
9月16日,雷军受中国质量协会邀请参加活动,作为产业界代表分享了小米做产品的质量管理思路。当天晚上,他还专门发了一条微博,表示AI技术确实带来了质量的新纪元,但数字化模拟终究替代不了真实测试。
小米这些年在研发和质量上的投入相当扎实。雷军透露,过去5年累计研发投入1050亿元,未来5年计划还要超过2000亿元。而且小米并非只靠砸钱,质量体系搭建得很早:2017年就成立了质量委员会,直接定下「质量优先、一票否决」的规矩,把质量和全公司绩效绑定;从2023年开始引入卓越绩效模式,覆盖全部业务流程。到现在,小米有802间自建实验室,以及与高校、科研机构合作的35间联合实验室。
落到造车上更有说头。五年前小米宣布造车时,网上质疑声不少,说互联网公司造车无非是PPT造势、靠AI仿真跑数据。这次雷军直接给出硬数据:这几年造车,小米前前后后投入了将近3000辆测试车。重点是,不是上市前跑完测试就完事,车子上市之后仍在持续做全生命周期的泛化测试。
销量也能侧面印证这份扎实:小米汽车上市两年多,已经交付80万辆,这个交付速度在国内新势力里属于第一梯队。一边冲交付、一边不松测试环节,这个节奏并不常见。
不久前,小米推出「龙甲电池安心保障计划」,做出两条重要承诺:一是若原厂电池质量问题导致车辆达到报废标准,直接赔同等配置的新车;二是碰撞事故、底盘剐蹭导致电池起火并达到报废条件,还会补偿车损险保额与保险赔偿金额之间的差额。终于有车企敢接电池的「锅」了——现在大家买电动车最焦虑的就是电池安全,尤其怕托底、怕磕碰留下隐患。小米敢给出这种赔付承诺,本质还是对自己的电池测试和品控有信心。
雷军特意强调真实环境测试,针对性很强。最近两年,AI仿真、数字孪生几乎成了造车圈的流量密码,不少车企发布会上都在讲用AI把研发周期缩短了多少、减少了多少辆实车测试,听起来又高效又高科技。但行业里的人都懂,模拟场景永远是有限的:现实里的极端暴雨冰雪、复杂城乡路况、车主千奇百怪的驾驶习惯,根本不可能靠算法全部推演出来。
在雷军的定位里,AI是用来提升质量管理效率、优化测试流程的,不是用来「省测试」的。有网友评论称,现在仍然能在路上看到小米各款车型的测试车,这才是对用户负责的态度。
AI肯定是好东西,能帮车企少走弯路、大幅提升研发效率,但前提是它是辅助而不是替代。汽车这种和人身安全强绑定的产品,再怎么强调真实测试都不过分。小米作为跨界进来的造车新势力,没有老牌车企几十年的经验沉淀,靠的就是用笨办法、砸硬投入,把每一个场景都跑遍、测透。从这两年的市场反馈来看,这条路至少走对了。
本文来源:https://www.toutiao.com/article/7686537969242538536/
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
On September 16, Lei Jun, invited by the China Association for Quality, attended an event and shared Xiaomi's quality management approach to product development as an industry representative. That evening, he also posted a dedicated Weibo message, saying AI technology has indeed ushered in a new era of quality, but digital simulation can never replace real-world testing.
Xiaomi's investment in R&D and quality over the years has been quite solid. Lei Jun revealed that cumulative R&D investment over the past five years reached 105 billion yuan, with plans to exceed 200 billion yuan in the next five years. And Xiaomi hasn't relied solely on spending money—its quality system was established early on: a Quality Committee was set up in 2017, which directly set the rule of "Quality First, Veto Power," tying quality to company-wide performance; starting in 2023, an Excellence Performance Model was introduced, covering all business processes. To date, Xiaomi has 802 self-built laboratories, along with 35 joint laboratories established in cooperation with universities and research institutions.
When it comes to car manufacturing, there's even more to discuss. Five years ago, when Xiaomi announced its entry into carmaking, there was considerable skepticism online, with claims that an internet company building cars was nothing more than PPT hype and running data through AI simulation. This time, Lei Jun directly provided hard numbers: over these years of carmaking, Xiaomi has invested nearly 3,000 test vehicles. The key point is that testing didn't end once the cars launched—after vehicles hit the market, the company has continued to conduct full-lifecycle generalized testing.
Sales figures also indirectly confirm this solid approach: in just over two years since Xiaomi's cars went on sale, it has delivered 800,000 vehicles, placing it in the top tier among domestic new forces in terms of delivery speed. Balancing delivery momentum while not easing up on testing is not a common pace.
Not long ago, Xiaomi launched the "Dragon Armor Battery Safety Assurance Plan," making two major commitments: first, if a factory-original battery quality issue causes the vehicle to meet scrappage standards, the company will directly compensate with a new car of equivalent configuration; second, if a collision accident or chassis scrape causes the battery to catch fire and meet scrappage conditions, the company will also compensate the difference between the vehicle damage insurance coverage amount and the insurance payout. Finally, a carmaker dares to take responsibility for the battery—these days, the biggest anxiety for EV buyers is battery safety, especially the fear of underside impacts and hidden damage from scrapes. Xiaomi's willingness to make such payout commitments essentially comes down to confidence in its battery testing and quality control.
Lei Jun's emphasis on real-environment testing is highly targeted. In the past two years, AI simulation and digital twins have almost become the traffic password of the carmaking world. Many carmakers at their launches talk about how AI has shortened R&D cycles and reduced the number of physical test vehicles, sounding both efficient and high-tech. But people in the industry all understand that simulated scenarios are always limited: the extreme rainstorms, ice and snow of the real world, complex urban and rural road conditions, and the wildly varied driving habits of owners simply cannot be fully extrapolated by algorithms.
In Lei Jun's positioning, AI is meant to improve quality management efficiency and optimize testing processes, not to "save on testing." Some netizens commented that Xiaomi test vehicles of various models can still be seen on the roads today—that is the responsible attitude toward users.
AI is certainly a good thing—it can help carmakers avoid detours and dramatically improve R&D efficiency—but the premise is that it serves as an aid rather than a replacement. For a product as tightly bound to personal safety as an automobile, no amount of emphasis on real-world testing is excessive. As a newcomer to carmaking from another industry, Xiaomi lacks the decades of experience accumulated by established automakers, so it relies on using painstaking methods and heavy investment to cover and thoroughly test every scenario. Judging from market feedback over the past two years, this path has at least proven to be the right one.