China Automotive Systems, Inc. (Nasdaq: CAAS)

Deep Dive Research Report

Sector: Consumer Cyclical / Auto Components Report date: 20 August 2026 Listing: Nasdaq Capital Market, ticker CAAS. Incorporated in the Cayman Islands since 11 September 2025 (previously Delaware). Files as a foreign private issuer on Forms 6-K and 20-F.

A note on this company's reporting cadence, because it changed mid-stream

CAAS was a US domestic filer reporting quarterly on Forms 10-Q until it completed a redomiciliation merger into a Cayman Islands entity effective 11 September 2025 (PR Newswire, 11 Sep 2025). It reported Q3 2025 on 12 November 2025, then moved to semi-annual reporting from 2026, a change CFO Jie Li confirmed on the FY2025 call. The most recent reporting period is therefore H1 2026, released 13 August 2026 with a conference call the same morning (Form 6-K, 13 Aug 2026). The six most recent reporting periods used throughout this report are:

1. What the company does

China Automotive Systems makes the parts that let a driver turn a car. Specifically, it makes steering gears: the mechanical and electromechanical box that sits between the steering wheel and the front wheels, converting a light turn of the wheel into enough force to swing 1,500 kilograms of vehicle around a corner. It also makes the pumps, hoses, columns, sensors, electric motors and control units that go with them. It sells almost all of this to vehicle manufacturers as original equipment, not to consumers.

The company is headquartered in Jingzhou, Hubei Province, and produced roughly 10.8 million power steering gears in 2024, up from 9.8 million in 2023 (FY2024 10-K summary). It supplies more than sixty vehicle manufacturers, including BYD, Geely, Chery, Changan, SAIC, FAW Group, Dongfeng, Beiqi Foton, Shaanxi Automobile Heavy Truck, Stellantis and Ford.

The founding story, and why it still shapes the company

The business did not start as a US-listed company. According to the company history compiled by Encyclopedia.com, Hanlin Chen established Ji Long Enterprise Investment Limited in Hong Kong in 1992 as a holding vehicle to build automotive component joint ventures on the mainland. The first operating venture was Shashi Jiulong Power Steering, which the FY2024 10-K dates to 1993 and which built integral recirculating-ball steering gears for heavy trucks. Output in the mid-1990s was roughly 200 sets a year, because China barely had a car industry to sell into.

The pivot came in 1997, when the group established Jingzhou Henglong Automotive Parts Co. Ltd. to make rack-and-pinion power steering for passenger cars and light vehicles, initially held 44.5% by Ji Long, with Hanlin Chen named chairman of Henglong. That is the structural decision that still defines the company: two parallel businesses, one built around heavy commercial vehicles and one around passenger cars, using genuinely different steering technologies, sharing a group R&D base and a supply chain.

The US listing came through a reverse merger in 2003, when Great Genesis Holdings merged into Visions-in-Glass, Inc., a dormant American shell that had previously distributed stained glass over the web. The combined entity was renamed China Automotive Systems and listed on Nasdaq. The company opened its first US office in Detroit in 2007, and its first overseas factory in São Paulo, Brazil in 2012, a fact management repeats often when it talks about South America (PR Newswire, 17 Dec 2025).

Three later decisions matter for understanding what the company is today. In 2021 it bought 40% of Sentient AB, a Gothenburg steering-software and steer-by-wire firm founded in 2009 (PR Newswire, 2021). In 2017-2018 Chairman Hanlin Chen, together with Wiselink Holdings Limited, his wife Liping Xie, and Chariot Company (Cayman) Limited (an affiliate of North Haven Private Equity Asia IV), proposed to take the company private at $5.45 per share, then withdrew the proposal in August 2018 citing market conditions (PR Newswire, 2018). And in September 2025 the company redomiciled to the Cayman Islands, a move CFO Jie Li told investors saves roughly $500,000 a year in listing and compliance expense.

The core value proposition

An automaker choosing a steering supplier is buying three things at once: a safety-critical mechanical component that must not fail over fifteen years, a piece of software that increasingly defines how the car feels, and a supplier relationship that will survive the four-to-six-year development cycle of a vehicle platform.

CAAS's proposition is that it can deliver all three at Chinese cost. Its historic edge was manufacturing hydraulic steering gears cheaply at very high volume for a domestic market that was growing 20% a year. Its current proposition is different and more interesting: it now argues it can deliver electronically controlled steering compatible with L2+ assisted driving at the same cost position, which is why electric power steering rose from 33.8% of sales in 2023 to 46.8% of sales in H1 2026.

The technical nature of the product

Steering is one of the few automotive systems where a failure is immediately catastrophic. That imposes a specific structure on the industry. A steering gear must survive several million cycles of load reversal without developing play. It must not seize. In an electric system, the assist motor and its electronic control unit must fail gracefully, so that if the electronics die the driver retains mechanical control. In a steer-by-wire system there is no mechanical column at all, so redundancy in power supply, signal path and actuation is not optional. China's new national standard, "Basic Requirements for Automotive Steering Systems," effective 1 July 2026, addresses precisely these scenarios: power supply failure, control transmission failure, and energy transmission failure (CarNewsChina, 26 Jan 2026).

This is why steering suppliers are few and old. The hard part is not designing a gear. It is accumulating enough validated field data, functional-safety process documentation and OEM qualification history that a carmaker will let you into a safety-critical position on a platform that will build a million vehicles.

What CAAS actually does for a customer, step by step

Take the European order announced in Q3 2025 and shipped in May 2026. A global automaker's European division decides a new vehicle platform needs rack-assist electric power steering, because the platform is heavy enough that column-assist cannot deliver the required rack force, and because the platform must support lane-keep and automatic parking.

CAAS's Jingzhou Henglong unit bids. It must demonstrate a rack-EPS design meeting the OEM's rack-force, noise, vibration and harshness targets; supply functional-safety documentation for the ECU and motor; commit to a per-unit price for the platform life; and pass the OEM's supplier audit for a plant in Hubei that will ship to Europe. The award, announced in Q3 2025, carried expected annual sales exceeding $100 million with mass production targeted for 2027 (PR Newswire, 12 Nov 2025).

Then comes two years of validation builds, tooling, line installation and pre-production sampling. In May 2026 the first production batch shipped from the CAAS factory, ahead of the stated 2027 timeline, for two new European models at roughly 300,000 units a year (PR Newswire, 18 May 2026). From that point CAAS ships to a release schedule for the life of the platform, absorbing annual price-down demands, and holds the position until the platform is replaced.

"This advanced intelligent electric steering motor presents new growth opportunities and represents a major breakthrough for high-torque steering motors in the global commercial vehicle markets."

• CEO Qizhou Wu, on the Hyoseong (Wuhan) 115-platform steering motor entering mass production, PR Newswire, 5 Dec 2025

2. Business segments

CAAS reports as a single line of business, power steering, but it does not run as one. Management discusses results by operating unit on every call, and each unit has a distinct technology, customer set and history. The FY2025 release identifies six principal profit-making subsidiaries: Henglong, Jiulong, Wuhu, Henglong KYB, Hubei Henglong and Brazil Henglong. Two further entities matter disproportionately: Hyoseong (Wuhan), 51% owned, and Sentient AB, 40% owned.

There is a second, orthogonal cut of the same revenue that management leads with: EPS versus traditional steering. In H1 2026 EPS was 46.8% of sales and traditional steering 53.2%, against 42.5% and 57.5% a year earlier. Both cuts are covered below.

2.1 Henglong (passenger vehicle steering)

What it does. Henglong is the passenger-car business, built around Jingzhou Henglong Automotive Parts Co. Ltd., founded 1997, and the wider Hubei Henglong Automotive System Group. It makes rack-and-pinion hydraulic gears, column-assist EPS (C-EPS), dual-pinion EPS (DP-EPS), rack EPS (R-EPS), steering columns, and, since 2025, active rear-wheel steering. Jingzhou Henglong's 2024 production and sales volume exceeded 5.0 million units, up 18.5% year over year.

Core capability. Volume rack-and-pinion manufacturing at Chinese cost, converted over roughly a decade into volume EPS manufacturing without losing the cost position. That transition is harder than it sounds: an EPS gear requires an in-house or tightly coupled motor and ECU supply, torque-sensor calibration, and software validation, none of which a hydraulic gear needs. Henglong did it by building motor capability in-house (via Hyoseong Wuhan) and buying software capability abroad (via Sentient).

Why it exists separately. Passenger-car steering and heavy-truck steering are different products built on different machines for different customers. A rack-and-pinion gear and a recirculating-ball gear share almost no tooling.

Competitive position. Domestically Henglong competes with Bosch Huayu Steering, ZF Shanghai Steering, NSK-Wanda and Zhejiang Shibao for Chinese OEM platforms, and with Nexteer and JTEKT for the higher-content programmes. It wins on price and on speed of development cycle with Chinese OEMs. It loses, historically, on premium European and Japanese platforms where the incumbents have decades of qualification history. The European R-EPS win in 2025 is the first real crack in that.

Fit in the group. Henglong is the scale engine. At roughly $205.7 million of H1 2026 sales it is about 49.9% of group revenue, and it grew 25.3% year over year. Every strategic initiative management describes (Europe, South America, Mexico, rear-wheel steering) runs through this unit.

2.2 Jiulong (commercial vehicle steering)

What it does. Shashi Jiulong Power Steering Gears Co., Ltd., founded 1993, is the heavy commercial vehicle business. Its core product is the integral recirculating-ball steering gear, the architecture used on trucks and buses where rack force requirements are far beyond what a rack-and-pinion gear can handle. Its customers are truck makers: Beiqi Foton, Shaanxi Automobile Heavy Truck, Dongfeng, Nanjing Iveco.

Core capability. Jiulong's differentiating asset is iRCB, the intelligent electro-hydraulic circulating-ball steering system. The second generation entered mass production in China in 2025 and is described by the company as China's first iRCB compatible with L2+ assisted driving (PR Newswire, 2025). This matters because a truck cannot easily go to pure electric assist; the forces are too high for a 12V system, so an electro-hydraulic architecture that adds electronic control on top of hydraulic muscle is the practical route to assisted driving in heavy vehicles. The company also has eRCB, a fully electric recirculating-ball variant, supported by the Hyoseong (Wuhan) 115-platform motor delivering over 20 N·m of torque.

Why it exists separately. Different physics, different customers, different regulatory environment, and a separate 1993 joint-venture history predating the passenger-car business.

Competitive position. The Chinese heavy-truck steering market is less penetrated by the global majors than the passenger market, because truck steering is lower-volume and more regionally specified. Jiulong's iRCB, if it holds a lead on L2+ compatibility for trucks, is arguably the most defensible technical position anywhere in the group. On the Q3 2025 call CEO Qizhou Wu said new iRCB orders in July 2025 were "at a record setting pace in the power steering industry" for the ramp to mass production.

Fit in the group. The growth surprise. Jiulong grew 28.9% in FY2025 and 42.9% in H1 2026 to roughly $61.7 million, about 15.0% of group revenue. It is small enough that its growth rate does not carry the group, but it is where the most differentiated technology sits.

2.3 Wuhu Henglong (the Chery unit)

What it does. Wuhu Henglong Auto Steering Systems is the operating unit aligned to Chery Automobile, whose home base is Wuhu, Anhui. Locating a plant next to a single major OEM is standard practice in automotive; it cuts logistics cost and puts engineering resource inside the customer's development loop.

Core capability. Proximity and dedicated capacity for one customer, which converts into launch speed and share of that customer's platforms.

Why it exists separately. It is a customer-dedicated plant. Chery is one of China's largest privately owned automakers and, critically, one of its largest exporters, which is why this unit is currently growing while the Chinese domestic market shrinks.

Competitive position. Its position is derivative of Chery's. It wins when Chery wins. Notably, the FY2025 release recorded that the company "achieved higher sales in all except for sales to Chery Auto," so this unit is volatile: it fell in FY2025 and then grew 40.3% in H1 2026 to about $22.7 million, roughly 5.5% of group sales.

Fit in the group. A concentrated bet on one customer. Small in absolute terms, high beta.

2.4 Brazil Henglong (South America)

What it does. The São Paulo factory, established 1997's Henglong lineage extended overseas in 2012, supplies South American vehicle production, principally Stellantis. As of the Q2 2025 call it had three production lines running at roughly 90% utilisation, with a fourth, EPS-focused, line targeted for completion by year-end 2025.

Core capability. Fourteen years of local manufacturing, local content and local OEM relationships in a market that is difficult to serve by export because of import duty structures and logistics distance.

Why it exists separately. Different currency, different regulatory regime, different customer base, and a manufacturing footprint requirement that cannot be satisfied from China.

Competitive position. Brazil is where CAAS is closest to a genuine local incumbent rather than a Chinese exporter. That position produced 34.7% growth in FY2025. It also produced a 5.1% decline in H1 2026, the only major unit to shrink, which management has not fully explained beyond noting it as the exception.

Fit in the group. About 7.9% of H1 2026 revenue at roughly $32.6 million. Strategically it is the beachhead: the December 2025 C-EPS award from a South American OEM for over 300,000 units a year, mass production early 2028, is the payoff for the 2012 investment.

2.5 Henglong KYB and Hubei Henglong

Henglong KYB is the joint venture with KYB Corporation, the Japanese hydraulic and shock-absorber specialist. On the FY2024 call management cited KYB EPS sales up 27.2%, and on the Q1 2025 call Henglong KYB EPS sales up 38.2%. The JV gives CAAS access to Japanese process discipline and, importantly, to KYB's global customer relationships. It is the vehicle through which the Malaysia expansion runs: in November 2025 Hubei Henglong signed a strategic cooperation MoU with KYB-UMW Sdn Bhd, a joint venture between KYB and Malaysia's UMW, which holds a 38% stake in Perodua, Malaysia's largest carmaker (PR Newswire, 3 Nov 2025). KYB-UMW's new SP25 plant was scheduled to complete in December 2025 and become operational in 2026, producing steering products using Hubei Henglong technology.

Hubei Henglong Automotive System Group is the principal mainland operating group and the contracting entity for the Sentient stake, the Malaysia MoU and the South American award. Its Dongfang Avenue plant received national-level "Green Factory" designation from China's Ministry of Industry and Information Technology in December 2023; Jingzhou Henglong received the same designation for 2025, announced 17 August 2026, giving the group two such plants, supported by an 8MW rooftop photovoltaic system, digital workshop retrofits, waste-heat recovery and wastewater reuse (PR Newswire, 17 Aug 2026).

2.6 Hyoseong (Wuhan) Motion Mechatronics, 51% owned

The captive motor business. It produces the 115-platform electric steering motor for eRCB commercial-vehicle systems, output torque above 20 N·m, the result of three years of R&D, with mass production launched mid-December 2025. Roughly ten global OEMs had tested and verified the motor for commercial production at announcement. The production line was co-developed with Wiselink Technology Co., Ltd., a name that also appears in the Chairman's ownership structure (Wiselink Holdings Limited), which is worth noting as a related-party pattern.

Why this unit matters more than its size: the motor is the single most expensive and most supply-constrained component in an electric steering system. Owning it converts a purchased input into an internal margin pool and removes a dependency.

2.7 Sentient AB, 40% owned

Acquired in 2021 by Hubei Henglong, Sentient is a Gothenburg-based steering and vehicle-motion software house founded in 2009, with seven facilities in Sweden and an office in Suzhou, holding ten patents in steering, steer-by-wire, motion control and braking. Its output is software functions for steering, vehicle motion control and autonomous driving, integrated with its own powerpack hardware (software, ECU, motor, housing) and complete steering gear systems.

Sentient is the group's answer to the single most important strategic question it faces: as steering becomes software, does a Chinese hardware manufacturer become a commodity supplier to whoever owns the control algorithms? On the Q1 2025 call management described Sentient targeting 40,000 EPS units annually for Volvo trucks, plus a BYD Song programme in 2025 and a Renault passenger-vehicle collaboration using fly-by-wire technology.

Segment comparison

3. Products and business detail

The full catalogue

Traditional (hydraulic) steering, roughly 53.2% of H1 2026 sales

• Rack-and-pinion hydraulic power steering gears for cars and light-duty vehicles. The legacy passenger-car product. A hydraulic pump driven off the engine pressurises fluid that assists the rack.
• Integral recirculating-ball power steering gears for heavy-duty vehicles. A ball-screw mechanism multiplies force enough to steer a loaded truck. Jiulong's original product.
• Power steering pumps, hoses and oil reservoirs. The fluid circuit.
• Steering columns and intermediate shafts. The mechanical link from wheel to gear, including collapse behaviour in a crash.

Traditional steering is not a dying line at CAAS. It grew 11.2% in H1 2026 and 12.6% in FY2025, because heavy commercial vehicles and much of the Chinese export mix still use hydraulic architectures.

Electric power steering, roughly 46.8% of H1 2026 sales

• C-EPS (column-assist EPS). Motor mounted on the steering column. Cheapest, lowest rack force, used on small and mid-size passenger cars. This is the architecture in the South American award (over 300,000 units annually, mass production early 2028).
• DP-EPS (dual-pinion EPS). Motor drives a second pinion on the rack. Higher force than C-EPS, better feel. Showcased at Auto Shanghai 2025.
• R-EPS (rack-assist EPS). Motor acts directly on the rack via a ball-screw. Highest rack force, required for heavy vehicles and for platforms with demanding ADAS requirements. This is the product behind the European win and the Nanjing Iveco programme, which entered production in 2025 supporting automatic parking, lane keep assist and lane follow assist.
• iRCB (intelligent electro-hydraulic circulating ball). Jiulong's flagship. Second generation in mass production in China in 2025; described as China's first iRCB compatible with L2+ assisted driving.
• eRCB (electric recirculating ball). Fully electric commercial-vehicle steering, supported by the Hyoseong 115-platform motor.
• Active / distributed rear-wheel steering (RAS). Launched by Jingzhou Henglong in 2025. Historically a luxury feature (it shortens turning circle at low speed and improves stability at high speed); CAAS is targeting the upper mass market for Chinese new-energy vehicles.
• Steer-by-wire (SBW). Featured at Auto Shanghai 2025; the technology path running through Sentient.

Components and adjacent

• Electric steering motors (Hyoseong Wuhan, 115-platform, >20 N·m)
• Sensor modules and torque sensors
• Electronic control units (ECUs)
• Automotive electronics, electromechanical integrated systems
• Polymer materials
• After-sales service, R&D support services, inspection and testing of automotive products
• Aftermarket parts distribution in North America (Henglong USA)

What makes this hard to make

Three things.

Functional safety documentation. Any electronically controlled steering system sold into a regulated market needs a full functional-safety case: hazard analysis, safety goals, ASIL decomposition, verification evidence. China's standard effective 1 July 2026 explicitly "clarifies functional safety requirements for electronic control systems" and modifies functional-safety verification tests. Building that documentation set is years of work and is not transferable between architectures.

Torque-sensor calibration and steering feel. Two mechanically identical EPS gears can feel completely different depending on the assist curve, damping and return-to-centre logic. This is tuning work done on a proving ground with OEM chassis engineers, and it is a large part of why a carmaker sticks with a supplier.

Motor supply. A permanent-magnet motor delivering smooth assist torque without cogging, in an automotive thermal and vibration environment, at automotive cost, is a genuine manufacturing problem. CAAS solved it by owning 51% of Hyoseong (Wuhan) and taking three years to bring the 115-platform motor to production.

Manufacturing and geography

Manufacturing is concentrated in Jingzhou and Wuhan, Hubei Province, with additional plants at Wuhu, Anhui (Chery), and São Paulo, Brazil. The FY2024 geographic revenue split was China 68.2%, United States 16.6%, other foreign markets 15.2%.

Management described its process investments on the H1 2026 call: integrated MES (manufacturing execution system) with autonomous guided vehicles, computer-vision Poka-Yoke error-proofing, and 100% robotic inspection. The Green Factory designations reflect the energy side: an 8MW rooftop solar array, waste-heat recovery, wastewater reuse.

The newest geography is Mexico. The largest single capital item in H1 2026 was $15.8 million for a Mexico land and facility project, with roughly $15 million more on capacity for EPS, ECU, eRCB and electric motors. Full-year 2026 capex is guided at approximately $50 million; excluding Mexico, roughly $35 million, in line with 2025. The Mexico logic is straightforward: Mexico raised vehicle tariffs on China-sourced imports to 50% from 20%, and Mexico-built content can enter the USMCA market under regional rules of origin. Building in Mexico converts a tariffed export into local content.

Milestones that changed the business

4. Customers

Who buys

CAAS sells to vehicle manufacturers, not consumers. The FY2024 10-K states business relationships with more than sixty vehicle manufacturers. Named customers across filings and calls include:

• Chinese private OEMs: BYD, Zhejiang Geely, Chery Automobile
• Chinese state OEMs: SAIC Motor, FAW Group, Dongfeng, Chongqing Changan
• Chinese commercial vehicle makers: Beiqi Foton, Shaanxi Automobile Heavy Truck, Nanjing Iveco
• Global OEMs: Stellantis (North America and Brazil), Ford
• Via Sentient: Volvo trucks, Renault (fly-by-wire collaboration)
• New in the last 18 months: an unnamed large European automaker (R-EPS, first order Q3 2025, first shipment May 2026), an unnamed leading South American OEM (C-EPS, December 2025), plus Guangzhou Auto and a bus manufacturer cited as R-EPS order adds on the Q1 2025 call

Who inside the customer decides, and on what

The decision sits with a sourcing committee, not an individual. Three functions each hold a veto.

Chassis engineering decides whether the gear meets rack-force, NVH, steering-feel and packaging targets, and whether the supplier's functional-safety case is credible. This is where a supplier is disqualified.

Purchasing decides price, and imposes an annual price-down over the platform life. In Chinese OEM programmes these are aggressive; the FY2024 10-K explicitly flags OEM price-cutting as a risk.

Quality and supplier development decides whether the plant can hold PPM defect rates and support the launch ramp. Green Factory status, 100% robotic inspection and MES integration are marketing to this audience, not to investors.

Sales cycle: typically two to four years from nomination to start of production. The European order illustrates the shape precisely: awarded in Q3 2025, mass production originally targeted 2027, first volume shipment May 2026.

Why they choose CAAS

Three specific reasons, in order of weight.

Price at qualified quality. CAAS competes as the cost option that nonetheless carries OEM-grade validation. That is the whole proposition against Bosch, JTEKT, ZF and Nexteer on a mid-market platform.

Development speed with Chinese OEMs. Chinese carmakers run compressed programme timelines that European supplier processes struggle to match. A supplier headquartered in Hubei with engineers who can be at a Hefei or Wuhu development centre the same day has a real advantage.

Vertical integration on the motor. Owning 51% of the motor supplier removes the single most common cause of EPS launch delay.

Switching costs

High, but asymmetric in time. Once a gear is designed into a platform, switching is effectively impossible until the platform is replaced: the gear is packaged into a specific subframe, the assist software is tuned to that vehicle's dynamics, and re-homologating a steering system mid-cycle would require re-running the functional-safety case and, in many markets, re-certification. Practically, the customer is locked for four to seven years.

The asymmetry is that the lock-in resets completely at platform changeover. Every model refresh is a fresh competition, and incumbency helps but does not decide. This is why the flow of new nominations is the number that matters more than current volume.

Concentration

Real and disclosed. The five largest customers accounted for 57.1% of sales in Q2 2025, and 54.0% in Q1 2025, with one customer individually above 10% of consolidated net product sales.

Is this risk or quality signal? Both, and the split is informative. Concentration in Chinese OEMs (BYD, Geely, Chery, Changan, FAW, Dongfeng) is largely a reflection of where Chinese vehicle volume sits, and is diversified across six or more groups. Concentration in Stellantis is different and sharper: the Q1 2025 call attributed a 10.3% North American sales decline specifically to reduced Stellantis demand, and the Q1 2025 Brazil increase specifically to higher Stellantis orders. One customer group therefore drives both of the company's non-China geographies in opposite directions in the same quarter. That is genuine single-customer exposure dressed up as geographic diversification.

Contract structure

Automotive OEM supply is not spot business and it is not a fixed-value contract. The structure is a platform nomination: the supplier is awarded the business for a named vehicle programme, at an agreed per-unit price with a contractual annual price-down, for the life of the platform, with volumes forecast by the OEM but not guaranteed. Revenue is therefore highly visible (you know which platforms you are on and roughly what they build) but not contracted (if the model sells badly, you ship less).

This produces a specific revenue profile: predictable within a year, step-functional across years as platforms launch and retire, and structurally deflationary on price, which means a supplier must reduce unit cost every year merely to hold margin. It also means announced awards, like the 300,000-unit European and South American programmes, are the leading indicator of revenue two to three years out.

Tariffs sit on top of this. On the FY2025 call CFO Jie Li said a US Supreme Court ruling had reduced the effective tariff rate on the company's US-bound exports from roughly 70% to roughly 60%, and that the ruling "did have a positive impact to our export-related business." On the Q1 2025 call management said tariff cost was being shared with North American customers, and that inventory had been built forward in the US to keep customer production running through September.

5. Competitive landscape

The structure of the industry

Global steering is an oligopoly at the top and fragmented underneath. JTEKT, Bosch and Nexteer together hold roughly 60% of global EPS market share, and adding ZF and NSK covers a dominant share of global production (Mordor Intelligence). These five have the OEM relationships, the functional-safety heritage and the global manufacturing footprint that lets them serve a platform built on three continents.

Below that tier sits a Chinese domestic layer: CAAS, Zhejiang Shibao, and a set of joint ventures between global majors and Chinese partners (Bosch Huayu Steering, ZF Shanghai Steering, NSK-Wanda). The JVs are a distinctive feature of the Chinese market and a genuine competitive problem for CAAS, because they combine global technology with local cost and a state-linked distribution channel into SAIC and its affiliates.

Named competitors

Robert Bosch GmbH (Germany, Private). The largest automotive supplier in the world by revenue and a top-three EPS player. Competes across every architecture. Against Bosch, CAAS wins on price and loses on global footprint and on premium-platform credibility. Bosch's Chinese exposure runs largely through Bosch Huayu Steering Systems, a JV that in December 2025 showcased 48V direct-drive steer-by-wire, targeted for mass production as early as 2027.

JTEKT Corporation (Japan, TSE: 6473, approx. ¥689.5bn / US$4.0bn, as of July 2026). Ships more power-steering units than any other supplier by its own account. Strong on Japanese OEM platforms and on column EPS. Against JTEKT, CAAS competes primarily in China and increasingly in ASEAN, where the KYB-UMW Malaysia arrangement puts CAAS technology into a Japanese-affiliated regional supply chain.

Nexteer Automotive Group (Hong Kong / United States, HKEX: 1316, approx. HK$9.7bn, as of late July 2026). The most direct comparison: a former General Motors steering business now Chinese-controlled, with a portfolio spanning EPS, HPS, steer-by-wire, rear-wheel steering, columns, driveline and brake-by-wire. Named as a major competitor in the CAAS 10-K. Nexteer has deeper North American OEM relationships and a genuinely global engineering footprint, including a new Mexican technical laboratory. Nexteer's market capitalisation fell roughly 37% in the year to July 2026, which says something about how the market is pricing the whole steering complex.

ZF Friedrichshafen AG (Germany, Private, foundation-owned). Competes at the top of the technology stack and has secured volume contracts with Chinese EV brands and German luxury OEMs. Its Chinese presence runs through ZF Shanghai Steering Systems, named in the CAAS 10-K as "Shanghai ZF," a direct competitor.

Zhejiang Shibao Company Limited (China, SZSE: 002703, approx. CN¥10.4bn, as of late July 2026). The closest domestic peer and the most direct price competitor on volume column-EPS programmes for Chinese OEMs. Where CAAS loses a Chinese mid-market platform on price, this is often who wins it.

Huayu Automotive Systems (China, SSE: 600741, approx. CN¥59.8bn, as of August 2026). Not a pure steering competitor, but the SAIC-affiliated components group whose Bosch Huayu Steering JV is a first-tier competitor in China, and whose scale and SAIC relationship make it structurally advantaged on SAIC platforms.

NSK Ltd. / NSK-Wanda (Japan / China, JV). Column EPS specialist with a Chinese JV serving domestic platforms.

Others named in filings: FAW FKS (a captive supplier to FAW), BWI Group (Chinese-owned, ex-Delphi chassis), Sakthi Auto Components (India), and Hyundai Mobis for Korean platforms.

Competitor comparison

Barriers to entry

Genuinely high for the safety-critical core, and rising. A new entrant needs functional-safety process certification, a validated design library, OEM supplier approval, and enough field history that a carmaker will accept the liability. The July 2026 Chinese steer-by-wire standard raises this further by codifying failure-mode requirements that a new entrant must now design to from day one.

But lower than they look on the manufacturing. Making a competent hydraulic gear is a solved problem, and Chinese machine-tool and casting capability is broadly available. This is why the hydraulic end of the market is price-competitive and why margin has historically been thin.

The real barrier is shifting from metal to software. As steering becomes a controls problem, the moat migrates to whoever owns the assist algorithms, the redundancy architecture and the ADAS interface. This is exactly the transition where an incumbent hardware supplier can be reduced to a contract manufacturer. CAAS's 40% stake in Sentient is a direct response to that risk, and it is a 40% stake, not control.

Market share and where CAAS actually sits

CAAS is not a global top-five steering supplier and its filings do not claim to be. The FY2024 10-K's own framing is "a leading supplier of power steering components and systems in China." At roughly 10.8 million steering gears a year it has meaningful Chinese domestic scale, particularly in commercial-vehicle recirculating-ball steering where the global majors are less present.

Where it is strong: Chinese commercial vehicles (iRCB has a genuine lead on L2+ compatibility), Chinese mid-market passenger platforms, Brazil (local incumbency), and increasingly ASEAN via KYB-UMW.

Where it is exposed: premium European and Japanese platforms, where it has one order and no track record; North America, where its position depends on Stellantis and on a tariff regime that has run at 60-70%; and the software layer, where its position is a minority stake in someone else's company.

Structural shifts under way

Three are moving at once. Electrification of steering is the tailwind CAAS has ridden from 33.8% EPS mix in 2023 to 46.8% in H1 2026. Steer-by-wire is the next discontinuity, now regulated in China from 1 July 2026, with Bosch Huayu targeting mass production as early as 2027 and the Nio ET9 already in market. Chinese OEM globalisation is a tailwind for Chinese suppliers who follow their customers abroad, and it is the single biggest reason CAAS grew 20% in a Chinese market that shrank.

This is not a wide-moat business. It is a scale-and-cost business with one genuinely differentiated product line (iRCB), an option on the software layer (Sentient), and a customer set that is currently exporting hard.

6. Industry

What drives demand

Vehicle production volume, first and last. Every vehicle needs exactly one steering gear. Content per vehicle can rise, but unit demand is a direct function of how many cars get built.

Content per vehicle, second. This is where growth above production volume comes from. A hydraulic rack costs less than a column EPS, which costs less than a rack EPS, which costs less than a steer-by-wire system with rear-axle steering. Each step up in ADAS capability forces a step up in steering content, because assisted driving requires the steering system to accept a torque or angle command from a computer, which a purely hydraulic system cannot do.

Regulation, third and increasingly. ADAS mandates and functional-safety standards force content that consumers did not ask for.

Size and growth

The global electric power steering market was worth roughly US$29-30 billion in 2025, with forecasts clustering around a 5% CAGR toward US$39-50 billion by the early to mid 2030s (Mordor Intelligence; Research Nester). Asia-Pacific accounted for over 50.5% of the market in 2025. Global production volume across the top suppliers exceeds 150 million units a year.

The steer-by-wire sub-segment is projected to grow from roughly US$4.61 billion in 2025 to US$7.21 billion by 2031.

Where CAAS sits in the supply chain

Tier 1. It sells complete steering systems directly to vehicle manufacturers. Below it sit Tier 2 suppliers of castings, forgings, bearings, semiconductors and magnets. Above it sit the OEMs. Its 51% stake in Hyoseong (Wuhan) pulls one Tier 2 input (the motor) inside the boundary; its 40% stake in Sentient reaches sideways into software.

The demand environment right now, which is unusual

China's auto market in H1 2026 is the most important context for reading this company, and it is bifurcated to an extreme degree. Per CAAM data:

• Total Chinese-made vehicle sales fell 4.1% to 15.017 million units in H1 2026
• Domestic sales fell 21% to 9.921 million units, a fifth consecutive month of decline
• Exports rose over 65% to 5.096 million units
• NEV sales rose 7.3% to 7.446 million, with domestic NEV down over 13% and NEV exports up 120%
• Monthly vehicle exports exceeded one million units for the first time in 2026

The domestic collapse is attributable to the withdrawal of policy incentives, the exhaustion of replacement demand pulled forward into 2024-2025, and the NEV purchase tax moving from full exemption to a 50% reduction.

This matters enormously for reading CAAS. Its H1 2026 sales rose 20.1% against a domestic market down 21%. That is not a share-gain story alone; it is a story about being attached to the customers who are exporting. Chery, whose dedicated unit grew 40.3%, is one of China's largest exporters. Kevin Theiss made the contrast explicit on the H1 2026 call, noting overall vehicle production and sales fell year over year while company sales grew.

Regulation and certification

China's "Basic Requirements for Automotive Steering Systems", effective 1 July 2026, is the most consequential piece of near-term regulation. It strengthens failure provisions for power-source, control-transmission and energy-transmission failure; establishes alarm mechanisms for ageing and degraded steer-by-wire components; clarifies functional-safety requirements for electronic control; and modifies functional-safety verification testing with technical guidance for test institutions. In practical terms it legalises and standardises steer-by-wire in China, which both opens the market and raises the bar.

Beyond that, suppliers face IATF 16949 quality-system certification as table stakes, ISO 26262 functional safety for electronic systems, and per-OEM supplier approval. In China, an additional incentive shapes behaviour: high-and-new-technology enterprise tax status requires sustained R&D intensity, which management has cited directly as a reason to hold R&D at roughly 5% of revenue.

Two further regulatory layers touch CAAS specifically. US tariffs on Chinese auto components, which management has described as running at an effective 60-70% rate on its exports. And Mexican tariffs, raised to 50% on China-sourced vehicles from 20%, which is the direct driver of the $15.8 million Mexico land and facility investment.

Cyclicality

Auto components are a classic cyclical: demand tracks vehicle production, which tracks credit conditions, consumer confidence and policy incentives. The 2024-2026 Chinese cycle is a textbook illustration of policy-driven pull-forward followed by an air pocket.

But the cyclicality is asymmetric for a supplier. Volumes swing with the cycle, while the annual OEM price-down does not pause in a downturn. Fixed manufacturing cost does not fall as fast as volume. And a platform nomination won in a downturn only converts to revenue two to three years later, which means the order cycle and the revenue cycle are offset.

Tailwinds and headwinds

Tailwinds: EPS penetration still rising globally; ADAS mandates forcing steering content; steer-by-wire commercialising with regulatory backing in China from July 2026; Chinese OEM export volumes at record levels, dragging their supply chains abroad with them; ASEAN and Latin American localisation creating room for new entrants.

Headwinds: Chinese domestic demand in a sustained decline; structural annual price-down; tariff walls between the largest producing country and the largest consuming markets; a technology transition that could relocate value from hardware to software; and, as Nexteer's roughly 37% market-capitalisation decline over a year suggests, an equity market that is not currently rewarding the steering complex.

7. Growth triggers

Every item below is drawn from one of the six concalls or from a company announcement dated within that window.

• European R-EPS programme moving from award to volume. The first R-EPS order from a large European automaker was announced with annual sales expectations exceeding $100 million and mass production targeted for 2027 (Q3 2025 concall, 12 Nov 2025). It was repeated on the FY2025 call (22 Apr 2026), and then pulled forward: the first batch shipped in May 2026 for two new European models at approximately 300,000 units annually (announcement 18 May 2026), and management confirmed shipments had begun on the H1 2026 call (13 Aug 2026).

  "First REPS product order from a large, well-known European automobile producer" with "annual sales expectations exceeding $100 million."

  • FY2025 concall, 22 April 2026

• South American C-EPS platform, mass production early 2028. Hubei Henglong won a C-EPS award from a leading South American OEM covering gasoline and hybrid passenger models, over 300,000 units a year, with mass production expected early 2028 (announcement 17 Dec 2025). Repeated and quantified on the H1 2026 concall (13 Aug 2026), where CFO Jie Li put the expected annual contribution at approximately $40 million.

• Mexico manufacturing facility. $15.8 million spent in H1 2026 on Mexico land and facility, the single largest capital item of the half, within a full-year capex plan of approximately $50 million (H1 2026 concall, 13 Aug 2026). New disclosure; no production date given.

• Malaysia and ASEAN via KYB-UMW. Strategic cooperation MoU signed with KYB-UMW Sdn Bhd for local production of EPS and mechanical steering gears using Hubei Henglong technology, with KYB-UMW's SP25 plant scheduled to complete December 2025 and become operational in 2026 (announced 3 Nov 2025; repeated on the FY2025 concall, 22 Apr 2026, and again on the H1 2026 concall, 13 Aug 2026).

• Second-generation iRCB ramping to mass production for heavy vehicles. China's first iRCB compatible with L2+ assisted driving entered mass production in 2025, with new orders in July 2025 described as record-setting (Q3 2025 concall, 12 Nov 2025). Repeated on the FY2025 concall (22 Apr 2026).

  "New orders in July, 2025 were at a record setting pace in the power steering industry."

  • CEO Qizhou Wu, Q3 2025 concall, 12 November 2025

• High-torque eRCB steering motor entering mass production. Hyoseong (Wuhan)'s 115-platform electric steering motor, output torque above 20 N·m, verified by approximately ten global OEMs, launched mass production mid-December 2025 (announcement 5 Dec 2025).

• Active rear-wheel steering entering the upper mass market. Jingzhou Henglong launched active rear-wheel steering in 2025, targeting Chinese new-energy vehicles in a segment previously reserved for luxury cars (FY2025 concall, 22 Apr 2026).

• Brazil fourth production line, EPS-focused. Three lines were running at roughly 90% utilisation with a fourth, EPS-dedicated line targeted for completion by year-end 2025 (Q2 2025 concall, 13 Aug 2025).

• R-EPS order expansion beyond the launch customer. Following the Nanjing Iveco R-EPS programme entering production, additional OEMs including Chery, Guangzhou Auto and a bus manufacturer placed R-EPS orders (Q1 2025 concall, 14 May 2025).

• EPS unit volume step-up. Management guided to roughly a 30% EPS volume increase in 2025, adding approximately 400,000 units (Q4/FY2024 concall, 28 Mar 2025).

• 2026-2030 strategic plan, "23144" framework. Effective 1 January 2026, targeting total sales exceeding 20 billion yuan including VAT by 2030, built on three pillars: a "China + N" manufacturing and supply chain network across North America, South America, Europe and Southeast Asia; R&D concentrated on steer-by-wire and rear-wheel steering; and platform-based, lean, automated manufacturing (announcement 6 Jan 2026; repeated in substance on the FY2025 and H1 2026 concalls).

  The plan "outlines the direction for the Group's development over the next five years."

  • CEO Qizhou Wu, 2026-2030 Strategic Plan announcement, 6 January 2026

• Expansion beyond steering into chassis via M&A. Management said it is evaluating merger-and-acquisition opportunities focused on new product categories rather than more of the current line, specifically chassis products such as suspension and braking to support autonomous driving (H1 2026 concall, 13 Aug 2026).

  Management's focus is on "new product expansion rather than buying more of the current product line."

  • H1 2026 concall, 13 August 2026

• Sentient-linked programmes. Sentient targeting approximately 40,000 EPS units annually for Volvo trucks, a BYD Song model launch in 2025, and a Renault passenger-vehicle collaboration using fly-by-wire technology (Q1 2025 concall, 14 May 2025).

Trigger summary

8. Key risks

1. The Chinese domestic market is contracting hard, and CAAS's growth currently depends on customers exporting

Mechanism. Chinese domestic vehicle sales fell 21% in H1 2026 to 9.921 million units, the fifth consecutive month of decline, driven by expired incentives, exhausted replacement demand, and the NEV purchase tax halving its exemption. CAAS still grew 20.1%, because Chinese OEM exports rose 65% and CAAS is attached to exporters. If export growth decelerates, whether from foreign tariff action, European Union measures, or simple saturation, CAAS loses the offset and is left facing the domestic decline directly with no cushion.

Calibration. High probability of some export deceleration over two to three years, given that 65% growth rates do not persist. Moderate to severe impact. This is the single most important risk in the report because it is the mechanism currently producing the company's entire growth rate.

2. Tariff walls on the largest export market

Mechanism. Management has stated the effective tariff on US-bound exports ran at roughly 70%, reduced to roughly 60% following a Supreme Court ruling. The United States was 16.6% of FY2024 sales. A 60% tariff is not a cost you absorb; it is a cost you either pass to the customer, share, or design around by localising. CAAS has chosen localisation, hence the Mexico investment, but Mexico also raised tariffs on China-sourced vehicles to 50% from 20%, and USMCA regional value-content rules require genuine local transformation, not transhipment. If the Mexico plant does not qualify content, the strategy fails and the North American business is structurally impaired.

Management's own framing (Q1 2025 concall, 14 May 2025):

"Overall, the tariff has very minimum impact to our business and the order flow."

That was said when inventory had been pre-built in the US to cover customer production through September. It is a statement about the near term, not about the structure. North American sales grew just 3.5% in H1 2026 against 77.3% in Q3 2025, which suggests the pre-build effect has washed through.

Calibration. High probability, ongoing, moderate impact today and potentially severe if Mexico does not qualify.

3. Stellantis concentration masquerading as geographic diversification

Mechanism. Both non-China geographies run through one customer group. The Q1 2025 call attributed a 10.3% North American decline to reduced Stellantis demand and a 30.2% Brazil increase to higher Stellantis orders, in the same quarter. Brazil then declined 5.1% in H1 2026, the only major unit to shrink. A single OEM's production planning is therefore the dominant variable in roughly a third of the business, and the five largest customers were 57.1% of sales in Q2 2025.

Calibration. Moderate probability of a material Stellantis programme change in any given two-year window; high impact when it happens, and it has already happened twice in the six-call window in opposite directions.

4. Value migrating from steering hardware to steering software, where CAAS owns 40% of an associate

Mechanism. Steer-by-wire removes the mechanical column entirely. What remains is a set of actuators plus the control software that decides how the car behaves. If OEMs insource the control layer, or if a software-first supplier owns it, a hardware manufacturer becomes a contract assembler at contract-assembler margins. CAAS's response is a 40% stake in Sentient AB, which holds ten patents in steering, steer-by-wire, motion control and braking. Forty percent is influence, not control. If Sentient's other owners take a different path, or if a competitor's software becomes the OEM standard, the option does not exercise.

Meanwhile Bosch Huayu showcased 48V direct-drive steer-by-wire in December 2025 with mass production targeted as early as 2027, and China's steer-by-wire standard took effect 1 July 2026.

Calibration. Moderate probability over five years, high impact. This is the risk that would change what kind of business this is.

5. Structural price-down with no pricing power

Mechanism. OEM supply contracts embed annual price reductions. The FY2024 10-K explicitly flags "OEM price-cutting" and "high dependency on few large OEMs who impose pricing pressures." A supplier must therefore reduce unit cost every single year merely to hold margin, funded by volume growth, mix shift to higher-content products, or productivity. CAAS is currently winning this by shifting mix toward EPS. When EPS mix stops rising, the price-down does not stop with it.

Calibration. Certainty, not probability. Continuous moderate drag.

6. Controlled-company governance and a history of a withdrawn take-private

Mechanism. Chairman Hanlin Chen owned 57.25% of the common stock as of 30 June 2025, and the FY2024 10-K flags "management control concentration limiting minority shareholder influence." In 2017-2018 Chen, Wiselink Holdings Limited, his wife Liping Xie, and an affiliate of North Haven Private Equity Asia IV proposed to buy out minority shareholders at $5.45 per share, then withdrew in August 2018. The redomiciliation to the Cayman Islands in September 2025 moved the company from Delaware corporate law, with its established minority-shareholder jurisprudence, to Cayman law.

There is also a related-party pattern worth naming: the Hyoseong (Wuhan) motor production line was co-developed with Wiselink Technology Co., Ltd., while Wiselink Holdings Limited appears in the Chairman's own Form 3 beneficial-ownership disclosure.

A further governance note, offered with explicit attribution and not verified against a current filing: the company history published by Encyclopedia.com records that Hanlin Chen claimed an MBA from Barrington University, described by the US Government Accountability Office's Office of Special Investigations as a diploma mill. This is a historical account, not a current disclosure, and I could not corroborate it in a filing I was able to retrieve.

Calibration. Low probability of an acute event, but a permanent structural discount on minority influence. The take-private history means an opportunistic bid at a depressed price is a live possibility rather than a theoretical one.

7. A promised buyback that has not happened

Mechanism. Management has now told investors across three consecutive calls that it intends to return capital via buyback, and has not done so. See Sections 9 and 10. The risk here is not financial; it is informational. When management repeats an intention across a year without executing, the reader must discount future stated intentions by the same factor.

Calibration. High probability of continued non-execution given the stated preference for reinvestment. Low direct financial impact, but it degrades the credibility of everything else management says it will do.

8. Capital intensity rising into an uncertain demand environment

Mechanism. Full-year 2026 capex is guided at approximately $50 million, up from $37.2 million in FY2025, of which $15.8 million went into Mexico land and facility in H1 alone. Simultaneously the company is building a fourth Brazil line, funding a Malaysia technology transfer, tooling for European and South American programmes, and running R&D at roughly 5% of revenue with H1 2026 R&D up 23.6%. All of this is committed spend against programmes that start production in 2027 and 2028, in a domestic market falling 21% a year. If the export offset fades before the new programmes start, the company carries the fixed cost of four new geographies through a trough.

Calibration. Moderate probability, moderate impact. This is the classic supplier trap: the capex cycle and the revenue cycle are offset by two to three years.

9. Reduced disclosure following the move to foreign private issuer status

Mechanism. The company now reports twice a year instead of four times, files 6-Ks and a 20-F instead of 10-Qs and a 10-K, and is exempt from US proxy rules. Its 10% beneficial owner, the Chairman, remains permanently exempt from Section 16 insider reporting even after the 2026 rule change (see Section 11). The practical effect is that an investor now waits up to six months for operating data and has structurally less visibility into the actions of the controlling shareholder. This is not a business risk; it is a monitoring risk, and it compounds every other risk in this list.

Calibration. Certainty. Permanent reduction in the reader's ability to detect deterioration early.

9. Walk the talk

The six calls used: Q4/FY2024 (28 March 2025), Q1 2025 (14 May 2025), Q2 2025 (13 August 2025), Q3 2025 (12 November 2025), Q4/FY2025 (22 April 2026), H1 2026 (13 August 2026). The most recent is seven days old.

The revenue guidance record, which is the cleanest test

Start with the oldest call. On 28 March 2025, management set full-year 2025 revenue guidance:

"Revenue guidance for the full year 2025, of $700 million."

• CFO Jie Li, Q4/FY2024 concall, 28 March 2025

On 14 May 2025, with Q1 in hand and North American sales down 10.3% on Stellantis weakness, management did not cut. It reaffirmed the $700 million target, citing "stable market conditions." That is the first useful data point: a management team that had a legitimate excuse to trim and did not take it.

On 13 August 2025, after Q2, they raised to $720 million, citing demand in Brazil and North America. On 12 November 2025, after Q3, they raised again to $730 million.

The year finished at $765.7 million, above even the twice-raised number, and above the original March guidance by a wide margin. Across four calls, guidance moved $700m to $720m to $730m, and the outcome beat the highest of those. This is the signature of a conservative guider, not an optimistic one.

The pattern then repeated immediately. On 22 April 2026, management guided 2026 to $810 million. On 13 August 2026, after one half, they raised to $850 million. Whether that lands is unknown, but the behavioural pattern is now four consecutive raises across two guidance cycles, with no cut, in a domestic market that fell 21%.

Where they were less accurate

EPS unit volume. On 28 March 2025 management guided to roughly a 30% EPS volume increase in 2025, adding approximately 400,000 units. FY2025 EPS product sales grew 25.5%. Units and revenue are not the same measure, and mix and price-down separate them, but as stated the outcome came in under the guided rate. Modest miss, not a material one, and not something management drew attention to afterwards.

Brazil. On 13 August 2025 management described three Brazil lines at roughly 90% utilisation and a fourth EPS-focused line targeted for completion by year-end 2025, in a unit that then grew 34.7% for the full year. It was a coherent, specific commitment. Brazil then declined 5.1% in H1 2026, the only major unit to fall. The fourth line has not been revisited on a subsequent call, and management's H1 2026 commentary named Brazil as the exception without explaining why. This is the clearest example in the six calls of a specific operational promise that has quietly stopped being discussed.

North America. On 12 November 2025 North American sales were up 77.3%, an extraordinary number that management attributed to underlying demand. On 13 August 2026 the same line was up 3.5%. Reading the two together with the Q1 2025 disclosure that inventory had been pre-built in the US ahead of tariffs, the Q3 2025 figure looks substantially like a pull-forward that management presented as demand. Nothing said was untrue. But the framing invited an extrapolation that did not hold.

Where they delivered ahead of what they promised

The European R-EPS programme. On 12 November 2025 management announced the first European R-EPS order with annual sales expectations exceeding $100 million and mass production by 2027. It was repeated on 22 April 2026 with the same 2027 date. Then on 18 May 2026 the first volume batch shipped, roughly a year ahead of the stated timeline, for two European models at approximately 300,000 units a year. Delivering an automotive programme early is uncommon. This is the strongest single credibility marker in the window.

The iRCB ramp. The second-generation L2+-compatible iRCB was described as entering mass production in 2025 and as generating record-pace July 2025 orders. It appears in FY2025 results as part of a Jiulong unit that grew 28.9% for the year and then 42.9% in H1 2026. The claim was made and the numbers followed.

Where they have repeatedly said one thing and done another

The buyback. This is the one clear failure of follow-through, and it runs across the entire window.

On 13 August 2025, management discussed a share buyback programme, explicitly citing the stock as undervalued. A $5 million authorisation with a $5.50 per share price cap, announced 18 November 2024, was live at the time and ran to 15 November 2025. The FY2025 release records no share buybacks in 2025.

On 22 April 2026, having let that authorisation lapse unused, the CFO said:

The company will "recommend to reinitiate share buyback program."

• CFO Jie Li, FY2025 concall, 22 April 2026

Four months later, on 13 August 2026, management said discussions on dividends and buybacks were "at board level" but that reinvestment was being prioritised given global expansion.

So: an authorisation that expired unused, followed by a commitment to reinitiate, followed by a softening back to "under discussion." Three calls, one direction of travel, zero shares repurchased. Note that this is separable from a capital-allocation judgement. Prioritising a Mexico plant over a buyback may well be the right call. The issue is that management said it would buy back stock and then did not, twice, without acknowledging the change.

Promise versus outcome

Assessment

On operations and revenue, this management does what it says, and slightly more. Four consecutive guidance raises with no cut, a European programme delivered a year early, and a new product line (iRCB) that produced the growth it was said to produce. The consistent conservatism is a genuine asset in a Chinese small cap, where the base rate runs the other way.

On capital returns, they do not do what they say, and they do not acknowledge it. The buyback has been promised in some form on three of the last four calls and executed on none of them.

On framing, they are selective rather than misleading. The Q3 2025 North American figure and the quiet disappearance of the Brazil fourth line are both cases where the reader had to notice an absence rather than catch a false statement. Nothing said was untrue; the emphasis simply moved away from the parts that stopped working.

The net read: credible on the business, unreliable on shareholder returns, and requiring the reader to track what stops being mentioned.

10. Shareholder friendliness index

Dividends. CAAS paid a special cash dividend of $0.80 per share declared 19 July 2024, record date 30 July 2024, paid 22 August 2024, aggregating roughly $22.4 million by the CFO's figure on the FY2024 call (the announcing release cited approximately $25 million) (PR Newswire, 19 Jul 2024). That is the only common dividend in the three-year window: FY2023 nil, FY2024 $0.80 (special, one-off), FY2025 nil, and nothing to date in FY2026 (stockanalysis.com dividend history). The FY2025 release does record approximately $2.2 million of dividends paid, but this is not a distribution to common shareholders of the parent; no per-share common dividend was declared for 2025, and management confirmed the position directly on the 22 April 2026 call: "we don't have a plan at the moment." On the 13 August 2026 call the position softened only to dividends being under discussion at board level, with reinvestment prioritised. There is no dividend policy, and the 2024 payment should be read as a one-off return of surplus cash rather than the start of a trend.

Buybacks and dilution. Two sources, two windows. For the last 90 days, there is no evidence of any repurchase: the H1 2026 release (13 August 2026) reports no buyback activity and the H1 2026 call described buybacks as still at board-discussion stage. For the older three-year period, web-searched filings and announcements show one authorisation and effectively no execution: the board approved a programme of up to $5 million on 18 November 2024, restricted to open-market purchases at prices not exceeding $5.50 per share, running to 15 November 2025 (PR Newswire, 18 Nov 2024). The FY2025 release states no share buybacks were made in 2025, and the authorisation lapsed. Treasury stock stood at 2,167,600 shares at 31 December 2025, a balance carried from earlier programmes rather than added to. On share count, weighted average diluted shares were 30,170,702 in FY2025 and the identical 30,170,702 in H1 2026 - flat, with no option-driven creep. So the count is neither shrinking nor growing: no dilution, but no retirement either. One caveat on the lapsed programme: with the price cap set at $5.50, a stock trading above that level for much of the window would have made execution mechanically impossible, which is a plausible explanation for zero purchases but is not one management has offered.

Verdict: Hoards Capital - one special dividend in three years, an authorised buyback that expired entirely unused, a flat share count, and a stated preference for reinvesting in Mexico, Malaysia and new programmes over returning cash.

11. Insider activities

An important structural point about what is and is not filed

CAAS's insider-disclosure regime changed twice inside the last twelve months, and reading the record without knowing that would produce a badly wrong conclusion.

1. Until 11 September 2025, CAAS was a US domestic filer. Directors, officers and 10% owners filed Forms 3, 4 and 5 under Section 16.
2. From 11 September 2025 to 17 March 2026, CAAS was a foreign private issuer and its directors and officers were exempt from Section 16 entirely under Exchange Act Rule 3a12-3(b). No Forms 4 were required and none were filed. This is a genuine reporting blackout, not an absence of activity.
3. From 18 March 2026, the Holding Foreign Insiders Accountable Act, signed into law 18 December 2025, removed the FPI exemption. FPI directors and officers must now file Forms 3, 4 and 5 (Morgan Lewis; Cleary Gottlieb). Critically, 10% beneficial owners of an FPI remain exempt from Section 16 in its entirety, and FPI directors and officers remain exempt from the short-swing profit rule and the short-sale prohibition.

Point 3 has a direct and permanent consequence for this company: Chairman Hanlin Chen, who held 57.25% of the common stock as of 30 June 2025 (per the Q2 2025 Form 10-Q), is a 10% owner and therefore is not required to report his transactions on Form 4 at all, even under the new law. He filed a Form 3 because he is also a director and officer, but the reader should not assume future Form 4 coverage of the controlling shareholder's dealings.

Filings located in the last 12 months (August 2025 to August 2026)

The only Section 16 filings on record for CAAS in this window are the six Form 3 initial statements of beneficial ownership filed on 18 March 2026, each carrying an event date of 11 September 2025 (the redomiciliation) and each signed 5 March 2026. These are catch-up initial statements triggered by the HFIAA, not transactions.

No Form 4 transaction filings were located for CAAS between August 2025 and August 2026. Aggregator data corroborates this, reporting zero insider buy transactions and zero insider sell transactions over the trailing six months. An older Form 4 does exist in the public record showing CFO Li Jie selling 10,000 shares at approximately $9.00, but that filing is dated 8 December and predates this window by several years; it is not recent activity and is excluded from the assessment above.

Buys - reading the signal

There are no open-market insider purchases to read. Not one director or officer bought CAAS stock in the open market in the last twelve months, on the available record.

This absence needs to be weighed carefully rather than dismissed, because it coincides with the period in which the company's own management repeatedly described the stock as undervalued. On the 13 August 2025 call, management justified the share buyback programme by explicitly citing undervaluation. If the people running the company believed the stock was cheap enough to justify spending corporate cash on it, the absence of any personal open-market purchase over the following twelve months is a meaningful non-signal. It is not bearish in itself, but it removes the strongest bullish evidence that could have existed.

One partial mitigant: for roughly half the window (11 September 2025 to 17 March 2026) insiders had no obligation to report, so a purchase in that period could in principle have gone undisclosed. That is speculation, not evidence, and it is not how I would weight it.

Sells - working out the why

There are no material insider sales to explain. No Form 4 sales, no disclosed block trades, no secondary offering, no sponsor exit. The controlling shareholder's Form 3 holdings (15,762,547 direct plus 1,502,925 via spouse plus 8,198 via Wiselink Holdings) are consistent in scale with the 57.25% stake reported at 30 June 2025, which indicates no large disposal took place across the redomiciliation.

The single most important caveat has already been stated and bears repeating: the Chairman, as a 10% owner of a foreign private issuer, is permanently outside the Form 4 regime. An absence of reported sales by him is therefore not positive evidence of an absence of sales. It is an absence of a reporting obligation. Any future stake change would surface only through a Schedule 13D/G amendment or the annual 20-F ownership table, both of which are far slower and coarser than a Form 4.

Net assessment

Insiders are neither net buyers nor net sellers on the reported record: the only Section 16 activity in twelve months is six housekeeping Form 3s filed on a single day because a new US statute compelled them.

Three things stand out. First, the activity is not concentrated in one or two people because there is no activity at all. Second, the one genuinely new datum is that independent director Liu Tao reported nil beneficial ownership as of his Form 3, which means at least one board member has no personal economic stake in the company. Third, and most consequential, the disclosure regime around the controlling 57%+ shareholder has become permanently weaker: he is exempt from Form 4 reporting, the company now reports twice a year rather than four times, and it has moved from Delaware to Cayman corporate law.

Read: neutral on transaction signal, mild concern on disclosure architecture. There is nothing bearish in what insiders did. The concern is what the reader can no longer see, layered on top of a controlled company whose chairman has previously attempted to take it private at a low price and withdrew.

12. Scenarios

Bull case

The export flywheel keeps turning and the localisation bet lands. Chinese OEMs keep taking share in Europe, Latin America, ASEAN and the Middle East, and their component suppliers travel with them, because a Chinese carmaker building 300,000 units in Brazil or Thailand wants a supplier who already knows its platforms and its price points. CAAS is that supplier. Chery keeps exporting, the Wuhu unit keeps compounding, and Jiulong's iRCB becomes the default electro-hydraulic architecture for Chinese heavy trucks as those trucks add L2+ capability to meet fleet-operator demand.

The four overseas bets convert in sequence rather than all at once, which is what makes the story credible. The European programme, already shipping a year ahead of its stated timeline, expands from two models to a wider platform family, and the reference customer does what reference customers do: it makes the next European bid easier. Malaysia's SP25 plant comes on line and puts Henglong technology inside a KYB-affiliated regional supply chain with a route to Perodua. The Mexico facility qualifies USMCA regional content, which turns a 60%-tariffed export into local supply and reopens North America properly rather than through pre-built inventory. Brazil's fourth line fills, and the South American C-EPS platform ramps into 2028 exactly as scheduled, because it was won on a design that already exists.

Underneath all of it, the mix keeps improving. EPS passes half of revenue and keeps climbing; rear-wheel steering moves from luxury into the Chinese new-energy upper-mass market; the Hyoseong motor gets designed into other people's systems as well as CAAS's own. The 2026-2030 plan's chassis ambition turns into a real acquisition in suspension or braking, and the company stops being a steering supplier and becomes a chassis supplier, which is a materially larger content-per-vehicle position. Sentient's software becomes the differentiator on steer-by-wire programmes rather than a defensive stake, helped by China's July 2026 standard giving domestic steer-by-wire a regulated runway. Management, having raised guidance four times running without a cut, finally reinitiates the buyback with the stock still unloved, and the reduced share count does what a reduced share count does.

Base case

The domestic Chinese market stays difficult and the export offset gradually normalises. Growth decelerates from the 20% pace of H1 2026 toward something lower but still positive, because the mix shift into EPS is a real and durable source of content growth independent of unit volume, and because the new programmes start contributing exactly when the export tailwind starts fading.

The European programme ramps to its roughly 300,000 units and delivers what management indicated, but it takes the full runway to get there and does not immediately multiply into other European wins, because European OEMs move slowly and one shipped programme is a reference, not a franchise. South America starts in early 2028 at roughly the scale described. Malaysia contributes modestly, because a technology-licensing and cooperation arrangement into a partner's plant is inherently lower-revenue than owning the plant. Mexico opens, absorbs its capital, and takes longer than hoped to qualify content and win nominations, so North America stabilises rather than re-accelerates.

Brazil recovers from its H1 2026 decline as Stellantis programme timing normalises, but the unit stays lumpy because it always has been. Jiulong keeps outgrowing the group on iRCB, and Henglong grinds forward on EPS mix. Capex stays elevated for two or three years while four geographies are built simultaneously, so free cash generation is thinner than the operating performance suggests. The annual OEM price-down keeps eating whatever productivity does not cover, and margin gains prove partly cyclical rather than structural, consistent with the CFO's own warning on 22 April 2026 that 2026 margin would be healthy but "not going to be as high as Q4 2025."

Capital returns stay where they have been: discussed at board level, deferred in favour of reinvestment, and no buyback executed. The share count stays flat. The company remains a controlled, semi-annually reporting Cayman entity with a 57% shareholder, and the market keeps applying whatever discount it applies to that structure, much as it has to Nexteer.

Bear case

The export offset fades before the new programmes arrive, and the two-to-three-year gap between the capex cycle and the revenue cycle opens up at exactly the wrong moment. Chinese vehicle exports, growing 65% in H1 2026, run into European trade measures, saturation, or local-content requirements in destination markets that favour building there rather than shipping from Hubei. Domestic Chinese demand, already down 21%, does not recover because the incentive pull-forward genuinely consumed future demand rather than merely shifting it. CAAS is left with a domestic market shrinking, an export channel flattening, and four half-built overseas footprints carrying fixed cost.

North America is the acute version of this. A 60% effective tariff is not survivable through cost-sharing indefinitely, and the H1 2026 slowdown to 3.5% growth after Q3 2025's 77.3% suggests the pre-build cushion has already been spent. If the Mexico facility fails to qualify USMCA regional value content, or qualifies it too slowly, the entire North American strategy is a $15.8 million-plus land purchase that solves nothing. Meanwhile Mexico's own tariff on China-sourced content has gone from 20% to 50%, so the input side of that plant is not cheap either.

Stellantis is the second acute risk, and it has already demonstrated its power twice inside six calls: it took North America down 10.3% in Q1 2025 and pushed Brazil up 30.2% in the same quarter, then Brazil fell 5.1% in H1 2026. A genuine Stellantis platform loss or production cut would hit both non-China geographies simultaneously, and the five largest customers were 57.1% of sales.

The slower, more damaging version is technological. Steer-by-wire arrives on schedule following China's July 2026 standard, Bosch Huayu's 48V direct-drive reaches production in 2027, and the value in a steering system migrates decisively to the control software and the redundancy architecture. CAAS owns 40% of Sentient, not 100%, and cannot direct it. OEMs insource the control layer or standardise on a competitor's stack, and CAAS's beautifully efficient, Green Factory-certified, 100%-robotically-inspected plants end up building actuators to someone else's specification at actuator margins, with the annual price-down still running.

The governance tail sits underneath all of it. A controlled company with a 57% chairman, semi-annual reporting, Cayman law, a controlling shareholder permanently exempt from Form 4 disclosure, related-party arrangements with Wiselink entities, and a prior take-private attempt at $5.45 per share that was withdrawn rather than abandoned in principle. In a scenario where the operating story deteriorates and the shares fall, the minority shareholder's position is structurally weak, and the historical precedent for what happens next is not reassuring.