Electric Cars

Xiaomi has revealed its inaugural in-house developed chips specifically engineered for artificial intelligence and self-driving functionalities. This marks a significant shift for the automaker, which has historically depended on Nvidia for its hardware solutions. The transition signifies Xiaomi's ambition to gain greater autonomy in the rapidly evolving landscape of automotive technology, allowing for bespoke hardware optimization tailored to its unique sensor configurations, software ecosystems, and driving models. This strategic investment not only positions Xiaomi to potentially mitigate rising AI chip costs but also grants it enhanced control over supply chains and development timelines, aligning with a broader trend among Chinese EV manufacturers to cultivate domestic silicon capabilities.

On a recent Monday, Xiaomi officially unveiled a new family of proprietary silicon, prominently featuring the Xring D100 chip, explicitly designed to facilitate 'intelligent driving'. This announcement underscores Xiaomi's commitment to joining the growing roster of electric vehicle manufacturers that are actively investing in their own AI inference technology, rather than exclusively relying on external market leaders such as Nvidia. This internal development strategy allows Xiaomi to fine-tune its hardware to seamlessly integrate with its distinct array of sensors, proprietary software, and unique driving algorithms. Moreover, it empowers the company with increased autonomy over financial outlays, procurement, and project scheduling, critical factors in maintaining a competitive edge within the automotive industry.

Previously, Xiaomi's electric vehicle models, including the SU7 and YU7, have incorporated Nvidia's hardware, starting with the older Drive Orin system and progressing to the more advanced Drive Thor. This reliance on an American GPU producer placed the Chinese brand in direct competition for chips in a global market characterized by an insatiable demand for AI-powered solutions. The development of in-house chips, therefore, serves as a crucial buffer against the escalating costs associated with AI chip acquisition, providing Xiaomi with a strategic advantage in managing its expenditures. Furthermore, designing its own hardware enables Xiaomi to optimize the chip's performance specifically for its vehicle architecture, ensuring tighter integration and potentially superior efficiency compared to off-the-shelf solutions.

An examination of Xiaomi's newly introduced silicon reveals impressive technical specifications. The D100 chip is equipped with a 20-core CPU and a 16-core Neural Processing Unit (NPU). Xiaomi claims that this chip is the first of its kind to be manufactured in mainland China using a 3-nanometer production process. This advanced manufacturing technique promises enhanced efficiency and greater processing density within a single chip. Additionally, the D100 supports an impressive 160 gigabytes of unified memory, sufficient to run a massive 200-billion-parameter model directly on the device. For context, Nvidia's current Thor hardware typically features a 14-core, 4-nanometer processor with up to 128GB of memory. Tesla's Hardware 4, in comparison, offers 16GB of memory on a 7nm process, with its upcoming Hardware 5 targeting an even more efficient 2nm process and an anticipated 144GB of memory. While these figures highlight Xiaomi's competitive entry into the chip manufacturing space, the true performance in terms of Trillions of Operations Per Second (TOPS), a key metric for AI inference speed, remains to be officially disclosed.

Xiaomi has committed over $3 billion to the development of its new hardware suite, which includes not only the Xring D100 but also the O100 AI acceleration chip and the O3 System-on-Chip (SoC) platform designed for mobile devices. This substantial investment underscores the company's belief that the benefits of achieving technological independence far outweigh the considerable research and development expenses. The O3 platform is slated for its debut in Xiaomi's forthcoming 18 Fold mobile phone, while both the D100 and O100 chips are scheduled for commercial deployment in various applications starting next year. This multi-faceted approach signifies Xiaomi's broad ambition to integrate its proprietary chip technology across its diverse product ecosystem, from personal electronics to advanced automotive systems.

Tesla's much-anticipated "Robotaxi" service is encountering a surge of negative feedback and operational challenges, particularly in Texas. Even dedicated supporters and long-time customers of the Tesla brand are expressing profound disappointment with their experiences. These issues range from excessively long travel times for short distances to inconvenient drop-offs far from designated destinations.

One prominent example involves Reggie Overton, a self-identified "Tesla Ambassador" and artificial intelligence professional, who chronicled a deeply frustrating encounter. Despite his destination being a mere 2.3 miles away, the Robotaxi initially projected a 66-minute travel time. After contacting support and receiving reassurances of a "bug" that temporarily reduced the estimate, the route never altered, and the time quickly reverted to an impractical 57 minutes. Ultimately, Overton and his companions abandoned the ride and opted for an alternative transportation service, criticizing the lack of effective support and the absence of an immediate refund. This incident gained significant traction online, prompting Tesla to issue a refund only after widespread public attention, underscoring a reactive rather than proactive customer service approach.

Further compounding these concerns, "TexasTSLA," a four-time Tesla owner and keen observer of vehicle performance, has meticulously documented a continuous series of failures within the Robotaxi service. Operating within the narrowly defined service areas, this user has reported numerous instances of the Robotaxi dropping passengers miles from their intended location, refusing service due to late hours, or experiencing significant glitches that render the vehicle inoperable. These recurring issues include cars getting stuck in the middle of roads, repeatedly bypassing designated pickup points, and requiring manual remote intervention from support staff to complete journeys. Such widespread operational inconsistencies, even within limited geographical zones, raise serious questions about the fundamental reliability and safety of the "unsupervised" service, especially when compared to competitors like Waymo, which boasts a significantly higher volume of successful driverless rides.

The consistent reports of unsatisfactory Robotaxi experiences from Tesla's most loyal customers serve as a crucial indicator of underlying systemic problems. While Tesla's autonomous driving technology may show signs of improvement in some areas, the critical supporting infrastructure – encompassing efficient routing, dependable pickups and drop-offs, and responsive human assistance – appears to be significantly lacking. For a service positioned to revolutionize urban transportation, these foundational weaknesses prevent it from offering a truly viable and premium experience that justifies its cost. Overcoming these challenges is essential for Tesla to realize its vision of a fully autonomous ride-hailing future and to regain the trust of its advocates and the broader public.