At the 2026 China Automotive Chongqing Forum held from June 12th to 13th, Deng Kun, Vice President of Application Engineering at BLACK SESAME Intelligence, delivered a speech.
Deng Kun noted that from a technical perspective, achieving a balance between high performance, low cost, and low power consumption remains the perpetual goal for chip companies and their automotive OEM clients. However, he cautioned that an overemphasis on cost reduction alone could hinder keeping pace with the development of large models and AI, potentially rendering chips mere prototypes unsuitable for mass production.
He argued that to strike this balance, companies should avoid pursuing high performance solely by stacking computational power, nor should they relentlessly cut costs by downgrading specifications. "A more viable path lies in innovative architectures, particularly focusing efforts on chip design technology and the development of proprietary chip IP," Deng Kun stated.
Furthermore, he highlighted that rising prices for memory chips pose a challenge to supply chain security. He proposed that future solutions should increasingly leverage near-memory computing architectures and in-memory computing technologies. These approaches would integrate storage overhead within the chip itself, thereby reducing reliance on foreign DDR chips and enhancing overall performance.
Key Areas for Chip Companies
Deng Kun emphasized that chip companies must place greater emphasis on automotive compliance, functional safety, and information security. Ensuring chips meet OEM requirements with guaranteed specifications in these areas is crucial. Simultaneously, production capacity must possess flexible and interactive capabilities, utilizing dynamic allocation across multiple domestic and international sites to satisfy OEM demands. Additionally, given the critical nature of automotive products, it is essential to ensure long-term capabilities for continuous interaction and iterative development.
Comments