AppLovin misses guidance despite 53% revenue growth in Q2

AppLovin reported revenue of $1.92bn for the quarter ended June 30, up 53% year on year.

The adtech giant posted a rise in adjusted EBITDA of 58% YoY to $1.61bn, but fell short of guidance for the first time in several quarters, with margin holding at 84%. Net income reached $1.27bn, up 55% YoY.

Despite the revenue growth and record profitability, AppLovin shares fell sharply in the aftermath, dropping more than 20% as investors focused on the guidance miss and margin pressure from higher AI training costs.

CEO Adam Foroughi attributed the shortfall to timing rather than any change in advertiser demand.

“This quarter came down to timing,” he said on the earnings call. “Our pace of meaningful model improvement was lighter than normal during the quarter, and the next step up in model performance landed just after quarter end.”

Foroughi said the company still believes its combination of gaming-model improvements and consumer expansion could deliver around 30% annual growth over time.

The company also pointed to continued growth outside gaming, with advertiser spending in its consumer advertising vertical hitting a new high, 28% above Q4 2025 levels. AppLovin said this segment remains too small to fully offset variability in its gaming business but expects its contribution to grow more meaningful over time.

CFO Matt Stumpf said the company is targeting adjusted EBITDA margins in the low-80s over the long term, with dips possible in the near term while the company continues to invest in AI and infrastructure.

Foroughi added that Max, AppLovin’s publisher-facing ad mediation platform, saw publisher earnings grow sequentially in the double digits, and said there was no indication during the quarter of weaker advertiser demand or a shifting competitive landscape.

AppLovin guided for revenue of $2.06bn to $2.09bn for the quarter ending September 30 – growth of 46-48% year on year – with adjusted EBITDA of $1.71bn to $1.74bn, around 83% margin. That outlook also factors in higher training and compute costs tied to the model improvements now live.