Horizon is preparing to go head-to-head with Tesla; countdown to mass production of Journey 7.
```
Author | Zhou Zhiyu
Horizon is beginning to stockpile ammunition for the next battle.
Wall Street Insights has learned that Horizon is accelerating the development of its next-generation intelligent driving chip, the Journey 7 series, based on a new Riemann architecture. Its performance aligns with Tesla’s AI5 and is planned for release and mass production next year. Meanwhile, Horizon will also launch an integrated cockpit-driving solution, expected to be released during this year’s Beijing Auto Show.
The urgency of these two moves is evident. The 2025 report card released on March 19 also confirms this: revenue of 3.76 billion yuan, up 57.7% year over year, and shipments of NOA chips surged nearly fivefold. The logic of simultaneous quantity and price increases has been fulfilled, but Horizon’s CEO Yu Kai devoted more attention at the earnings conference to the future. The reason is simple: Journey 6P is still ramping up, with an ASP of only about $60; releases of high-value products are just beginning.
The industry’s changes over the past year have been quicker than Horizon’s own. Nvidia’s Thor enters vehicles, Nio and Xpeng’s self-developed chips go into mass production, and Qualcomm is expanding from cockpit to intelligent driving. Intelligent driving chips are evolving from the traditional “chip company supplies, automaker purchases” single model into a dual-track structure where automaker self-development and third-party suppliers coexist. Horizon needs to seize the window of the next tightening cycle.
Holding the Cards
Wall Street Insights has learned that Horizon internally has made the integrated cockpit-driving intelligent solution its most important strategic product of the year. April’s Beijing Auto Show will be the first public reveal for this plan.
Horizon already leads domestic chips in the intelligent driving domain—holding 47.7% market share for ADAS in domestic brands for top position, and 44% in the NOA market for models under 200,000 yuan. But sticking to intelligent driving alone has its ceiling. In 2025, total chip shipments will exceed 4 million, with 1.8 million mid- to high-tier chips supporting NOA—nearly five times last year, but only 45% of total shipments.
Yu Kai did the math himself: The $60 ASP has ten times the upside compared to the average price of urban NOA products. Shipments are rising, but the share of truly high value-added products remains low.
The cockpit domain is another world. Qualcomm monopolizes with over 70% market share; Snapdragon 8295 is almost synonymous with Chinese cockpit chips, and the next-gen Snapdragon 8797 already has Leapmotor’s D series and SAIC-GM. In recent years, several domestic companies tried a direct challenge to Qualcomm, all failed.
Yu Kai also admitted at the earnings meeting: "We have no intention of hand-to-hand combat with Qualcomm in the mature cockpit market."
What he aims to do is change dimensions—combine intelligent driving and cockpit domain controls into one chip, using “chip + intelligent driving software + intelligent OS" full stack capability to target a joint area unreachable by both Qualcomm and Nvidia.
Qualcomm’s weakness is in intelligent driving software ecosystem; advanced intelligent driving solutions largely rely on third-party companies like Yuan Rong and Momenta for algorithms, lacking its own complete software stack. Nvidia’s weakness is in mass production of integrated cockpit-driving—they have pushed a joint-project with a luxury overseas carmaker for years with no mass production yet. Horizon believes it is currently the only player able to fully connect chip, intelligent driving software, and intelligent OS.
Yu Kai’s view is that entering cockpit from intelligent driving is a “more disruptive” strategy than the opposite.
Integrated cockpit-driving is becoming a clear industry trend. Tesla’s FSD+Grok shows the possibilities of fusing cockpit AI and autonomous driving. Li Auto and Xpeng have merged their intelligent driving and cockpit teams organizationally. In the low-end market, ArcFox under BAIC has equipped models priced at just 100,000 yuan with integrated cockpit-driving solutions.
Industry insiders told Wall Street Insights that if current intelligent driving suppliers don’t secure a place in this trend, they may lose both intelligent driving and cockpit markets in the future.
The cost calculation can stand. Integrated cockpit-driving can eliminate wiring harness, cooling, PCB area, domain control casing and other hardware costs. Yu Kai specifically mentioned the surging price of memory chips this year—just memory alone could save automakers thousands of yuan. Amid industry-wide cost reduction, it’s a calculation that works out.
However, cockpit is after all Qualcomm’s home turf, cultivated over five product generations. Snapdragon digital chassis covers over 350 million vehicles globally, with mature multimedia ecosystem and toolchain—not something one new chip can close the gap on.
Distant Battles
Journey 7 addresses the next cycle’s challenges.
Wall Street Insights learned that Journey 7 is progressing smoothly, with high internal confidence about matching Tesla’s AI5 performance. Previously, Yu Kai also stated to Wall Street Insights that Journey 7 has “very strong advantages” compared to Tesla’s next-gen AI5 in specifications.
Unlike Journey 6, mainly led by the chip team, Journey 7’s product definition is driven more by the algorithm team. In-car models are evolving quickly, expanding from millions of parameters to tens of billions—the chip must reserve space for model iteration over the next two to three years. Having algorithms define chips is itself a signal.
Competitors are visible. Chip computational demand is moving from hundreds of TOPS toward thousands. Horizon’s joint venture with Volkswagen, CoreTron, previously revealed that their C7H chip (based on Journey 7 architecture) will deliver 500-700 TOPS per chip, using 3-4nm process. By Horizon’s product naming, the superior C7P should have at least 1000 TOPS.
Moreover, industry focus is shifting. Stacking peak computing power is no longer the only benchmark; memory bandwidth, data transfer efficiency, inference accuracy might be the real bottlenecks for deploying large models in cars. Tesla’s FSD uses 144 TOPS on HW3.0 but remains a benchmark for efficiency—deep software-hardware collaboration is more critical than pure computing capacity.
The time window is narrow. Musk says AI5 will tape out soon, and mass production in 2027. Journey 7 is aiming for the same timeframe. Before that, Horizon must use the Journey 6P+HSD’s mass production to demonstrate software iteration capabilities.
Financial reports show Horizon’s R&D expenses in 2025 reached 5.15 billion yuan, up 63.3% year over year and far higher than revenue of 3.76 billion—Yu Kai’s stance is not afraid of R&D investment, as HSD is not only the core for winning urban intelligent driving, but also the foundation for L3/L4 technology.
HSD only began mass production last November, with 22,000 sets shipped by year-end; this year’s goal is 400,000. Models with HSD as core configuration have the top-spec version accounting for 83% of sales, breaking the industry norm of “low-spec accounts for volume, high-spec accounts for style.” Over Spring Festival, user intelligent driving mileage share reached 41%—close to the 50% threshold where machine-driven mileage surpasses human. These numbers show consumers are willing to pay for useful intelligent driving, but high-value products are still limited to a few models; whether this can expand from several cars to dozens is key this year.
Yu Kai’s MPI roadmap is: this year increase from tens to hundreds of kilometers, next year to thousands, with some regions surpassing 10,000 kilometers and meeting L4 Robotaxi operational requirements.
In the second half, Horizon plans to launch Robotaxi pilot projects with ecosystem partners. This path is quite similar to Tesla: FSD has already reached over 1,000km integrated MPI in the US, with some regions having Robotaxi tests underway. Horizon’s pace is about one year behind, but the direction is the same.
Who Stays at the Table
A larger change is underway.
Competition in intelligent driving chips is evolving from “chip companies competing” to a multi-party battle. On one hand, Nio, Xpeng, and Li Auto’s in-house chips are successively in production, BYD and Xiaomi also have plans. Former clients are now developing their own capabilities and the space for third-party chip companies in the high end is shrinking.
On the other hand, algorithm companies are extending into the chip domain. The first auxiliary driving chip from Momenta’s chip venture, Xinhangtu, is expected to be mass produced for vehicles this year, securing SAIC Volkswagen and Beijing Hyundai as clients. In the future, Momenta’s designated clients may shift to using this chip.
Momenta founder Cao Xudong predicted “intelligent driving competition will end by 2026, with only three participants winning." Huawei Qian Kun is already solid in the high-end market, while Horizon and Momenta are just beginning their head-to-head competition in the under-150,000 yuan segment.
Horizon’s advantage is its unique ecosystem. It is currently the largest domestic intelligent driving platform, with most shipments and cooperation with carmakers; over 95% of deliveries are via ecosystem partners. But this open ecosystem also comes with its tension. As HSD extends from chips to software, Horizon does both hardware and software, its role has blurred. Previously, Journey 6P was exclusive to HSD; now it’s open to partners like Yuan Rong and Bosch—more partners, but some are also reaching out to Nvidia and other chip companies.
Going overseas is another variable. Horizon has secured mass production designation from 11 automakers and over 40 exported models, plus designation from 3 international automakers for markets outside China—full lifecycle shipments have reached 10 million sets. But Momenta entered the global automaker system with “globalized algorithms” from the outset, covering over 160 models including GM and Nissan. Overseas competition is also accelerating.
Yu Kai revealed that income growth forecasts for the next few years are raised from 50% to 60%, backed by robust pipelines and ramp-up in Journey 6 series production. But he’s aware that what Horizon faces isn’t just a product race, but an ecosystem and timing competition as well.
A fact that cannot be ignored is: Horizon’s R&D expenditure for 2025 is 5.15 billion yuan, up 63.3% year over year, with an adjusted operating loss of 2.37 billion. In chips, converting R&D to product competitiveness takes three to four years, but intelligent driving model upgrades are every six months. The battle for Journey 6 isn’t over, yet Journey 7’s development must proceed full speed. The resulting pressure from fighting on two fronts isn’t easy for any company.
The intelligent driving chip table is taking shape. A period of lag may mean a structural gap. Horizon must accelerate its sprint to maintain its hard-won domestic lead among intelligent driving solution providers.
Risk Warning and DisclaimerThe market has risks, investment must be prudent. This article does not constitute personal investment advice, nor does it take into account individual users' special investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their circumstances. Investing based on this is at your own risk. ```