not just through advertising but increasingly through Google Cloud, which grew 63% year-over-year in Q1 2026 to $20 billion — a revenue stream that belongs analytically in IT but is classified in Communication Services because advertising still dominates Alphabet’s total revenue.
Microsoft Copilot has 420 million monthly active users and $14 billion in annualized Copilot revenue, with enterprise users averaging 11.3 interactions per workday. Copilot’s investment exposure already exists inside VGT through Microsoft at 10.2% of the fund. Every Copilot enterprise seat sold at $30 per user per month compounds Microsoft’s commercial cloud revenue today — no IPO required, no classification uncertainty.
Claude has been the most dramatic market share mover in early 2026, jumping from under 2% to 10% of US mobile chatbot daily active user share in a single quarter. Annualized revenue scaled from $5 billion to $47 billion in under a year — arguably the fastest revenue ramp in enterprise software history. Despite a smaller consumer footprint than ChatGPT, Claude wins approximately 70% of head-to-head enterprise deals against OpenAI — reflecting the deliberate professional positioning that distinguishes it from ChatGPT’s consumer mass market approach.
The Big Four AI Tools: User Metrics and Investment Destinations
The Competitive Dynamic That VGT Investors Will Own
One data point from the 2026 competitive landscape deserves particular attention. Despite ChatGPT’s 900 million weekly users dwarfing Claude’s consumer footprint, Anthropic wins approximately 70% of head-to-head enterprise deals against OpenAI. This divergence reflects the same distinction that defines our two AI revenue categories.
ChatGPT optimized for consumer scale — maximum user reach, general purpose capability, accessible to everyone. Claude optimized for professional and enterprise depth — reliability, precision, and the kind of extended analytical work that generates willingness to pay at enterprise rates. Enterprise win rates translate directly into revenue quality. A $1 million annual enterprise contract is more valuable than 50,000 free consumer users — it is recurring, predictable, and difficult to switch away from once embedded in workflows.
Whether that enterprise advantage persists after both companies go public — when OpenAI’s access to public capital accelerates its enterprise buildout — is an open question. What is clear is that this competition will play out inside VGT rather than between VGT and VOX. Passive investors who own VGT will own both sides of the frontier model competition simultaneously, without having to predict which company wins.
What VGT Looks Like After Both IPOs
If both OpenAI and Anthropic list as expected and receive IT classifications, VGT’s composition shifts in a structurally important way. For the first time, the fund would own not just the infrastructure layer of the AI economy but the frontier model layer as well.
The current VGT is a bet on AI infrastructure: Nvidia sells the GPUs that run every frontier model. Microsoft sells the Azure cloud that hosts them. Broadcom sells the networking chips that connect data centers. Applied Materials and Lam Research make the semiconductor fabrication equipment. These companies benefit from AI regardless of which consumer applications win, because every AI application requires their products.
Post-IPO VGT adds a second layer: the frontier AI model companies themselves. This is qualitatively different exposure. Anthropic and OpenAI benefit not from building the infrastructure but from using it — and their business models depend on maintaining a capability advantage that justifies enterprise customers paying premium prices for their specific models rather than building in-house or switching to cheaper alternatives.
Whether that capability advantage is durable over a decade is the central unanswered question in AI investing. If model commoditization accelerates as open-source models improve and inference costs fall, the pricing power of frontier model companies compresses. If frontier capability continues to compound in ways that open-source cannot match, Anthropic and OpenAI could become some of the most valuable software companies in history. VGT holders will own both scenarios passively, without having to make that prediction.
The Bottom Line: VGT as the Full AI Stack
The 2026 AI IPO wave, if it proceeds as currently structured, would transform VGT from an AI infrastructure fund into something more complete: a passive vehicle that owns the entire AI technology stack from foundational infrastructure to frontier model applications.
The foundational layer is already there. Nvidia’s GPUs power every frontier model. Broadcom’s networking chips connect the data centers. Microsoft’s Azure hosts the workloads. The semiconductor equipment companies make the chips that make everything else possible. VGT has owned this layer since Nvidia’s AI transformation began compounding.
The application layer is about to arrive. Anthropic’s pure enterprise software model routes it cleanly into IT and into VGT. OpenAI’s classification is likely IT at IPO based on current revenue mix, with a real but manageable risk of future reclassification if its advertising pivot succeeds beyond current projections. Microsoft’s Copilot enterprise revenue is already inside VGT and compounding at $14 billion annualized.
The one layer VGT does not own is the consumer advertising application layer — Meta’s AI-powered ad targeting, Alphabet’s AI search monetization, the consumer platforms that monetize user attention rather than selling technology directly. That layer sits in VOX, classified in Communication Services, held alongside Verizon and AT&T.
The companies building and selling the AI economy’s capabilities are converging on one passive vehicle. That convergence is not a strategy — it is the mechanical consequence of a revenue model classification system routing companies to their sectors based on where the money comes from. And in 2026, the money at OpenAI and Anthropic comes from selling technology, not selling audiences.
