Big Tech’s AI costs have surged. Can ad revenues keep up?
Analysis
Earnings by Big Tech companies Alphabet, Amazon and Meta in Q2 displayed strong advertising revenue growth — but also staggering increases in capital expenditure amid the rush to invest in AI development.
The three American giants collectively account for roughly two-thirds of the UK ad market and the majority of the global ad market outside China. Their earnings — which continue to grow at a double-digit pace despite their already massive scale — are increasingly being used to subsidise the construction of data centres and the purchase of chips.
The question, one which has begun to spook investors this year, is whether the rapidly expanding costs associated with AI will ever deliver profitable results for these businesses, particularly a business like Meta that is not a cloud host and is entirely reliant on advertising revenue, a market in which it is already a dominant figure.
The three companies collectively have forecast capital expenditures of between $545bn and $570bn this year alone. In Q2, Alphabet increased its capex forecast this year to to between $195bn and $205bn, Meta increased its capex forecast to between $130bn and $145bn, and Amazon increased its capex forecast to $220bn.
Those costs will have to be balanced with real business results. Nimmi Shah, a consultant finance director for media, PR and marketing businesses, told The Media Leader “investors, including their clients, are looking for an actual return on their investment in AI now, not just in theory.”
Citing reports from S&P Global, Shah warned that it is unclear how much longer frontier AI model developers such as Google, OpenAI and Anthropic “can keep subsidising their customers” before needing to deliver profitability. That could mean higher costs are soon to come, with downstream effects on all businesses, including the tech giants but also media agencies, that have placed big bets on AI use driving more efficient workflows.
On the Unfiltered Media podcast last week, The Media Leader columnist Ian Whittaker agreed Big Tech’s capex numbers are “going to come under more scrutiny.”
“Yes, numbers are being delivered, they’re strong, advertising growth continues to perform quite nicely,” he said. “But investors are always looking ahead.”
The following are takeaways from the Q2 earnings results of the Big Three public ad sellers.
Alphabet
Alphabet reported 24% year-on-year revenue growth to $119.8bn, the company’s 12th consecutive quarter of double-digit overall revenue growth.
Total advertising revenue grew 14.4% to $81.6bn. A majority of the company’s revenue is still derived from Google Search, which grew 17% to $63.3bn. YouTube ad revenue also grew 13% to $11.1bn.
According to Whittaker, Alphabet’s capex now equals more than 40% of its annualised revenues, a ratio that is more typical of telecommunications and utilities companies than traditional software businesses. The increased costs have alarmed some investors.
While Alphabet’s capital expenditures ballooned, its cloud revenue also grew 82% to $24.8bn. Even with growing capex, the backlog for its cloud services expanded to $514bn, reflecting strong demand for its AI infrastructure.
Meanwhile, Gemini Enterprise, Google’s AI platform for businesses, has received “wide adoption”, including use by nearly 90% of Fortune 100 companies, Google CEO Sundar Pichai said on the company’s earnings call.
Google launches AI marketing assistant as it expands search and agentic commerce advertising options
The degree to which AI will drive business results for other segments of Google’s business remains an open question. Pichai said AI is driving “an expansionary moment” in search, and Google launched a number of new agentic tools for marketers during Q2.
Google revamped its traditional search experience in May, moving away from the “ten blue links” that have characterised the product for decades in favour of expanding AI Overviews and AI Mode features. The decision has had a profound knock-on effect for publishers by reducing clickthrough rates.
As the search experience changes, so too does the ad experience, with new ad formats tailored to conversational search results, such as Highlighted Answers and Conversational Discovery Ads.
The tech giant is also using AI to expand into commerce, rolling out shoppability to AI Mode and YouTube, as well as options to book travel and food delivery services.
Ashish Gupta, Google’s VP/GM of merchant shopping, previously claimed consumers “are really embracing AI for shopping, driving a transition towards agentic commerce where AI actually takes action on behalf of shoppers.”
Meta
Meta reported 28% year-on-year revenue growth to $60.8bn. This was driven by a 14% increase in ad impressions and a 12% rise in average price per ad.
However, such strong growth was substantially outpaced by Meta’s spending. Costs and expenses rose at double the pace of revenue in Q2, jumping 55% year-on-year to $42bn. The majority of costs were attributed to AI development, and the figure also notably includes $2.4bn in legal charges as the social media giant has fought a torrent of lawsuits this quarter, many of which accuse the company of creating harmfully addictive platforms.
To put that in perspective, Meta spent three times more in legal fees in Q2 alone than Reddit earned in total revenue in the same quarter.
It is the second consecutive quarter in which Meta’s AI strategy appears to have disappointed investors. Shares of Meta fell 8% in trading the day following the report; a similar drop occurred in Q1. Shares of Meta are down over 30% over the past year.
On the company’s Q2 earnings call, CEO Mark Zuckerberg brushed off concerns about Meta’s increased capex, arguing that AI is “improving the experience for people using our apps, driving better performance for advertisers, and helping our teams build new experiences and ship faster.”
The concern for Meta is it’s not clear the extent to which the high costs of AI development will ever become profitable, especially as Meta lacks a cloud business.
Zuckerberg argued that AI investments for advertisers “are paying off”, with large language models (LLMs) improving ad relevance and driving increased conversions on Facebook and Instagram. He also argued that its growing line of smart glasses, which have picked up the moniker “pervert glasses” over privacy concerns, are also dependent on AI features.
On the other hand, Meta has a steep hill to climb to get consumers and businesses to engage with Meta’s AI search and enterprise solutions. While Zuckerberg believes there is a “large enterprise opportunity” to sell advertisers business agents and other companies access to compute, both enterprise and consumer demand for Meta’s AI products has to this point been largely outstripped by competitors like Google, OpenAI and Anthropic. In other words, the AI market is as crowded as it is expensive.
Zuckerberg nevertheless believes Meta’s Muse brand of image and video generation tools — which allow users to create what has colloquially become known as AI slop — will create “a whole new and nearly infinite universe of personalised content” that will “make our services a lot more useful and engaging for people.”
Meanwhile, Zuckerberg said the refreshed Meta AI personal assistant has seen a 60% increase in the number of people “interacting with the assistant each day”, though it’s not clear what constitutes an interaction nor from what base this growth is from.
As Wall Street Journal writer Asa Fitch noted in reaction to Meta’s earnings, the social giant is “the most financially stretched of the biggest tech companies”, with its current ability to generate returns on AI investment hinging “solely on ad sales”.
Amazon
Amazon’s advertising segment grew 26% year-on-year to $19.8bn in the second quarter. Advertising now comprises roughly 10% of Amazon’s net sales, though it is a much larger driver of profit given the relatively high margin in its advertising business.
CEO Andy Jassy explained Amazon’s Sponsored Products offering continues to comprise the core of Amazon Ads’ business and is a “key driver of growth”.
For Amazon, AI investment is primarily benefitting its cloud business, with Amazon Web Services (AWS) growing 36.7% year-on-year to $42.2bn — its fastest pace in four-and-a-half years. Like Alphabet, despite Amazon’s $220bn in capex commitment this year, AWS still won’t have enough capacity to service demand. The cloud provider now counts a backlog of $496bn in demand.
As Jassy explained on the earnings call, AWS is now a $169bn annualised revenue run rate business. If it were a standalone company, he noted, it would place 24th on the Fortune 500 list.
Beyond cloud, Amazon’s leadership has insisted AI is helping driving growth in Amazon’s advertising business. Amazon has expanded its Ads Agent, an AI tool aimed at simplifying planning and creative development, to 11 new countries this year, including the UK in February.
The company claims brands using the Ads Agent have on average reduced their cost per impression by 8% and cost per acquisition by 6% by developing more efficient campaigns.