China's cloud computing sector sees massive AI model price cuts by Alibaba and Baidu: Here's why

livemint.com:

Alibaba's cloud division announced substantial price cuts, slashing costs by up to 97 percent on various Tongyi Qwen LLMs. For example, the Qwen-Long model's price dropped to just 0.0005 yuan per 1,000 tokens, a significant reduction from the previous rate of 0.02 yuan per 1,000 tokens. Baidu quickly followed suit, declaring that its Ernie Speed and Ernie Lite models would be available at no cost for all business users, just hours after Alibaba's announcement.

The cloud computing market in China has seen a fierce price battle over recent months, with Alibaba and Tencent both reducing prices for their cloud services. AI chatbot services, bolstered by investments in LLMs following the successful launch of OpenAI's ChatGPT in late 2022, have been a key sales driver for many Chinese cloud providers. The price war has now extended to the LLMs behind these chatbots, potentially impacting profit margins for these companies.

Previously, businesses were charged to use Baidu's Ernie Lite and Ernie Speed models, which were introduced in March. Bytedance recently joined the fray, announcing last week that the primary model of its Doubao LLMs would be priced 99.3 percent lower than the industry average for business users. Chinese LLM developers have primarily targeted businesses for monetizing their investments in these models, although some have also begun addressing individual users.

For instance, Chinese startup Moonshot introduced a tipping feature that allows both businesses and individuals to pay for prioritized access to its chatbot services. Baidu was the first in China to offer LLM products to individual consumers, charging 59 yuan per month for access to its advanced Ernie 4 model. (With inputs from Reuters) Milestone