Zelos

1.1 Identity, Product Scope, and Geographic Footprint

Zelos presents itself as a full-stack autonomous logistics company rather than a narrow software supplier. The homepage and about materials say the company builds the technology stack in-house, uses map-free L4 autonomy, and operates purpose-built RoboVan platforms for logistics use cases. Public company materials also support a dual-footprint story. QCC and the Suzhou Industrial Park administration anchor the core PRC operating entity as Jiushi (Suzhou) Intelligent Technology Co., Ltd., founded in 2021 in Suzhou Industrial Park, while the official about page and April 2026 Singapore partnership releases show an active Singapore operating address and commercial team. Scale disclosures are large but not perfectly harmonized across sources: the current official site says 25,000+ vehicles across 350+ cities and 20+ countries with 100M+ autonomous kilometers, whereas April 2026 partner releases still use 20,000+ vehicles and 300+ cities. The strongest read is that Zelos has real global deployment breadth, but the company updates headline metrics across surfaces over time, so chapter-1 should preserve both the current official numbers and the lower April 2026 corroboration rather than collapse them into one falsely precise benchmark.[CO001, CO002, CO005, CO006, CO007, CO009]

1.2 Founders, Leadership, and Governance Surface

Public leadership evidence is directionally useful but not fully clean. QCC identifies Cui Xiao as legal representative and executive director of the Suzhou operating entity and Kong Qi as manager, while the DHL launch release names Sean Zhang as co-founder and COO. The April 2026 Zelos and SingPost partnership releases then describe Sean Zhang as Co-Founder & CEO, Global Business, and Terry Zhou as Managing Director, Zelos Singapore. China Daily separately names Zhuang Li as a co-founder, and Taibo says Zhuang Li, Kong Qi, and Zhu Weicheng previously came from JD’s autonomous vehicle team. That does not amount to a clean current org chart. Instead, it suggests founder continuity plus regional operating leadership, but with role labels that vary across registry, partner, and media sources. For diligence, the bigger issue is governance opacity: no retained source gives the full current board roster, observer rights, ownership splits, or the post-Cainiao/post-February-2026 control structure. That means the company can be described as founder-linked and operationally visible, but not as a transparently governed late-stage company on public evidence alone.[CO008, CO012, CO013, CO014, CO015, CO016]

1.3 Funding History, Investors, and the Cainiao Link

Zelos’s capital story is strong in headline terms but weaker on mechanics. Gasgoo reported a $100 million B1 round in November 2024 led by CDH BAIFU and Blue Lake, and Blue Lake later said the company completed nearly $300 million of Series B financing after a fresh $100 million B3 round in April 2025, with Asia Investment Capital and Baidu Ventures participating. Reuters-derived U.S. News reporting added a separate strategic twist in January 2026: Cainiao would contribute its autonomous-driving business and cash for a stake in Zelos, without taking control, at an approximately $2 billion robovan-business valuation according to the Wall Street Journal. Multiple February 2026 secondary outlets then described a new financing of more than $300 million and a valuation above RMB 10 billion, roughly $1.4 billion. That unicorn-level narrative is well corroborated in press, so it should not be dismissed, but the exact valuation mechanics still rest on secondary coverage rather than retained primary financing documents. Chapter-1 therefore can support a large late-stage financing story and a meaningful Cainiao strategic relationship, while explicitly preserving cap- table, post-money, governance-rights, and ownership uncertainties.[CO018, CO019, CO020, CO021, CO022, CO023]

1.4 Milestones, Commercial Deployments, and Adverse Signals

The milestone record shows why Zelos is notable. SIPAC said the company was founded in Suzhou Industrial Park in 2021 and secured Singapore’s first unmanned transport vehicle license on 2024-05-30. By July 2024 it claimed 100+ cities and 2M+ kilometers; by April 2025 China Daily described six countries and 200+ regions; and by April 2026 Zelos and SingPost were using 20,000+ vehicles and 300+ cities as externally corroborated scale markers. Commercial proof is not limited to one pilot. DHL launched a Singapore deployment in 2025, FairPrice received public-road approval in October 2025 and planned close to 30 vehicles, SingPost expanded its MOU in April 2026, and Zelos separately used 365Group and Intertraffic Amsterdam announcements to signal Middle East and European ambitions. The chapter should still stay disciplined about weaker evidence. Jiemian’s China Post order, market-share, factory-count, and annual-capacity claims are useful but secondary. Taibo’s report that JD alleged IP infringement and referred the matter to public security is the clearest retained adverse source and cannot be waved away, especially when Yahoo also depicts a still-fluid IPO path. The main underwriting posture is therefore positive on deployment proof, cautious on financial transparency, and explicitly alert to legal and governance risk.[CO024, CO025, CO027, CO028, CO029, CO030]

1.5 Exhibits

2.1 Market Boundary and Demand Base

Zelos should be analyzed inside China autonomous logistics rather than inside the whole autonomous-vehicle market or the whole e-commerce market. The relevant boundary is autonomous movement of parcels and freight across repeatable logistics workflows: campus and community last-mile relay, urban public-road distribution, depot-to-store middle-mile, and park or hub circulation where vehicles, dispatch software, and relay robotics work together. That included spend covers autonomous delivery vehicles, software and fleet operations, relay robots, and the site integration needed to turn a route into a repeatable operating line. It excludes passenger robo-taxis, consumer delivery gadgets, and generic parcel labor or non-robotic transport spend. Demand is unquestionably large. China's express-delivery network handled 174.5 billion parcels in 2024, 199 billion in 2025, and 47.73 billion in the first quarter of 2026 alone. Official sources and industry media also show how that demand is being operationally absorbed: automated sorting centers, drones, indoor relay robots, and unmanned delivery vehicles are being deployed in mainstream parcel operations rather than confined to labs. The important analytical caution is that demand volume is not the same thing as robovan revenue. Public evidence supports a very large parcel economy, a large logistics-automation adjacency, and a fast-growing global delivery-robot and delivery-drone market; it does not support one clean China-only robovan TAM. That is why this chapter preserves proxy-based sizing and explicit evidence gaps. [CM001, CM002, CM003, CM004, CM005, CM006]

2.2 Policy, Pilot Zones, and Deployment Readiness

China's policy picture is moving from experimentation toward more standardized commercialization, but deployment still depends heavily on where a route sits and which regulator owns it. Suzhou and broader Jiangsu are leading current ground-delivery operations, while Beijing's high-level autonomous-driving demonstration zone is becoming a benchmark for what large-scale public-road readiness looks like. By 2025, Beijing had already issued permits to 33 companies covering about 900 vehicles and more than 32 million autonomous-test kilometers, and the city plans to expand its zone to roughly 3,000 square kilometers. Beijing's new autonomous-vehicle regulation, effective April 2025, explicitly opens a road-application pilot path for urban operational services including logistics. National policy is also supportive. The 2026 government work report named the low-altitude economy as an emerging pillar industry, which matters because China increasingly treats autonomous logistics as an ecosystem question touching roads, depots, drones, vehicle-road-cloud infrastructure, and industrial standards rather than as one narrow product category. Still, the source pack shows that commercialization remains jurisdiction-specific. The presence of Beijing's pilot path, Jiangsu's early legislation, and Meituan's national low-altitude certificate is strong evidence that policy friction is falling, but not that open-road deployment is turnkey nationwide. Investors should underwrite geography and route eligibility as hard market filters, not as administrative details. [CM013, CM014, CM015, CM016, CM017, CM018]

2.3 Buyer Segments and Workflow Economics

The buyer map is workflow-led. Postal and express operators buy autonomous logistics to defend parcel throughput and last-mile economics; platform operators buy it to keep delivery fees low and integrate autonomous capacity into existing digital networks; grocers and supply-chain operators buy it to move palletized or boxed goods between depots; and large cross-border or contract-logistics operators test it first in controlled facilities before seeking public-road approval. The same technology stack therefore sells into two different budget logics: last-mile relay is usually justified by parcel economics and labor substitution, while middle-mile and depot-transfer use cases are justified by route reliability, DC productivity, and vehicle utilization. The evidence suggests that buyers care most about cost-to-serve and public-road readiness, not abstract autonomy scores. SingPost frames public-road deployment as the lever for reducing last-mile cost. FairPrice and DHL both tie AV adoption to moving goods between distribution centers and to reducing manual routine tasks. KR-Asia and 36Kr report that unmanned vehicles can halve cost per parcel in some settings, reduce customer operating cost materially, and cut remote labor share after deployment matures. Product specs also show why last-mile and middle-mile should not be conflated: Zelos' Z5 is a smaller urban-delivery workhorse, while the Z10 pushes into heavier, longer-range middle-mile freight. That segmentation matters because the buyer, route, and ROI threshold change with vehicle class. [CM022, CM023, CM024, CM025, CM026, CM027]

2.4 Competitive Structure, Barriers, and Sizing Gaps

China's autonomous-logistics market is no longer one-company-one-scenario. The competitive structure now includes specialist robovan companies such as Zelos, Neolix, and White Rhino; platform-backed operators such as Cainiao, JD, and Meituan; and adjacent automotive or supplier entrants moving in from 2025 onward. Cainiao's decision to fold its autonomous-driving business into Zelos shows how distribution power and route access can matter as much as vehicle hardware. China Post's 7,000-vehicle order likewise suggests that large procurement rounds can quickly concentrate volume with scaled vendors. The barriers are equally real. Public-road commercialization still depends on pilot permissions, route qualification, remote-operations economics, and adverse-weather or cold-start performance. Taibo's account of JD's IP complaint adds a second barrier category: legal and trade-secret risk in a market where founding teams and engineers have migrated from large incumbents. The source pack also preserves a more mundane but important analytical problem: public sources do not agree on Zelos' current fleet scale, with partner materials citing more than 20,000 deployed vehicles and product pages citing more than 25,000 active vehicles. Combined with the absence of a clean published China-only robovan TAM, that means valuation work should underwrite adoption with multiple proxies and explicit diligence asks rather than with one headline market-size slide. [CM031, CM032, CM033, CM034, CM035, CM038]

3.1 China Direct Operators and Peer Classes

The direct China field is not a single homogeneous peer set. Zelos positions itself as a global RoboVan platform with more than 25,000 autonomous logistics vehicles, over 100 million kilometers, and operations in 20-plus countries and 350-plus cities, which is larger than any other disclosed robo-van fleet in this source set. The closest like-for-like competition comes from Neolix/New Stone and White Rhino, which both market purpose-built autonomous delivery vehicles for dense urban logistics. Neolix claims more than 15,000 vehicles deployed, a RoboVan-as-a-Service model, and a dense Qingdao benchmark fleet, while White Rhino emphasizes same-city grocery, supermarket, and instant-delivery operations with Huolala and SF Same-City. Cainiao Xiaomanlv and JD Logistics are structurally different but strategically important because they combine autonomy with existing parcel or e-commerce infrastructure. Cainiao already claims L4 public-road capability, 7x24 service, and more than 10 million delivered parcels, while JD has shown both emergency last-mile deployment and a procurement plan measured in millions of robots, vehicles, and drones. WeRide belongs in the direct set only for heavier open-road cargo benchmarking: its Robovan W5 is more truck-like in payload and range than campus or sidewalk robots, but it proves that another China autonomy company is commercializing urban logistics on public roads with public-company disclosure.[CP001, CP002, CP003, CP004, CP006, CP007]

3.2 Operating Model and Global Benchmark Comparison

The global benchmark set is useful mainly for operating-model contrast, not for one-to-one Chinese substitution. Starship and Serve prove that autonomous delivery can reach public scale, but the evidence is strongest in sidewalk, campus, or tightly geofenced formats rather than heavier public-road robo-vans. Starship discloses more than 10 million deliveries and 2,700-plus robots, while Serve discloses tens of thousands of deliveries and a signed plan to deploy up to 2,000 robots on Uber Eats. Nuro is the clearest cautionary benchmark: it still reports years of driverless deployments and more than 1.7 million autonomous miles, but its 2025 fundraise came at a lower valuation and alongside a shift away from operating goods-only vehicles toward licensing autonomy. Gatik sits outside last-mile delivery but is still relevant because it shows a different path to enterprise adoption: autonomous transportation as a service on repeatable middle-mile routes, combined with heavy stakeholder engagement and structured safety validation. Against those benchmarks, Zelos looks less like a consumer app network and more like a multi-scenario robo-van platform that still relies on bespoke enterprise packaging rather than transparent list pricing. Public sources reveal service models, payload domains, and support structures, but not clean apples-to-apples contract economics, which makes capability and channel access more observable than gross margin or realized price.[CP019, CP020, CP021, CP022, CP023, CP029]

3.3 Regulatory Access and Channel Power

The most important competitive pattern in public evidence is that channel ownership and regulatory access often matter more than autonomy software in isolation. Cainiao combines robo-van development with parcel-station density, campus traffic, and broader e-commerce logistics infrastructure; Zelos itself says Cainiao injected its RoboVan business and became a shareholder, which can help distribution but also raises platform-dependence questions. JD brings an even larger logistics network and has already demonstrated public-road delivery robots, indoor hospital robots, and an automation procurement plan that dwarfs what specialist vendors disclose. Meituan is not a direct robo-van vendor, but it is the clearest proof that Chinese logistics incumbents can turn regulatory approval into a structural moat: by April 2025 it had secured the country's first nationwide low-altitude logistics operating certificate, after already building 53 routes and more than 450,000 deliveries. AutoX and DeepRoute.ai matter because they show adjacent autonomy stacks with permits, OEM backing, or mass-produced smart-driving deployments that could move into logistics if the commercial case improves. For Zelos, the implication is that winning in China likely requires not only strong robo-van technology, but also privileged route access, order aggregation, and operating partners that can match the embedded advantages of Cainiao, JD, or Meituan.[CP005, CP012, CP013, CP024, CP025, CP026]

3.4 Moat Durability and Displacement Risks

Zelos has real differentiation in the public record: a broad claimed fleet footprint, multi-country operations, a purpose-built robo-van product line, and open-road plus closed-road deployment language that is more expansive than most private China specialists disclose. That matters because many competitors either focus on a narrower operating domain, such as White Rhino's retail same-city use cases, or on different autonomy categories, such as Meituan drones, AutoX robotaxis, or Gatik middle-mile trucks. The durability question is whether Zelos can turn those product and scale claims into defensible access. Public evidence points to four main risks. First, incumbent distribution power from Cainiao, JD, and Meituan can compress the value of independent vehicle vendors. Second, exact customer pricing and realized economics remain opaque across nearly all peers, so the market still lacks hard evidence on who can profitably operate robo-vans at scale. Third, the best-documented global scale benchmarks today are Starship and Serve, which run lighter, more geofenced formats than Zelos. Fourth, Nuro's down-round and licensing pivot show that capital intensity can force even well-funded pioneers to reframe their go-to-market model. The chapter therefore supports a tempered conclusion: Zelos appears competitively credible, but its moat is not yet publicly proven without clearer win-loss data, pricing evidence, and direct disclosure on how Cainiao, regulators, and enterprise customers shape recurring demand.[CP001, CP002, CP003, CP033, CP041, CP042]

4.1 Revenue model, pricing visibility, and GTM motion

Zelos does not look like a pure software company in the retained record. The clearest monetization evidence comes from a mix of registry scope, investor communication, and deployment announcements. Qichacha and Aiqicha show a business scope that includes AI software and hardware, IoT devices, vehicle and equipment rental, battery and charging equipment sales, and related logistics services. Blue Lake’s April 2025 financing note goes further by explicitly describing a “turnkey vehicle sales + full lifecycle operation and maintenance” model, while a later Sohu/Cyber-car article says Zelos first pushed low-cost hardware plus software installments before more mature products moved back toward fixed-price selling. What is still absent is equally important: none of the retained official product, partner, or financing pages discloses a public sticker price or standard contract value for the Z5, Z10, or enterprise service package. The go-to-market motion also appears enterprise-led rather than self-serve. SingPost, FairPrice, DHL, and China Post all point to large-account deployments or procurement programs, and Blue Lake’s B4 note adds a national distributor and service-provider network that looks more like channelized industrial sales than app-like subscription acquisition.[CI001, CI002, CI003, CI004, CI005, CI011]

4.2 Unit economics, cost structure, and public traction proxies

Public unit economics are directional rather than complete, but the pattern is consistent: Zelos is selling or operating heavy physical assets with a visible service tail, not an asset-light software SKU. The product pages show Z5 and Z10 payload, battery, and range specifications and include illustrative savings calculators, but both pages explicitly say those savings are examples rather than guaranteed realized outcomes. More useful operating proxies come from partner and media surfaces. Cainiao’s own unmanned-vehicle page says the platform can cut labor cost by 20%–30%, drive per-delivery cost to RMB0.1, and includes free maintenance, high-value insurance, and a five-year warranty on core parts. Blue Lake’s April 2025 note claims the turnkey model cuts customer operating cost by more than 50%, while Gasgoo cited a 62% reduction in November 2024. KrASIA’s translated 36Kr coverage then added a sharper but still secondary operating proxy: franchisees reportedly spend RMB2,000–3,000 per vehicle per month, with cost per parcel falling from RMB0.2 to RMB0.1. On the cost side, Jiemian’s report of six factories and 45,000 units of annual capacity, plus the March 2026 Xinghui materials investment, both reinforce that supply chain and vehicle economics matter at least as much as software margins.[CI006, CI007, CI008, CI009, CI010, CI014]

4.3 Funding history, investor base, and capital commitments

Zelos’s financing chronology is one of the better-supported parts of the public record, although exact lifetime capital still depends on secondary summaries. Gasgoo reported a $100 million B1 in November 2024 led by CDH BAIFU and Blue Lake. Blue Lake then said its April 2025 B3 round added another $100 million and brought Series B financing to nearly $300 million, with Asia Investment Capital and Baidu Ventures participating. The strongest new primary-like funding anchor is Blue Lake’s October 2025 B4 post, corroborated by EqualOcean, which says Ant Group led a fresh $100 million round and Series B cumulative funding reached $400 million. After that, the evidence becomes more strategic and more secondary. Reuters-derived U.S. News coverage said Cainiao would inject its autonomous-driving business plus cash, would not take control, and that the combined robovan business was valued around $2 billion by the Wall Street Journal. February 2026 press then converged around another >$300 million financing and valuation above RMB10 billion, while April 2026 Chinese financial media circulated a roughly $600 million Hong Kong IPO plan that remained unconfirmed. The net effect is clear access to capital, but weaker evidence on post-money mechanics and proceeds already consumed.[CI019, CI020, CI021, CI026, CI027, CI028]

4.4 Financial verdict and diligence blockers

The financial verdict is mixed but investable only with management support. Zelos has enough public deployment proof, financing history, and enterprise customer evidence that a zero-revenue or purely promotional story is not credible. The company plainly appears to monetize through some combination of vehicle sales or leases, service and maintenance, and channel or fleet operations support. It also appears materially capital intensive: the product is hardware-heavy, maintenance obligations are explicit, national service networks are large, factories and annual capacity are discussed in press, and 2026 filings around Xinghui point to upstream supply-chain positioning rather than a purely outsourced model. But the underwriting blockers remain severe. No retained source discloses revenue, ARR, gross margin, burn, runway, debt, or customer concentration, and even the most repeated valuation markers are still press-level rather than financing-document-level evidence. The adverse overlay matters too. Taibo’s JD intellectual-property report is material because it links legal uncertainty directly to valuation and IPO eligibility. Investors can therefore underwrite that Zelos is scaled, financed, and still funding-dependent; they cannot yet underwrite revenue quality, margin path, or capital adequacy with conviction from public sources alone.[CI005, CI022, CI023, CI024, CI025, CI028]

4.5 Exhibits

5.1 Product family and customer workflow fit

Zelos’ current official product surface is built around logistics-specific RoboVan variants rather than a single generic autonomous van. The CMS-backed product catalog exposes five currently named Z-series SKUs: Z5, Z5-Cold Chain, Z5 Multiple Locker, Z10, and Z10-Cold Chain, each tied to a distinct workflow and payload profile. The smaller Z5 is positioned for urban and campus delivery, reverse pickup, and shuttle operations; the Z10 moves into heavier middle-mile or depot-transfer work; the cold-chain and locker variants adapt the same autonomy base into temperature-controlled or multi-drop last-mile configurations. The scenario layer is equally workflow-led. Zelos’ public solutions materials map the platform into FMCG, pharmaceutical, business-park, parcel, and retail-replenishment jobs, while partner-detail cards add named operating examples such as Yum store replenishment, Sinopharm hospital-campus delivery, DHL warehouse logistics, and FairPrice public-road pallet transfer. The most important underwriting distinction is maturity by branch. The Z family has clear published specs and the strongest deployment proof. By contrast, 36Kr and supplier pages point to newer E-series and L-series branches, but those newer lines are less fully documented on Zelos’ own product catalog. The right chapter-5 read is therefore that Zelos already has a real, differentiated RoboVan portfolio for B2B logistics, but investors should separate the mature, spec-visible Z family from the newer E/L roadmap branches whose exact trims, economics, and certification status are not yet equally transparent.[CE001, CE002, CE003, CE004, CE005, CE006]

5.2 Autonomy stack and safety architecture

Zelos publishes a more concrete autonomy stack than many private AV logistics peers, but the evidence is still uneven between architecture claims and formal safety-case disclosure. The company homepage and about materials consistently describe a map-free, proprietary, full-stack L4 system built in-house from vehicle engineering through autonomous driving software. The technical modules are more specific: one module states that the perception stack fuses lidar, radars, cameras, and ultrasonic sensors into 360° awareness with millimeter-level precision; another says a next-generation end-to-end model uses a bird’s-eye-view representation to jointly model roads, vehicles, and pedestrians while removing HD-map dependence; and a third claims the onboard L4 compute unit operates below 100 W. Partner sources deepen the hardware view. Hesai said the 2024 Z5 used four AT128 long-range lidars, while Seyond later said the newer E6 uses two Lingque W lidars plus millimeter-wave radar, ultrasonic radar, and multi-camera sensing. On the operations side, DHL’s Singapore deployment adds advanced navigation, real-time monitoring, built-in safety protocols, and warehouse-management- system integration, while SingPost’s expanded partnership explicitly covers fleet monitoring, management, maintenance, and training. That is enough to conclude that Zelos is selling a full operating system for logistics autonomy, not just an AV kit. What is still missing is the hard underwriting layer: retained sources do not publish max speed, braking distance, MTBF, disengagement, remote-ops staffing ratios, or a public ISO 26262 / SOTIF / ISO 21434 style assurance package. So chapter 5 can support a credible stack narrative and specific safety mechanisms, but not a complete public safety case.[CE014, CE015, CE016, CE017, CE018, CE019]

5.3 Deployment model, fleet operations, and ODD

Zelos’ operational-design-domain story is best understood as a phased expansion from controlled or private logistics environments into designated public-road freight routes, with Singapore as the clearest template market. Customer and official materials show the same pattern repeatedly. DHL’s first phase is in-plant movement inside the Asia Pacific Advanced Regional Centre, with public-road expansion only in phase two and only subject to regulatory approval. FairPrice’s 2024 trial similarly progressed from test drives with a safety vehicle to a second phase without an accompanying safety vehicle, and Zelos’ own October 2025 news item says the year-long pilot achieved zero takeovers and zero incidents before nearly 30 Z10s were slated for scaled rollout. SingPost adds another stage gate: hub operations and training first, then public-road driverless deployment with explicit fleet-management and maintenance roles. Malaysia’s Pos Malaysia PoC starts from the National Mail Centre before any broader operational or public-road rollout. The workflow detail behind these deployments matters. Zelos is solving repetitive B2B logistics jobs—DC-to-DC pallet transfer, warehouse material movement, parcel relay, pharma campus replenishment, and park logistics—not generalized urban autonomy. That focus likely narrows the ODD to structured freight routes with repeatable loading patterns and operational oversight, which is a strength for commercialization. It also means investors should underwrite public-road autonomy as a route-qualified, regulator-dependent capability rather than as a universally available city-driving product. The real product is vehicle plus dispatch, fleet monitoring, training, and compliance progression.[CE010, CE012, CE013, CE022, CE023, CE035]

5.4 Manufacturing, supplier dependencies, roadmap, and open gaps

Zelos has stronger manufacturing-maturity evidence than an average robotics startup, but its next-generation platform roadmap is still only partly transparent. The official milestone record says the first prototype launched in May 2022 to validate self-driving systems and hardware, Z5 mass production started in June 2023, and the all-scenario Z-series platform launched in June 2024. Those milestones, together with supplier announcements from Hesai and later Seyond, support a real production program rather than a lab-only prototype effort. The Cainiao integration then moves the story beyond a standalone vehicle OEM: Zelos says Cainiao injected its RoboVan business, joined the shareholder base, and contributed management, manufacturing, operations, and VLM/world- model research capabilities. Official materials also say L-series models and L4.5 technology were launched in September 2025, while 36Kr describes an emerging three-branch portfolio: Z for standard products, E for lighter and lower-cost use cases, and L for heavier-load demand. The problem is that the newest branches are not yet documented with the same clarity as the current Z catalog. Public sources also suggest sensor-stack evolution rather than a stable bill of materials: Z5 2024 used four Hesai lidars, partner reporting on E6 highlights dual Seyond lidars, and a March 2026 media item says the latest-gen RoboVan lineup will standardize on a 200,000-unit Hesai lidar order. That does not invalidate the roadmap, but it does create diligence questions on trim management, supplier concentration, and whether different markets are using different hardware generations. Public patent visibility is also thin: retained sources do not enumerate patent numbers or technical papers. So the roadmap is credible, but still not fully underwritable without deeper factory, BOM, and safety documentation.[CE024, CE025, CE026, CE027, CE028, CE029]

5.5 Exhibits

6.1 Customer Segments and Scenario Map

Public evidence shows Zelos sells into workflows rather than one narrow customer archetype. The cleanest segment split is between postal and express operators, grocery and retail supply chains, contract-logistics or industrial operators, and ecosystem channels that can resell or operationalize vehicles at scale. Zelos’ own surfaces repeatedly highlight last-mile parcel, retail, campus, industrial-park, intralogistics, and pharmaceutical use cases, while third-party coverage expands that map into fresh-food, dairy, electronics, grain-and-oil transport, tires, auto parts, and hotel-linen logistics. Buyer, user, and payer also shift by segment: postal networks and courier operators care about depot-to-station relay and route density; retailers care about distribution-centre transfers and shelf-availability; industrial hubs care about closed-scenario yard productivity; and channel partners care about standardizing a fleet-and-service layer across many end accounts. That scenario-first framing matters because it explains why Zelos can show proof in campuses, communities, industrial parks, and logistics hubs without implying one homogeneous customer base.[CU001, CU002, CU003, CU004, CU005, CU006]

6.2 Named Deployments and Public Adoption Proof

The strongest public customer proof sits in Singapore and postal logistics. FairPrice provides the clearest production-like public-road reference: a customer press release says it received LTA approval, signed a 2025 collaboration agreement, and aimed to add close to 30 Zelos vehicles for distribution-centre transfers. SingPost is the best durability proxy because both sides describe a sequence from hub trials since December to an expanded 2026 MoU focused on public-road commercialization, fleet management, and distribution business development. DHL and Infineon show a different pattern: an in-plant deployment with a defined phase-two path to public roads and more customer operations. Outside Singapore, Pos Malaysia is still in PoC mode, while KEZAD is an industrial-hub deployment rather than a broad public-road network. China Post is the opposite extreme: large procurement scale, but limited public disclosure on realized utilization and economics. The chapter therefore supports real adoption proof, but the proof varies sharply by customer and maturity stage.[CU005, CU006, CU007, CU008, CU009, CU010]

6.3 Procurement, Go-to-Market, and the Cainiao / Alibaba Channel

Zelos does not appear to rely on a single go-to-market pattern. Jiemian’s ADS-center reporting suggests the company is deliberately widening procurement options across direct sale, leasing, and fully managed operations, which fits a market where some customers do not want vehicle ownership or operational burden. Customer proof in Singapore supports that reading: SingPost’s MoU adds team training, seed vehicles, and distribution-business development rather than only hardware supply. The Cainiao relationship deepens the model further. Zelos’ official release, Gasgoo, and Reuters all describe a business-and-equity integration in which Cainiao stops directly manufacturing or selling vehicles, Zelos operates both brands, and the Cainiao brand focuses on logistics scenarios and large enterprise clients. That should help enterprise access and international deployment speed, especially within Alibaba’s ecosystem and warehouse network. But it also blurs the boundary between customer, shareholder, and channel. For diligence, the key underwriting question is not whether Cainiao helps — it clearly does — but how much of future volume becomes dependent on one ecosystem-controlled scenario funnel.[CU022, CU023, CU024, CU025, CU026, CU027]

6.4 Durability, Expansion Proxies, and Customer Risks

Public durability evidence is directional, not complete. The positive side is real: FairPrice moved from early cooperation to approved public-road rollout; SingPost moved from hub trials to a broader strategic partnership; DHL framed its current deployment as phase one with additional customer operations planned; and Pos Malaysia plus KEZAD show the model translating into postal and industrial export settings. Those are credible expansion proxies. The missing side is equally important: public sources do not disclose contract length, renewal rates, churn, NRR, GRR, top-customer revenue share, or overall customer count. That means the chapter can defend adoption breadth and channel momentum, but not true retention or concentration comfort. Customer risk is also not hypothetical. Securities Times reported multiple 2026 incidents involving Zelos-branded delivery vehicles, including collision and route-compliance allegations, which could affect customer trust, deployment permissions, and community acceptance if repeated. The result is a balanced judgment: customer proof is materially stronger than a logo wall, yet still weaker than a fully underwritable enterprise-retention dossier.[CU018, CU019, CU020, CU021, CU031, CU032]

6.5 Exhibits

7.1 Ranked risk view and thesis-break framing

The public record supports a clear ordering of Zelos's risk stack. The first bucket is regulatory and legal: national Chinese rules for unmanned delivery vehicles are still incomplete, city-by-city route-right and insurance obligations remain decisive, Singapore scale-up is still subject to approvals, and JD's IP allegation remains unresolved on primary public documents. The second bucket is partner and customer dependence. Cainiao's asset-plus-cash integration undeniably improves scale and scenario access, but it also pulls Zelos closer to one ecosystem's priorities at the same moment that China Post's 7,000-vehicle tender appears large enough to become a material execution burden. Third is capital intensity and operating leverage: public sources describe 25,000+ deployed vehicles, multi-factory manufacturing, and breakeven only at much larger fleet scale, implying little room for route-right or utilization disappointment. Fourth is operational safety and execution. Sector-level legal reporting shows that accidents, privacy, and remote-safety responsibilities are still being worked out in public roads. The correct underwriting stance is therefore not to reject Zelos's traction, but to treat regulatory, platform, concentration, and IP diligence as thesis-critical rather than incidental.[CR001, CR010, CR016, CR020, CR024, CR025]

7.2 Regulatory, legal, and policy risk

China's unmanned-delivery policy surface is still supportive but not fully normalized. Xinhua and CCTV both frame the core problem as unfinished legal status, with local pilot rules doing most of the real work while national qualification, liability, and vehicle-classification questions remain open. That matters because Zelos is trying to scale a cross-city and cross-border fleet, not a single-zone demo. Beijing's and Shenzhen's local rules impose insurance, monitoring, emergency-response, and designated-route controls; national standard-setting is still advancing around safety, cybersecurity, and operating conditions rather than offering a single settled national regime. Singapore is more commercially promising but legally similar in one important sense: FairPrice, SingPost, DHL, and Zelos itself all describe a phased path in which larger public-road deployments remain approval- dependent. The second legal thread is JD's complaint. Taibo and QQ News say JD reported suspected IP infringement to public security and that the matter remained under investigation, while 36Kr argues that a formal escalation could become material to IPO timing. That does not prove wrongdoing. It does show that Zelos enters any financing or listing process with an unresolved, specifically alleged, potentially high-severity legal overhang.[CR001, CR002, CR003, CR005, CR006, CR007]

7.3 Partner dependency, operations, and safety

Zelos's strongest commercial proof also highlights its dependency risk. Reuters and Chinese press agree that Cainiao is not just a passive investor: it is contributing assets, integrating teams, licensing a brand, and moving key-account and logistics-scene capability into the combined structure. That should help route acquisition, data scale, and overseas rollout, but it also makes Zelos more exposed to one ecosystem's capital allocation, platform priorities, and strategic shifts. Customer concentration is similarly real but hard to quantify. China Post's 7,000-vehicle order is described as the largest procurement in the segment, and third-party coverage says Zelos will shoulder roughly half of the new capacity. SingPost, FairPrice, and DHL are credible overseas references, but all three are phase-based programs rather than disclosed long-dated revenue annuities. Operationally, the company is attempting to run multi-scenario, map-free L4 vehicles across 300+ cities, 20+ countries, and a growing set of anchor accounts. Sector legal reporting confirms that open-road incidents, privacy handling, emergency response, route compliance, and remote-safety responsibility are still live issues. A company-specific safety pattern is not proven on retained primary records, but 36Kr's reported incidents are enough to justify diligence on loss runs, disengagements, and root-cause controls before treating fleet scale as pure strength.[CR016, CR017, CR018, CR019, CR020, CR021]

7.4 Capital intensity, people, and downside scenarios

Zelos is no longer a lightweight software thesis. Public sources describe a business with tens of thousands of vehicles, multiple factories, aggressive capacity targets, international expansion, and a model that still appears to need far larger fleet scale before full breakeven including R&D. That makes capital efficiency and organizational execution central risks. Jiemian's interview material suggests the company is trying to expand capacity, lower unit costs, and preserve rapid growth simultaneously while also integrating Cainiao and pursuing overseas markets. Sohu and 21jingji show why that can become structurally difficult: large customers are under intense cost pressure, route-rights remain fragmented, and sector competition is pushing low upfront pricing plus recurring software subscriptions. People and public- acceptance risk are therefore subtler than an outright consumer backlash. Xinhua says social acceptance is improving, but local rules, safety staffing, partner training, and compliance operations still require significant human systems around the vehicles. The evidence-based downside scenarios are not hypothetical abstractions: a formal IP escalation, a serious public-road incident, a route-right slowdown, or a failure to diversify beyond Cainiao and mega-accounts would all transmit quickly into deployment pace, financing leverage, and valuation support.[CR023, CR024, CR025, CR026, CR027, CR028]

7.5 Exhibits

8.1 Valuation evidence and confidence

Zelos enters valuation work with real strategic momentum but weak pricing precision. The strongest positive evidence is that the company has accumulated a meaningful financing history, pulled Cainiao’s robovan business into its orbit, and is consistently described in February 2026 coverage as having raised more than $300 million at a low-to-mid unicorn price. The problem is that the retained sources do not resolve to one exact figure. Reuters-derived reporting anchors a roughly $2 billion valuation for the robovan business associated with the Cainiao transaction, while multiple February 2026 Chinese and English reports cluster around “above RMB 10 billion,” which is roughly $1.4 billion. That is not a trivial discrepancy. It means the public record supports the direction of a unicorn mark, but not the exact number. Because the same source set also fails to disclose revenue, gross margin, or financing terms, confidence in the precise February 2026 price must stay low even though confidence in Zelos’s strategic relevance is much higher.[CV003, CV004, CV005, CV006, CV007, CV008]

8.2 Thesis, anti-thesis, and comparable set

The positive case for Zelos is that it may already control the most scaled independent robo-van footprint disclosed in this source set: the company says 25,000-plus vehicles and 100 million-plus kilometers, partner and media sources still corroborate a 20,000-plus vehicle base, and a reported China Post order suggests the platform can win national-scale procurement. Cainiao’s asset injection also matters because it is a stronger signal than ordinary venture backing: a major logistics platform chose combination over in-house continuation. The anti-thesis is equally important. Nearly every valuation-supporting fact is company-led, partner-led, or secondary-media-led rather than audited. Public comps such as WeRide and Serve prove investors will support rich autonomy multiples, but those businesses also publish financials and filings that Zelos does not. Nuro’s down-round is the more sobering lesson: when autonomy leaders cannot show a capital-efficient path, the market can re-rate quickly. Zelos therefore screens as strategically credible but disclosure-poor, which is why the thesis must stay price-sensitive rather than celebratory.[CV001, CV002, CV005, CV014, CV016, CV017]

8.3 Scenario range and entry discipline

Because Zelos does not disclose a live revenue denominator, the valuation range has to be constructed from evidence-supported anchors instead of false-precision multiple math. The key anchors are the conflicting February 2026 private marks, the stronger strategic signal from Cainiao’s contribution, the China Post order, and the public valuation envelopes visible in WeRide, Serve, and Nuro. In the bear case, investors eventually discover that the fleet headline overstates monetized scale or that the round terms contain heavy protection, forcing a steep discount to the strategic headline. In the base case, the RMB 10 billion cluster is directionally right, but the company still deserves a discount to the higher strategic mark until revenue and margin are auditable. In the bull case, the Cainiao combination, China Post rollout, and IPO path convert from story to disclosed operating proof. That framework implies entry discipline, not categorical rejection: the reported unicorn price is not obviously impossible, but it is too weakly documented to pay a premium without a data room.[CV014, CV015, CV016, CV024, CV028, CV034]

8.4 Recommendation, hard-stop triggers, and final diligence asks

The recommendation is TRACK / RESEARCH-MORE with low confidence and high risk. That stance is not a denial of Zelos’s progress; it is the consequence of missing underwriting variables. The retained record is strong enough to say that Zelos is more than a slide-deck company: it has visible partner deployments, a real Cainiao link, and a credible case for why strategic investors would want a robo-van champion. It is not strong enough to say that an exact February 2026 valuation is cleanly supported. A buyer should treat the current mark as workable only if diligence can convert the story into numbers—audited revenue, gross margin, ordinary investor terms, and customer economics on the China Post and Cainiao-backed programs. Until then, the right posture is to define hard-stop triggers rather than rationalize ambiguity. If the company cannot evidence monetization, if legal issues escalate, or if the IPO narrative weakens, the valuation case should compress quickly. If disclosure sharpens and repeat large-account economics hold up, the same recommendation can move upward fast.[CV015, CV017, CV018, CV019, CV020, CV040]