ByteDance's Seedance 2.0 hits 10 billion yuan in monthly MaaS revenue

ByteDance's video-generation model is proving AI can generate real revenue at scale.

ByteDance's Volcano Engine raised its 2026 MaaS revenue target to 150 billion yuan, a tenfold increase from last year's actual revenue, driven by Seedance 2.0 — a video-generation model generating over 10 billion yuan monthly.

"The model's ability to reduce short-drama production costs from weeks to hours has driven adoption faster than we anticipated," a person familiar with Volcano Engine's operations said.

The 150 billion yuan target, up from 100 billion yuan set in late 2025, compares with just 15 billion yuan in actual MaaS revenue last year. Seedance 2.0, launched in February, now contributes more than 10 billion yuan in single-month revenue, primarily from China's short-drama industry where studios use AI to generate establishing shots, complex transitions and mid-range scenes.

The milestone validates the model-as-a-service business model at a scale no Chinese AI company has achieved, putting pressure on competitors including Kuaishou's Kling, which generated 650 million yuan in the first quarter, and Alibaba's HappyHorse. ByteDance's ability to monetize video generation could reshape the competitive dynamics of China's AI sector.

How Short Dramas Became the Killer Use Case

Short-drama studios have become Seedance 2.0's largest customer segment. Traditional production requires location shoots, extras coordination, post-production effects and intensive editing, with a single episode costing tens of thousands of yuan. By replacing real footage with AI-generated scenes, studios have compressed production timelines from days to hours of prompt engineering, while eliminating location rental, travel and staffing costs.

The economics work even at current API pricing. A short-drama company executive said the cost savings from reduced production overhead more than offset the inference fees, making AI-generated content a permanent fixture in the production pipeline rather than an experimental tool.

Global Competition Intensifies

Seedance 2.0's success comes as global rivals slash prices. Google's Veo 3.1 Lite charges $0.05 per second of video, down from $0.50 for Veo 2 in December 2024. Alibaba's HappyHorse prices 1080p generation at about $0.78 per second, slightly below Seedance 2.0's roughly $0.84 per second and ByteDance's own Seedance 2.0 Lite expected at $0.50 per second.

Kuaishou's Kling, which generated 1.04 billion yuan in all of 2025, faces a different challenge. Without a cloud business to subsidize inference costs, Kling relies on subscription fees and creator incentives, spending 10.3 billion yuan on sales and marketing in the first quarter alone. Kuaishou plans to invest an additional 11 billion yuan in large models including Kling this year.

Seedance 2.0's monthly revenue of more than 10 billion yuan already exceeds Kling's full-year 2025 revenue by nearly ten times, highlighting the advantage of being embedded in ByteDance's cloud ecosystem where inference costs can be offset against broader computing and storage revenue.

What It Means for Investors

The revenue milestone shifts the narrative around AI monetization from promise to proof. ByteDance, valued at about $268 billion in private markets, has demonstrated that video-generation models can generate meaningful revenue within months of launch — a feat that has eluded OpenAI's Sora, which saw user retention drop to 1 percent by day 30.

The key question is whether Seedance 2.0's revenue growth can outpace the rising cost of compute. Volcano Engine's aggressive target implies confidence that inference costs will decline faster than revenue growth, a bet that depends on continued efficiency improvements in model architecture and hardware utilization. Competitors without cloud infrastructure to absorb losses — particularly Kuaishou's Kling — face an increasingly difficult path to profitability.

This article is for informational purposes only and does not constitute investment advice.