(Yicai) Sept. 1 -- Cambricon Technologies’ shares slumped despite the chairman of the Chinese semiconductor giant reportedly saying its research and development team can still innovate and develop new products despite being downsized.

After tumbling by as much as 8.7 percent in Shanghai trading earlier today, Cambricon [SHA: 688256] ended 6.3 percent lower at CNY148.83 (USD20.52) a share.

Cambricon’s R&D team has a reasonable structure and sufficient skills to support the firm’s technological innovation and product R&D, Chen Tianshi said on its first-half earnings conference call yesterday, according to The Paper.

The Beijing-based company slashed its R&D headcount by 19 percent to 980 in the six months ended June 30 from a year ago, while R&D spending plunged 23 percent. But outlays on R&D were still more than five times revenue at CNY482 million (USD66.5 million).

Cambricon took into full consideration industry and market trends before making these adjustments, said Chen, who is also general manager. It is the first time Cambricon has spoken about the matter.

The company’s first-half loss shrank 12.3 percent to CNY545 million (USD75 million), according to its latest earnings report released on Aug. 30. Revenue plunged 33 percent to CNY114 million.

The narrower loss was due to a decrease in share-based payments from the same period last year, the chipmaker said. The firm has also improved its R&D efficiency, optimized resource allocation, shrunk employee compensation and trimmed laboratory testing and processing fees. The slump in revenue had to do with problems in the supply chain, it added.

Since the beginning of the year, many large language models have appeared, leading to higher demand for computing power. Cambricon has optimized the performance of its artificial intelligence-generated content and LLMs through cloud products, and has fostered technology and product tie-ups with many industry clients and software vendors, Chen said.

Cambricon’s sixth-generation intelligent processor microarchitecture and instruction set are under development, and will focus on optimizing scenarios such as natural language-processing LLMs to enhance product competitiveness, he added.

Editor: Kim Taylor