Venture Capitalist: Chinese AI Models Nearly on Par with US
A venture capitalist says Chinese AI models are now very close to US capabilities, noting open-source models help build user trust and market share. He says valuation gaps reflect business performance, not technology, and that China can meet inference chip demand without cutting-edge GPUs.
Kuang Ziping, founding managing partner of Qiming Venture Partners, said Chinese AI model capabilities are now very close to those of the United States. He noted that China's focus on open-source models serves as an effective marketing tool, helping to accumulate user numbers and market trust more easily.
Kuang said the current valuation gap between major US and Chinese AI model companies does not stem from the open-source nature of Chinese models. Instead, he argued, it reflects differences in business operations, revenue and profitability among Chinese AI enterprises.
Turning to the semiconductor sector, Kuang acknowledged that investment is currently very heated, with market attention focused on high-end GPUs. However, he pointed out that AI inference requires substantial computing power that does not need the most complex or advanced chips, and that China is fully capable of meeting that demand. As AI models become smaller and edge capabilities grow more powerful, more computing can be handled offline on devices, making edge AI chips a focus of increasing attention.
Hao Yekang, partner and co-chairman of Silver Lake Capital Management, said AI will bring transformative change and that institutional management must fully commit to deploying it. He cited a chief executive who noted that in research and development departments, where programming efficiency is relatively easy to measure, frontier models can boost the productivity of the top-performing 20 per cent of a team by 100 times.