In Conversation with William Li: NIO's Average Selling Price Surpasses BBA, AI Deals a Triple Blow to the Auto Industry

A Counterintuitive Cost Phenomenon: When "Economies of Scale" Start to Fail

For the past century, the auto industry has believed in one iron law: as volume goes up, costs come down. NIO founder William Li recently proposed a judgment that contradicts this — now that volume is up, costs may actually go up instead.

He calls this phenomenon "diseconomies of scale." The root cause is that the AI industry is currently in a phase of investing regardless of cost, and AIDCs (AI data centers) have enormous demand for memory, PCBs, copper, aluminum, and other resources. Once an automaker's procurement volume rises, it directly enters a resource competition with the AIDC industry.

"If you ask for less, they can still give you a discount; if you ask for more, sorry, the price goes up instead." Li told Leiphone. "This is quite interesting."

This is not a metaphor. NIO's per-unit cost in the second quarter rose by 14,000 yuan compared to the end of last year. Calculated at that quarter's sales volume, that's roughly 1.5 billion yuan in extra costs out of thin air in a single quarter. Li expects costs to continue rising by another two to three thousand yuan in the second half of the year, and there is "absolutely no room for negotiation."

A Counterintuitive Earnings Report: Revenue Growth Outpaces Sales, Gross Profit Growth Outpaces Revenue

Against such severe cost pressures, NIO delivered a counterintuitive half-year report.

In the first half of the year, cumulative deliveries reached 191,100 units, up 67.4% year-over-year; revenue was 57.67 billion yuan, up 85.8%; total gross profit was 10.77 billion yuan, up 282.2%; and the overall gross margin reached 18.7%, up about 10 percentage points year-over-year. Revenue growth outpaced sales, and gross profit growth outpaced revenue. Second-quarter operating profit was 207 million yuan, marking the third consecutive quarter of profitability.

On a horizontal comparison, according to CPCA data, passenger vehicle retail sales fell 20.2% year-over-year in the first half, and the profit margin of vehicle manufacturing was only about 1.5%, nearly halved year-over-year — converted at an average selling price of 202,000 yuan per vehicle, net profit per vehicle was less than 3,000 yuan.

There are doubts, however: NIO's three consecutive quarters of positive operating profit use a Non-GAAP basis, excluding non-cash items such as equity-based compensation, so the "hardness" of the profit remains open to discussion. Li's response is that operating quality should be judged by operating profit, not by GAAP net profit, which is distorted by non-operating factors such as interest and investment income.

The difference between Non-GAAP and GAAP: GAAP (Generally Accepted Accounting Principles) requires all expenses related to current-period operations to be included in profit, including non-cash items such as equity-based compensation and asset impairments; Non-GAAP allows these items to be excluded to reflect "operating" profit after removing one-time or non-cash effects. The disagreement between the two lies here: although equity-based compensation is a book expense, it dilutes shareholders' equity and remains a real cost over the long term. Therefore, using Non-GAAP to measure profitability is very common among tech and EV startup companies, but different companies exclude different items, so horizontal comparisons must be made cautiously. When judging earnings quality, it is usually also necessary to look at operating cash flow, per-vehicle gross profit, and expense ratios to avoid being misled by differences in accounting basis.

Average Selling Price of 406,000 Yuan Surpasses BBA: How Brand Premium Is Forged

What most directly reflects operating quality is the transaction price. In the second quarter of 2026, NIO's average transaction price reached 406,000 yuan, rising further to 434,600 yuan in July, comprehensively surpassing BBA — about 180,000 yuan more than Audi, about 70,000 to 80,000 yuan more than BMW, and tens of thousands of yuan more than Mercedes-Benz.

NIO co-founder and President Qin Lihong added one data point: the brand factor now accounts for more than 30% of NIO users' car-purchasing decisions, ranking second, and Li expects it to rise to first next year.

What supports this brand premium is NIO's sustained investment over the past few years in full-stack technology, its battery-swapping network, and its service system. Li believes that in the market above 300,000 to 400,000 yuan, there are not many companies that can truly compete with NIO at the level of a systematic brand.

AI's Triple "Blow" to the Auto Industry

Li summarizes AI's impact on the auto industry as a triple squeeze:

The first is resource competition. Resources such as memory, PCBs, copper, aluminum, and even batteries are being heavily absorbed by AIDCs. Of the 16,000 to 17,000 yuan increase in per-vehicle cost, about 10,000 yuan comes from memory price increases, 3,000 to 4,000 yuan from batteries, and the rest from materials such as copper and aluminum.

The second is a talent siphon. New fields such as embodied intelligence are attracting large numbers of people from the auto industry.

The third is capital diversion. The capital market's attention has been captured by the AI field, and valuations in the auto sector are under pressure.

But he also points out a risk that is easy to overlook: the advantage of latecomers. If the computing power built today for 1 trillion US dollars can be completed three years later for only 300 billion US dollars, then the extra 700 billion US dollars invested earlier may not create corresponding value. He expects cost pressure in the auto industry to continue for about two more years, and by 2028 there may be more ample supply of memory and some bulk materials.

In NIO's view, the car itself is an important carrier of AI. Take the ET9 as an example: calculating all the semiconductors on the vehicle together, their value is about 105,000 yuan, already exceeding that of the battery pack. "The car itself is currently one of the highest-value AI products."

AIDC (AI Data Center): Refers to large data centers built specifically for AI training and inference workloads. Compared with traditional data centers, their single-rack power density can reach dozens or even hundreds of kilowatts, and their demand for high-bandwidth memory (HBM), high-end PCBs, liquid cooling, copper, aluminum, and other materials grows exponentially. Because AI computing power investment has exploded in the past two years, AIDC manufacturers are willing to lock in upstream capacity at higher prices, creating a "crowding-out effect" on general materials such as memory, copper-clad laminates, and copper foil. Although the auto industry is also a major user of these materials, its procurement bargaining power is not superior to that of AI manufacturers — this is the industrial backdrop to Li's statement that "if you ask for more, the price goes up instead."

The CBU Mechanism: Breaking a Car Down into 17,000 to 18,000 Parts to Do the Accounting

The underlying support for NIO's improvement in operating quality is a management mechanism called CBU (operating unit). The core logic is to break the business down to the smallest unit, from per-vehicle CBU, per-store CBU, and per-market-activity CBU all the way down to the individual part level.

Li revealed that NIO is breaking a car down into 17,000 to 18,000 parts and then using AI to find opportunities to improve efficiency and reduce costs. This approach is in the same lineage as ontology and software engineering logic in AI.

On the R&D side, the CBU mechanism is not simply about cutting expenses. Qin Lihong said R&D expenses still account for more than 7% of sales revenue, which is not low; the key is that efficiency is improving. Li put it more directly: "Now 2 billion yuan can do the work that others need 3 billion yuan to do."

At the organizational level, NIO emphasizes a bottom-up operating logic — what empowerment and support the front line needs, and regional expenses starting from front-line operating results, rather than the traditional top-down breakdown of targets. Li estimates that building the capability for deep integration of AI and CBU will take about another year.

The connection between ontology and CBU: In AI and knowledge engineering, ontology refers to the formal definition of concepts, attributes, and their relationships within a domain, providing a unified "language" for data modeling, semantic reasoning, and automated decision-making. NIO's breakdown of the entire vehicle into 17,000 to 18,000 parts is essentially building an enterprise-level ontology of materials and operations: each part, process, and cost item has a unique identifier and attribution relationship, based on which AI identifies waste and optimizes procurement and design. This is consistent with modularization and the single-responsibility principle in software engineering — only when objects are defined in sufficient detail and relationships are clear enough can automated cost reduction and efficiency improvement be scaled up.

The Second Growth Curve Has Already Emerged: Service and Community Business Revenue Reached 5.8 Billion Yuan in Half a Year

Regarding the second growth curve, Li's answer surprises many people — it has already appeared.

Last year, service and community business revenue was about 10 billion yuan. In the first half of this year, it reached 5.8 billion yuan, including 2.7 billion yuan in the first quarter and 3.1 billion yuan in the second quarter, and it has remained consistently profitable. There are already more than a dozen business lines around service and community, including subscription services and intelligent assisted driving subscriptions. Intelligent driving subscriptions from used-car users alone have already generated several tens of millions of yuan in revenue this year.

Charging and battery-swapping volume grew about 40% year-over-year this year. At present, the installed base eligible for intelligent driving subscriptions is about 40,000 vehicles, of which about 8,000 are regular subscribers, leaving substantial room for growth.

As for the third growth curve, Li believes it may come from the energy business or physical AI, but that is further off in time. He emphasizes that in a core market like Shanghai, if NIO's share reaches 8%, sales can triple — before accomplishing these things, talking about a second curve is aiming too high.

Focus and Restraint: "Can Do" and "Should Do" Are Two Different Things

At a time when automakers are eagerly expanding into robotics and energy storage, NIO is instead emphasizing focus.

Li divides robots into three categories: performance robots, productivity robots, and general-purpose service robots. He believes the space for productivity robots is huge, but they do not necessarily have to be humanoid — assisted driving in a car is itself a kind of productivity robot, "equivalent to buying a car and getting a driver for free."

For embodied intelligence, NIO's strategy is to invest rather than enter the field itself. Li said the robotics industry chain is still at a very early stage, and NIO has already invested in several companies, but at this stage its focus remains on doing cars well.

"In terms of capability, of course we can do energy storage, software, and various intelligent electric products, but being able to do something and being supposed to do it are two different things. Just as Apple has many capabilities, it does not do everything just because it can."

This restraint is also reflected in pacing: technology roadmaps are considered on a ten-year horizon, product planning on a five-year horizon, and product definition on a two-year horizon. A car is measured over at least five years from R&D through its entire life cycle, but the accounting must be clear at every stage.

Industry Judgment: Pure Electric Becomes Mainstream, Another Three to Five Years in the Middle-to-Late Stage of the Marathon

Regarding industry trends, Li and Qin offered several key judgments.

Pure electric penetration reached 44.3% in July, 1.5 times that of extended-range plus plug-in hybrid; within NEVs, pure electric already accounts for 68%, twice that of plug-in hybrid plus extended-range. Li expects pure electric penetration to reach 50% by the end of the year, and the keyword of the year could be "pure electric becomes mainstream."

On the competitive landscape, Li believes there are still three to five years left in the middle-to-late stage of the auto industry's marathon, and the later stage may never truly end. Industry concentration will certainly rise, but not to the extent many people imagine — geopolitics is an important variable.

On technology routes, NIO is clearly focused on pure electric and is actively promoting battery cell standardization, with the direction being 46-series large cylindrical long-life cells and a future model more likely to be "in-house development plus contract manufacturing." On chips, Shenji is expected to generate several hundred million yuan in external revenue this year.

When asked whether the auto industry's status will be replaced by AI, Li's answer was blunt: "Absolutely unshakable. China's auto industry is about 11 trillion yuan in scale and affects an enormous number of jobs."

One-line Summary

William Li says AI has triggered diseconomies of scale, and NIO is growing against the trend through premiumization and the CBU mechanism

Paragraph Summary

NIO founder William Li proposes the counterintuitive idea of "diseconomies of scale": AI data centers are absorbing large amounts of memory, PCBs, copper, aluminum, and other resources, so the more cars an automaker sells, the higher its costs become. NIO's per-vehicle cost rose by 14,000 yuan in the second quarter, adding about 1.5 billion yuan in costs out of thin air for the quarter. Yet NIO delivered a counterintuitive earnings report in the same period: 191,100 vehicles delivered in the first half, revenue of 57.67 billion yuan, an overall gross margin of 18.7%, and an average selling price of 406,000 yuan, surpassing BBA. Li summarizes AI's impact on the auto industry as a triple blow of resource competition, talent siphon, and capital diversion, and expects cost pressure to last about two more years. NIO uses the CBU mechanism to break the entire vehicle down to the parts level for accounting, and its service and community business has already formed a second growth curve at the tens-of-billions level.