OpenAI wants $100bn of ad revenue. But it can’t ‘guarantee’ brand safety
The Media Leader Interview | Cannes Lions 2026
Benji Shomair, OpenAI’s VP of monetisation, speaks to The Media Leader about how OpenAI is rapidly iterating its fledgling ad platform.
It has been 19 weeks since OpenAI launched its advertising model.
“It’s deliberate to count weeks because it’s been a very short time,” Benji Shomair, OpenAI’s VP of monetisation, tells The Media Leader from a room with a view in the Majestic Barrière Hotel in Cannes.
In that time, the AI giant has rapidly pivoted its testing model from high-cost per thousand impressions (CPMs) to lower-cost CPMs and now to cost per click (CPC). It was “pragmatic” to start with CPMs, Shomair says, because the company “wanted to get something out to the world, water running through the pipes”.
“We aspire to a much more sophisticated adtech stack,” he adds, calling CPC a “step in the right direction” but still not the full-fledged outcome measurement that OpenAI intends to bring to market “as soon as possible”.
Shomair continues: “When we put that measurement plumbing in place, you can start to see conversions, which is even closer to what [marketers] really want.” OpenAI is currently testing conversion optimisation.
The company now counts “thousands” of advertisers across the seven markets in which its ads pilot is live: the US, Canada, UK, Australia, New Zealand, Japan and South Korea. Brazil and Mexico are slated to be next, “in the coming weeks”.
ChatGPT, meanwhile, now has over 900m users globally, and the company’s commercial leadership claims over one-fifth (22%) “show commercial intent” when they interact with the chatbot.
Mission alignment and desire for speed
Shomair shares that OpenAI’s philosophy for its ads business is “user value and privacy first, advertiser value second”. The team, he insists, is at great pains to protect the ChatGPT user experience and not undermine trust in the chatbot even as advertising is introduced.
Why get into advertising in the first place? OpenAI has been criticised for commercialising its product through ads, with competitors such as Anthropic running Super Bowl ads promising to keep their chatbots ad-free. (“That was a fun weekend,” Shomair joked of his reaction to Anthropic’s campaign.)
But Shomair insists ads are “mission-aligned”. OpenAI’s ostensible mission, born out of when the company operated solely as a not-for-profit, is to “bring the benefits of artificial general intelligence (AGI) to all of humanity.” (Artificial general intelligence is defined as AI that can outstrip human ability at most or all tasks, something that even AI moguls have not yet concluded has been demonstrated.)
“If you want to bring the frontier capabilities to more and more people, you actually need a model like ads,” Shomair reasons. “Because otherwise it’s gated by a subscription.
“If you think of the swaths of humanity across the world, there are many people who don’t have the means or the interest to pay for those subscriptions, and that’s fine. We need to find a way to give them access to these tools, if they are as beneficial and powerful as we believe.”
Another reason to embrace advertising? OpenAI needs cash — fast.
According to a leak of OpenAI’s 2025 earnings published earlier this month by blogger Ed Zitron and the Financial Times, OpenAI earned just $13.7bn in revenue last year, compared with $34bn in costs and expenses. The AI giant has also reportedly told investors it expects to spend $600bn on infrastructure commitments by 2030 (this was revised down from $1.4tn).
It is not clear whether OpenAI will ever be able to generate enough revenue to break even from its tremendous costs. Investment bank HSBC has predicted the company won’t become profitable by 2030, and in April, The Wall Street Journal reported that OpenAI had missed its own targets on revenue and user growth. CFO Sarah Friar was reported to have privately warned that the company would be unable to pay for future computing contracts without faster revenue growth.
Axios has reported OpenAI expects to generate $2.5bn in ad revenue this year, with expectations that this will scale rapidly to $100bn by 2030.
OpenAI has not made any public comments about its financial health. Shomair offers that, rather than cost pressure, the desire to move fast on its ads business is simply “because we can”.
“We know where we need to go and we have the tools now, so why would you slow down?” he asks rhetorically. “We know that we need to get to this place of a more sophisticated ad platform for our clients, and so we’re running there as fast as possible in service of them.”
‘What is a guarantee?’
That all sounds very “move fast and break things”, and not just because Shomair (like other recent additions to OpenAI’s leadership team) was once an employee of Facebook.
“We’re not breaking things,” Shomair tells The Media Leader, referring to OpenAI’s ads business. “When you build an adtech stack, you don’t want to break things.”
Still, ChatGPT was the first significant large-language model (LLM) of its kind to launch to consumers in 2023, sparking ongoing concerns about copyright infringement, user safety, and the spread of misinformation.
At the WAN-IFRA World Media Congress this month, The New York Times publisher AG Sulzberger said, “AI companies’ data takes without consent or compensation”, adding that “OpenAI has acknowledged that it would be impossible to train the most popular AI models without using copyright-protected material.”
Shomair declined to answer when asked by The Media Leader whether OpenAI would be open to returning some of the ad revenue derived from copyrighted material back to publishers cited by ChatGPT. At WAN-IFRA, OpenAI’s VP of media partnerships, Varun Shetty, said the company has no plans to share ad revenue with publishers.
New York Times publisher warns journalism faces its ‘Napster moment’
Meanwhile, several bereaved families have accused ChatGPT of encouraging their loved ones to commit suicide.
When asked by The Media Leader how OpenAI is working to ensure ads don’t appear in inappropriate contexts, such as conversations about suicidal ideation, Shomair insists that OpenAI has taken a “deliberate and measured approach” that puts users’ interests first.
The reason for developing an in-house adtech stack, he adds, is to “have full control over the messages that are appearing within ChatGPT to make sure they are showing up in the right places and showing up in a positive, additive way.”
Shomair continues: “Relative to some other platforms, we are in some ways more concerned about these issues [of brand safety] than the advertisers,” Shomair claims. “And the advertisers are concerned, you’re right. But we’re actually even more concerned. In this intimate shared space, who is it worse for if we show an inappropriate ad there? It’s actually worse for us, because we have damaged the trust that users have placed in us.”
OpenAI’s ad policy states that conversations deemed “sensitive” or otherwise containing “prohibited contexts” are ineligible to receive ads.
“You can guarantee that?” The Media Leader asks.
“What is a guarantee?” Shomair responds.
“Can you actually give 100% certainty that no ad will be shown there?” The Media Leader asks.
“Just like any other system, there’s automation that drives this,” says Shomair. “All other platforms have similar things. And they aspire as much as possible to capture everything.”
Making decisions
Advocates for greater action on online safety, such as Conscious Advertising Network co-founder Jake Dubbins, have warned that tech platforms have perverse incentives for keeping users on their platforms. “The longer we are all on those platforms, the more money that can be made,” Dubbins said at Advertising Week Europe this spring.
The logic, therefore, is that platforms will always optimise to keep users on their apps or sites for as long as possible, even if this requires harmfully addictive design, serving whatever content maintains engagement, or — in the case of chatbots — designing for sycophancy.
Does this not apply to ChatGPT now that it is embracing ads?
“We’re not a social platform,” Shomair says, acknowledging that such companies “are generally incentivised to have you stay there longer and longer”. But people come to ChatGPT, he argues, “to make decisions”.
“So it isn’t actually a system that is focused on time spent; it is a system focused on getting you the answer and the information you need and making a better-informed decision.”
‘Right diagnosis, wrong prescription’? Adland torn as Starmer announces under-16s social media ban
Repeating that OpenAI’s “North Star” is “user value and user privacy”, Shomair says ChatGPT users have appeared to have responded positively to ads for this reason. He describes how OpenAI uses a wealth of data on user behaviour to test their responses to ads, such as whether users click the X button on an ad or how often they return to ChatGPT before and after they began seeing ads.
When asked whether OpenAI considers market share relative to competitors as one measure of tracking people’s feedback of being served ads, Shomair plainly says: “No.”
Rising competition from Google’s Gemini and Anthropic’s Claude, however, has caused ChatGPT to fall below 50% market share for the first time this month.
‘A lot of optimism and excitement around AI’
Questions abound for OpenAI ahead of an expected initial public offering this year. Can it make enough money to stay in business? How fast can it scale its ad business? Can it stave off competition from products that are highly similar? Will it put safety before profit? After all, its mission is to create technology for the “benefit of all of humanity”.
Can it do so without further wrecking the global environment, already engulfed in a climate crisis, by causing water and land shortages and a massive carbon footprint?
According to the United Nations University, AI data centres’ electricity use in 2025 had a carbon footprint that would require 3.2bn tree seedlings grown over 10 years to offset, roughly equivalent to the total number of trees in the entire United Kingdom. Data centres’ electricity consumption last year was equivalent to meeting the annual basic domestic water needs for over 600m people in Sub-Saharan Africa.
Advertisers don’t appear concerned. When asked by The Media Leader whether marketers have sought information on the environmental impact of their adspend on ChatGPT, especially given the widely reported negative externalities AI development has had on the environment, Shomair says: “It’s not a conversation I’ve had.”
Instead, speaking in between back-to-back meetings with high-value clients amid a record-setting European heatwave, Shomair’s sense is that there is “a lot of optimism and excitement around AI within the ad industry”.
“It’s not search, it’s not display, it’s not social. It’s a decision layer,” he says of ChatGPT, one that is neither “lean back” nor “lean forward”, but rather an “upper-mid funnel” opportunity for users toward the start of their product consideration journey.
“There’s always been this consideration phase. What’s unique is that it was taking place offline in so many different places that it was hard for a marketer to participate.
“Now that it’s come to ChatGPT, and now that we’re launching ChatGPT Ads, we believe it’s one of the first opportunities for a marketer to participate in those moments of discovery.”