FB Pixel no scriptNio's William Li on R&D management: Innovation involves facing many failures | KrASIA

Nio’s William Li on R&D management: Innovation involves facing many failures

Written by KrASIA Connection Published on   14 mins read

“As long as you do it seriously, the company will give you the opportunity,” the CEO of Nio said.

As the de facto “CTO” of Nio, William Li (also known as Li Bin) is planning to deliver an outstanding performance through the development of the ET9, the company’s latest flagship sedan.

During last year’s Nio Day event, the Chinese electric vehicle automaker unveiled the ET9, which is priced as high as RMB 800,000 (USD 112,500), commencing its presale. Measuring over 5.3 meters in length, this executive flagship model encapsulates nearly all of Nio’s latest R&D achievements and is expected to be delivered in the first quarter of 2025.

Photo of the ET9, Nio's executive flagship vehicle model.
Nio unveiled the ET9, its executive flagship vehicle model, during the Nio Day 2023 event on December 21, 2023. Photo and header photo courtesy of Nio.

From 2020 to the present, Nio has invested a cumulative RMB 27.4 billion (USD 3.85 billion) in R&D. In the past two years, the company has introduced over ten new car models and featured various upgrades to its software and hardware systems. With the ET9, attention will be paid to the technologies integrated into the model, including a self-developed autonomous driving chip, a 900V high-voltage architecture, a 46105 large cylindrical battery, and an intelligent chassis system. According to Li, the ET9 will feature “17 globally debuted technologies and 52 leading technologies.”

The 900V high-voltage architecture and the intelligent chassis system represent Nio’s capabilities in their respective fields.

In the past two years, the EV industry has seen a surge in charging speeds. While many 800V new cars have been launched this year, Nio has raised the bar further to 925V, achieving a peak charging power of 600 kilowatts per hour.

Nio has been independently developing electric drive systems since its inception, and this time, it has iterated on technologies such as motors and silicon carbide power modules, extending its research to upstream battery cells as it pursues further enhancements.

According to Nio, the self-developed 46105 large cylindrical battery boasts an energy density of 292 watt-hours per kilogram and a resistance of 1.6 milliohms, only half of other large cylindrical batteries. The assembled battery pack will be 120 kWh, supporting faster charge rates of up to 5C.

Zeng Shizhe, vice president of battery systems at Nio, said that the large cylindrical battery has high safety and functional upper limits. However, there remains room for improvement as the industry’s understanding and application of the technology remain relatively early. This also affords Nio the opportunity to establish platform advantages.

Nevertheless, the peak of large cylindrical battery technology is yet to be reached. According to an engineer from Contemporary Amperex Technology (CATL), achieving 5C fast-charging requires not only addressing internal resistance but also proper management of heat dissipation. Large cylindrical batteries typically use radiator coils for heat dissipation—a method ostensibly less efficient than the equivalent for square batteries.

Nio had previously partnered with Svolt to jointly develop large cylindrical batteries. However, the progress of this collaboration remains unclear. Even Tesla, a pioneer in large cylindrical batteries, has seemingly encountered challenges in the industrialization of 5C fast-charging technology.

The newly released intelligent chassis system integrates technologies such as steer-by-wire, rear-wheel steering, and fully active suspension, making the ET9 both large and agile, ensuring smooth rides.

To demonstrate this capability, Nio showcased a video during the press conference: three layers of “champagne towers” were placed atop the ET9. As the vehicle started and drove over consecutive speed bumps, none of the glasses of champagne spilled. Nio named this chassis the SkyRide system, promoting it as the backbone of a “smooth and comfortable experience, similar to riding an aircraft traveling in the stratosphere.”

It is essential to note that technologies like full steer-by-wire and fully active suspension systems have not yet been mass-produced in the industry. Some industry professionals believe that achieving mass production of fully active suspension systems could take at least three more years. Nonetheless, similar to chips and batteries, Nio’s entrance into these domains suggests that efforts to tighten its foothold in the upstream industry chain may now be coming to fruition.

Shenji NX9031, the name of the new chip, will be manufactured using a 5-nanometer automotive manufacturing process and will consist of over 500 billion transistors. Although its computing power has yet to be specifically disclosed, Li said that the goal is to replace four industry-leading chips solely with the Shenji NX9031.

It’s worth noting that the ET9 will be equipped with two Shenji NX9031 chips to achieve computing redundancy.

Safety is a crucial consideration behind the design of the ET9, with backups incorporated in all seven critical systems of the vehicle: drive, brake, steering, perception, computation, communication, and distribution. In extreme cases of main system failure, the backup system can immediately take over.

The ET9 is scheduled for delivery in the first quarter of 2025. In other words, Nio’s announcement of the new vehicle has arrived a year ahead of time. Regardless, having invested over RMB 40 billion (USD 5.62 billion) in R&D, there’s an expectation that success is just around the corner for the company. Li summed up his experience in R&D with Nio, emphasizing that the company has largely centered its progress on one fundamental principle: whether it can create value for its target users.

In the R&D process, Li believes that the team cannot be overly demanding and avoid failures. “Innovation inherently involves facing many failures. As long as you do it seriously, the company will give you the opportunity. No one is infallible. Many of the decisions I make are wrong. How can you expect everyone to be completely right?”

However, he also acknowledged a bottom line in R&D management: to be practical and solve problems. “My requirement is simple: problems cannot be concealed. Concealing problems is something I cannot accept,” Li said.

As Nio’s branches of technological investment continue to spread, Li estimated that the company’s main business in China will break even when monthly sales reach 30,000 units and a 20% gross profit margin is achieved.

The following excerpts are from an exchange between various members of the media with William Li, CEO of Nio, Qin Lihong, president of Nio, Bai Jian, vice president of hardware and head of mobile business at Nio, and Zeng Shizhe, vice president of battery systems at Nio. These excerpts were prepared by 36Kr and have been edited and consolidated for brevity and clarity.

Premium brands must stay true to their DNA

Q: The ET9 won’t be delivered until 2025, and there has been speculation about why the information about this car has been released so early.

William Li (WL): Based on Nio’s previous experiences, the timing of this release is not considered early. The company typically unveils significant updates every Nio Day, and waiting until the next year isn’t practical when the vehicle is set to hit the market in the first quarter of 2025. The development progress of this model is much faster than everyone imagined. This car has been in development for two years and represents a completely redesigned vehicle in terms of its “soul, brain, and body.” Only the design DNA has not undergone a radical alteration.

Have Porsche or Mercedes-Benz ever significantly changed their design DNA? No. What kind of company would? Kia, Hyundai—they are always changing their designs, which differ from generation to generation. Should we be more like Kia and Hyundai? I’m not saying they are not good—they are successful. I’m just pointing out that every company has its own choices.

Regardless, the timeline is around the same. Of course, we could disclose the information only in February next year and start deliveries the following month, but is there truly a need to go to this extent? After all, as an executive flagship, it requires an extended pre-launch period with many aspects yet to be revealed.

Q: People are surprised by the design of the ET9, as it doesn’t fit the conventional appearance of an executive flagship model. How is Nio communicating this to potential customers?

WL: The design is segmented into exterior styling and interior features. From a styling perspective, three objectives were considered:

  • First, when we defined the styling two years ago, we took into account the user profile. If the users are innovation-driven yet relatively conservative consumers, it is unlikely that they would opt for our cars.
  • Second, smart EV technology has introduced breakthroughs in various aspects, exemplified by the ET9’s 200-millimeter ground clearance, a feat achievable only with a fully active suspension. The advantages of a high seating position become apparent with a fully active suspension.
  • Third, it’s crucial to emphasize that Nio’s design DNA revolves around iterative evolution, not revolution.

Revolutionary changes may not resonate with everyone. Times have changed. Unlike the past when a formal suit might have been universally desired, today’s preferences may lean more towards business casual attire. Discovering a new sense of appropriateness and seamlessly integrating it into contemporary business settings poses a challenge. For an executive flagship model, we believe the ET9 strikes the right balance, making it a suitable fit for the current landscape.

Staying ahead of the curve with self-developed technology

Q: What’s the computing power of the Shenji NX9031, and how does it compare to Nvidia’s chips?

Bai Jian (BJ): In terms of computing power, there is ample support for Nio’s intelligent driving capabilities. While having sufficient computing power addresses the basic requirements, we are now delving into more nuanced considerations.

Having come from the mobile industry, where power consumption is crucial due to its impact on battery life, I comprehend its significance. Historically, the automotive sector didn’t prioritize power consumption. For instance, the autonomous driving system of the NT1 model could run 24 hours a day, and I wouldn’t notice it because the power consumption was remarkably low. However, the landscape has evolved. With the increasing computing power of autonomous driving systems, there is a demand for more efficient power consumption.

Hence, we have implemented power gating in the neural system, with each “block” being meticulously designed to function independently, allowing it to be activated or put to sleep instantly. As a system ensuring high security and reliability, a vehicle must achieve this — there can be no compromise on safety at any point in the user experience.

Furthermore, we are delving into aspects like latency, ISP (image signal processor), and reliability. As intelligent driving perception technology advances, latency becomes increasingly crucial. Currently, a camera captures 30 frames per second, but this rate is expected to rise in the future. If data throughput and transmission are slow, it will extend the algorithm processing time. Even with substantial computing power, efficient utilization is impossible without adequate investment in this area.

Additionally, there is the challenge of achieving safety redundancy in the coming years, as mentioned yesterday. My goal is to achieve a failure in time (FIT) value in the hundreds (per billion). However, even with a lower value, we cannot guarantee error prevention. In the event of an error, can my backup system seamlessly take over within milliseconds? This requires collaboration not only within the chip system but also across the entire spectrum of operating systems and self-driving algorithms. A cohesive effort is necessary to achieve success.

Q: What are your thoughts on the trend of cockpit integration? Will you develop a cockpit chip?

WL: The challenge lies not in a technical dilemma but in defining the situation. The more integrated the system, the more it impacts the efficiency of handling core tasks, making cockpit chips a lower priority at present. However, our Central Computing Cluster (CCC), the car’s supercomputing platform, is highly adaptable. Whether the chips are integrated or separate is a matter of definition.

BJ: Integrated cockpits involve using a single chip to manage both self-driving and cockpit functions, assuming they are two distinct lines without software integration. However, in the architecture of the ET9, the CCC cockpit and driving functions are inherently integrated, allowing mutual access to their computing power.

Achieving cross-domain calling in this industry is challenging, involving not only technical obstacles but also organizational integration issues. Typically, autonomous driving and cockpit development are managed by separate B2B teams. However, within Nio’s R&D system, under the guidance of SkyOS, they are inherently integrated. Whether to opt for one chip or two chips is a decision that won’t significantly impact the user experience.

Q: The ET9 is equipped with three LiDAR sensors and two self-developed chips. Will this be the standard hardware configuration for the next generation of Nio’s autonomous driving platform?

WL: Many individuals have suggested that Nio should adopt a single motor in its vehicles and eliminate LiDAR sensors. From the perspective of the Nio brand, it is essential for everyone to acknowledge the value that computing power and sensor technology bring to users.

Attempting to set things in opposition, such as comparing pure computing vision solutions to LiDAR—as if the former is unquestionably superior to the latter—may not be the most accurate approach. Pure computing vision solutions are foundational, but it does not render LiDAR obsolete. The decision to use LiDAR or not involves weighing tradeoffs rather than being a purely technical issue.

Instead of making direct comparisons, it is more productive to understand the advantages of each technology to minimize the occurrence of negative outcomes. With the launch of the ET7, we achieved a certain level of redundancy in terms of computing vision. However, the question is, when does it come into play? Its role is limited to very few situations. Nonetheless, technological advancements address issues in these rare situations. I often use the analogy of airbags. You cannot assert that in 99.999% of cases, braking and seat belts will suffice, so airbags are optional.

Q: Nio used square cells before—why have large cylindrical cells been utilized for this generation?

Zeng Shizhe (ZS): For years, our research has focused on determining the most suitable battery cell shape for the next-generation platform, with several crucial considerations. The primary focus is on safety, and in addition, we aimed for battery cells not only to supply energy but also to support additional functionalities that enhance the overall vehicle experience. Subsequently, we recognized certain advantages associated with large cylindrical cells. In a more extensive single product, large cylindrical batteries can address challenges that have remained unresolved by others.

The delayed recognition of the strength of large cylindrical cells can be attributed to the absence of automotive manufacturers supporting the scale of large cylindrical battery production lines until now. Additionally, diverse requirements, whether 3C, 4C, or 5C, among others, further contribute to the challenge. Through manufacturing large cylindrical cells in-house, we seize the opportunity to broaden the platform and optimize the benefits it can deliver.

The difficulty in achieving 5C charging with large cylindrical cells revolves around its structural aspects. Our comprehensive testing involved practically every large cylindrical battery available in the market, along with samples provided by various factories. From my perspective, the application and understanding of large cylindrical cells across the industry are still in the early stages.

Q: There are claims within the industry that large cylindrical batteries are difficult to achieve 5C charging. What technology has Nio adopted to achieve 5C charging capability?

ZS: Nio’s self-developed 46105 large cylindrical battery boasts an internal resistance of 1.6 milliohms, whereas other large cylindrical battery cells commonly found in the industry typically hover around 3 milliohms, effectively halving the resistance.

Significant breakthroughs can still be achieved within the battery realm itself, exemplified by our earlier efforts with the 150 kWh semi-solid-state battery. Despite skepticism from many, we successfully brought this concept to fruition.

The tendency for many to rely on what suppliers offer without delving into the details can result in overlooking substantial areas for improvement. By prioritizing attention to detail, we have identified ample room for enhancements, spanning material selection, welding methods, component connection methods, and more. These refinements enable us to establish platform advantages concerning energy density and beyond.

Becoming the top ten automakers globally

Q: As “CTO,” what technological changes have you implemented at Nio?

WL: Technology comprises numerous layers, and my specialization lies in large-scale architecture. During my college years, I successfully cleared the system analyst examination. I believe I am the sole humanities student who achieved this feat, and I was also the first Microsoft-certified system engineer in China. My system analyst thesis delved into rapid prototyping, focusing on the expedited construction of software prototypes to enhance the iterative process. This approach is deeply ingrained in me—I am passionate about the architectural logic of technology and the interconnections between components.

For instance, I scrutinize the value of a mobile phone from a panoramic perspective, exploring its contributions to and synergy with the in-car experience, the underlying logic, and the overarching architecture. I approach technology architecture from a higher level, spanning domains and encompassing the full stack. However, when it comes to specific code, it has been a while since I wrote any, and it is not advisable for me to spend my time on coding.

Examining the recent organizational changes over the past few years, you can observe that our vision has broadened compared to before. Initially, we segmented the company into three major blocks: intelligence, electric, and automotive. Subsequently, I sensed that this division was not optimal, resembling three imposing mountains with too many connection points between them. Consequently, a few years ago, we implemented a substantial organizational change, breaking down these areas into several components. This restructuring eliminates the need for departments to navigate significant hurdles in communication—it’s more akin to overcoming small hills now. This transformation has significantly enhanced our collaboration and cooperation.

Q: How does Nio evaluate the direction of its R&D? Is there room for failure?

WL: Firstly, can you generate value for your target users? That is foundational. For instance, in the automotive sector, we believe that instilling a sense of security is a crucial value for the executive flagship model.

Secondly, the focus on systematic efficiency should not be confined to local, short-term costs and efficiency but should encompass global, full-life costs and efficiency. When we invest in R&D, there must be a sound rationale behind it. Simply engaging in activities for the sake of doing them is devoid of meaning.

Thirdly, from the standpoint of short-term execution, we also assess the equilibrium of various facets of the company’s operations. While R&D falls under the purview of the CTO’s decision logic, manufacturing belongs to the supply chain. What R&D is firmly committed to achieving may not necessarily require in-house implementation by the manufacturing team. Chips and battery cells constitute substantial assets and operations, and manufacturing them in-house is not an absolute necessity.

In the industry, each company needs to chart its own course of evolution. The ultimate competition will be all-encompassing. Over the past two years, we have been establishing battery swapping facilities, and now other automakers are also doing the same. We have been engaged in full-stack self-development, and other automakers may be pursuing similar strategies.

Different companies possess distinct DNA. We have steadfastly adhered to our approach since day one, and of course, this might involve some internal risks. Nio certainly allows for a margin of error. I convey to individuals within the company that innovation inherently involves facing many failures. If you approach your work diligently, the company will afford you opportunities. No one is infallible. No one is infallible. Many of the decisions I make are wrong. How can you expect everyone to be completely right?

It’s unrealistic to think that immediate success accompanies every endeavor—that would be overly simplistic. Nio has faced numerous challenges over the past few years. The crux lies in whether we’ve gleaned insights and if the organization has evolved. You have to be pragmatic, solve problems when you have them. But you can’t conceal problems. My requirement is simple: problems cannot be concealed. Concealing problems is something I cannot accept.

Q: Xiang Li (CEO of Li Auto) once said that only four automakers will remain in the future: Tesla, BYD, Huawei, and Xpeng Motors, without mentioning Nio.

WL: Nio rarely discusses who is or isn’t on this list. The automotive industry has never been characterized by a single winner taking all. Take, for instance, two companies currently performing well: BYD and the underestimated Chery. In 2019, both companies encountered significant challenges. BYD was once barely selling 40,000–50,000 units, and just a few years ago, no one would have imagined it reaching the milestone of selling 3 million units in a year, a feat achieved only in the past two or three years. Chery has been gaining strength over the last two years, although it has been engaged in globalization for over a decade.

Competition in the automotive industry will be protracted. We mention that, starting in 2025, we will enter the finals, but I have never specified when the finals will conclude because they might not have an endpoint. This concept extends beyond the automotive industry; there is no end to doing business.

I believe that by 2035, Chinese brands will represent five of the top ten global automakers. Different individuals will have diverse opinions on which five these will be, but Nio aspires to be one of them.

KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Peng Suping for 36Kr.


Auto loading next article...