$100,000, no need to buy a car, no need to build a technical team… can do the unmanned delivery capacity business in a city.

After two years, Ji Shi Intelligence held a strategic launch event again. This time, it threw out the "franchise" olive branch.

On September 10, at Ji Shi's 2026 strategic launch event, Founder & CEO Kong Qi, Co-founder & CCO Zhou Qing, and VP of Capacity Operations Fu Yulong took the stage one after another, and the entire event was extremely high in information density.

Covering 20+ countries, landing in 300+ cities, actual operating mileage exceeding 250 million km, fleet size exceeding 30,000 units... After announcing a series of achievements, Kong Qi revealed that Ji Shi began to think: for Level 4 autonomous driving, what exactly is the product?

Ji Shi CEO Kong Qi announces company strategic upgrade

His answer is: city-level Physical AI solutions. Because customers truly buy, is service.

Thus, Ji Shi announced the completion of the "City-level Physical AI" strategic upgrade, starting to move from vehicles to capacity. At the same time, launching the "Ji Shi Car Rental" business targeting express logistics, commercial distribution, urban shuttle and other scenarios, providing flexible rental methods such as daily, monthly, and annual rentals, and opening up city unmanned capacity franchise nationwide.

Kong Qi specially reminded that under the current trend of unmanned freight moving towards large-scale implementation, the operating cost of high-level autonomous driving vehicles is very heavy. "It's not like we buy a Tesla or other passenger vehicles, where assisted driving is on or off, for users and manufacturers, there is absolutely no cost."

Whether comprehensive operating costs can be pressed down low enough directly determines whether the Level 4 business chain is positive. Currently, Ji Shi's comprehensive operating costs, including network, remote safety officers, etc., have been pressed very low, and can achieve positive cash flow and gross margin per vehicle.

Thus, Ji Shi launched the "Ji Shi Car Rental" business.

Ji Shi Car Rental Business Map

"Customers don't need to own a vehicle first to have stable capacity." Zhou Qing introduced, Ji Shi developed two major platforms for this: Capacity Management Platform, from intelligent dispatch, flexible pricing to orders, vehicles and settlements, etc., stringing together key links in unmanned capacity operations to form a complete operational closed loop; Operations Monitoring Platform, to grasp vehicle location, status, etc., in real time. "It can be understood as the front-end vehicle is responsible for running, the back-end system is responsible for managing."

He further explained the confidence to do this: Technologically, launching L4-level scaled production without maps solution, deployment completed in the fastest one day; In hardware, conducting L4 system design from two dimensions of functional safety and performance safety, cumulative verified mileage exceeding 100,000 km; In products, from five-party, ten-party to refrigerated series, then to cooperation models with Yutong, Dongfeng, etc., forming a product matrix covering different capacities, different scenarios.

So far, Ji Shi Car Rental has been piloted for a period of time, completing 300 clients' rental model pilot in 45 cities, cumulative 1,500+ orders.

The model is running through, next is to attract more people to join. Fu Yulong gave three numbers: urban distribution logistics annual revenue about 3 trillion, urban distribution vehicle stock 30 million vehicles, and the current penetration rate of China's unmanned delivery is only 1%.

"We have 99% of market space waiting to be dug out." He announced, Ji Shi Car Rental partner minimum $100,000 can join, rent-sale ratio 3:1, 4 vehicles can start, and supporting flexible exit, helping you acquire customers, operations management system, operation handbook and other support. "Ji Shi is looking for not the person with the strongest capital strength, but the person who is truly interested in this career, this track, willing to run it as a long-term career."

Ji Shi Partner Investment Policy

Before the launch, Cyber Auto dialogued with Zhou Qing. From the underlying logic and specific tactics of the rental business, to the cooperation model of Zelos Inside, to the landing progress of the mapless solution, the pace of overseas market expansion, and his judgment on the Physical AI competition landscape, he gave quite honest answers.

In the process, Zhou Qing repeatedly mentioned a word: threshold. Lower the threshold for customers to use unmanned driving, lower the threshold for franchisees to enter the capacity ecosystem, lower the threshold for AI to enter the real physical world, this is almost the common clue of Ji Shi's business model iteration every time in the past five years.

Following is the interview record of Zhou Qing in this exclusive interview (slightly edited).

01

Ji Shi Car Rental: Let Unmanned Driving Change from "Product" to "Productivity"

Q: Can you introduce the specific situation of the rental business?

Zhou Qing: We launched Ji Shi Car Rental this time, the core goal is to let more users be able to use unmanned driving capabilities and services more flexibly. For individuals and enterprises with distribution transport needs, including supermarket delivery, fresh food cold chain, park logistics, express last-mile and other scenarios, are all very suitable for planned, semi-planned and other relatively stable transport needs.

In the past, when people used unmanned vehicles, they often first thought of buying a car or renting a car. Now we hope to lower the threshold for users to use unmanned driving, let unmanned vehicles change from a product that needs to be purchased, gradually into a flexible productivity that can be used.

We not only provide full rental and time-sharing rental, but also provide order rental, rental by order, even rental by demand. Can meet daytime delivery, nighttime delivery, including idle capacity, can also be used for time-sensitive logistics delivery.

In Ji Shi Car Rental's operation model, there will be both Ji Shi's self-operated fleet and franchise fleet. Franchise fleets do not need to build their own technical teams, through our platform can obtain the same management capability as the self-operated fleet.

Q: Mainly based on our own built platform? Or also support suppliers to build platform?

Zhou Qing: Planned, semi-planned logistics are all on our platform; idle capacity or some part-time capacity, we can put them on other platforms.

Q: Why choose this time point to launch to the market?

Zhou Qing: We have a customer, agent Ah Gang in Kunshan, now is also a Ji Shi super partner. Early 2025, he bought four cars, and tried to rent the cars out, didn't expect this model ran up quickly. From four to dozens, then to the current 200+, he reorganized his capacity in excess of self-use through pooling orders, pooling time, now his overall business income every month is 5 times of the past.

In this process, we found that a part of customers are like Ah Gang, the concern is not "what car I want to buy", but "whether I have stable transport capability every day, whether this car can work according to my needs". So we thought, by building a platform, and then more partners like Ah Gang, turn unmanned capacity into the infrastructure of city services. Because we ourselves have technology, have platform, Ah Gang and others themselves have the motivation to do these things.

In the past period of time, we did not rush to push Ji Shi Car Rental to the market, but carried out pilots in different cities. To date, we have already carried out rental model pilots in 45 cities and 300 clients, completing more than 1,500 rental orders in total.

These real operational data are more important to us than any theoretical deduction. They verified a key conclusion: when the threshold for using unmanned driving is lowered, market demand will emerge quickly. Some customers may not want to buy, also don't want to rent full, they focus more on whether they can obtain stable, continuous transport capability in a more flexible way.

We are now launching Ji Shi Car Rental, also not just to do a simple unmanned vehicle rental service, but hope to further explore how to let unmanned driving capability be obtained, matched, organized, used more flexibly like other production factors. We also believe, this is the important step for unmanned driving to go from product to productivity.

Q: What kind of partners does the rental business need?

Zhou Qing: For our self-operated fleet, also including our franchisees, the whole capacity productivity ecosystem, to do the speculation, We need a group of people who truly understand the local market, understand customer needs, can organize transport services. For example, certain local customer resources and logistics scenarios, can find real transport demands.

Secondly, there needs to be certain operation and service capabilities, can do vehicle dispatch, daily management and customer maintenance well. In Ji Shi Car Rental platform, franchisees including self-operated fleet cities, can also achieve unmanned vehicle management through capacity platform management, such as dispatch, pricing and other full-link closed loops. Through capacity monitoring platform, we can also achieve safe and efficient operations, monitoring unmanned vehicles and remote intervention when necessary. We believe, in the future with the increase of franchisee quantity, city unmanned capacity will also operate more efficiently, at the same time idle vehicles can also be fully utilized, can also let more customers obtain unmanned driving capability with a lower threshold.

Q: Is the threshold for building a platform high? If other families want to imitate, is it difficult?

Zhou Qing: Actually, the difficulty of doing this platform is not high. Look at we now itself have vehicle sales, also have car rental. If car rental has a high threshold, it is also impossible to be with vehicle sales as the same type. To put it plainly, our different models face different customer groups, so there will not be too much competition inside.

The core of this rental model is still to find suitable franchisees, this is a core competitiveness. I have platform, not necessarily have franchisees who can trust Ji Shi's technology, willing to join. So the key, First is the manufacturer's technical capability, second is the platform's applicability, third is we have a batch of franchisees who trust technology, and at the same time can understand the local market.

Q: Can it be understood that the cost of building a platform is not particularly high, technically relatively not so difficult, more is to find the right person?

Zhou Qing: Yes.

Q: How to find so many suitable people?

Zhou Qing: For customers who are just starting to contact unmanned driving, he may first care about whether this technology can help me save people, can help me improve efficiency. At this time we hope is he can start using first, run first, use real operation results to verify this value. When customers truly start using it, transport demand becomes more and more stable, vehicle utilization rate becomes higher and higher, he will discover using this model is actually more appropriate. So we hope to form a relatively complete customer path: first lower the use threshold, then verify value, finally according to own business scale, choose the most appropriate way.

Q: Ji Shi quickly obtained large number of user support, what is the core point?

Zhou Qing: First most basic, we must have enough quantity, enough good technical strength, enough stable and safe technology. When car use is not stable enough or not safe enough, everything is very void, this is the most basic.

The second is that our platform must be good enough. For franchisees, if this platform cannot provide me with enough tools, I still need to invest more people to do this matter, this business cannot calculate the account, then I definitely will not do. For customers, actually is the concern he cares about, can I solve my transport problem with the lowest cost or not.

Q: After the rental business lands, startup cost is relatively low, will there be some customers who are original vehicle purchase or monthly rent, cutting into the relatively light asset investment?

Zhou Qing: Yes, will be. Before only could buy, this is suitable for customers with sufficient cash flow and certain profit margin, he wants to continue to use long-term investment to reduce cost, improve efficiency, this can be car special use. But this kind of customer is after all few. Most customers still may be: I can just solve my one time transport demand, each time fixed time transport demand I can solve, usually I don't need this car, using lower cost model is actually better choice.

Q: Behind this does not mean, Ji Shi itself needs to bear the cost will be heavier?

Zhou Qing: Yes, but we cannot just look at one vehicle's book depreciation, also need to look at how much value one vehicle can create in the whole lifecycle. If vehicles are just sold out, after the transaction is complete, value comes mainly from one-time sales, then actually is only one-time accounting, follow-up has no continuous income. For manufacturers, follow-up maintenance is definitely not so active, because I this time profit is actually calculated out.

But if entering rental and capacity service system, we need to further improve vehicle usage efficiency. We also hope it can continue to generate income in a longer lifecycle. In this case, cars run more and more full, fixed asset utilization rate will be higher and higher, created value will be higher, and it created value is definitely higher than our one-time car sales created value.

So, with fleet scale expansion, vehicles' BOM cost, maintenance, spare parts, including operation and other links will also gradually form economies of scale, our single vehicle operating cost will also be further reduced. So the main goal of rental business, is still through resource integration and scientific operation method, turn this asset into production data that can continuously create value.

Q: What is the business logic of Ji Shi Car Rental? Rental fee, or commission per order?

Zhou Qing: We afterwards may be commission, is what you just said full rental, including time-sharing by hour to collect, then including commission per order, actually is charging by order. Roughly these several kinds, follow-up will be diversified.

Q: From buying cars to renting, what is the logic behind Ji Shi's business model change?

Zhou Qing: If looking back at Ji Shi's business model changes in the past few years, every iteration is solving a very specific problem, rather than doing changes for the sake of pursuing business model innovation.

At the beginning we did vehicle sales, because customers first need a really good unmanned vehicle that can run well and work on real roads, solving the problem of product can be used. When products gradually mature, we started to explore financial services, including FSD subscription and other models, essentially is lowering customers' one-time investment threshold, let more customers be able to use Ji Shi's autonomous driving capability, solving the problem of wanting to use but entry threshold still high.

Until today to launch Ji Shi Car Rental, we see is further demand: some customers even don't need to own an unmanned vehicle, they truly need is stable, flexible transport productivity.

So these several changes look like from selling cars to subscription, then to car rental, but behind logic is actually the same, is constantly lowering the threshold for customers to use unmanned driving, let more customers be able to truly use it, and obtain value from use.

Future we also will not limit ourselves to a certain business model inside. Customers need to buy cars, we provide products; buying cars and using cars costs too high, we provide FSD subscription model; customers don't need to own this car, don't need car property rights, also don't need car special use, then we provide through flexible transport capability, provide car rental and capacity, is Ji Shi Car Rental this service. So for us, these services long-term view, may finally this service for Ji Shi obtained value higher.

Q: Are the several models mutually convertible? For example, from buy to rent, from rent to buy?

Zhou Qing: Yes, we can turn. For example I rent for a long time, then I can through depreciation method buy this car. Buy change to rent is the same, even we can provide combination, you can buy a part also can rent a part.

02

Zelos Inside: Technical base treat everyone equally, landing form dynamic adaptation

Q: This time also announced cooperation with GAC, overall logic is same with Dongfeng? Or will there be some differentiated things?

Zhou Qing: Our cooperation with GAC, in underlying logic is consistent with Dongfeng. Core is putting Ji Shi's mature L4-level autonomous driving capability, namely our Ji Shi brain and behind algorithms and software and hardware integrated solutions, open up, empower excellent automobile industry partners.

In this point, Ji Shi provides technical base and safety standards completely is treating everyone equally. However in specific landing form and production division, indeed will according to each car enterprise's own manufacturing advantages, capacity layout and supply chain characteristics do dynamic adaptation.

Both sides in product definition, technical development, engineering verification, actually have all done deep collaboration. Ji Shi also has mature L4-level autonomous driving technology and scaled mass production experience, car factories have mature vehicle whole research development manufacturing and quality system, both sides are complementary advantages. Just this industry inside different models, technical solutions and cooperation depth differences, now still don't have a unified standard. We finally still must take product safety quality and customer needs as premise. Not necessarily same, but overall large cycle, large time nodes may be about the same, details will definitely have some differences.

Q: Cooperation with Dongfeng from signing to delivery is 8 months, reason for quick landing?

Zhou Qing: Yes, we have already possessed mature L4 autonomous driving technology, also have scaled mass production experience. Dongfeng also has mature vehicle whole research, manufacturing and quality system. So in this case, we can reduce traditional cooperation mode inside repeated communication and development links. This speed actually cannot be simply copied, it extremely relies on both sides teams' efficient mutual trust and agile collaboration.

Traditional models are taking year as iteration time. But for autonomous driving, it more focuses on autonomous driving technology, actually it is a software iteration, this speed is faster than traditional cars. Because our chassis may not have to become very big, but our technology can do continuous iteration.

Q: Our this set technology regardless of with Dongfeng, GAC, or other families, can reach this speed?

Zhou Qing: Logic is the same, correct.

Q: Joint development, verification, delivery mode, but sounds very simple, but executing is not easy, how to deal with car enterprises always relatively strong attitude?

Zhou Qing: Yes, actually we need to change is also car enterprises' traditional ideas. Car enterprises traditionally, vehicle core carrier is still car. But for autonomous driving, unmanned cars, software and hardware actually is a status, chassis important, but its autonomous driving technology is also equally important. After such many years of industry development, each car enterprise also knows what they are more good at, actually we are also a particularly good complementary advantage resource for them.

So for us, Now many like Dongfeng, Yutong, GAC already found this point, already started taking us as an equal partner, come jointly do joint development together. Future I think this kind of car enterprises will be more and more.

Q: In landing process have some difficulties overcome experience can share?

Zhou Qing: This block actually has very many. Like chassis adaptation not to say, many chassis adaptations actually are normal work process. Like we do airport, need to focus on safety problems near aircraft, this for algorithms is also a very big challenge, because aircraft are much taller than cars, it cannot see the car's full picture. Actually we need to process in algorithms is special scenario dedicated cars encountering some long-tail problems, is not any obstacle, but needs enough many technical capabilities to polish.

Q: Safety, maintenance have unified rules?

Zhou Qing: Safety responsibility aspect, L2 and L4 responsibility logic is somewhat different. L2 stage still belongs to driving assistance, so drivers need to bear driving subject responsibility. And L4 unmanned driving core change, is vehicle assumes dynamic driving tasks under limited operation scenarios, therefore in responsibility specific division, will according to vehicle, autonomous driving system and operation and other links responsibilities do clarification.

We also will in cooperation initial stage clarify both sides responsibility boundaries with car factories, and establish corresponding safety and service mechanisms. Different car factories' product architecture and cooperation methods may be different, but Ji Shi in autonomous driving system research development, verification and safety management aspect standards and requirements are consistent, also will not because cooperation partner differences have different standards.

Q: For Inside mode, hope it finally made what appearance?

Zhou Qing: From longer term view, we more hope through Zelos Inside put Ji Shi's technology accumulated in autonomous driving field, software and scaled mass production capabilities, with host factory mature vehicle whole research, manufacturing, supply chain and service systems combined together, jointly create unmanned driving products facing different business scenarios.

We also hope Zelos Inside finally not just a single model cooperation mode, but able to form a complete set mature, reproducible industry cooperation system, further promote unmanned cars from point applications to scaled landing, truly realize let unmanned driving empower thousands of industries.

Q: Previously Inside mode like Huawei, latter with many host factories have capital level binding, we will have similar arrangements?

Zhou Qing: We and Huawei mode in cooperation concepts have certain similarities, are all hoping with host factories to conduct deeper joint development, then jointly create safe, competitive and cost-effective products.

But Ji Shi does not need rely on capital binding to push landing. We believe long-term cooperation finally still must return to products and customer values. Host factories and autonomous driving companies can continue cooperation, core lies in both sides can exert respective advantages, do products well, do markets well, and continuously create value for customers. Ji Shi also hopes rely on self in L4 autonomous driving technology, product mass production and scaled operation etc. capabilities, and host factories form long-term stable cooperation relations.

Of course, with both sides cooperation continuously deepening, future also not exclude explore deeper level cooperation methods, including establish joint ventures, capital operations these. Regardless adopting what kind of mode, product power and customer values is able to long-term continuous foundation.

Q: Huawei HI mode is not also completely equal, leading to in products time effect have differences, L4 not this will happen?

Zhou Qing: L2 is customers themselves actively choose whether to use, at the same time customers have bottoming for this system, it core still needs to rely on drivers. But for L4 opposite is not the same, completely is smart driving companies to do drivers decisions, it in customer experience on reflected is this car of mine opens well or not, safe or not, steady or not. For passenger vehicles maybe still not that different.

Q: If is all same, will there be some car enterprises mutual competition, exclusion agreements?

Zhou Qing: Won't, it after all and passenger vehicle logic still not that same, commercial vehicles plainly speak is I this goods actually can steady, on time fulfill.

Q: Can understand as, ultimately speaking it is a business behavior. We can provide same things, price is also same, but models have competitiveness, rely on car enterprises themselves to control costs, carry out market behavior, is correct?

Zhou Qing: Yes, yes, correct.

03

Strategic Upgrade: From Unmanned Delivery Cars to "City-level Physical AI Commercialization Creator"

Q: Strategy layer upgrade to "City-level Physical AI Commercialization Creator", why carry out dimension upgrade at this time point?

Zhou Qing: Ji Shi is not on today suddenly decide from unmanned delivery turn to physical AI, but past 5 years technical accumulation and commercial practice development to certain stage after, we to self positioning one time re-generalize.

In the past 5 years, Ji Shi first solved is a very specific problem, is let unmanned driving truly walk into real logistics scenarios. We from put technology make out, to realize full scenario, full time, full road condition operations, then to form global large-scale L4 commercial operation system. At the same time, we also discovered, Ji Shi truly accumulated down is not just unmanned delivery cars this one product, but a set of let AI perceive real world, understand real world, and in real world execute tasks capabilities.

This change actually has very specific business verification. On one hand, Ji Shi's Zelos Inside let Ji Shi's L4 capabilities start from logistics delivery walk into mine wells, airports, security, sanitation and other more industries. On the other hand, we with Dongfeng, Yutong, GAC and other automobile and industry partners together cooperate, also further verified Ji Shi's autonomous driving capabilities can through open cooperation enter more types of vehicles and scenarios.

Also because Ji Shi already from doing one unmanned car, walked to output a set can enter real physical world, serve different scenarios AI capabilities new stage. In this time node propose city-level Physical AI commercialization, also hope to further clarify Ji Shi next stage development direction, let AI through smart vehicles enter real cities, real industries and real commercial scenarios, and continuously create value, product forms will also from single unmanned delivery cars gradually to more city logistics and commercial service scenarios to expand.

Q: How to persuade market, this is real route upgrade, rather than a brand talk?

Zhou Qing: We are through unmanned delivery products, gradually accumulated AI in various scenarios perception, decision and task execution capabilities, this actually is a process of natural consequence.

Q: Future business also will further expand, for example towards humans direction go extension?

Zhou Qing: Short-term won't, because logistics this market already very big, has urban distribution, has trunk lines, we to do matters already very many. This market finish after, if have extra ability, we again to consider passenger carrying.

Q: Under new definition, market competition logic have changes?

Zhou Qing: I think competition logic indeed is currently changing. At the beginning everyone more compares is single item technical indicators, perception capabilities, algorithm capabilities or autonomous driving capabilities. Later everyone starts to compare autonomous driving commercialization capabilities, just compare who deliver quantity more. When technologies gradually walk into maturity, more important problems is can AI truly deploy to real world, and form scaled, sustainable commercial closed loop.

Ji Shi's advantages lie in we are not from laboratory directly walk to concept products, but already established from AI algorithms, autonomous driving systems, then to unmanned vehicle products, scaled mass production, then to real city operations complete link. Large amount of real roads and commercial scenarios also continuously for algorithms and products iteration provide feedback.

So we truly hope established competencies is not just a certain item technology lead, but technology, products, manufacturing, operations, data comprehensive capabilities. We are able to let AI faster, more stable, larger scaled deploy to city real scenarios, I think this is next stage Physical AI competition key.

Q: In strategic upgrade process, possible encountered biggest difficulties will be what?

Zhou Qing: This difficulties may be how I put technology, products, operations, data etc. comprehensive capabilities do improvement, is how maintain good technology, polish good products, at the same time how and operations combine do more efficient delivery, let users satisfied, at the same time we can through data to feed back my technology improvement. Actually whole cycle is a very key competitiveness.

04

Break-even: 50,000 units is phase goal, business model health degree more important

Q: Previously you mentioned, reaching 50,000 units around, can realize including R&D break-even. Currently still so?

Zhou Qing: Our current operating logic and previously actually is consistent, is with vehicle ownership and commercial scale continuous growth, economies of scale are also becoming more and more obvious.

Previously we mentioned about 50,000 units, is based on then business structure, R&D investment and scaled cost calculation get a phase goal, is not an absolutely fixed number. Now Ji Shi global operated vehicles already 30,000 units, distance this goal scale still have certain distance.

But more importantly, with product iteration, supply chain's continuous maturity and scaled manufacturing capability improvement, our single vehicle cost and R&D cost amortization efficiency actually are continuously improving. We current stage still will put R&D investment placed in very important position. For a company in rapid development stage tech company, short-term profits is not only goal, more importantly continuously build technology and products advantages, and through scaled, commercialization, finally realize healthy sustainable development.

Q: Can understand as, for example previously need 50,000 units reach break-even, now may based on our better business model, fewer than this number can realize?

Zhou Qing: Yes, correct.

Q: Vehicle costs also have space for decline?

Zhou Qing: Have. From cost side view, economies of scale and technology iteration actually also are simultaneously play acting role. With product sales and ownership quantity improvement, like parts procurement, manufacturing supply chain and R&D investment unit amortization actually will further decline. On the other hand, autonomous driving software and hardware also continuously iterate, will also continuously push sensors, computing platforms etc. core hardware cost optimization, at the same time improve algorithm running efficiency and vehicle running efficiency. Scale says, our future definitely will further reduce costs.

Decline space says, specific data maybe not particularly good to evaluate. But its decline amplitude is definitely we can expect, definitely is a year decline, and year will have a relatively considerable decline ratio.

Q: Overseas market progress how? For future growth expectations how is?

Zhou Qing: Now Ji Shi's overseas business currently still is in rapid expansion stage, current overseas ratio and growth also conform to expectations. For example we in Singapore market, our unmanned cars already and local largest supermarkets FairPrice reached hundred unit scale cooperation.

In global market current stage, we more focus on different markets actual commercialization conditions and customer needs. So we now globalization strategy generally can summarize to two paths: on one hand is to target overseas market own logistics and commercial demands establish localization cooperation; On the other hand is according to domestic customers and partners jointly do good globalization.

Of course, different countries may in road right regulations and environment needs also exist some differences, therefore we also will not use one set solutions replicate to all markets, but will according to local different localization demands do corresponding product logic.

For example in UAE, we with UAE Post established joint ventures. On one hand post itself has demands, on the other hand it also has expand road right capabilities. In South Korea, we are with Lotte Group in parks inside first to do cooperation. Different countries maybe still all not that different.

Q: Future overseas market growth pace, can further share?

Zhou Qing: I side have not specific data, now still is in expansion stage, but we are certain is, It future definitely is ratio will become larger and larger, because overseas after all still is a profit better market.

05

Mapless Solutions: Landing ratio near 90%, difficulties still long-tail

Q: Mapless solutions maturity and landing situation how?

Zhou Qing: Ji Shi's L4 mapless solutions have entered scaled application stage. Currently our new routes adopt mapless solutions ratio already near 90%, by end of month estimate can reach 95%. So for Ji Shi, mapless solutions on one hand greatly lowered map costs, originally adopting HD map solutions, map costs per kilometer just about several hundreds, now we lowered to approaching zero. On the other hand, unmanned vehicles possess stronger environment understanding and autonomous adaptation capabilities, thus further lowered entering new areas and new routes deployment threshold.

Q: You just said 90-95% is vehicle coverage, or route coverage?

Zhou Qing: Route coverage.

Q: Some routes no realized mapless reasons is what?

Zhou Qing: Some routes maybe scenarios are longer tail or more complex. For example, our mapless solutions for Chongqing this kind of mountain city maybe support degree doesn't so full, need iteration and upgrade after, can to further promote.

Q: Mapless solutions difficulties?

Zhou Qing: Actually it hardest place is all of us entering real physical world intelligent systems all will face these long-tail scenarios problems. Actually for this block says, all smart driving, whether L2, L3 below assist driving passenger car systems, or like our L4 level this kind of cargo carrying autonomous driving systems say, actually have a very big challenge.

Q: Got to quickly push mapless landing reasons?

Zhou Qing: Actually main still is data problems. Why like pure L4, actually did mapless not many, is because its data quantity is not enough. Like L2 or L3 say, actually it do mapless, even if it bottoms can't, it can lower up give people to do takeover. But for L4 say, cars already need to do last final bottoming. For this block says, is a relatively core problem.

Ji Shi's advantages lie in our road unmanned cars already enough many, so we can have enough many data. At the same time, we also established a set able to continuously discover, process and digest this kind long-tail problems system. With vehicle scale and actual operating mileage continuous growth, we every day will encounter large amount different roads, weather, traffic participants and special scenarios, then these real data will continuously feed back algorithms and products.

At the same time, we are also through simulation data mining and targeted training, put small amount probability extremely low scenarios into can continuously train and verify problems, put long-tail problems continuously converge, and let systems with scaled operations continuously evolve. Current Ji Shi already entered such a positive cycle.

Q: Track players continuously flood in, Ji Shi core advantages is what?

Zhou Qing: I think Ji Shi most core barriers is still scaled closed loop capabilities. Like data, technology, customers and ecosystems, actually are not mutually independent. Ji Shi through L4 autonomous driving technology and product capabilities obtain customers, then through scaled operations obtain large amount real road data. These data further promote algorithms and products iteration, product capability improvement after then able to enter more customers and scenarios. At the same time, with host factories, logistics enterprises and other industry partners cooperation, we also further perfected from R&D manufacturing to operations industrial chain links.

In AI industry, single direction advantages often difficult to form competition barriers. So technology, products then to mass production operation data then feedback to iteration, this complete closed loop is competition key. Scale is bigger, closed loop efficiency is higher, this is also we believe future truly able to long-term form competition barriers a point.

Of course, technology definitely is foundation.

Q: How to view new entrants?

Zhou Qing: We think now more and more excellent enterprises already started entering this field, this actually we are very welcome, this also proves this industry is walk towards maturity one very important performance. Like now new entrant these enterprises actually all have their own technical accumulation and industry advantages, so more participants come in, also will promote entire industrial chain further maturity.

Future we still will insist own pace, do products well, do customer services well, at the same time continuously improve technology and scaled capabilities. Industry finally compete still is long-term for customers create values capabilities, we always think so.

Q: Since Ji Shi founded, you feel which decisions are very key?

Zhou Qing: Looking back past 5 years, I think there are three decisions actually very key.

First is firmly chose L4 autonomous driving this track, and we long-term investment. We very early judged logistics and other business scenarios are L4 autonomous driving most likely first scaled landing fields, therefore didn't because short-term market changes change technology directions.

Second point is put scaled commercial operations as goals, rather than stay in technology verification stage. Only truly enter real cities, real roads and real customer scenarios, technologies have continuous iteration and verification opportunities.

Third point is continuously open industry cooperation, from with logistics enterprises cooperation to with host factories explore Inside mode. We also gradually realized, autonomous driving industry final is not one company's solo play, is technology, vehicle whole manufacturing, supply chain and operational capabilities a overall synergy.

If say these three decisions have any common points, that is always around one goal, is let L4 autonomous driving truly enter real world, and form scaled commercial values.

Q: Future hope others mention Ji Shi when, to its definition is what?

Zhou Qing: From underlying capabilities look, we hope become city-level physical foundation capabilities providers, through autonomous driving technology, intelligent vehicles and AI capabilities, empower host factories, logistics enterprises and more industry partners. At the same time, in some need deep participation business scenarios, we also through operations services help customers truly put unmanned cars use up.

Future Ji Shi will be a technology and product capabilities, plus industry synergy capabilities both have companies. We provide underlying intelligence capabilities, but final purpose not is own how many cars, but let AI truly enter city logistics, commercial and service systems.

If say future 5 years determine Ji Shi fate key variables, I think still is AI entering physical world efficiencies. Technologies can continue evolve, products can continue scaled, then our business models can continuously create customer values, will all return to a problem, is we can let AI safer, more efficient, and larger scaled in real world go running.

Q: This sees Ji Shi development situation, also sees entire industry technology iteration speeds?

Zhou Qing: Yes.

Q: Currently world have you relatively appreciate, you think Ji Shi can infinite close or surpass enterprises are?

Zhou Qing: We are hope can make a new, of course past templates inside like Huawei this kind of manufacturers actually still is relatively successful. But we also don't hope paste any kind label, more hope is still can make own poetry and distant places, walk own paths.