Tencent can shoulder latest US tech export restrictions as internet giant has large stockpile of much-coveted Nvidia AI chips

arifnsn November 16, 2023

0 80 2 minutes read

[ad_1]

“We have enough chips to continue our development for Hunyuan for at least a couple more generations,” Tencent president Martin Lau Chi-ping said at a conference call with analysts on Wednesday after the Shenzhen-based company’s third-quarter earnings results announcement.

Tencent, which runs the world’s largest video gaming business by revenue and China’s biggest social media operation, has one of the largest inventories of AI chips in the country, according to Lau. He said Tencent was the first to put in orders for the H800 chip, the China-export version of Nvidia’s H100 graphics processing unit (GPU) that was developed to comply with earlier US tech export curbs.

Nvidia Corp’s H800 graphics processing unit, based on its H100 chip and designed for export to China. Photo: Nvidia

Lau said the updated tech export curbs will affect enterprise clients’ ability to lease use of AI chips via the company’s Tencent Cloud service in future. As such, Tencent will optimise use of its existing inventory of Nvidia GPUs.

“Going forward, we will have to figure out ways to make usage of our AI chips more efficient,” he said. “We’ll try to see whether we can offload a lot of the inference capability to lower-performance chips, so that we can retain the majority of our high-performance AI chips for training purposes.” He also pointed out that Tencent will “try to look for a domestic source for these training chips”.

The stakes are high for Tencent, as it looks to AI technology to become the new growth engine for its business.

Tencent now using ChatGPT-like AI model in 180 services

The Hong Kong-listed company on Wednesday reported a 10 per cent year-on-year revenue increase to 154.6 billion yuan (US$21.5 billion) for the third quarter. Profit was down 9 per cent to 36.2 billion yuan.

Tencent has taken a more cautious approach as a relatively latecomer to the AI frenzy on the mainland, where 238 large language models (LLMs) – the technology used to train intelligent chatbots like ChatGPT and similar services – have been launched as of October.

LLMs are deep-learning AI algorithms that can recognise, summarise, translate, predict and generate content using very large data sets.

Nvidia to tailor another set of new chips for China clients

Tencent and other Chinese tech companies, meanwhile, are awaiting Nvidia’s update on its new server, the HGX H20, and new L20 and L2 GPUs – meant to replace the A800 and H800 chips – that complies with the updated US tech export restrictions. The development of these new products for Nvidia’s mainland clients was reported by the Post last week.

Before the updated US tech export curbs last month, a fast-growing market for smuggled GPUs from Nvidia and Advanced Micro Devices has apparently thrived on the mainland, according to a Post report in June.

Baidu CEO slams China tech firms’ frenzy over AI models as ‘waste of resources’

Tencent said it was still early to talk about monetising Hunyuan and integrating the LLM’s capabilities into social media apps WeChat and QQ.

“It’s still in a very early stage of concept design,” Lau said during Wednesday’s call. “Definitely not yet at the stage of product design and definitely not at the stage of thinking about monetisation.”

While the Hunyuan chatbot is not available for public use, Tencent has opened the Hunyuan LLM to various enterprise clients. The AI model has also been integrated into more than 180 in-house services, including conferencing app Tencent Meeting and web-based word processor Tencent Docs, along with the company’s online advertising business and WeChat search.

[ad_2]

Source link