ByteDance, the Chinese tech giant behind TikTok, is reportedly pre-training an artificial intelligence model with as many as 10 trillion parameters, according to the Financial Times. If accurate, the system would be one of the largest disclosed AI models in the world, potentially rivaling Anthropic's Mythos in scale.
Parameters are the adjustable numbers an AI model learns from data. They are an imperfect proxy for capability, but they do signal how much computing power, data, and engineering a company is willing to invest. A 10-trillion-parameter model would represent a massive leap over most existing systems.
Context: China's AI race heats up
ByteDance's reported move would put it far ahead of China's current biggest disclosed models. Moonshot AI, a Chinese AI startup, has said its Kimi K3 model has 2.8 trillion parameters. Meituan, a Chinese tech giant, and DeepSeek, an AI lab, have previously pointed to models around 1.6 trillion parameters. A 10-trillion-parameter model would be a step change in scale.
The report comes amid rising US-China tensions over AI technology. The US has imposed export controls on advanced chips, which could complicate ByteDance's ability to source the necessary hardware. However, the company has been building out its own AI infrastructure and has access to domestic chip alternatives.
ByteDance's founder has previously urged staff to focus on long-term AI development rather than taking shortcuts, as US-China tensions rise. This new report suggests the company is pushing ahead with ambitious plans despite the geopolitical headwinds.
What it means for investors
For everyday investors, this news is a reminder that the AI race is not just a US story. Chinese companies like ByteDance are investing heavily in frontier AI, which could have implications for global technology competition and the companies that supply the underlying infrastructure.
Large-scale AI models require enormous computing resources, which drives demand for data centers, chips, and energy. This is part of a broader trend where Big Tech's AI data center leases total $1.16 trillion, according to Reuters. That spending is a key driver for companies like Amazon's AWS, which has seen accelerated growth on AI demand.
For investors, the key takeaway is that AI development is a capital-intensive race with no clear winner yet. While scale is one measure of progress, it doesn't guarantee commercial success. Many experts caution that parameter counts are not the only factor in model quality; training data, architecture, and efficiency also matter.
ByteDance's reported move also highlights the competitive pressure on US AI leaders. If Chinese models can match or exceed US capabilities, it could affect the pricing power and market share of companies like OpenAI, Anthropic, and Google. However, it's still early days, and regulatory and geopolitical factors could slow or alter the trajectory.
What to watch next
Investors should watch for official confirmation from ByteDance, as well as any details about the model's performance and deployment. The company has not commented on the report, and pre-training is an early stage; many models never make it to production.
Also worth monitoring is how US export controls evolve. If ByteDance cannot access the most advanced chips, it may struggle to train and run such a large model efficiently. Conversely, if it succeeds, it could signal that China's domestic chip industry is maturing faster than expected.
Finally, keep an eye on the broader AI investment cycle. The massive spending on AI infrastructure has boosted tech stocks, but it also carries risks if returns don't materialize. As Alibaba's open-weight AI model could shake up US tech giants, the competitive landscape is shifting quickly.
For now, the ByteDance report is a reminder that the AI race is global, expensive, and far from decided. Investors should stay informed but avoid making hasty decisions based on unconfirmed reports.
