Xpeng, a key player in the burgeoning electric vehicle market, is making significant strides towards self-sufficiency in autonomous driving technology. According to reports from the China-based tech news outlet 36kr, Xpeng’s proprietary artificial intelligence chip, named Turing, is slated to enter mass production by the second quarter of 2025. Initially, this advanced chip will be integrated into a new Xpeng model, underscoring the company’s strategic intent to stand out in the competitive autonomous driving arena.
An exceptionally potent chip, the Turing boasts two in-house developed neural network processing units. This architecture achieves 20% greater efficiency compared to conventional automotive chips, and is equipped to handle vast AI models with up to 30 billion parameters. For context, Li Auto’s current VLA (Visual Language Action) model processes around 2.2 billion parameters, highlighting Turing’s ambitious technical prowess. Despite potential latency challenges with such expansive models, the Turing chip’s processing capacity reaches an impressive 700 TOPS (trillion operations per second), rivalling Nvidia’s Drive Thor platform. While Nvidia’s Thor chip, unveiled in 2022, targets 2,000 TOPS, its current market offering achieves 750 TOPS.
Extending beyond vehicles, Xpeng’s CEO, He Xiaopeng, revealed plans for the Turing chip’s application in other innovative mobility solutions, such as robots and flying vehicles. The chip is engineered to enhance both smart cockpit systems and autonomous driving software.
In an era where Chinese technology and automotive companies prioritize vertical integration, the focus is on reducing external reliance and fostering in-house capabilities. For instance, Nio has begun mass production of its Shenji NX9031 chip, utilizing a 5 nm manufacturing process, in its latest model, the ET9. Li Auto is also nearing production with its proprietary chip. Xpeng’s Turing chip entry marks a transformative shift in the electric vehicle sector. As dependence on tech giants like Nvidia wanes, manufacturers are cultivating internal hardware ecosystems, accelerating innovation and expediting model updates.
.jpg)
Yorum Yap