Deep learning: Chinese state-backed funds bet on GPU developers

Data InsightDealspeak 6 June

Deep learning: Chinese state-backed funds bet on GPU developers

Chinese funds are betting heavily on national developers of graphic processing controlling units (GPUs), according to Mergermarket and AVCJ data.

The trend is being driven by growing demand for artificial intelligence (AI) applications, particularly deep learning, which involves massive amounts of data and is dependent on GPU capacity.

Last week, Chinese chip developer Jingjia Microelectronics [SHE:300474] announced plans to raise CNY 4.2bn (USD 591m) through a private placement of shares to invest in two projects to develop chips for GPUs.

On the other hand, US chip sanctions have added urgency to talk of Chinese tech self-sufficiency. The complex geopolitical situation makes it difficult to extract comprehensive data about transactions related to this sector in China.

Trade and investment figures in the semiconductor industry have become a matter of national security. The potential disruption that could be brought by additional US sanctions is convincing Chinese policymakers and businesses alike to keep under wraps the full extent of efforts in the field.

State funds come to the rescue

A downturn in the semiconductor industry – triggered by weak global demand and US measures aimed at reshaping the tech supply chain – appears to have taken a toll on Chinese GPU frontrunners.

Shanghai-based startup Biren Technology was said to be looking to raise USD 314m at a valuation of USD 2.7bn in March 2022 but no update has been given to the public since then.

The blow to GPU developers has been softened by the intervention of domestic funds. In particular, state-owned institutions playing a crucial role in fundraising efforts to offset risks stemming from US measures and a downturn in the semiconductor industry. Three fundraisings carried out by Iluvatar CoreX, MooreThreads and Metax exceeded the CNY 1bn (USD 143.9m) investment threshold in 2022.

State-owned funds include Financial Street Holding [SHE:000402], which led the Series C+ round in Iluvatar CoreX; and China Mobile Digital New Economy Industry Fund and Hexie Health Insurance, which led the Series B round in MooreThreads. Meanwhile, CMG Media Convergence Industry Investment Fund co-led the latest fundraising in Metax.

Silicon rally?

Mainland Chinese companies face several challenges in building competitive products. The country’s shortcomings in chip manufacturing pose the biggest hurdle as overseas contract chipmakers such as TSMC [TPE:2330] are restricted by Washington in servicing Chinese GPU developers.

Last year, Chinese startup Biren launched a 7nm GPU chip, which the company said it had broken records for computing ability and would have been used in large-scale, cloud-based AI training in data centres.

However, a new round of US sanctions later denied Biren access to TSMC chip manufacturing process, forcing the company to tweak its advanced chip designs to reduce processing speeds to the threshold imposed by the sanctions.

Meanwhile, Chinese tech companies such as Huawei, Baidu [NASDAQ:BIDU; HKG:9888] and Alibaba [NYSE:BABA; HKG:9988] are finding workarounds to train the large language models necessary for AI bots like ChatGPT with fewer or less powerful semiconductors, while US chip giant Nvidia Corp [NASDAQ:NVDA] was also forced to offer products with lower performance.

Although Chinese GPU developers prefer state-backed investors and the largest deals tend to be in renminbi, the sector isn't completely averse to foreign capital. According to AVCJ, some of the best-funded startups – such as Biren, Enflame, Moore Threads, and Iluvatar CoreX – have received capital from US dollar funds managed by the likes of Sequoia Capital China, Primavera Capital, IDG Capital, GGV Capital, and Centurium Capital despite the regulatory issues.

Increasing Chinese interest in non-Western countries such as Brazil and Malaysia could lead to co-investment and cooperation on supply chains in the future.

An extended version of this analysis can be found here.

Analytics by Helen Lee and Manu Rajput

Did you enjoy this article?

Add the following topics to your interests and we'll recommend articles based on these interests.


Your M&A Future. Today.

Next-generation Mergermarket brings together human insights and machine intelligence to deliver groundbreaking predictive analytics.

Be the first to know with next-generation Mergermarket

Book a demo today