Tiktok owner ByteDance takes AI arms race to trillion-parameter scale

TikTok owner bets on massive new model as China’s technology groups race to close the gap with US AI leaders ByteDance, the Chinese technology group behind TikTok, is escalating its artificial-intelligence ambitions with plans to train a model on a scale that could reshape the competitive landscape between China and the US. The company is reportedly working towards an AI model with as many as 10-trillion parameters, according to the Financial Times, putting ByteDance among the technology companies pursuing the most resource-intensive systems in the rapidly expanding AI industry. The move represents a significant shift for ByteDance, whose global success has largely been built around TikTok and its recommendation algorithms. AI is increasingly becoming a strategic business in its own right, with the company investing heavily in foundation models, generative video, image creation and AI agents. ByteDance's AI research is being driven by its Seed division, which has emerged as one of China's more prominent AI research operations. The company has been releasing increasingly sophisticated models while competing with Chinese rivals such as Alibaba, Tencent and DeepSeek, as well as US technology companies including OpenAI, Google, Meta and Anthropic. The scale of ByteDance's latest project illustrates the enormous computing requirements of the AI race. Training frontier models requires vast amounts of advanced processors, data-centre capacity and electricity, turning access to computing infrastructure into a strategic business issue. For Chinese technology companies, that challenge has been compounded by US restrictions on exports of advanced AI chips to China. Those restrictions have forced companies to find ways of extracting greater performance from available hardware while developing domestic alternatives. The result has been a parallel race in China to improve model efficiency and reduce the amount of computing power required to deliver increasingly capable AI systems. ByteDance's strategy is particularly significant because of its access to one of the world's largest digital consumer ecosystems. The company can potentially deploy its AI technology across TikTok, advertising, search, e-commerce, content creation and other services, giving it multiple routes to monetise the substantial investment required to develop frontier models. Its Seed division has already moved aggressively into generative video. ByteDance's Seedance technology is aimed at producing increasingly sophisticated video content from text, images and other inputs, positioning the company to compete in one of the most commercially promising areas of generative AI. The company is also reportedly encouraging researchers to develop its models independently rather than relying heavily on techniques that allow smaller systems to learn from competing AI models. That reflects a broader change in the AI market. The competitive advantage is shifting from simply having access to generative AI tools to owning the underlying models, computing infrastructure and research capabilities. For ByteDance, the investment carries substantial financial and strategic risks. The cost of training and operating frontier models is rising even as competition pushes down the price of AI services. Companies must therefore demonstrate that increasingly expensive models can generate sufficient revenue through advertising, subscriptions, enterprise services, cloud infrastructure or AI-powered consumer products. ByteDance has one advantage that many AI start-ups lack: scale. Its experience processing enormous volumes of consumer data and operating algorithm-driven platforms provides a potentially powerful foundation for deploying AI products to hundreds of millions of users. But the company is also operating under intense geopolitical pressure. TikTok's future in the US remains politically sensitive, while Washington's restrictions on advanced semiconductor exports continue to complicate the ability of Chinese technology companies to access the latest AI hardware. ByteDance's AI investment therefore sits at the intersection of two races — a commercial battle for leadership in generative AI and a broader technological contest between China and the US. The company that succeeds will not necessarily be the one with the largest model. Increasingly, the advantage may belong to companies that can combine powerful models with affordable computing, proprietary data, distribution and commercially viable applications. ByteDance is betting that its experience building TikTok's recommendation engine gives it a head start. The next test will be whether it can turn that algorithmic advantage into a profitable AI business capable of competing at the global frontier.