ByteDance’s Volcano Engine President Tells Baidu VP Not to Speculate on Motive of LLMs Price Cuts
TMTPOST -- Chinese companies in the AI sector should learn from DeepSeek and focus on the development of AI, instead of making baseless assumptions on the motive of price cuts in large language models (LLMs), said Tan Dai, president of ByteDance’s cloud service platform Volcano Engine.
Tan’s remarks are made in response, for the first time, to Shen Dou, executive vice president of Baidu Group and president of Baidu Smart Cloud, regarding aggressive pricing competition in China's AI sector since last year..
Shen recently said in an internal meeting that last year’s "malicious" price wars in China's large model industry led to significantly lower overall revenues compared to international markets.
He pointed out that DeepSeek's rapid rise poses a direct challenge to ByteDance’s AI chatbot, Doubao, due to its high training and traffic acquisition costs.
In response, Tan posted on China’s social media WeChat Moments that price reductions in LLMs are driven by technological innovation, saying that developing AI should be the focus of all Chinese AI companies.
Tan said that Doubao 1.5 Pro's pre-training and inference costs are lower than those of DeepSeek V3 and significantly lower than other domestic models, maintaining strong profitability even at current pricing levels.
"Tech companies worldwide are using innovation to reduce model costs. We have merely achieved the price level of Gemini 2.0 Flash, which is entirely possible through technological progress," Tan said in Mandarin, as translated by AsianFin. He urged industry players to focus on innovation and fundamental improvements rather than external attributions.
Doubao-1.5-pro-32k charges 0.8 yuan per million token inputs, 0.16 yuan for cached tokens, and two yuan per million token outputs, with a disclosed gross profit margin of over 50%.
Doubao's AI model usage has surged over the past year, with its daily token usage surpassing 4 trillion in December 2023—33 times higher than when it was first launched in May.
Volcano Engine has integrated DeepSeek’s model into its offerings. The Doubao team on Wednesday unveiled a new sparse model architecture, UltraMem, which improves inference speed by 2-6 times compared to the MoE (Mixture of Experts) architecture while reducing inference costs by up to 83%, reinforcing its commitment to continuous technological innovation.
Baidu announced on Thursday that its Ernie AI chatbot will be free for individual users starting in April. This decision could help the Chinese tech giant regain momentum in the competitive race to develop next-generation artificial intelligence.
The chatbot, modeled after OpenAI’s ChatGPT, will be available to all users free of charge. Following the announcement, Baidu's stock surged more than 12% in Hong Kong, marking its biggest intraday gain since October.
Baidu’s move is expected to intensify the ongoing AI price war, as major Chinese players, including Alibaba and ByteDance, have been slashing prices since the rapid acceleration of AI development in 2023. While Baidu's Ernie was one of the first generative AI platforms to secure approval for public use in China, competitors like ByteDance’s Doubao have quickly amaessed a larger user base.
In early 2025, DeepSeek disrupted the AI landscape with a model that was trained at a fraction of the cost of its US and Chinese counterparts, gaining increasing adoption among Chinese entities. Additionally, Alibaba has attracted investors and users with its enhanced Qwen model, which has outperformed models like Meta’s Llama and DeepSeek's V3 in several benchmark tests.
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
