Development of the flagship video generation model Seedance 2.0, intended to compete with OpenAI's Sora and other similar systems, has slowed down. Internal sources report that several key ByteDance divisions, including teams from Douyin (TikTok) and other products, have simultaneously requested access to computing clusters to train their own AI models. This has created a shortage, preventing the Seedance 2.0 project from obtaining the necessary volumes of GPU power on schedule. The problem is exacerbated by the global chip shortage and sanctions restrictions, which make it difficult for Chinese companies to access the most advanced processors.

This situation highlights a key problem in the modern AI race: access to computing resources is becoming a strategic advantage, while their shortage is a major brake on innovation. ByteDance, despite possessing vast data and talented engineers, is now hitting a physical limitation—a lack of 'hardware.' This calls into question the company's ability to quickly catch up with leaders in generative video, where OpenAI, Google, and Meta are already setting the tone. The delay with Seedance 2.0 could mean missed market opportunities and a weakening of positions in one of AI's most promising areas.

Technically, training models like Seedance 2.0 requires thousands of specialized GPUs (e.g., NVIDIA H100) working for weeks or months. Each video generation request is a colossal computational task. Internal competition for resources at ByteDance is particularly intense, as the company is developing many AI directions simultaneously: from recommendation algorithms for TikTok and Douyin to models for cloud services and creative studios. Parallel to the infrastructure crisis, a legal one is growing. Rights holders, including media companies and individual authors, are accusing ByteDance of using their content (videos, images, texts) to train AI without licenses or compensation payments. These claims could lead to costly litigation and a need to revise data collection strategy.

Market experts note that ByteDance's problems are typical for the Chinese AI industry as a whole, which, despite state support, faces a 'bottleneck' in chips and growing international legal risks. Analysts warn that delays in internal projects could weaken the company's competitive position in the global market, where the pace of innovation is extremely high. Investors are beginning to assess more skeptically the ability of even tech giants like ByteDance to independently break through to the forefront in fundamental AI development without unimpeded access to critical resources.

For the industry as a whole, the ByteDance story is a warning signal. It demonstrates that the era when breakthroughs were determined solely by model architecture and data volume is ending. Control over computational infrastructure and a clean legal basis for training data are coming to the fore. For users, the slowed development of Seedance 2.0 could mean later access to advanced 'Sora-for-all' video generation tools. However, it could also spur the search for more efficient, less resource-intensive AI training methods, which in the long term could make the technology cheaper and more democratic.

The prospects for the Seedance 2.0 project now depend on ByteDance's ability to solve two fundamental problems. First, the company will be forced to aggressively invest in building its own computing capacity and possibly developing specialized chips, as well as seeking workarounds for technology access. Second, it must develop a transparent data policy, potentially making deals with rights holders or betting on synthetic and legally cleared data. The open question remains whether the company can overcome these barriers quickly enough to avoid falling behind in a race that is becoming increasingly expensive and complex, not only technologically but also legally.