Airbnb CEO Brian Chesky this week said that the company’s spending on inference “pales in comparison” to the returns it generated in Q2 2026. He added that the firm plans to ramp up its AI spending across the months to come, becoming an "AI-native company."
This came during an exclusive interview with CNBC after a 17.4% jump in the company’s stock following its second-quarter results, closing at $178.07. The company’s revenue is likewise up 17%, according to the latest shareholder letter, reaching $3.6 billion, with net income of $816 million.
Rather than treating AI as incidental to the company’s success, Chesky has framed this as "probably the number one explanation for our results." His explanation is that even a small increase in the number of customers completing a transaction can cover the cost of a large number of tokens - the billing unit for AI models, roughly three-quarters of a word each.
The shape of Airbnb’s AI adoption
The majority of major technology firms have integrated AI into their development and customer service workflows. However, Chesky’s interview draws an unusually direct causal link between AI spending and Airbnb’s financial performance.
Chesky attributed Airbnb’s strong performance to its internal AI adoption, and in turn to the arrival of new CTO Ahmad Al-Dahle, who joined the company in January after previously heading up Meta AI’s Llama model program. Al-Dahle has overseen a significant expansion of Airbnb’s internal AI adoption. Airbnb now uses more than a dozen models, including OpenAI’s Codex and Anthropic’s Claude Code.
As a result, Airbnb has stated that its development timelines have been shortened by as much as 60%. It also claims to have shipped 80% more new features and upgrades in the first half of this year than during the same period in 2025. Crucially, it adds that its AI initiatives have not led to layoffs; instead, roughly the same number of employees have been able to ramp up their output.
On the customer-facing side, Airbnb reports that its support costs per booking fell by roughly 16% after it began delegating support tasks to an AI agent. Around 45% of the customers who spoke to this agent did not need to be transferred to a human, it added. Competitors Booking Holdings (operator of Booking.com) and Expedia have also integrated AI into their workflows.
On an earnings call, Booking CEO Glenn Fogel acknowledged that its net return on AI investments is already positive; its own customer-service costs are also down by a double-digit amount.
However, he added that referrals from external LLMs account for less than 1% of total room nights, a metric it seeks to grow going forward. Expedia likewise told investors that its natural-language trip planning does not yet convert bookings but captures valuable information about traveler intent.
The scale of Airbnb’s AI spending is still unknown
Airbnb filed its quarterly report with the SEC, but this does not go into granular detail on some aspects of expenditure. The report splits spending into five functional categories. It does not separately disclose AI or token spending, making its total cost impossible to isolate.
What is clear, however, is that Airbnb’s AI spending is only set to increase. CFO Ellie Mertz said on an earnings call that its updated guidance assumes a “material increase” throughout the rest of this year.
Mertz pointed to lower customer-service costs and slower headcount growth as two areas already offsetting the additional expense. This gives Airbnb’s AI experiment a clearer test in the quarters ahead: whether the company can continue increasing its inference bill while simultaneously widening margins and accelerating revenue growth.