Thank you. Good evening, ladies and gentlemen.

Welcome to the investor conference call and audio webcast hosted by Horizon Robotics. On today's call, we have Dr.

Kai Yu, Horizon's founder, CEO, and the chairman of the board, and Mr. Lei Wang, CFO and head of fixed capital markets.

We'll begin with Dr. Kai sharing our latest strategic and business developments.

This will be followed by Lei's review of the company's financial performance in first half of 2026. We will then proceed to the Q&A session.

Before we start, we want to remind you that this call may include forward-looking statements which are underlined by a number of risks and uncertainties, and may not be realized in the future for various reasons. Information about general market conditions comes from a variety of resources outside the Horizon Robotics.

This presentation also contains some unaudited non-IFRS financial measures that could should be considered in addition to the company's financial papers. I will now turn the call over to our CEO, Dr.

Yu. 各位投资人朋友们，大家晚上好。

Dear investors, good evening.

我是地平线创始人、CEO余凯。很高兴在8月的最后一天，和大家在线上相聚，向各位汇报我们上半年的经营业绩情况。

I'm Yu Kai, founder and the CEO of Horizon Robotics. I'm delighted to be with all of you online on the last day of August to report on our operating results for the first half of the year.

我先说一下行业。坦率的讲，2026年上半年，中国汽车行业呈现出明显的内需承压、出口高增的结构性分化。据乘联会统计，上半年国内乘用车销量显著变弱，零售同比下滑超过了20%，尤其是有一些客户，下滑超过了30%。但另外一头，出口数据很强劲，上半年整车出口同比增长超过65%，创下历史新高。这么复杂的内外外部环境，对每一家企业的经营都是挑战。在这样的大背景下，地平线上半年确认收入20.55亿元，同比增长32.9%，毛利率保持在66%的高水平。在维持饱和研发投入的同时，我们也注意逐步提升经营效率。上半年地平线研发、销售和管理三项费用的合计增速是23.3%，低于收入和毛利的增速。这说明我们的经营杠杆在收入高增长的同时，开始逐步释放。更细节的数字一会王雷会跟大家做详细的介绍。我今天想跟大家分享的，更多的是一些近期的思考。 Let's start with the industry.

[foreign language]

Go ahead. Let's start with-- Thank you.

Let's start with the industry. Frankly speaking, China's automotive industry in the first half of 2026 showed a clear structural divergence of domestic demand under pressure as well as high-growth export.

According to CPCA data, domestic passenger vehicle sales weakened significantly in the first half with retail sales down by more than 20% year-on-year. As far as known that for some domestic automakers, they have seen a decline by more than 30%.

On the other hand, export figures were very strong with vehicle export by more than 65% year-on-year in the first half, setting an all-time high. Such a complex internal and external environment poses challenges to every company's operations.

Against this backdrop, Horizon Robotics recognized the revenue of 2.055 billion RMB in the first half, up by 32.9% year-on-year, with gross margin at high level of 66%. While maintaining saturated R&D investment, we have also focused on steadily improving our operating efficiency.

The combined growth rate of Horizon Robotics R&D, selling, and administrative expenses in the first half was 23.3%, lower than the growth rate of revenue and the gross margin. This shows that our operating leverage is emerging as revenue grows rapidly.

Mr. Wang Lei will walk you through more detailed numbers shortly.

What I would like to share with you today is more about some of my recent reflections.

第一，关于高阶智驾市场。即使上半年市场整体偏弱，汽车智能化率其实还是在继续的攀升。我们看到的数据显示，中国乘用车智驾渗透率已经达到了76.1%，较2025年的全年水平进一步提升了8.5个百分点。其中，自主品牌固然是智能化的核心玩家，但是大家可能没注意到，今年合资车企的智驾渗透率也第一次站上了80%的极高水平。说明智驾已经成为中国乘用车市场的一个刚需配置。关于合资企业的情况，晚点儿我还会跟大家专门讲一下我们的进展。至于高阶智驾，那么搭载城区NOA功能的车型的渗透率达到了23%左右。请注意，这里的分母不是智能车的销量，而是所有乘用车的销量。也就是说，今年上半年所有的销售的新车中，已经有接近四分之一搭载了城区NOA的功能。而我们看行业数据，高阶行业到了5到6月份的时候，渗透率就已经超过了去年年底的最高水平了。高阶智驾展现出了非常强劲的增长动能。

First, on the advanced intelligent assisted driving market, even though the overall auto market is relatively weak in the first half, the automotive intelligence penetration rate continued to rise. The data we see shows that the intelligent assisted driving penetration rate for passenger vehicles in China has already hit 76.1%, reaching yet another new level compared with 2025, up by 8.5-- 8.5 percentage points.

Domestic brands are, of course, the primary driver of this growth, but you may not have noticed that this year, the intelligent assisted driving penetration rate among joint venture automakers topped an extremely high level of 80% for the first time, indicating that intelligent assisted driving has become a necessity in China's passenger vehicle market. As for the situation with joint ventures, I will talk specifically about our progress there a little bit later.

As for the advanced assistance systems, the penetration rate of models equipped with urban NOA function has reached about 23%. Please note that the denominator here is not the sales of intelligent assisted vehicles, but sales of all passenger vehicles.

In other words, among all new cars sold in the first half of this year, nearly 1/4 were already equipped with urban NOA function. Looking at the industry data, during May and June, the penetration rate of advanced intelligent assisted driving already surpassed the peak level reached at the end of last year, showing that advanced intelligent assisted driving has very strong growth momentum.

我觉得上半年资本市场一直在担心智驾行业的竞争格局，也关心我们的看法。在高阶智驾这个赛道，其实地平线并非是防守方，而是新入局的进攻方。我们去年才开始大规模的量产高速NOA芯片解决方案，那么到去年底才开始量产城区NOA的软件解决方案HSD。但是大家可以去看，仅仅过去6个月，今年上半年支持城区NOA的智驾芯片方案市场上，我们在中国自主品牌的车企里的份额已经提升到23%左右，提升了大概5个百分点，排名从半年内从第三名跃升到第二名，目前仅次于一家全球科技巨头。这家公司虽然维持住了第一名，但它的市场份额比去年全年降低了将近10个百分点。

I think the capital market may make some confusion about the role we played in the advanced intelligent and driving during the first half of this year. In this arena, Horizon is on the offensive, not the defensive.

We only began mass production of highway NOA solution last year and didn't start mass producing urban NOA solution until the end of the year. Take a look, in just the past six months, in the first half of this year, our market share of computing platform featuring urban NOA function rose to around 23% among Chinese domestic brands, an increase of roughly 5 percentage points, and our ranking jumped from the third to the second place within half a year, now only after one US tech company.

Although this US tech company stood at the first place, its market share fell by nearly 10 percentage points compared with the full year of last year.

我们一直很喜欢跟大家讲地平线的Wintel模式，但是最近我们也常常跟大家讲我们的新商业模式，就是我们叫ARM+Android的模式。那我先讲讲Wintel模式。其实就是征程芯片、星空芯片的销售模式，它就像Intel，是通过卖芯片来获取收入。那么地平线的软件，HSD和卡卡鲨，它就像微软的商业模式，是通过直接销售软件来获取收入。当然我们相信未来到L4时代，软件的销售的商业模式会转成服务的订阅模式。那么Wintel对应了地平线的直接市占率，也是我们自己产品所占据的市场份额。而ARM+Android模式代表了地平线的延展市占率，也就是我们通过赋能合作伙伴、赋能客户，让他们采用我们的技术IP来开发芯片、开发软件产品，所间接带来的市场份额。以地平线的直接市场份额加上延展的市场份额，当然是会大于大家所看到的地平线直接卖芯片和软件所占有的市场份额的。那比如说，地平线支持车企自研芯片或自研软件，其实就属于比较典型的延展市场的概念。直接的市占率高，就直接市场的这个市占率高，是地平线它的芯片以及地平线的直接的售卖的软件，它的这个市占率高。但是延展市占率高，其实也很有战略性的意义。因为它不仅带来可观的IP授权收入，而且有助于构建我们与客户长期的并且极强粘性的伙伴关系，而且客观上面有助于排除竞争。

Next, I would like to share with our our business model. First is about the Wintel, and also our extension business model, ARM plus Android.

When it comes to Wintel, we have our SoC and also our software. This is just like the sales model of Intel, that is our direct revenue.

In the software platform, we have HSD and also Kakashark. The sales model is more like the Microsoft.

When we evolve to L4 and L5, and then we believe that our business model and will more becoming the subscription model. It means that in addition to the Wintel, and we also have the ARM plus Android, the extension business model.

It means that we can use our IP to earn our revenue. That is our more indirect revenue.

In wrap up, we have both direct revenue plus our indirect extended revenue, coming from our IP licensing. In terms of the direct revenue, and we have our SoC plus our software, our market share will continue to rise.

At the same time, we have our extension and business model through our IP licensing and it can help us to secure long-term partnership and stickiness from our customers. At the same time, and we can continue to keep our extremely highly competitiveness.

这两年高阶智驾芯片的市场的竞争格局，正是我们直接市场份额和延展市场份额的一起的提升，从而去压缩了我们的竞争对手的市场份额和他们的发展空间。那么这个演化的速度会非常快。这让我想起了过去几年ADAS市场的份额的变化。大家知道4年前，2022年，我们的份额是多少？不到5%。那么今年上半年是多少？首次突破了50%，并且市占率是第二名的两倍。从不到5%到50%，我们只用了4年。我们觉得高阶市场也会是类似这样的一个发展。到明年，地平线的高阶智驾芯片，加上地平线IP所赋能的我们的客户的自研的芯片，就是我们所说的叫直接市占率加延展市占率，应该会变成高阶智驾市场的第一名。

We believe that our competitive edge will come from both our direct market share and also our extended market share. We will continue to capture the market share from our competitors, and I also believe that the growth rate will be very high.

This reminds me of our share shifting in the ADAS market over the past few years. Do you know what our share was four years ago in 2022?

Less than 5%. What about the first half of this year?

It broke through 50% for the first time, and our market share is twice that of the number two player. Going from under 5% to 50% took us only four years.

I think it will be the same in the advanced market. By next year, Horizon Robotics SoC, that is our direct market, plus the automakers' in-house chips that Horizon Robotics IP sports, that is our extension market share, will continue to take the number one position in the advanced market.

短期来看，地平线和地平线的车企客户一起去拿下更多份额；长期来看，这个行业的格局会怎么演变？我们的判断是这样的，如果一个行业的天花板比较低，技术曲线很快的趋于饱和，那么头部玩家的技术会很快的均质化，白尺竿头难再进一步。那么这个行业的最终格局，大概是行业头部玩家也会被后面的追赶者所追上，那么格局是分散的。大家都不能搭建起有效的护城河，从而陷入内卷，最终只能拼价格。而如果一个行业是真正的长坡厚雪，技术曲线非常陡峭，并且上限很高，那么领先企业的优势就会不断扩大。原因很简单，头部企业每年的领先优势乘上一个更长的周期，最终会搭建起一个后来者无法逾越的深厚的护城河。 In the short term, Horizon Robotics and the automakers will together capture more shares, and in the long run, how will this industry landscape evolve?

My observation is this: if an industry has a relatively low ceiling where the technology curve quickly plateaus, and the leading player's technology soon approaches perfection and it becomes hard to advance any further from the top, then that industry's ultimate landscape will most likely be fragmented. No one can build an effective moat, but everyone falls into cutthroat competition and ultimately can only compete on price.

If an industry truly offers a long slope with thick snow, with a very steep technology curve and a very high ceiling, then the advantages of the leading companies will keep expanding. The reason is simple: the annual lead of the top company multiplied over a longer period ultimately builds a deep moat that latecomers cannot cross.

我们认为自动驾驶就是这么一个行业。从2019年到2020年，地平线推出征程2和征程3芯片，主要聚焦在ADAS产品。2021年到2023年，我们推出首款超过100 TOPS的征程5芯片，开始基于规则算法来做高阶智驾的解决方案，那么研发费用加起来超过50个亿，累计。2024年到2026年，我们密集推出了5款征程6系列的芯片和1款舱驾一体的智能体芯片，推出了HSD 1.0、2.0，以及首个智能体Agent卡卡鲨。研发费用累计加起来快接近150个亿。那么大家看我们现在是什么状态？L3到L5自动驾驶还未实现，这是一个行业现状。那么我们的竞争对手其实已经越来越稀少。所以这并不是一个高度竞争的行业，其实到最后的话呢，卡尖的玩家是越来越少的。那么从长远来看，我们认为智驾的玩家要留在赛场上，首先需要世界顶级的芯片和算法能力，然后每一个玩家还需要长跑的耐力、长期地坚定地投入的实力，以及持续迭代迭代的能力。平心而论，这是非常不容易的。这些条件地平线都具备。所以我们对地平线在整个智能驾驶，无论是在低阶、中阶还是在高阶市场的终局拿下更大的份额是充满信心的。这是我们对行业长期的看法。

I believe intelligent assisted driving is exactly such an industry. From 2019 to 2020, Horizon launched the Journey 2 and Journey 3 SoCs, focusing mainly on ADAS products.

Then from 2021 to 2023, we launched the Journey 5, our first SoC exceeding 100 TOPS, and began building AD solutions based on rule-based algorithms with R&D expenses totally more than 5 billion RMB. From 2024 to 2026, we have rolled out five Journey 6 series SoCs and one integrated cockpit-driving fusion agent chip in rapid succession, along with HSD 1.0 and 2.0 and our first agent, Kakashark, with R&D expenses totally nearing 15 billion RMB.

Look at where we stand today: L3 to L5 autonomous driving has yet to be realized, and the number of competitors has already dwindled. It means that it's not that highly competitive landscape.

In the long run, I believe that for an intelligent driving player to stay in the game, it first needs world-class SoC and algorithm capabilities, and then every player also needs a stamina for a long-distance race, the strength for long-term investment, and the ability to iterate continuously. Frankly speaking, this is far from easy.

Horizon Robotics possesses all of these attributes, and I am very confident that Horizon will capture larger share in the end game of the advanced market. This is my long-term view of the industry.

我们刚刚说到头部企业的进步曲线非常陡峭，才能在长周期里面最终胜出。大家可能知道，其实地平线跟行业里面的其他技术团队比，包括车厂的自研的自动驾驶软件团队以及第三方的自动驾驶软件团队相比，我们算是进入高阶市场最入局最晚的一个。但是入局虽晚，我们却是厚积薄发。我们在高阶智驾领域的饱和投入，已经在今年开始给我们带来显著的技术优势。例如最近在某头部新能源车厂客户的横向评测中，我们跟竞品相比取得了大幅度的领先。接管率要远远的显著好、好于竞品。去年11月推送的HSD version 1.0和今年6月底推送的HSD V2.0展现出了强大的全场景通行能力，博弈收放自如，给用户带来平顺的驾乘感受。我们前段时间在北京、上海、深圳为投资人朋友们组织了一系列的试乘试驾活动，支持大家在最复杂的市中心自己选择路线。我们通常是说把这个车交给来试乘试驾的投资者，只要后来把车还回来就可以了。那么我听说有一些投资人朋友挑早高峰，还专门去开了开到最繁忙的医院、学校，就是这样的最有挑战的这种街区或路口。还有一些选择在巨鹿路这样的窄路上面调头。还有一些投资人则开到了去深圳的城中村。那么相信体验过的朋友都能体会到我们HSD和市场上其他的高阶智驾方案的区别。我觉得究其原因，是地平线在过去11年积累了世界顶级的芯片能力和算法研发能力。这在全行业看来其实都是独特和稀缺的。这样的软硬结合综合能力，恐怕只有特斯拉可以媲美。这也是为什么行业里面有时候把地平线称之为不造车的特斯拉的原因。

Okay.

sorry. [foreign language] Go ahead.

It has been just mentioned that a leading company's progression curve must be very steep in order to ultimately win out over the long term. As you may know that compared with other in-house and third-party software and algorithm suppliers, Horizon was arguably the last to enter the advanced market.

We entered late, but we have built up strength over time, and now we are delivering with full force. According to some third parties' evaluation, and we are already taking the leadership over other competitors.

For example, in terms of take-over ratio, and we are already better than our competitors. HSD version 1.0, released last November, and the HSD version 2.0, released at the end of this June, have demonstrated powerful, full-scenario driving capabilities, competing with ease and delivering a very smooth driving and riding experience for our users.

A while back, we organized a series of test ride and test drive events for our investors in Beijing, Shanghai, and Shenzhen, allowing everyone to choose their route in the most complex downtown areas. I heard that some people picked the morning rush hour to drive through the busiest hospitals.

Some chose to make U-turns on very narrow streets, like Julu Road, and others drove to Shenzhen's urban villages. I believe anyone who has experienced it can appreciate the difference between our HSD and other intelligent driving solutions on the market.

I think the underlying reason is that over the past 11 years, Horizon Robotics has accumulated world-class SoC and algorithms capabilities which are unique and scarce across the entire industry. This kind of comprehensive capability combining software and hardware is perhaps matched only by Tesla.

That's why Horizon Robotics is sometimes called the Tesla that doesn't build cars.

我知道大家非常关心，地平线在高阶智驾的芯片，以及高阶智驾的软件HSD，跟国内的主要的车厂、自主品牌的合作。我们现在能透露的是进展相当不错。那么征程6M去年已经切入了最大的新能源车企，到今年的下半年，会占到其入门级的AD的出货的七成到八成。那么最新的进展是，我们的大算力芯片，就是征程6P和征程6H，以及我们的高阶智驾的软件HSD，也获得了这家车企客户的高阶智驾平台的新定点，并且在今年内开始量产交付。那么大家知道在媒体上面能够看到的是，地平线的HSD在奇瑞和长安都实现了量产。那么最新的进展是，地平线的HSD，它已经赢得了中国销量最大的头部的5家自主品牌的全部的定点。那么这些量产，从今年下半年就会开始交付，那么到明年的话呢，将会放量地展现出规模落地的这样一个结果。所以综合来看的话呢，地平线的这个HSD已经拿下了中国销量最大的前五大的所有的这个前五大的自主品牌。那么包括合资品牌里面销量最大的是丰田和大众，那我们其实HSD都获得了定点。所以大家可以看到，今天取得的这个结果，是过去几年我们持续高强度的研发投入以及领先的技术实力它所种下的果。那么今年地平线无论是在HSD的研发，以及大算力芯片、新一代的大算力芯片的这个研发，那么实际上包括我们的软件方方面面，包括我们新拿的这些平台性的定点，是今年种下的因，会成为后续更大的这个成果的落地，做好铺垫。所以我们预计的话呢，明年的话在高阶智驾上面的话，我们会继续的进一步的放量落地。

As for the collaboration with the different domestic, international brands, that everyone is very interested in, especially for HSD. What we can reveal now is that the progress is pretty good.

By the end of the year, J6M will account for 70 to 80% of the entry-level AD shipments at the largest new energy vehicle manufacturer. Our high computing power SoC, first HSD, will also begin mass production and delivery on their advanced platform within this year.

Take this year for example, and we have already seen that our HSD has been installed for Chery, and also Changan. Actually, our HSD is already available on the top five highest-volume Chinese brands.

Also for the joint ventures, you know that the largest joint ventures are Toyota and also Volkswagen. Our HSD is has also become the design win for these two joint ventures.

By the end of this year, we are going to see the mass production and delivery in all these Chinese domestic brands and also the JV brands. We can see that today's results is the fruit of our input, investment, and also state-of-the-art technologies.

We also believe that this year, because of our HSD, thanks to our R&D investments and high-level SoCs and the software and also the design and wins, will continue to bear more fruit going forward. It is also estimated that starting from next year, thanks to the mass production, and we are going to see our scale up in more advanced market.

还有一个问题，我相信大家也非常关心，随着车企之间的价格战不断加剧，我们可以看到行业里面的话呢，车价持续地下滑，整个行业的利润空间也在不断地被压缩。大家会关心在这样的行业背景下，地平线如何保持继续高质量的增长？而我们认为在价格战不断加剧的环境下，恰恰是对地平线最有利的发展机会。因为地平线能够给车企提供最具成本优势、更好性能、更优的用户体验的产品。并且会让大部分车企，无论是芯片自研还是软件自研，都会更加地回归理性。因为我们始终相信，产业终局只有20%或甚至都不到20%的塔尖车厂，才会保持自研的能力和实力。而绝大部分的车厂都会采用像地平线这样的独立第三方所提供的技术和产品。那么当前的价格战会加速这样的一中局的到来。

Currently, a lot of investors are also very concerned about the current price war. We have seen the selling price of vehicles is also dropping and also the profitability is also dropping.

Under such circumstances, how Horizon can keep high-quality growth? As a matter of fact, we believe that the current price war is providing a very good opportunity for Horizon.

Because we have our cost-effectiveness and we can provide some high-performance technologies, and also very excellent customer experience. We believe that with our SoCs and the software, and the market can become more rational.

We also firmly believe that in the end game, only 20% or maybe less than 20% of the leading companies can have their in-house R&D capabilities. A great majority of the OEMs will adopt the technologies or products from the independent third parties, like Horizon.

We believe that such a price war will accelerating this process, which is beneficial for Horizon.

我想分享的第二点，是关于合资和出海的。合资企业这几年大家知道面临自主品牌的激烈竞争，但是大家没有太注意到的是，合资品牌的市场份额其实已经稳定在35%左右，并且保持了快两年。现存的合资品牌加上特斯拉，大概75%左右都是在价位段是这叫大众市场，或者叫mass-market市场的车型。所有的合资的豪华车加起来大概20%多。合资阵营也有着迫切的智能化升级的需求，并且和过去的海外技术中国落地的路径不同，这一轮智能化技术升级更多来自于中国的供应商或中国的技术伙伴。那么地平线的合作的伙伴的生态在国内是很有口碑的，一直是行业里面最繁荣最活跃的生态。我们在面对合资品牌和他们出和以及出海的机会的时候，也一样是维持了生态之王的地位。我们通过博世、电装、本田旗下的爱斯泰莫，通过和原来在大陆，现在原来大陆集团现在叫欧摩威成立的合资公司智驾新程，通过我们和大众成立的合资公司酷睿程，以及国内大量的生态伙伴，我们成为了合资车企在中国以及中国车企全球化的智能化的首选伙伴。

Next, I would like to share about the JV and also overseas expansion. In recent years, as and JV have faced fierce competition from domestic brands.

What hasn't received much attention is that the market share of JV brands has actually stabilized at about 35% and has held there for nearly two years. Among the existing JV brands plus Tesla, roughly 75% are mass-market models, while all the JV luxury vehicles combined account for a little bit more than 20%.

The JV brands also have an urgent need for intelligent upgrading. Unlike the past path of overseas technologies implemented in China, this round of intelligent technology upgrading comes largely from Chinese suppliers or Chinese partners.

Horizon's partner ecosystem enjoys a very strong reputation in China and has already been the most thriving and active in the market. As we face opportunities with JV brands and overseas markets, and we have likewise maintained our position as the king of the ecosystem.

Now we are working with the former Continental, now OVM, and and we also have the partnership with other Tier 1 suppliers like Astemo, and also Aptiv, and also Bosch, and also CARIZON, that is the JV we established with Volkswagen, and also Denso. We believe that we are becoming the preferred supplier for many vehicle manufacturers.

在合资车企里，我们通过酷睿程基于地平线征程6系列芯片以及HSD的IP的知识产权打造的高阶智驾方案，今年会搭载于一汽大众、上汽大众、安徽大众的7款全新车型，并于明年开始更广泛的拓展至大众的主流的CEA架构以及更多的将近20款车型。大家如果去看近期官宣的车型价格，15万以内就可以买到一辆带城区自驾、带地平线大算力芯片的大众品牌汽车。即使和自主品牌相比，也非常的具有价格竞争力。我们还基于地平线的人工智能基础模型赋能酷睿程，未来在L3以及L4级别的自动驾驶能力的落地。前段时间我们双方在资本层面也做了深化的合作。地平线赎回了大众持有的部分可转债，为全体股东减少了4.9%的股权稀释。而大众提前转股，成为我们持股比例9.9%的重要战略股东，并且自愿锁定12个月。我们和大众的合作的深度和广度一直在延续。

Next, among these JV automakers, through CARIZON, our advanced intelligent assisted driving solutions built upon Horizon's Journey 6 series SoC and core intellectual property will be deployed this year in seven all-new models from FAW-Volkswagen and SAIC Volkswagen and Volkswagen Anhui, and will begin expanding more broadly to Volkswagen's mainstream CEA architecture next year. Altogether will be about 20 models.

In the future, you will see that the models from Volkswagen with the price range about RMB140,000 to 150,000 will be using the Horizon Robotics urban NOA. You can see even in terms of pricing, they are also going to be more competitive.

Also building on our foundation model, we are also empowering CARIZON to be able to bring Level 3 and Level 4 autonomous driving capabilities in the market next year. Some time ago, both sides also deepened our cooperation at the capital level.

Horizon redeemed a portion of convertible bonds held by Volkswagen, reducing equity dilution by 4.9% for all shareholders. Then Volkswagen's CARIAD meanwhile converted its shares early to become an important strategic shareholder with an 9.9% stake and voluntarily agreed to a 12-month lockup.

You can see the depth and the breadth of our cooperation with Volkswagen continue to grow.

和丰田的合作，基于征程6系列芯片也已经在上半年量产落地，首发搭载于广汽丰田最高量的入门级车型上。后续通过和电装的合作，我们还会服务丰田其他最主流的车型平台，并落地中高阶的智驾方案，预计从后年开始贡献可观的销量。大众和丰田加上本田，占据了所有合资车企销量的50%，而我们通过酷睿程、电装、爱斯泰莫等合作伙伴，已经打开了这部分的增量空间。 Also

our cooperation with Toyota based on the Journey 6 also entered the mass production in the first half of the year, firstly launched on GAC Toyota's highest-volume entry-level models. Going forward, through Denso, we will also serve its most mainstream vehicle platforms and deliver advanced intelligent driving solutions, which are expected to begin contributing to sales the year after next.

Volkswagen, Toyota together with Honda account for 50% of the sales volume for all JV automakers. Through partners such as CARIZON, Denso, and Astemo, we have already opened up this avenue of growth.

再说出海。这可能是今年上半年汽车圈最重要的趋势。出口销量前六大的车企集团集中度很高，合计份额超过了70%。而能支持中国车企出海的智驾供应商，基本上就只有三家，地平线是其中之一。集中度更高，我们已经拿下了24个品牌，将近60款出海车型，覆盖了出口量前六大的所有的车企集团。从量纲上看，我们在过去12个月拿下的出口新车型，已经占到了未来出口总量里面相当大的比重。大家最近看博世和智驾新程的官宣，他们已经基于地平线J6B芯片各自打造了全新一代入门级产品，并且斩获了千万级和数百万级的定点项目，覆盖中国、欧洲、澳洲、新西兰、拉美、中东和亚太等全球主要市场。截至目前，征程6B生命订单生命周期的订单已经突破2000万套，成为普惠级智能辅助驾驶的核心的芯片算力基座。地平线为什么这么重视海外的入门级智驾市场？首先因为ADAS是目前海外市场需求的主流，我们一直重视的就是放量的市场。第二个是因为技术生态是有粘性的，我们通过入门级智驾产品先牢牢占据市场，建立客户之间的信任，为后续海外市场向中高阶智驾的升级打下统一的软硬件底座的基础，占住客户导入的核心路口。

Next we can talk about going global, which may be very an important trend in the auto industry in the first half of this year. The top six auto groups by export sales are highly concentrated.

Altogether, they account for more than 70% of the total share. Meanwhile, there are basically only three intelligent driving suppliers capable of supporting these Chinese automakers going global, and Horizon is one of them.

You can see this is an even higher level of concentration. We have already secured 24 brands and nearly 60 export vehicle models, covering the top six auto groups by export volume.

In terms of scale, for the new export models we have won over the past 12 months, will already account for a very considerable share of our future total export. As you may also seen in recent announcements by the Bosch and the new HCT, they have each built a brand new generation of entry-level products based on the J6B SoC, gaining design wins in the tens of millions and millions of units respectively.

They have also covered major global markets, including China, Europe, Australia, New Zealand, Latin America, Middle East, and Asia Pacific. For the J6B, we have already secured the total number of the order of 20 million units.

Why does Horizon place such importance on overseas entry-level intelligent driving market? Because in the overseas market, ADAS is still the mainstream.

We want to secure this mainstream customer base. Also, because technology ecosystems are also very sticky.

We hope that by firmly capture the market with entry-level intelligent driving products and we can gain the trust from the customers, and also we can establish a unified software and hardware foundation for the future upgrade of the overseas market to advanced intelligent driving. Then we can secure the core gateway for customer onboarding.

更重要的是，以前大家总觉得海外并没有高阶的智能驾驶的需求，但最近我们观察到海外也开始出现向高阶智能驾驶、向AD升级的趋势。我们的J6E和J6M智驾芯片已经被多款明星车型定点，并随车销往海外。虽然海外市场整体来讲，高阶智驾的量产的节奏较国内市场要慢，但是我们积极布局，在法国、德国、日本都已经跑起样车。地平线作为最国际化的公司之一，依托众多的海外生态伙伴，在许多海外主流车企的合作序列里是排在第一顺位的。这是我想分享的第二部分思考。

Also, more importantly, people used to think that the overseas market had no advanced intelligent driving demand. Recently we have observed that this trend towards upgrading to advanced intelligent driving, or AD, is beginning to emerge in overseas market as well.

Our J6E and the J6M SoC have already been designed into multiple nominated models that are being sold overseas. Although in the overseas market, they are relatively slower than that of the Chinese market, but we can still see some overseas automakers are beginning to plan AD related vehicle models, such as in France, in Germany, and also in Japan.

As one of the most international company, Horizon, supported by overseas ecosystem partners, ranks the first in the cooperation line-ups of many mainstream overseas automakers. This is the second part of my observations that I would like to share with you.

我想分享的第三点，是前面提到的Wintel，就是Windows+Intel，以及ARM+Android的商业模式，以及依托这种商业模式的商业拓展。地平线有芯片平台和软件平台。芯片平台就像Intel的商业模式一样，软件平台里面的HSD和卡卡鲨，其实就是新一代的汽车软件操作系统，那么销售模式更像Windows。当HSD演进到L4到L5的自动驾驶，那么卡卡鲨也会成为车主出行生活中的贴心的助手。我相信汽车软件的商业模式将变成用户里程收费或订阅的模式。就像今天微软的Microsoft 365以及很多产品的订阅模式收费模式一样。

The third one, I would like to talk about the Wintel, as I mentioned just now, as well as the ARM plus Android business model. We would like to expand beyond the automobile based on such business model.

In terms of the Wintel, Horizon has both the SoC platform and the software platform, that is more like the business model of Intel. Then we also have our HSD and also Kakashark for our software business model.

That is more like the new-generation operating system for automotive. Their business model is more like that of Microsoft.

Once HSD evolves to L4/L5 autonomous driving and Kakashark becomes the helpful assistant in car owner's daily life, I believe the business model for automotive software will shift to charging users by mileage or on a subscription basis, just like pricing models of Microsoft 365 and other software products today.

至于ARM+Android的IP授权模式，我们多年来持续向多家世界级的汽车客户授权技术，支持有实力的车企客户自研。其中包括了世界最大的新能源车企、世界最大的零部件集团之一、中国最大的合资车企。这个商业模式的粘性很强，会让客户的产品深深地根植于地平线的技术生态，一代一代产品都持续与我们合作。就像我们和大众的合资公司酷睿程，以及刚刚提到的全球最大的汽车零部件集团之一，在2026年上半年都持续从地平线去授权算法、软件和服务，依然是我们的最大的客户之一。我一直觉得技术和生态两个维度同等重要。在技术上面，我们要追求的是通破天花板的能力，做到行业顶尖，引领行业的计算的标准。在生态上，我们坚持商业模式开放灵活、全方位利他，让合作伙伴能共同受益、共同成长。伙伴愿意跟着你走，是因为你能在芯片、软件工具链上面真正赋能他们。这样的话才能一起的共同成长，把蛋糕做大。 Then

for our ARM plus Android IP licensing model, and over the past many years, we have continuously licensed our technologies to a number of world-class automotive customers, supporting capable customers in their in-house development efforts. These include the world's largest new energy vehicle maker, one of the largest auto parts groups in the world, and China's largest JV automaker.

This business model has very strong stickiness, which roots our customer's products deeply in Horizon's technology ecosystem. They keep working with us for products generation after generation.

For example, CARIZON, our joint venture with Volkswagen, and the aforementioned one of the world's largest auto parts groups, have both continued to license algorithms, services, and software from Horizon in the first half of this year and they remain among our top customers. I have always believed that the ecosystem should be all the strong because we believe that we have very strong technologies, which can help us to play a leading role.

At the same time, equally important is ecosystem, which should be open and also helping our partners in order to achieve an all-win results. Then partners are willing to work with you because you can genuinely empower them across chips, software.

Then this is the only way we can grow the pie together.

而且在汽车业务以外，我们发现基于自动驾驶研发的底层技术有很强的适用性和延展性。它可以在终端实现各种各样的生动声色的具体应用。比如我们帮助具身智能公司赋能了各种各样的人形机器人、轮式机器人、四足机器人等等，帮助家庭智能体公司实现家里面的贾维斯成为真正的家庭助手，帮助无人物流车公司提升物流效率等等。举个更具体的例子。我们已经通过重要的联营公司地瓜机器人赋能了超过100种下游机器人品类、400余家客户，并且在中国的具身智能的阵营中覆盖了半数以上的具身智能的企业。从地平线聚焦于BPU和AI基础模型，从汽车到机器人，通过ARM+Android模式对外授权。地平线和地平线的关联子公司孵化的公司地瓜机器人，已经一跃成为国产机器人芯片加模型底座的最大公约数。我们的开源具身模型目前已经被英伟达、Meta、微软、字节跳动、阿里巴巴、宇树科技、清华大学、浙江大学、斯坦福大学等超过100家学术机构和企业测试或使用。半年内我们还将发布融合世界模型和通用理解模型的下一代大脑小脑模型。

Beyond automobiles, we have found that the underlying technologies developed for vehicles have strong applicability and extensibility. They can enable all kinds of vivid, real-world applications on terminal devices.

For instance, we have helped embodied intelligence companies empower humanoid robots, wheeled robots, and also the quadruped robots, and more; helped home intelligent agent companies create a household Jarvis that becomes a true home assistant; and also we helped unmanned logistics vehicles improve logistics efficiency. To give you a more concrete example, through our associate, D-Robotics, we have already empowered over 100 downstream robot categories, serving over 400 customers, and covering more than half of the embodied intelligence companies in China.

Horizon Robotics focuses on the BPU and AI foundation models from automobiles to robots, licensing them externally under an ARM plus Android model. D-Robotics, meanwhile, faces customers across thousands of industries to achieve commercial monetization.

Horizon Robotics and D-Robotics have rapidly become the greatest common denominator of domestic robot computing SoCs plus model foundations. Our open-source embodied models have so far been tested or used by more than 100 academic institutions and companies, including NVIDIA, Meta, Microsoft, ByteDance, Alibaba, Unitree Robotics, Tsinghua University, Zhejiang University, and Stanford University.

Within six months, we will also release the next generation of brain and cerebellum models that integrate a world model with a general-purpose understanding model.

展望2026年下半年，我们有信心在核心客户的入门级AD车型平台上达到七成到八成的份额，并并在他们的高级高阶的城区的NOA的车型上面完成大算力芯片和HSD的量产交付。基于星空芯片和卡卡鲨舱驾一体解决方案也会在今年的四季度如约量产，保持我们持续的在行业里面最快的从芯片推出到整车量产的纪录的速度纪录。 Go ahead.

Go ahead, please.

Looking ahead to the second half of 2026, we are confident of reaching a 70 to 80% share on our core customers' entry-level AD vehicle models, and of completing mass production and delivery of our high-compute SoCs plus HSD on their urban NOA models. The cockpit-driving integrated solution based on the Starry chip and Kakashark will also enter mass production as scheduled in the fourth quarter of this year, maintaining our record from the very initial design to the mass production in a very quick, in a very few months.

很多朋友关心征程7芯片的进展，这是我们面向L3和L4级别自动驾驶场景的战略级产品。在技术层面，征程7在架构设计上面实现了多项的关键突破，不仅算力大幅提升，更重要的是针对我们下一代更大参数量的自动驾驶模型、HSD模型，以及本地化面向座舱区部署的大语言模型都做了原生性的优化。目前研发进展顺利，预计2027年推向市场。即使现在产品还处于研发阶段，有意思的是，征程7已经在市场端获得了巨大的关注。多家头部车企和Tier 1主动和我们接触，表达了要首发量产的合作意向。这从侧面印证了市场对这颗行业最先进的智驾芯片的高度期待。我们也有信心和所有的合作伙伴一道，把征程7打造成自动驾驶芯片全世界的新标杆。

Many friends are interested in the progress of Journey 7 SoC, which is our strategic product targeting L3/L4 autonomous driving scenario. On the technical front, Journey 7 has achieved several key breakthroughs in architectural design.

Not only has its computing power increased substantially, but more importantly, it has been natively optimized for the local deployment of our next-generation HSD intelligent driving model with a larger number of parameters, as well as our local LLM cockpit model. R&D is currently progressing smoothly, and we expect to complete tape-out early in the second quarter of next year.

Even though the product is still in the R&D stage, Journey 7 has already attracted tremendous attention in the market, with a number of leading automakers and Tier 1 suppliers proactively reaching out to us to express their interest in collaboration. This indirectly confirms the market's high expectation for this SoC, which represents the most advanced level in the industry.

We are confident that together with these partners, we can make Journey 7 a new world benchmark for autonomous driving chips.

除了针对L3、L4级别自动驾驶场景的芯片的研发布局，我们今年内也携手一家领先的科技平台在若干城市进行L4级的Robotaxi的实验项目，进一步验证并拓展地平线软件层面在高阶自动驾驶场景的应用潜力。其实我们认为当自动驾驶技术真正做到L4级别以后，究竟是私家车载着你从A点到B点，还是无人出租车载着你从A点到B点，其实已经不重要了。因为在L4这个层级上面，底层技术是同源的。

Actually, in terms of L3 and L4 SoC, actually we are also working with another partner to test our L4 Robotaxi. We hope that we can continue to test the potential of the application of this advanced intelligent assisted driving software.

Actually, when it comes to L4, whether it is a private car taking you from A to B or it is a Robotaxi taking you from A to B, it doesn't matter at all, because in terms of the underlying foundation, they are coming from the same source.

即便行业环境依然充满挑战，但是基于我们对下半年业务局势的判断，我们有信心实现50多亿的全年收入，并构建一个非常强大的明年的pipeline。使得从2027年开始，就是明年，地平线的高阶智驾芯片加上地平线支持的我们IP授权支持的车企的自研的高阶智驾芯片，直接份额加延展的份额将做到中国高阶智驾智驾芯片市场的第一名。

Even though the industry environment remains full of challenges, based on our assessment of the business landscape in the second half of the year, we are confident of achieving full-year revenue of more than 5 billion RMB. We will also build a solid pipeline for next year.

Starting in 2027, that is next year, Horizon SoCs together with the automakers' in-house chips supported by Horizon IP, which means the direct plus our extended market share, will collectively rank the first in the market share in the advanced segment. 当地平线的高阶智驾芯片加上地平线延展的支持的高阶智驾芯片取得市场份额第一的市场地位以后，那么在其上的整个应用软件生态，那地平线将会有强大的影响力。

Here you can imagine that Horizon's own advanced SoC plus the extended market share, which is the automakers' in-house chips supported by our IP, and also coupled with our applications going forward, and the entire ecosystem put together, which means that Horizon Robotics will be even more influential in the market.

我们11年前从一个很小的想法出发，走到今天，证明了这条路能走通。前面的路还很长，我们还是要继续地保持谦逊、开放、利他，继续地把该做的事情做扎实。谢谢大家。下面我们请地平线CFO王雷来给大家简述一下2026年上半年业绩更细节的数字情况。 All right.

Thank you, Dr. Yu, and Xuan.

I will now-- Sorry, go ahead. 11 years-- [laughter]

11 years ago, we set out from a very small idea, and having come this far today, we have proven that this pathway works. The road ahead of us is still long, we need to remain humble and open-minded and altruistic, and continue to do what needs to be done in a solid and thorough way.

Thank you all. Next, we would like to invite our CFO, Mr.

Wang Lei, to give us a brief overview of more detailed numbers for the first half of 2026. The floor is yours.

Thank you. Thanks to the translator.

Thank you, Dr. Yu and Xuan.

I will now present the key financial results for the six months ended June 30, 2026. To be mindful of the length of this call and focus on main items in our discussion today, I encourage participants to refer to our earnings release, which was posted on our investor relationship website for additional details.

Back in July, we published the preliminary range of our first-half financial results. I'm pleased to announce today that both of the revenue and gross profit for the reported period hit upper ends for the guidance that were previously provided.

Specifically, for the first half of the year, we recognized revenue slightly less than RMB2.1 billion, representing a year-over-year growth of 32.9% compared to the same period last time. Our gross profit for the first half came in around 1.4 billion RMB, also reflecting year-over-year growth of 32.9%.

This implies that, consistent with the same period last year, we remain and sustained a robust gross margin of 66%. Our revenue is composed with revenue from product solutions plus license and services.

Despite a 20.2% year-over-year decline in domestic passenger vehicles retail in the first half of the year, our shipment volume bucked this headwind, growing by 12.1% to reach approximately 2.2 million units, driven by our gains in the market shares. As a result, revenue from our product solutions reached RMB225.6 million, representing year-over-year growth of 14.8%.

Turning to our license and service revenue, CARIZON and one of the world's largest automotive parts company continued to license our algorithms and software, and both remain among our top five clients during the period of time. Our license and service revenue reached RMB1.1 billion, representing year-over-year growth of 52.7%.

As we mentioned earlier, our overall gross margin stayed at 66%. Breaking this down by segments, gross margin for product solutions was 36.2%, while license and services delivered a gross margin of 90.4%.

As we shared with you before, in the near term, we have been bundling the sales of domain control units together with our core products in order to accelerate the mass production ramp-up for HSD at early clients. This bundling of non-core products has has had a dilutive impact on our gross margin.

However, this is a one-time impact. Excluding this one-time effect, the actual gross margin for our product solutions was 48.1%, an improvement of 3 percentage points compared with the same period last time.

Now, let's turn to major expenses. The R&D expenses, including SBC, came in slightly more than RMB2.7 billion.

Total combined SG&A expenses were RMB634 million. Together, three expenses amounted to RMB3.3 billion, increased of approximately RMB650 million compared with the same period last year.

Essentially, the vast majority of this increase was driven by the cloud-related training costs. This resulted in operating loss of RMB1.7 billion.

If we add back non-cash items such as SBC, we arrived at an adjusted operating loss, under non-IFRS measurements, of RMB1.3 billion. Finally, let's turn to the bottom line.

As we noted in our previous early pre-announcements, the fair value change arising from the CARIAD convertible loan resulted in a substantial gain for us. Under IFRS, the company reported profit of RMB3.8 billion for the reporting period.

If we exclude the fair value change from this convertible loan, adding back the impact of SBC, and deduct the one-time impact from the deconsolidation of D-Robotics, our adjusted net loss came to RMB1.67 billion. Now, this concludes our management's presentation.

I will now turn the call over to the operator to facilitate Q&A sessions. Please.

管理层你好，感谢接受我的提问。我是Morgan Stanley的Tim。我有两个问题，第一个问题想请问一下就是下半年这个项目跟产品。因为今年我看公司这边芯片出货会集中在下半年。那我想刚才这个余博士在简报里面稍稍提到了，就是在主要头部新能源客户中会有这个走量的平台。想问一下就征程6M跟6P，就是这个中阶、高阶哪一个级别的平台收入可能增长会更加突出？然后往下半年去看的话，除了季节性这个增长之外，就这些新增定点还有供应链这个地位提升或者其他方案的这个替代是不是会带来带来较为显显著的一个增量？ So my first question is about the project wins into second half, because with the chip shipment weighted to second half this year, could management share a bit more details about project wins at the major EV OEM platforms that management just briefly touched on?

In the meantime, could you also double-confirm that which product chains would draw would drive stronger revenue growth? Would it be like a Journey 6M mid-range or the high-end 6P?

Any additional colors about the upside coming from the new design breakthroughs would be highly appreciated. That's my first question.

好，谢谢，谢谢Tim。如刚刚跟大家汇报的，我们今年下半年会在头部的车企新能源车企它的这个入门级的这个AD的平台，那么会显著的放量。所以的话，J6M将会成为一个从这个volume来讲的话呢，会是一个显著的迅猛的一个增长。那么销量增长只是我们收入增长的一部分，进一步驱动我们收入实现高增长的，还有我们就是单价更高价值的产品，大算力的征程6芯片，征程6P和征程6H，以及这个HSD的软件解决方案。那么这个的话呢，其实也会关联到我们新增的定点。那么我们的大算力的芯片跟这个HSD 600，其实最近有非常振奋人心的进展。我们近期的话呢，又突破了两家中国自主品牌的头部车企的定点。那么我们的HSD现在可以说已经覆盖了中国销量最大的头五家车企，以及销量最大的头两家合资车系。那么可以说几乎覆盖了这个市场里面可以说是最重要的这个客户群。那如果你去看整个行业，能同时在自主品牌的龙头，top

5的龙头企业和合资品牌的这个top 2的这个车企上面，在这两条线上门都能做到这样的一个定点和布局，我觉得在行业里面应该没有第二家能够做到。地平线是独一无二的这样一个地位。那么这些定点的话呢，从今年年底会开始这个量产。比如说我们其实看到刚刚发布的大众的ID.ERA 6，也是市场上这个大家都非常看好的一个热门车型。其实就是地平线的这个大算力芯片，然后这个HSD赋能的软件方案。那么领头羊加持的话呢，我相信会带给我们不仅说是今年年底，包括明年以及未来更长周期会带来显著的收入贡献。谢谢。 Thank you very much for

Thank you very much for your question, Tim. As we just reported to everyone, in the second half of this year, we will see the volume ramp-up on the entry-level AD platform of a leading NEV company, and the J6M will be the main driver in terms of the volume.

Fast volume growth is only part of our revenue growth. What further drives our high revenue growth is our higher-value product, high-compute Journey 6 SoC, and also 6P in particular, and also 6H, and also our HSD software solution.

It is also related with our new design win, and with high-compute chips plus an HSD 600, there has been exciting progress. We recently made breakthroughs with two leading Chinese automakers.

Our HSD platform now covers the top five Chinese domestic brands and the top two joint ventures, covering almost all of the highest-volume customer groups in this market. If you look at the entire industry, I believe no other company can achieve such positioning across both the leading domestic brands, top five, and the joint-venture giants, top two, at the same time.

Horizon is very unique. These design wins will begin mass production at the end of the year, and we expect them to deliver significant revenue contributions next year.

According to the recent media reports, you can see Volkswagen has launched their newest model, ID.6, which is also equipped with our chips and also HSD software solutions. If we put them together, this year, next year, and also going forward, we will see significant ramp-up of our revenue.

Thank you.

感谢管理层。我的第二个问题是有关投报投报率跟盈利能力。因为考虑到公司2026年可能会进一步加码车端与云端大模型训练以及下一代芯片的研发。那我们如何平衡这个投入产出？预估何时我们的盈利能够见到一个比较明显的改善？ My second question is about the return on investment and also profitability as Horizon stepped up 2026 budget for higher spending, um, both on vehicle and cloud-side large model training and the next-generation chip development.

How would you balance the return on investment, and when should we expect a visible inflection in Horizon's profitability? That's my second question.

Thank you.

好的，谢谢Tim。关于投入的问题，我想跟大家分享一下我们的思考和逻辑。首先的话，我们对于我们的产品，我们会保持比较高的毛利率水平。这样的话呢，一方面它体现我们在市场里面的这个竞争地位，另外一方面的话就是为后续的高强度的投入啊留下足够的空间。这是我们持续投入能力的根基。没有这个根基，后面的长期投入都是空话。那么自驾行业其实是一个长周期的行业，那么它的每一代的产品，从投入研发到最后量产落地，其实都要好几年的时间。今天我们看到的收入，就是我们刚刚汇报的2026年上半年的收入，它实际上一定是过去几年持续的研发投入所种下的果实。换句话说，财报反映的其实都是过去投入的兑现结果。但是真正决定未来的我们的业绩报表的上限的，是我们今天在研发上面的投入强度。如果今天投入不足，那我们后续增长的天花板就被限制住了。正因如此，我们还会继续保持高强度饱和式的投入节奏。我们对于未来锁定最终的胜局，不能抱有任何的侥幸心理，不能去假设他轻松的投入就能够去成为未来智能驾驶的王者。所以最终的胜利是靠持续扎实的投入一点点堆出来的。我们认为这不这对于一个每年就是说轻轻松松的研发，比如说这个没有高强度的研发投入，就能够达到终局的行业。反过来看，如果这个行业真的是靠每年这个不高的研发投入，比如说每年10亿、20亿的研发投入，就能够实现终局的话，那地平线也不需要从事这个行业了。因为这个行业的门槛其实并不高。所有有一定规模以上的车企其实都能够轻松的去完成这样的研发投入，那么这个行业就成了一个低天花板的行业。所以正是因为自驾行业长坡厚雪，所以要取得最终的胜利，需要有长跑的耐力，长期投入的实力，持续迭代的能力。那么也就是说，地平线如果今天我们不去投入高强度的研发，其实并不会影响地平线今年的收入规模。那么从P&L角度来讲的话，也许会比较好看。但是由于这是一个高速增长的这么样的一个技术的领域，所以地平线必须要为未来去持续的投入。那我给大家举个例子，假设我们做到100亿的收入，60%的毛利率，那么也就是60亿的毛利。在这个体量上面的话，假设我们把这些毛利都投到芯片和中家大模型的研发上面去，那么大家可以看到，这样的一个研发的投入的体量，其实已经超过了绝大部分车企都能够承受的研发投入水平。那么这个到底是好事还是坏事呢？那是好事。因为当车企的话的研发都不能承受的时候，那么实际上这些智能驾驶的技术产品都会交给地平线这样的独立第三方来做。那么其实这个对地平线是有利的。那么关于这个盈利的改善，我们还是坚持我们过去的这样定目标，就是在2028年左右实现盈亏平衡。这是我们综合考虑了收入增长节奏、毛利率水平和研发投入的强度之后，认为可以实现的时间点。好的，谢谢。

We're expecting the translators.

Thank you once again for your question, Tim. Regarding the investment and we will continue to maintain our gross margin level, and our logic and our observation is that we need to continue gross margin at a very high level in order to keep our competitive edge, in order to lay the foundation for continued high investment going forward.

Because this industry is a long-cycle industry, and now we have received our H1 revenue. That is thanks to our R&D investment and also our innovation.

Today's revenue is the fruit by investment made over the past few years. The financial statement reflects the realization of past investment, while R&D investment determines the ceiling of future statements.

We must maintain high intensity and a saturated investment, and we have no illusion of luck when it comes to securing ultimate victory in the future. Because if we do not make a high level investment, in the future we will be restricted.

We will keep very high level and a saturated investment in the long term. It is not an easy job.

If we do not make a very high investment in R&D, for example, if we only make investment of just 1 billion or 2 billion in R&D, we cannot be a very high bar or high threshold in this industry. It also means that it is going to be a very easy game for everyone.

That's why we believe that our industry has a very steep tech of technology curve, so that requires resilient strength and iteration. For example, with um R&D, um so if we only have low on R&D, maybe it is not going to affect our P&L, maybe it is very good-looking in the short term.

Actually, this is a very fast-moving technology, we need to make investment for our future. For example, with RMB10 billion in revenue and a 60% gross margin, that means uh 6 billion in our gross profit.

If we make all these gross margin into our R&D for our SoC chips and also AI, then it is going to exceed that of the vast majority of other automakers. Is it good or bad?

Of course it is good for us, because if other automakers cannot bear it, then it means that the technologies will be dominated by the independent third party, like Horizon. That is beneficial and conducive for us.

This is the first part of our investment. Secondly is about how to make improvement on our profitability.

We would like to align this objective like before. We will continue to work hard to reach the breakeven about 2028.

Thank you. 感谢管理层。

好，谢谢。谢谢Tim。

好，感谢管理层，余博、雷总，我是宇琛。两个问题。第一个关于HSD，我们看到核心大客户的这个进展感觉都还挺顺利的。那我们接下来应该怎么理解26和27年的出货规模，接下来放量最关键的里程碑是什么？ 第二个问题想请教关于J7，如果SOC预计明年二季度完成流片的话，从流片到客户SOP再到规模化放量，时间线和收入贡献节奏大概怎么理解？

Two questions. First on HSD, with strong major OEM progress, how should we think about the HSD volume in '26 and '27, and what are the key milestones for that ramp?

Second question is on Journey 7. Scheduled SoC tape-out completed scheduled in second quarter next year, how should we think about the timeline and the contribution from tape-out to a customer SOP and the meaning for revenue and volume ramp?

好的，谢谢丁总。随着HSD的逐步量产，以及技术表现力的迅速提升，那么HSD已经成为地平线面向整车客户的一个核心的行业名片，市场的认可度持续提升。在国内自主车企阵营，我们已经实现了前五大销量最大的自主品牌车企的全部的定点突破。那么这个其实覆盖了自主品牌的绝大部分的这个销量，完成了主流本土头部车企的全覆盖。其中在某新能源头部车企，我们实现了从入门级的AD到城区高阶AD平台的全覆盖，并从今年年底开始逐步放量。同时在合资品牌领域，我们也前面也说成功拿下了top 2的两家合资车企的核心定点项目，核心项目的定点。那么客户的结构的话呢，也进一步的拓展和多元化，不再单纯的依赖本土的自主品牌，那么合资市场也从而实现了关键的落地。我想我可以给一下就是说关于我们芯片的这个出货量的一个大概的指引。我们今年的话应该会突破500多万颗芯片的出货。那么我们有信心明年的话呢，出货量会达到700万颗左右。关于这个征程7—— 700万左右。 哦，对。嗯，好的。

Thank you for your question. Regarding the HSD, we can see that with the gradual mass production of HSD and rapid improvements in its technology performance, HSD has already become Horizon's core industry name card for OEM customers, and the market reputation continues to rise.

With the domestic Chinese OEM brands, we have already achieved design win breakthroughs with all the top five domestic automakers, covering the most the bulk the majority of the domestic brand sales volume, completing full coverage of mainstream leading local automakers. Actually for one of the leading NEVs, actually we have already covered their entry-level AD and also their urban AD.

They have already entered the mass production step by step. At the same time, in the JV segment, we have also successfully secured core projects design wins with two leading JV automakers, as mentioned the top two.

Further diversifying our customers mix so that we no longer only relying solely on local domestic brands. We have achieved a key breakthrough in the JV market.

Then in terms of chip shipments, I can give you a little bit of guideline. This year, we are going to achieve around 5 million units this year, and also next year, we try to hit more than 7 million units next year.

Thank you.

关于征程7芯片的这个量产的进展，那目前呢它处在研发阶段，进展顺利，预计2027年推向市场。那么即使现在产品还处在研发阶段，征程7在市场端已经获得了巨大的关注。多家头部车企和Tier 1已经主动和我们接触，都表达了要首发量产的这个争取做首发量产的合作意向。这个从侧面印证了市场对这个行业最先进的水平的芯片的一个高度期待期待。我们也有信心和合作伙伴一道，把征程7打造成自动驾驶芯片的这个世界级的标杆。谢谢。 Then

regarding the second question, currently the R&D is moving forward very smoothly, it is expected to be launched next year in 2027. Currently, and it has already attracted great attention from the leading automakers and also Tier 1 suppliers.

It is highly expected, and also we are very confident that we can work with our partners to launch J7. Also some of them are going to launch their model in their debut models.

We hope that we can build it as the world benchmark in the advanced market. Thank you.

感谢余总。 好，谢谢丁总。谢谢。

谢谢大家。

Thank you. Due to time constraints, we will now conclude today's call.

Thank you again for joining us today. If you have any further questions, please feel free to contact our IR team.

Thank you.