Alibaba said it developed a new chip for artificial intelligence (AI) inference that speeds up machine learning tasks on its platforms by 12 times the previous rate. The chip is called Hanguang 800 after a legendary sword in ancient China, and it’s already in use across Alibaba’s e-commerce sites.

Other cloud service providers including Amazon Web Services (AWS), Google Cloud, and Facebook have similar products either already in use or under development.

Alibaba said it will use the new chip in its cloud data centers to improve computing performance for its customers. It does not, however, plan to sell the chip as a standalone product.

In a statement, Alibaba CTO Jeff Zhang explained why the AI inference chip is important to the company. It will “drive both our current and emerging businesses, while improving energy efficiency. This is the first time that we’ve used our own self-developed hardware framework, while incorporating our algorithms, in the chip design,” he said. “While it can take traditional [graphics processing units] an hour to run through and categorize the 1 billion new product images that are uploaded onto Taobao each day, the Hanguang 800 can complete the same task in five minutes.”

Taobao is the company’s online shopping site.

Alibaba has invested heavily in its cloud computing arm, Alibaba Cloud, in recent years. And Zhang has said that Alibaba Cloud will be the group’s “main business” in the future.

It’s the market leader in China, and while Alibaba Cloud still isn’t as big as AWS, Microsoft Azure, and Google in the rest of the Asia Pacific region or globally, it continues to outpace the market and grow its market share, according to Synergy Research Group’s latest cloud service spending report.

Still, Chirag Dekate, a senior director and analyst at Gartner, said he expects the global impact of Alibaba’s new AI chip to be “relatively minimal. The only potential impact would be relatively less need in the long term for generalized GPU architectures at least in the Chinese ecosystem.” He does expect the new chip to help Alibaba Cloud compete more aggressively in Southeast Asia, however.

This is more of a long-term play for the Chinese tech giant, Dekate added. “They are trying to go head-to-head against Amazon and Google,” he said. “But getting a chip out there is easy. Trying to get the right software around it and making sure applications can be easily ported to take advantage of it is relatively hard.”

Google, as part of its end-to-end AI stack, first announced its AI-focused tensor processing units (TPU) in 2016 and has launched two updated versions.

AWS last November debuted its Inferentia machine learning chip for inference workloads. And Intel said in January that it is working with Facebook to develop AI chips, also built for AI inference workloads.

Alibaba is still playing catch up with these U.S.-based cloud giants. But developing its own AI inference chip aligns well with the “mission that the Chinese ecosystem has to develop indigenous technology by 2025,” Dekate said.

Baidu last year unveiled an AI chip for training and inference, and Huawei recently developed its own AI semiconductor as well.

The move also points to the growing customer demand for compute power that supports AI and deep-learning workloads, Dekate said. “This is an indicator that there is going to be increased activity in the AI and deep learning space in the next year, and you should anticipate more chip announcements especially for inference workloads.”