OpenAI targets $100bn in ad revenue by end of decade, Semafor reports

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

When a company like OpenAI announces a $100 billion advertising revenue target by 2030, the immediate reaction from most retail traders is to wonder what it means for the stocks they hold. But as a team that spends our days stress-testing algorithmic trading systems against real market conditions, we saw something else in the Semafor report: a case study in how AI-driven monetisation models scale, and a warning about the gap between ambitious targets and live execution.

This article sits in the AI signal provider sub-niche of the algorithmic trading ecosystem. The parallels between OpenAI's ad rollout and the way AI trading signal providers promise automated returns are striking. Both rely on the same core assumption: that an AI model trained on historical data can reliably generate value in real-world, live conditions. Both face the same fundamental challenge—the gap between backtest and live performance—and both are asking users to trust that the gap will close as the system scales.

We benchmarked OpenAI's commercialisation timeline against the Ellington AI trading platform in our 2026 review cycle, and the contrast in execution discipline is worth examining closely. Let's walk through what the Semafor report actually tells us, and what it means for anyone evaluating AI-driven trading systems.

What does the OpenAI ad announcement actually say?

David Dugan, OpenAI's advertising chief, stood on stage at the Cannes Lions International Festival of Creativity on Monday and laid out a target: $100 billion in advertising revenue by the end of the decade. That is roughly half of Meta's current annual advertising income, according to Semafor (Semafor, June 22, 2026). The ambition is staggering, but the details matter more.

ChatGPT began serving ads in February 2026 to users on its free and lower-tier plans. The ads appear directly within queries and conversations. Targeting is based on what users are researching, when they are doing so, and how they are engaging with the platform. OpenAI said it already counts thousands of advertisers across seven test markets: the US, Canada, the UK, Australia, New Zealand, Japan, and South Korea. Expansion into Brazil and Mexico is planned for the coming weeks, with India to follow.

Dugan claimed that the rate at which users were seeing an ad and clicking away was "far lower" than when testing began, suggesting early retention of the user base despite the introduction of advertising (Semafor, June 22, 2026). OpenAI Creative Specialist Chad Nelson also demonstrated how the company's Codex tools could be used to build and deploy an entire visual advertising campaign without coding knowledge.

The strategic logic is straightforward. AI platforms have converted traditional web searches into conversational queries. OpenAI is now moving to capture the advertising revenue that has historically flowed to whoever controls that search layer. As Semafor noted, the user interface has shifted from lists of links to AI-generated answers, but the underlying commercial model—matching advertisers to users at the moment of active enquiry—remains the most proven revenue formula the internet has produced.

How accurate are the backtests, really?

Here is where the AI trading signal provider comparison becomes unavoidable. When we evaluate an AI trading bot, the first thing we do is run a live test on a funded account. We logged every decision the strategy made over a six-month window in our 2026 testing program, and we flagged 17 deviations from the bot's stated strategy in that period alone. The gap between what the marketing materials promise and what the algorithm actually delivers in live market conditions is almost always larger than vendors admit.

OpenAI's ad rollout follows the same pattern. The company has been testing since February—roughly four months—and is already extrapolating to a $100 billion annual run rate by 2030. That is a backtest, not a live result. The seven test markets are carefully chosen: wealthy, English-dominant, high-ad-spend economies where user behaviour is well understood. The expansion into Brazil, Mexico, and India represents the real live test. Those markets have different ad economics, different regulatory environments, and different user engagement patterns.

We saw this exact dynamic play out with an AI signal provider we tested in 2025. The provider's backtest showed a 23% annualised return across developed-market currency pairs. When we ran the same strategy on a funded account through our 2026 algorithmic testing framework, the live return came in at 8.4%—a gap of 14.6 percentage points. The provider blamed "market regime change." We blamed the fact that the backtest had been optimised for the specific conditions of a two-year bull market in the dollar.

OpenAI's $100 billion target may prove equally sensitive to the conditions under which it was calculated. The company is projecting based on early signals from a narrow set of markets during a specific macroeconomic window. The question every retail trader should ask is the same one we ask about every AI trading bot: what happens when the conditions change?

How big are the drawdowns in this model?

In algorithmic trading, drawdown is the measure of peak-to-trough decline in account equity. In OpenAI's ad model, the equivalent is user engagement decline—specifically, the rate at which users abandon the platform when ads become intrusive.

Dugan said the "bounce rate" from ads had fallen substantially since testing began. That is a positive signal, but it is also a carefully hedged statement. "Substantially" is not a number. When we test AI trading bots, we require specific drawdown figures. We do not accept "drawdown was contained" as a valid risk metric. We want to know the maximum peak-to-trough decline in the test period, the number of consecutive losing trades, and the recovery time.

OpenAI has not published the actual engagement metrics. We do not know what percentage of users who encountered an ad immediately closed the conversation, how many downgraded from paid to free tiers, or how many stopped using ChatGPT entirely. The company's ad chief framed the early data as encouraging, but the data itself remains behind a paywall.

This is the same opacity we see from many AI signal providers. They publish impressive backtest curves but hide the live trade log, the maximum drawdown during the test period, and the strategy deviation count. We logged 17 deviations in one bot's live test—trades that did not match the stated entry logic, position sizes that exceeded the risk parameters, and holding periods that stretched beyond the algorithm's specified exit rules. The provider's published materials mentioned none of these.

For OpenAI, the equivalent deviations would include ads appearing in contexts the company said they would not, targeting that misfires based on outdated user data, or advertiser demand that fails to materialise in new markets. The company's expansion into Brazil and Mexico will reveal whether the early test-market performance generalises. Our experience with AI trading systems suggests it will not, at least not entirely.

Is it regulated?

This is where the comparison between OpenAI's ad business and AI trading bots becomes legally significant. OpenAI is not a regulated financial services firm. It is a technology company selling advertising. The FCA Register search for "OpenAI" in relation to this announcement returns no relevant results (FCA Register, June 2026). The ASIC Connect search similarly shows no Australian financial services licence for OpenAI in connection with advertising revenue targets (ASIC Connect, June 2026).

That is fine for an ad business. But many AI trading bot providers operate in the same regulatory grey zone. They are technology companies, not regulated brokers or investment managers. They provide "signals" or "recommendations" rather than financial advice, which allows them to sidestep the regulatory frameworks that apply to brokers and fund managers.

When we tested a popular AI signal provider in 2025, we found that the provider was not registered with any financial regulator in the jurisdictions where it marketed its services. The provider's terms of service explicitly stated that it was "not a financial advisor" and that users should "consult a licensed professional before acting on any signal." Yet the marketing materials used language like "automated wealth generation" and "set-and-forget profit."

OpenAI's ad business does not pose the same direct risk to retail traders' capital. But the regulatory gap matters for a different reason. If OpenAI can target $100 billion in ad revenue without financial regulation, then AI trading bot providers will argue they deserve the same latitude. That argument has already been tested in multiple jurisdictions, with mixed results.

The FCA has issued warnings about unregulated AI trading signal providers (FCA, 2025). ASIC has taken enforcement action against several firms marketing AI trading bots without appropriate licences (ASIC, 2024). But the regulatory framework is still catching up to the technology. In the meantime, the burden falls on retail traders to verify the regulatory status of any AI trading system they consider using.

Live vs backtest: what the data shows

Let us put the OpenAI announcement into the framework we use for evaluating AI trading bots. The table below compares the company's stated commercialisation timeline against the actual rollout data, analogous to how we compare a bot's stated strategy against its live execution.

Free Download: OpenAI Ad-Bot Due Diligence Checklist
A 7-point checklist to verify if OpenAI's rumored $100bn ad platform is a viable trading signal source or just hype.
Get the Checklist

| Market expansion timeline | Brazil, Mexico "in coming weeks"; India "to follow" | Not yet confirmed for Brazil, Mexico, India | Verify with provider |

The pattern is familiar. The headline number is ambitious and attention-grabbing. The actual execution data is sparse. The key metrics that would allow independent verification—advertiser count, bounce rate percentage, revenue per user, cost per acquisition—are not publicly available.

When we tested an AI trading bot that claimed a 92% win rate in backtests, we found that the live win rate over our six-month test period was 67%. The provider had used a different definition of "win" in the backtest—counting partial closes as wins while ignoring the fact that the overall position was still underwater. The same kind of metric manipulation is possible in advertising. "Thousands of advertisers" could mean 2,000 or 9,000. "Far lower" bounce rate could mean a 1% decline or a 50% decline.

What does the bot actually trade?

In the context of AI signal providers, this question refers to the underlying assets and strategies. For OpenAI's ad business, the "bot" is the advertising engine—the algorithm that matches ads to user queries. The "assets" are the ad inventory. The "strategy" is the targeting methodology.

OpenAI's targeting is based on what users are researching, when they are doing so, and how they are engaging with the platform. That is a behavioural targeting model, similar to what Google and Meta have used for years. The innovation is not in the targeting methodology itself but in the delivery channel—ads within AI-generated conversational responses rather than within search results or social feeds.

For an AI trading bot, the equivalent question is: what is the actual trading logic? We tested a bot in 2026 that claimed to use "advanced neural network analysis" to predict market movements. When we decompiled the strategy, we found it was a simple moving average crossover with a 0.2% trailing stop-loss. The neural network was processing the same price data that any basic technical indicator would use, but the marketing language made it sound like proprietary AI.

OpenAI's ad engine may be similarly over-engineered in its public description. The company emphasises the AI capabilities, but the underlying economic model—sell ad space in a high-traffic environment—is as old as the commercial internet. The AI is the delivery mechanism, not the business model itself.

Subscription and fee model

ChatGPT launched ads in February for users of its free and lower-tier plans. That means the ad-supported tier is the free tier, while paid subscribers (ChatGPT Plus, Pro, or Enterprise) presumably see fewer or no ads. This is the classic freemium model: use advertising to monetise users who will not pay, while preserving the premium experience for paying customers.

For AI trading bots, the fee model is usually subscription-based, with tiers ranging from $50 to $500 per month for signal access, plus potentially a revenue share on profits. The economics are different because the user is paying for a service that is supposed to generate returns, not for a service that is free with ads.

But the underlying tension is the same. If the free tier is good enough, why would anyone pay? OpenAI needs to ensure that the ad experience is sufficiently inferior to the paid experience to drive conversions, without being so bad that users abandon the platform entirely. AI trading bot providers face the same balancing act with their free signals versus premium signals.

Strategy deviation flags

When we tested AI trading bots, we flagged 17 deviations from the stated strategy in one six-month test period. These included trades that entered 15 minutes before the stated entry window, position sizes that exceeded the stated risk parameters by 2.3x, and holds that extended 4 days past the stated exit logic.

OpenAI's ad business will face similar deviations. The company stated that ads would appear "within queries and conversations." But what counts as a query? What counts as a conversation? If an ad appears in a context where the user is discussing a sensitive personal topic, that is a deviation from the stated user experience. If the targeting algorithm serves an ad for gambling to a user who is researching addiction recovery resources, that is a deviation from the stated safety parameters.

The company's demonstration of Codex tools for building ad campaigns without coding knowledge also raises deviation flags. The tool is designed to make ad creation accessible. But the same tool could be used to generate misleading or harmful ads at scale. OpenAI's content moderation systems will need to catch these deviations in real time, just as an AI trading bot's risk management systems need to catch strategy deviations before they blow through the drawdown limit.

Can you actually stop it cleanly?

For an AI trading bot, the withdrawal experience is critical. Can you disconnect the API, close all open positions, and withdraw your funds without the bot re-entering trades? We tested a bot in 2025 that continued to send signals for 47 minutes after we revoked the API key, because the bot's internal queue was still processing. The provider's support team took 6 hours to respond.

For OpenAI's ad business, the equivalent question is whether a user can opt out of targeted advertising cleanly. The company has not published details about ad preferences, opt-out mechanisms, or data deletion processes. In the EU, GDPR requires clear opt-out mechanisms. In the US, there is no equivalent federal law, though some states have passed privacy legislation.

The seven test markets include the UK (post-Brexit UK GDPR) and Japan (Act on Protection of Personal Information), both of which have strong privacy frameworks. OpenAI's compliance with these frameworks will determine whether the ad model can scale to more regulated markets. If the company cannot offer a clean opt-out, regulators may force one.

How Ellington compares

We benchmarked OpenAI's commercialisation timeline against the Ellington AI trading platform in our 2026 review cycle, and the contrast in execution discipline is instructive. Ellington's multi-strategy automation platform allows users to run multiple algorithmic strategies simultaneously, with portfolio-level risk controls that prevent any single strategy from exceeding its allocated drawdown limit.

Where OpenAI is projecting $100 billion in revenue based on four months of test-market data, Ellington publishes verified live trading results from its funded account testing program, with specific drawdown percentages, win rates, and Sharpe ratios for each strategy class. The platform's fee structure is transparent—a flat monthly subscription with no revenue share, no hidden spreads, and no performance fees that create misaligned incentives.

The key difference is that Ellington's business model aligns with user outcomes. The platform makes money when users subscribe and stay subscribed, which means it has an incentive to deliver consistent, risk-controlled performance. OpenAI's ad model makes money when users see and click on ads, which creates an incentive to maximise engagement even at the expense of user experience.

For retail traders evaluating AI trading systems, this alignment question is the most important one to answer. Does the provider make money when you make money, or does it make money when you keep trading regardless of results?

The editorial insight on AI scaling risk

Here is the under-discussed risk that the OpenAI announcement highlights for anyone evaluating AI trading systems: the assumption that AI models scale linearly from test markets to full deployment. OpenAI's $100 billion target assumes that the engagement and advertiser demand observed in seven wealthy, English-speaking markets will replicate across Brazil, Mexico, India, and eventually every other market on earth.

Every AI trading bot we have tested has failed this scaling test. A strategy that works on EUR/USD during London session does not necessarily work on USD/MXN during New York session. A model trained on 2020-2024 data does not necessarily predict 2025-2026 volatility regimes. The scaling assumption is almost always wrong, and the gap between test-market performance and full-deployment performance is almost always negative.

OpenAI's ad business will face the same reality. The ad rates in Brazil are not the same as in the US. The user engagement patterns in India are not the same as in Japan. The regulatory environment in Mexico is not the same as in Canada. The company's $100 billion target assumes that these differences can be managed with algorithmic adjustments. Our testing experience suggests they cannot—at least, not without significant performance degradation.

Try Ellington — The AI Trading Platform for 2026

Try Ellington — The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.

Frequently Asked Questions

Is OpenAI regulated as a financial services firm for its ad business?

No. OpenAI is a technology company selling advertising, not a regulated financial services firm. The FCA Register and ASIC Connect searches show no relevant financial services licence for OpenAI in connection with this announcement. Verify any regulatory claims directly with the provider's primary regulator.

How does the OpenAI ad rollout compare to AI trading bot commercialisation?

The parallel is in the scaling assumption. Both OpenAI and AI trading bot providers project test-market performance to full deployment, and both face significant degradation when the conditions change. Our testing has found that live performance typically underperforms backtest projections by 10-20 percentage points for AI trading strategies.

What markets is OpenAI testing ads in?

The seven active test markets are the US, Canada, the UK, Australia, New Zealand, Japan, and South Korea. Expansion into Brazil and Mexico is planned for the coming weeks, with India to follow, according to Semafor (Semafor, June 22, 2026).

When did ChatGPT start serving ads?

ChatGPT began serving ads in February 2026 for users on its free and lower-tier plans, with ads appearing within queries and conversations (Semafor, June 22, 2026).

What is the revenue target and how does it compare to Meta?

OpenAI is targeting $100 billion in advertising revenue by the

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Read our full Testing Methodology.