China's New Energy Vehicle Independent R&D Has Finally Come to the Moment to "Show Its Hand"

Written by | AUTO Xinqiu

Author | AUTO Xinqiu Team

Just now, the new Onvo L60 was launched.

There are two things worth looking at together around this launch event. Counting back one day, NIO participated in the formulation of Singapore's battery swapping standards. On the day of the launch, an autonomous driving video of Onvo L60 crossing an extremely narrow karst cave was released to the outside world, causing quite a discussion on overseas social media.

But what truly made the industry notice this was the price. The Shenji NX9031 self-developed chip, full-domain 900V high-voltage platform, SkyOS Tianxu operating system — these three things appeared simultaneously on a car with a starting price of just over 200,000 (CNY). And these three technologies were positioned as flagship-level on previous NIO brand models.

This cannot be explained by a product iteration alone.

In the new energy vehicle industry these past two years, everyone is shouting "independent R&D". Battery, motor, and control system self-research, chip self-research, system self-research, intelligent driving self-research — this term has become somewhat inflated. But even with the same "self-research", some is a light asset approach of "defining specifications, outsourcing design", while others is a heavy asset route of "doing the entire process from architecture to tape-out themselves". The difference between the two is not visible in the short term, but vastly different in the long term.

The emergence of the new Onvo L60 basically brings this problem to the table: after the independent R&D system is truly proven, what exactly does it bring?

The True Divide of Independent R&D, Starting with a Chip

To figure out what makes NIO's independent R&D different from others, the best entry point is the intelligent driving chip.

Shenji NX9031, the team behind this project has over 600 people. Li Bin specifically emphasized one sentence at the media communication meeting: "It is a complete design capability from front-end to back-end." This is not said off the cuff.

There are two paths for making intelligent driving chips in the industry: one is to define parameters well, find a design company to do it for you, saving money and fast, but the disadvantage is subsequent iterations are constrained by others. The other is to build a team yourself, pressing the entire process from architecture design to back-end tape-out in your own hands, burning money, slow, but chips, algorithms, sensors, and operating systems can be coupled end-to-end.

This chip is not that chip. Just like building cars, outsourcing to an engineering design company and doing full-stack self-research yourself both result in something called "car", but the foundation is completely different.

Li Bin mentioned three specific comparison dimensions at the communication meeting.

The first is memory bandwidth, the Shenji NX9031's memory bandwidth is "basically 2 times that of industry peers" — this indicator determines how large a model can run on the car and how the inference speed is.

The second is multi-modal signal processing capability, also known as ISP. Li Bin mentioned that the good performance in crossing karst caves has something to do with this, because image processing capability is key in extremely dark environments.

The third is dual-chip millisecond-level hot switching. Li Bin said this is very important for L3, L4 level intelligent driving.

These three indicators alone are technical parameters, but together they point to the same thing: only when chips, algorithms, and operating systems are all your own does the initiative of product definition truly lie in your hands.Otherwise, you are just making choices on a supplier's quote sheet.

But the dividends of chip independent R&D extend beyond single-point indicators. Because the chip is your own, and sensor layout was planned uniformly in advance, like NIO decided on a hardware baseline of "7 8-megapixel cameras + 7 surround cameras" starting from the ET7 era, continuing all the way to Onvo L60, so the hardware architecture is naturally reusable.

Hardware reusability means data reusability; data reusability means toolchain reusability. Here, Li Bin revealed a detail: all intelligent driving versions of NIO's second-generation platform models can achieve "simultaneous release within a time window of no more than a few weeks plus or minus". Behind this rhythm, after the underlying architecture is unified, one set of code can cover multiple car models.

Assume there are 5 cars under the brand, if the intelligent driving software for each needs to be developed, tested, and iterated separately, maintenance costs are multiplicative. After the underlying architecture is unified, adding one more car to share R&D costs, the marginal cost is so low it can be ignored.

The new Onvo L60 can use flagship intelligent driving, not because NIO is generous, but because the cost structure changed itself after hardware unification.

The Moat of Independent R&D Has Already Been Dug Open

If the chip story explains what independent R&D can bring, then lightweighting and battery swapping explain that some things cannot be done at all without independent R&D.

Let's talk about lightweighting first.

Li Bin spent a large portion of the launch event speaking about lightweighting, which is uncommon in the industry. Lightweighting is a "invisible" indicator. Users do not pay for a low body-in-white lightweighting coefficient, and car companies usually do not take it as a selling point.

But Li Bin broke it down very finely.

The new Onvo L60 achieved a body-in-white lightweighting coefficient of 2.22. So, how was this done? There are four keys, mostly requiring independent R&D support.

The first is the battery. The battery is the heaviest component of the whole vehicle. If you focus on stacking range, stacking large-capacity LFP batteries, the weight goes up. Onvo's strategy is restrained: 85 kWh uses ternary lithium batteries, which is over 100 kg lighter than LFP of the same capacity. But ternary lithium is expensive. Li Bin's original words were: "This is a large sum of money."

The second is systemic reconstruction. Onvo L60's "smart fuse", which sounds like just changing a part, actually requires the entire vehicle's high-voltage and low-voltage architecture to be redesigned. Using traditional fuses, circuit design is standard parts; switching to smart fuses, everything from distribution logic to harness routing must be redone. Integrated design is the same, packing more functions into fewer modules, saving not just weight but also space and BOM costs. But each is an independent R&D project; without independent chassis and independent electronic/electrical architecture teams, you can't do it alone.

The third is full-domain 900V. Many people haven't noticed the contribution of high-voltage platforms to lightweighting — the higher the voltage, the smaller the current under the same power; the smaller the current, the thinner the harness; the thinner the harness, the lighter the whole vehicle. This loop sounds simple, but the threshold of 900V itself is extremely high. From power semiconductors to insulation design to thermal management, all are tough challenges.

Finally is engineering capability. A lightweighting coefficient of 2.22 tests the ability to "achieve the best structure with the least material while ensuring crash safety". Li Bin made an analogy: "Sometimes overusing materials yields no safety benefit, just like some houses look very thick walls, it does not mean they are structurally solid." This capability cannot rely on suppliers; it relies on the data, simulation, testing, and trial-and-error accumulated by the whole vehicle engineering team over the years.

After going through these four things, the lightweighting matter is clear: it is not a single-point technical indicator, but the comprehensive result of several independent R&D systems working collaboratively: chassis, core electric systems, electronic/electrical architecture, and engineering simulation.

Let's talk about battery swapping next.

On the same day the new Onvo L60 launched, NIO participated in the formulation of Singapore's battery swapping standards. Adding to this, over 60 battery swapping stations already operational in Europe, this model which was repeatedly questioned in the past is turning from a corporate choice into an industry option.

But the battery swapping matter is independent R&D from start to finish. Currently, among Onvo L60 users, the penetration rate of the battery leasing scheme is over 90%.

Obviously, once users accept "separation of vehicle and battery", the risks of battery life, anxiety about resale value, and concerns about charging convenience are all transferred to the operator, which is NIO itself.

And NIO dares to take this responsibility, there is only one reason: the battery swapping network is part of the entire independent R&D system. Battery charging/discharging strategies, life management algorithms, residual value assessment models are all run by their own team. If you outsource battery swapping to third-party operators, you simply dare not promise users "battery rental covers lifetime warranty" because you cannot control the real health status of the battery.

Chip independent R&D lets you modify, lightweighting lets you save, battery swapping lets you guarantee — these three things put together form a complete independent R&D puzzle. And the new Onvo L60 is the first time these three things are fulfilled simultaneously on a car in the 200,000 level.

The Most Difficult Path, Highest Barrier Instead

Speaking of here, there is an unavoidable question: Independent R&D is so money-burning, who is paying the bill?

In 2026, the supply chain pressure is quite real. Automotive-grade memory prices are rising sharply, and lithium prices are also rebounding. For one Onvo L60, just the cost end increased by over 10,000 yuan. If converted to the final selling price including taxes, the impact is about 15,000.

Here, Li Bin spoke a big truth: "Onvo L60 still has gross profit, but it's quite miserable."

But under this pressure, the Onvo L60 starting price was still placed just over 200,000. However, 90% of orders selected the high-spec version with Shenji chips and LiDAR. Average selling price for the whole brand in the first quarter was 240,000 — this is higher than many traditional luxury brands. From January to May, NIO overall growth was 68%, higher than internal targets.

Looking at these two sets of numbers together, the conclusion might be somewhat counter-intuitive — Independent R&D burns money, but Independent R&D is exactly the way to digest costs.

The logic here is not complex. The same technical system covers three brands: NIO, Onvo, Firefly; R&D expenses are spread across more cars; the same hardware architecture serves multiple models, and after procurement volume goes up, negotiation power also increases. After hardware unification, data reuse, and toolchain sharing are proven, for every additional car, marginal costs will drop significantly.

The new Onvo L60 presses flagship technology down to the 200,000 level, not because NIO is doing charity, but because this economic model is operating on its own.

But if you stretch the timeline, the real account of independent R&D is not calculated by quarter.

If you look at a quarter — others are competing in sales volume, he is repairing battery swapping stations. Others buy chips to compete in computing power, he builds a team from scratch for tape-out. Others stack big batteries to compete in range, he calculates lightweighting coefficients. Every move seen individually looks like engaging in an unprofessional business.

But looking at 3 to 5 years, you will find 3,900 battery swapping stations have been laid out, chips have run through end-to-end coupling, lightweighting coefficients reached 2.22 — these single points begin to link into a system.

Looking at ten years, you will find this company built a complete set of infrastructure on core electric systems, intelligent driving, energy replenishment, and electronic/electrical architecture that others cannot bypass. This is the core barrier brought by independent R&D, and also the key reason why consumers are increasingly paying for NIO.

Written at the End

Back to the new Onvo L60.

The significance of this car is not that it is 50,000-60,000 cheaper than Model Y, nor is it about using flagship chips and LiDAR, let alone how much it sold for.

Its significance lies in a 200,000 level car paired with a technical combination that only a full-stack independent R&D system can support. China's new energy vehicle independent R&D has entered the stage of "can it work well, is it worth the account" from the stage of "can it be made".

In the table of China's new energy vehicles, players willing to build infrastructure for ten years, trade time for space, trade deep plowing for barriers can be counted on one hand. The emergence of the new Onvo L60 is simply saying one thing: the investments that everyone thought were quite silly back then are now beginning to be redeemed line by line on reports, products, and industry standards.

The marathon is not yet finished. But those running ahead are always the ones who dug the deepest foundation before the starting line.

* All images in this article are from the internet

Focusing on intelligent cars, assisting key decisions.