Black Sesame International is a Chinese fabless chip designer, meaning it creates the automotive compute system-on-chips (SoCs) behind advanced driver-assistance (ADAS) and autonomous driving but outsources their manufacturing, then sells them bundled with software and solutions. The report rates it Hold. Driving-assistance products and solutions are the core business and the bulk of revenue, with smaller intelligent-imaging and newly added embodied-AI lines around it. In 2025, revenue rose 73.4% to RMB822.3 million and group gross margin held at 41.0%, evidence that commercialization is real and the product mix is stable rather than deteriorating.
The fundamentals still describe a company that spends like a platform before it earns like one. In 2025, R&D alone reached RMB1.417 billion, roughly 1.7 times revenue, which is why the company stays deeply loss-making despite the growth. Operating cash flow was a net outflow of RMB985.4 million, still larger than gross profit, and owner earnings remain negative, meaning the business does not yet generate distributable cash. To fund the gap, Black Sesame leaned on a February 2025 placing and two further share subscriptions in 2026, so dilution has become a recurring feature of the story rather than a one-off event.
The moat is real but narrow. A1000 chips are already designed into models from Geely, BYD, Dongfeng, and FAW, and automotive qualification makes that kind of design-in slow and costly to displace. Yet the report sees no scale moat against the larger domestic leader Horizon Robotics, no installed-base or cash-conversion moat against the global benchmark Mobileye, and treats the embodied-AI push as optionality rather than a proven second growth curve; Ambarella serves only as an adjacent edge-AI reference point.
The stock has already rerated hard from its HK$28 IPO price to HK$10.38, and now trades at roughly 7.8x 2025 sales (P/S). The report calls that no longer an IPO-euphoria multiple though not distressed either, one that still assumes substantial future scaling. It finds no margin of safety at the current price and places a true entry only at an Ideal Buy Price of HK$5.2 to HK$5.8. The biggest risks it flags are intelligent driving moving downmarket faster than the cost structure and compressing margins, continued cash burn forcing more discounted equity raises, and the next-generation A2000 platform still having to prove mass-production economics, not just designated projects.
The report's stance is patience over enthusiasm: a credible challenger that has crossed the technical-credibility threshold but not the cash-conversion one, where the current price offers only a modest cushion until execution and financing discipline improve. The above is a summary of the report's views and does not constitute investment advice. Markets carry risk; invest with caution.
Prices in the article are as of publication; see the valuation band above for the live price.
Meta
- Ticker: 02533.HK
- Company: Black Sesame International Holding Limited
- Price & market cap: HK$10.38 close as of 2026-06-26; market cap about HK$7.4 billion on market-data services around the same date
- Currency: HKD; financials reported in RMB are translated at 1 RMB = HK$1.1519 using the CNY/HKD close on 2026-06-26
- Report date: 2026-06-28
- Industry: Semiconductors
- One-line positioning: A loss-making automotive compute-SoC designer that sells ADAS chips and solutions, with 2025 revenue of RMB822.3 million and continuing heavy R&D investment.
Research summary
This report uses a general-investment lens, a 12-month and 3–5-year horizon, and a balanced risk posture. The base currency is HKD because the stock trades in Hong Kong, but the business still has to be understood first in RMB, because that is how the company reports revenue, costs, cash burn, and R&D. The central question is not whether Black Sesame is exposed to a good industry. It plainly is. The real question is whether the company has crossed the line from being a technically credible autonomous-driving chip hopeful into becoming a durable commercial supplier that can finance its own growth without repeatedly leaning on equity markets. The evidence, as of 2026-06-28, says that it has moved part of the way, but not all the way.
What kind of company is it, really? In plain terms, Black Sesame is no longer just a chip design story. It is a bundled automotive-compute supplier selling driving-assistance products and solutions, with smaller intelligent-imaging and newly added embodied-AI businesses around that core. In 2025, driving-assistance products and solutions contributed RMB686.9 million of revenue, or about 84% of total sales; intelligent imaging contributed RMB39.2 million; embodied AI and solutions, which barely existed as a revenue line a year earlier, contributed RMB96.3 million. Gross margin at the group level held at 41.0%, but that stability masks a more interesting truth: margins are far stronger in imaging and decent in embodied AI, while the big revenue engine remains the automotive business, where scale matters and price pressure is real. The company still loses money because it spends far more than gross profit on R&D and SG&A. In 2025, R&D alone was RMB1.417 billion, roughly 1.7 times revenue.
The market is mainly trading two narratives at once. The first is the obvious one: China wants domestic alternatives in intelligent-driving compute, especially as smart-driving functions move down from luxury vehicles into the mass market. The second is more speculative: A2000 and SesameX might turn Black Sesame from a narrow ADAS chip vendor into a broader on-device AI platform company spanning cars, robotics, and embodied AI. Management leaned into both themes in the 2025 annual report, tying A2000 to L2/L3 expansion, Apollo Go robotaxi collaboration, overseas opportunities, and a newly launched embodied-intelligence platform. Those claims are not fictional. The company really did say them, and it really did post its first meaningful embodied-AI revenue. But the market has already seen enough China-semiconductor stories to distinguish between “promising platform” and “proven shipment curve.” Black Sesame is still closer to the former than the latter.
The past share-price path makes that divide visible. The stock came to market on 2024-08-08 at HK$28, carrying the scarcity premium of a specialist-technology listing and the publicity of being presented as the first listed Chinese autonomous-driving compute-chip company. That premium did not survive contact with the usual post-IPO realities: thin earnings quality, continuing cash burn, lock-up expiries in 2025, and then repeated equity injections in 2025 and 2026. By 2026-06-26, the stock had fallen to HK$10.38, while market data services showed a 52-week high of HK$26.38 and a 52-week low of HK$10.34. In other words, the market has already rerated the name from “scarcity growth asset” to “execution-required funding story.”
The biggest bull-bear disagreement is simple. Bulls think the company has already done the hardest part: getting automotive-grade chips into production vehicles, proving that Chinese OEMs and Tier 1s will buy from it, and extending the product map to A2000 and embodied AI before the domestic market fully tips into high-volume smart-driving adoption. Bears think the difficult part is still ahead: A1000 proved technical credibility, but A2000 still has to prove large-scale economics; embodied AI may be exciting, yet it also risks distracting a still-loss-making car-chip company; and the need for a 2025 placing plus two 2026 subscriptions suggests that internally generated cash is still too far away. Both sides have evidence. Revenue rose 73.4% in 2025, and the company says A2000 completed relevant review work, began deep adaptation with partners such as DeepRoute.AI and Nullmax, and expects mass production. At the same time, 2025 operating cash outflow was RMB985.4 million, and the company relied on a February 2025 placing and two new-share subscriptions in early 2026 to extend its runway.
On fundamentals alone, the company sits in a middle state that public markets often find awkward. It is too commercialized to be valued like a pure concept stock, but not yet commercialized enough to be valued like a self-funding semiconductor compounder. That is why 2024 headline profitability was largely a mirage: the reported profit was driven by a one-off fair-value gain tied to financial instruments issued to investors, mainly because preferred shares converted into ordinary shares upon listing. Once that effect dropped out, the underlying picture came back into view in 2025: stronger revenue, stable gross margin, lower adjusted loss, but still a large operating deficit. The right label is therefore not “high-quality growth” and not “valuation bubble.” It is a company in transition, moving from validation to commercialization and from single-track automotive compute toward a broader on-device AI ambition.
That label matters because companies in transition do not usually earn the market’s full trust all at once. They earn it step by step. For Black Sesame, the steps that matter are now obvious. The market wants proof that A2000 can move from designated projects to meaningful mass-production revenue; proof that embodied AI becomes a real second leg rather than a presentation deck; proof that gross margin can hold even as intelligent-driving functions move into lower-priced vehicles; and proof that the next capital raise is not around the corner. Until those proofs arrive, the stock will keep trading less on the size of the total addressable market and more on quarterly evidence that the business machine is becoming more real.
The qualitative portrait, then, is clear enough. Black Sesame is a company in transition with genuine technical credibility, improving commercialization, and a still-unfinished capital-markets burden. It is not hard to classify because the business is mysterious. It is hard to classify because two versions of the company coexist at once: the rising domestic automotive-compute supplier that has already won real business, and the capital-hungry pre-profit platform aspirant that still has to prove durable cash conversion. The market is discounting both at the same time.
Company vertical history
Origins and listing path
Black Sesame was founded on 2016-07-15. The founder profile explains a great deal about the company’s shape. Chairman and CEO Shan Jizhang came from OmniVision, where he spent nearly two decades and ended as a vice president in software engineering; the annual report says he holds more than 100 patents in visual perception. Co-founder Liu Weihong brought the other half of the puzzle: senior roles at Bosch Automotive Products, General Motors China, and Chassis Brakes International. One founder came from imaging semiconductors and core chip R&D; the other came from the realities of automotive procurement, systems, and industrialization. That is why Black Sesame did not emerge as a generic AI start-up. It was built from the start around automotive-grade compute, where semiconductors and vehicle-industry process discipline meet.
The company listed on HKEX’s Main Board on 2024-08-08. The interim report confirms the listing date, while the prospectus materials show that the global offering comprised 37 million shares priced at HK$28 each, implying gross proceeds of about HK$1.04 billion. Public-facing deal materials also framed the company as a Specialist Technology Company under Chapter 18C and highlighted its standing, according to the prospectus, as the world’s third-largest supplier by 2023 shipments of automotive-grade high-computing-power SoCs. That was the IPO story in one sentence: scarce domestic autonomous-driving silicon, now available in listed form.
The pre-IPO equity structure already hinted at what public investors would later face. Prospectus-era lock-up materials show a six-month unlock window ending 2025-02-07 for pathfinder SIIs and a longer 12-month window ending 2025-08-07 for another group of pre-IPO holders. That does not prove those dates caused later price declines on their own, but it does show that supply pressure was built into the calendar from the start.
Stages of development
The first stage, from founding through roughly 2021, was a product-validation period. The company’s job was to prove that it could design automotive-grade compute silicon instead of just talking about it. Revenue was tiny, just RMB60.5 million in 2021, but the foundation was already visible: proprietary ISP and NPU work, automotive-grade software processes, and the early construction of a bundled chip-plus-solution model. Financially, this was the classic deep-tech phase: modest sales, large operating losses, negative equity, and heavy dependence on shareholder capital.
The second stage, roughly 2022 through 2023, was the first commercialization turn. The A1000 family entered mass production in 2022, with more than 25,000 units shipped that year, and by the end of 2023 total A1000 family shipments had exceeded 152,000 units. Revenue rose from RMB165.4 million in 2022 to RMB312.4 million in 2023. The company also began offering autonomous-driving solutions in 2020 and, by its own prospectus-based description, was among the earliest Chinese companies to generate revenue from such sales. This stage mattered because it moved Black Sesame from “can build” to “can ship.” It did not solve profitability, but it changed the discussion.
The third stage was the IPO and normalization phase, spanning 2024. Revenue rose another 51.8% to RMB474.3 million, and group gross margin improved to 41.1%. Yet the most dramatic financial line in 2024 was the reported profit of RMB313.3 million, which came mainly from a RMB2.047 billion fair-value gain in financial instruments issued to investors. That was accounting clean-up tied to the capital-structure transition around listing, not evidence that the operating model had suddenly become profitable. The market is usually quick to see through that kind of accounting. It did. 2024 made the company more visible, but it did not settle the question of self-funding viability.
The fourth stage, beginning in 2025 and continuing into 2026, is the current one: scale-up with a second-curve experiment. On the core automotive side, 2025 revenue jumped to RMB822.3 million. The company says A1000 chips are already integrated into models from Geely, Dongfeng, BYD, and FAW, while A2000 secured designated projects from leading vehicle manufacturers and was expected to start mass production. On the adjacent side, SesameX was launched for embodied AI, helping create nearly RMB100 million of first-year embodied-AI revenue. At the same time, Black Sesame raised more external capital: a February 2025 placing, then a January 2026 first subscription and a March 2026 second subscription. This is the stage where the business is visibly broadening, but also where the capital-markets burden remains very real.
Financial vertical review
The long financial arc is straightforward. Revenue has compounded from RMB60.5 million in 2021 to RMB822.3 million in 2025, a more than thirteen-fold increase over four years. Gross profit rose from RMB21.9 million in 2021 to RMB337.1 million in 2025. But the operating-loss line tells the opposing story: RMB722.7 million in 2021, RMB1.05 billion in 2022, RMB1.70 billion in 2023, RMB1.75 billion in 2024, then some improvement to RMB1.45 billion in 2025. This is not a company whose costs have failed to scale at all. Selling and administrative expenses fell in 2025, and adjusted net loss improved. It is a company whose R&D burden remains large enough to overpower gross profit even after commercialization begins.
| Metric | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| Revenue RMB mn | 60.5 | 165.4 | 312.4 | 474.3 | 822.3 |
| Gross profit RMB mn | 21.9 | 48.6 | 77.1 | 194.7 | 337.1 |
| Operating loss RMB mn | -722.7 | -1,052.8 | -1,696.9 | -1,754.0 | -1,448.3 |
| Reported net profit or loss RMB mn | -2,356.5 | -2,753.9 | -4,855.1 | 313.3 | -1,424.7 |
| Adjusted net loss RMB mn | -613.6 | -700.3 | -1,254.2 | -1,304.3 | -1,075.7 |
The table shows a business that is commercializing, but slowly. What improved in 2025 was not just top-line scale. Cost discipline modestly improved too: selling expenses and G&A fell, while R&D stayed roughly flat in absolute terms at RMB1.417 billion rather than continuing to accelerate. That helped operating loss narrow. Still, “roughly flat” R&D at RMB1.4 billion on RMB822 million of revenue is another way of saying the company remains in investment mode. This is why the right valuation anchor is sales and future cash-generation capability, not current earnings.
Cash flow is the key discipline test, and it still says “capital consumer.” In 2025, net cash used in operating activities was RMB985.4 million, an improvement from RMB1.19 billion in 2024 but still very large against the revenue base. Cash and cash equivalents plus current financial assets at fair value through profit or loss totaled RMB1.531 billion at 2025 year-end. Management said that cash and funding from financing were sufficient for the next 12 months. That statement was reasonable only because financing kept arriving: the February 2025 placing generated net proceeds of HK$1.237 billion, of which HK$512.7 million remained unused at 2025 year-end; the first 2026 subscription added HK$538.1 million net; the second, once completed, added another HK$632.4 million net. My inference from those disclosures is that Black Sesame had pushed out near-term liquidity risk, but had not escaped the financing cycle.
Balance-sheet quality improved sharply after listing, but some caution remains appropriate. Total equity turned positive in 2024 because the preferred-share overhang disappeared with listing, and it remained positive in 2025. There was no auditor change after listing, and PwC remained auditor through the 2025 annual report. That lowers governance risk relative to many speculative technology names. At the same time, the business still has meaningful share-based compensation, a large pre-IPO option pool, and a voting-trust structure that amplifies founder influence beyond direct economic ownership.
Price and valuation history
Black Sesame’s valuation history has been short but revealing. At IPO, the market paid for scarcity, policy alignment, and the symbolic value of being a domestic autonomous-driving compute-chip listing. By mid-2026, that initial premium had largely washed out. The stock traded around HK$10.38 on 2026-06-26, down sharply from its HK$28 IPO price and far below its 52-week high of HK$26.38. The market did not simply turn bearish on intelligent driving. If it had, Horizon would not still carry a market cap above HK$50 billion. What changed was the center of gravity investors applied to Black Sesame specifically: from “could become important” toward “must now prove gross-profit durability and financing discipline.”
At the current price, the stock trades on roughly 7.8 times 2025 sales, using the latest close and translated 2025 revenue. That is no longer an IPO-euphoria multiple, but it is not distressed either. It still assumes substantial future scaling. The reason the multiple has compressed is not that the core business stopped growing. It grew. The reason is that the market now wants growth that spends less, dilutes less, and survives more than one product cycle.
Business model, industry and peers
How the business machine works
Black Sesame now runs three revenue lines. The core is driving-assistance products and solutions: SoCs, domain controllers, and intelligent front cameras sold into vehicle programs, often with software, middleware, and support around them. That business delivered RMB686.9 million in 2025 revenue and carried a 37.4% gross margin. The second line is intelligent imaging, much smaller at RMB39.2 million but far more profitable, with an 84.7% gross margin. The third is embodied AI and solutions, which produced RMB96.3 million in 2025 at a 48.7% gross margin after the SesameX platform launch. Put differently, automotive compute buys the company relevance; imaging supports margin; embodied AI offers optionality.
That revenue mix also explains the company’s operating leverage problem. Semiconductor design can produce strong leverage once a platform is standardized and shipping in volume. But automotive-grade design is front-loaded with validation, tooling, software adaptation, safety qualification, and customer-specific engineering. In Black Sesame’s case, fixed costs remain very high because winning a vehicle program is not the end of the work. It starts the long tail of ramp support and versioning. This is why H1 2025 looked so poor on gross margin. Revenue rose 40.4% to RMB252.9 million, but overall gross margin fell from 50.0% to 24.8% as application scenarios broadened and hardware and labor costs rose. The more encouraging part of the story came later: because full-year 2025 gross profit reached RMB337.1 million, the implied H2 2025 gross margin rebounded to about 48%. That suggests H1 was a transition trough, not a permanent new floor.
The real moats are narrower than the marketing language. The first genuine moat is automotive qualification and design-in stickiness. Once a chip family is qualified, integrated, and shipped into vehicle programs, replacing it is expensive and slow. The annual report says A1000 has a lifecycle of more than five years and has already been integrated into models from Geely, Dongfeng, BYD, and FAW. That is real evidence of embedment. The second moat is the integrated stack: proprietary IP cores, middleware, and developer tooling reduce customer work and can improve attachment rates. The third is founder and engineering fit: this is one of the few Chinese listed companies whose management pedigree clearly spans both vehicle-industry industrialization and deep chip development.
The weaker moats are just as important to name. Black Sesame does not yet have a scale moat against Horizon, much less against global players. It does not have a capital moat, which repeated placements and subscriptions make obvious. It does not have a proven data moat of the Mobileye kind, built over hundreds of millions of deployed vehicles. And its embodied-AI position is too young to call a moat at all. That business may become valuable; today it is better described as an option.
Governance is better than the average emerging-technology microcap, but not pristine. PwC has remained auditor since listing, and the 2025 annual report states there had been no auditor change. Related-party transactions disclosed in the statements did not, according to the company, constitute connected transactions under Chapter 14A except exempt remuneration matters. Those are positives. The discounts come elsewhere: a sizeable pre-IPO option pool, continued share-based compensation expense, and voting-trust arrangements giving Shan influence beyond his direct holding. That structure does not automatically make governance bad, but it does justify a governance discount relative to cleaner one-share-one-influence ownership models.
Industry structure and cycle
Black Sesame operates in the part of the automotive supply chain where two forces collide: rising intelligent-driving penetration and brutal cost-down pressure. The demand side is favorable. ADAS and higher-level assisted driving are moving from premium trims into mainstream vehicles, and Chinese OEMs are pushing “smart driving equality” into lower price points. Reuters reported that BYD planned to equip all models above RMB100,000 with its “God’s Eye” smart-driving system and even some below that threshold, a strong sign that assisted-driving capabilities are becoming table stakes rather than luxury add-ons. That broadens the addressable market for automotive compute. It also changes the economics: as features go mass-market, cost constraints get sharper.
The cycle here is not a classic memory-chip boom-bust. It is a technology-iteration cycle mixed with a policy and price-war cycle. Volumes can rise while per-unit economics tighten. Suppliers that win will not necessarily be those with the absolute highest compute; they will be the ones that offer adequate compute, automotive reliability, software maturity, and acceptable cost for the target vehicle class. That makes this industry unusually unforgiving for smaller fabless players. The prize is large, but so is the risk of becoming the “good enough but not default” vendor.
Geopolitics matters, but in a lopsided way. Black Sesame says the A2000 gained global market access advantages by passing relevant U.S. reviews and that the group mitigates geopolitical supply risks via dual sourcing, strategic inventory, and regional supplier proximity. Those are useful safeguards. They do not erase the structural fact that automotive semiconductors remain exposed to foundry concentration, export-control surprises, and validation regimes that can shift faster than product cycles. This is a long-duration risk, not just a headline risk.
Horizontal competitor analysis
The closest listed peer is Horizon Robotics. Horizon is what Black Sesame might look like if execution, ecosystem reach, and design-win depth scale up faster. Horizon reported 2025 revenue of RMB3.76 billion, up 57.7%, with a 64.5% gross margin. It also reported that by the end of the 2025 interim period it had accumulated nearly 400 vehicle-model designations, including more than 100 with highway-assist or higher-level functions. Customers buy Horizon because it has become a local default option: enough software, enough hardware, enough customer support, enough ecosystem density. Black Sesame is smaller and more exposed to proving each generation. Customers buy it where they want an alternative supplier with automotive-grade local capability, but they are not yet treating it as the default ecosystem choice.
Mobileye is the global benchmark for what mature ADAS commercialization looks like. In 2025 Mobileye generated $1.894 billion of revenue, $602 million of operating cash flow, and ended the year with $1.8 billion in cash. Its eight-year future expected automotive revenue pipeline reached $24.5 billion, and it has now shipped EyeQ technology into more than 230 million vehicles. Customers choose Mobileye for validated safety architecture, global OEM trust, and a deep base of production experience. Black Sesame cannot match that scale today. Its opportunity is more regional and more tied to the Chinese domestic stack.
Ambarella is not a direct ADAS challenger in the same way, but it is a useful adjacent reference because it shows how edge-AI silicon companies are being valued when they have meaningful automotive and robotics exposure without yet becoming dominant auto-compute platforms. Ambarella reported fiscal 2026 revenue of $390.7 million, GAAP gross margin of 59.2%, and cash plus marketable debt securities of $312.6 million. That is a cleaner edge-AI silicon profile, with less direct dependence on China passenger-car ADAS scale-up. Customers buy Ambarella for low-power edge vision and a broad physical-AI stack across endpoints, not primarily for becoming the main compute brain of Chinese NOA vehicles.
NVIDIA and Qualcomm sit in the outer ring of the comparison set. They are not valuation peers in any narrow sense, but they define the competitive ceiling. NVIDIA’s fiscal 2026 automotive revenue grew 39% year on year, while Qualcomm’s fiscal 2025 results emphasized automotive and IoT growth as part of a diversification story. Customers use those companies when they want global software ecosystems, huge balance sheets, and multi-market platform continuity. Black Sesame’s competitive path is narrower: it wins not by outspending them, but by being local, automotive-specific, cost-aware, and open enough to fit Chinese OEM needs.
| Dimension | Black Sesame | Horizon Robotics | Mobileye | Ambarella |
|---|---|---|---|---|
| Latest full-year revenue | RMB822.3m | RMB3.76bn | $1.894bn | $390.7m |
| Revenue growth | 73.4% | 57.7% | 15.0% | 37.2% |
| Gross margin | 41.0% | 64.5% | 47.7% | 59.2% |
| Market cap near 2026-06-26 | ~HK$7.4bn | HK$52.6bn | $6.38bn | $2.71bn |
| Approx. price-to-sales | ~7.8x | ~12.2x | ~3.4x | ~6.9x |
The numbers explain the market’s group portrait. Horizon gets the premium because it already looks like a scaled local winner. Mobileye gets a lower sales multiple because it is larger, slower-growing, and globally established; investors can underwrite its cash conversion rather than dream about it. Ambarella sits in the middle, rewarded for clean edge-AI exposure but without the same strategic scarcity within Chinese car autonomy. Black Sesame is valued above Mobileye on sales despite much weaker cash conversion, and below Horizon because it has not yet proven comparable depth. That feels directionally right.
Ecologically, Black Sesame is a challenger. It is not the platform leader. It is not a follower either, because few listed companies in China can point to comparable automotive-grade compute shipments and real vehicle integration. Its niche is the narrow strip between imported/global ceiling players and the strongest domestic incumbent. If the industry shifts toward higher penetration but also harsher price competition, that niche can either widen or disappear. It widens if OEMs want a second qualified domestic source. It narrows if those OEMs consolidate around one domestic standard stack plus their own in-house silicon efforts.
Current fundamentals and valuation
What is actually happening now
The most important recent operating fact is that 2025 looked much better than H1 2025 suggested. In the first half of 2025, revenue rose 40.4% to RMB252.9 million, but gross margin collapsed to 24.8% from 50.0%, and the company swung from a reported H1 2024 profit to a reported H1 2025 loss because the 2024 period still contained the non-cash fair-value gain from pre-IPO instruments. That presentation could easily have created a bearish impression. But the full-year result showed something else: revenue up 73.4% to RMB822.3 million, overall gross margin stable at 41.0%, and adjusted net loss improving to RMB1.076 billion from RMB1.304 billion. Back-solving from those figures implies that H2 2025 revenue accelerated to roughly RMB569.4 million and gross margin rebounded to about 48%. This is why the stock now trades on the dispute over whether H2 represented true normalization or just a brief mix benefit.
The segment mix also improved in a way that supports the commercialization case. The core driving-assistance business reached RMB686.9 million in 2025, up 56.8%. Embodied AI added RMB96.3 million in first meaningful revenue, and management explicitly described SesameX as a second growth engine. That matters because the company has long needed a way to make its inference-chip and software stack relevant beyond one automotive product family. The risk is obvious too: every second curve consumes management attention. On-device AI and robotics are natural adjacencies for the technology base, but they do not deserve the same valuation weight yet as shipping automotive programs.
Liquidity is better than the share-price chart implies, though not because the business is already self-funding. At the end of 2025, the company had RMB1.531 billion of cash and cash equivalents plus current FVPL assets, and management said liquidity plus financing would cover at least the next 12 months. Since then it completed the first 2026 subscription, and the annual report described the second 2026 subscription as pending at the latest practicable date; that second subscription was later completed on 2026-05-08, adding 33.5 million shares at HK$18.88 each. My inference is that, after accounting for the unused portion of the 2025 placing and the two 2026 subscriptions, gross liquidity was materially extended. The price paid was dilution.
Customer concentration is not the main near-term risk. For 2025, the top five customers contributed 38.1% of revenue and the largest customer 9.1%. That is meaningful, but not existentially concentrated for a company at this stage. The more important commercial question is design-win quality: whether A2000 programs become real shipment curves and whether the A1000 installed base becomes sticky enough to generate follow-on platforms.
Bull and bear divergence
The bullish case rests first on proof of life. Black Sesame is no longer pre-revenue, and no longer just showing slides. It has a production history with A1000, 2025 revenue growth above 70%, a stable full-year gross margin, and named integration into models tied to Geely, Dongfeng, BYD, and FAW. The second bullish point is that the product map is widening at the right time. A2000 is aimed at the part of the market where Chinese OEMs want domestic compute options, and management says it has passed relevant U.S. reviews, is being deeply adapted with partners, and is expected to enter mass production. The third point is that liquidity risk, while not gone, has been pushed out by recent fundraisings, buying time for the next commercial step.
The bearish case is more uncomfortable because it is also well grounded. First, Black Sesame still spends like a platform company before earning like one. R&D was RMB1.417 billion in 2025 against revenue of RMB822 million, and 2025 operating cash burn was still nearly RMB1.0 billion. Second, dilution risk has become part of the business model, at least for now: February 2025 placing, January 2026 first subscription, March 2026 second subscription. Third, the embodied-AI story may help multiple support, but it also risks giving investors an excuse to overvalue unproven optionality while the core automotive economics remain unsettled. Fourth, the stock is no longer expensive in the old IPO sense, but it still is not cheap versus cash conversion.
Valuation analysis
The first discipline is to ignore headline P/E. 2024 reported profit was inflated by a non-cash fair-value gain tied to investor instruments, and 2025 swung back to a large reported loss once that effect disappeared. Owner earnings remain negative. Operating cash flow was negative in both 2024 and 2025, and the gap between accounting profit and cash in 2024 was especially misleading. For this company, owner-earnings-based P/E is not just less useful than EV/Sales or market-cap-to-sales. It is actively misleading.
The current market value of roughly HK$7.4 billion equates to about 7.8 times 2025 sales after currency conversion. That places Black Sesame between Ambarella and Horizon on a simple sales-multiple basis, and clearly above Mobileye. I think that relative position is directionally fair. The company deserves a premium to slow-growth global incumbents only if investors believe Chinese domestic ADAS compute still has a steep ramp ahead and that Black Sesame remains one of the few listed ways to access it. It deserves a discount to Horizon because Horizon already has broader ecosystem scale, more design wins, and a larger installed commercial footprint.
The valuation that fits best is a scenario-based sales multiple, not a point estimate. The next 12 months will be driven by revenue credibility, program ramps, and dilution risk more than by near-term profit. I use translated HKD-equivalent sales and a range of price-to-sales assumptions that sit below Horizon, around Ambarella, and above Mobileye where growth warrants it. This is valuation-scenario analysis within a research framework, not investment advice.
| Dimension | Conservative | Base | Optimistic |
|---|---|---|---|
| Revenue and margin assumptions | 2026 revenue around RMB1.00–1.05bn; A2000 ramp slow; group gross margin around high-30s to 40% | 2026 revenue around RMB1.15–1.25bn; A1000 steady and A2000 starts meaningful contribution; gross margin around 40%–42% | 2026 revenue around RMB1.35–1.45bn; A2000 ramps well and embodied AI scales; gross margin 42%+ |
| Cash-flow assumptions | Operating cash burn remains heavy; another financing window remains possible | Burn moderates as revenue scale improves; no immediate new equity needed | Burn narrows sharply; market begins to price a path toward self-funding |
| Multiple assumptions | 4.0x forward sales | 5.5x–6.0x forward sales | 7.0x–8.0x forward sales |
| Implied value per share | about HK$6.5–6.8 | about HK$10.6–12.1 | about HK$15.3–18.7 |
| Key catalysts | A1000 steady shipments, no margin collapse | A2000 commercial proof, stable margins, slower burn | Visible A2000 scale-up, embodied-AI traction, no further dilution |
| Permanent-loss trigger | A2000 slips, gross margin stays in H1-2025 territory, cash burn forces another discounted raise | Mixed ramp and continued funding dependence | Growth disappoints after a sentiment-led rerating |
The expectation gap is no longer about whether autonomous driving is a large market. The market already knows that. The gap is whether Black Sesame can move from niche domestic challenger to financially sturdier commercial platform before shareholders tire of funding the journey. The next data that matter most are A2000 shipment evidence, the persistence of the H2 2025 gross-margin recovery, and whether 2026 needs yet another significant capital raise.
The margin-of-safety check is not flattering. At HK$10.38, the current share price is well above the conservative-case value. That means margin of safety is zero on a conservative valuation basis. If revenue merely flattens for several years rather than compounding hard, the likely annualized return from the current price is poor, and certainly not compelling against a Hong Kong 10-year government-bond yield around 3.3% on 2026-06-26. This is the clearest sign that Black Sesame is a good story with only a modest price cushion.
Margin-of-safety sufficiency verdict: none.
Risk variables, catalysts, and tracking dashboard
The most serious business risk is not generic “competition.” It is a specific version of competition: Chinese intelligent-driving compute moving downmarket faster than Black Sesame’s cost structure does. If OEMs decide that smart-driving hardware must fit ever-cheaper vehicles, suppliers can win volumes and still lose economics. Probability is medium to high; impact is high; the observable signal is another period like H1 2025, when driving-assistance gross margin compressed sharply because costs rose with broader application scenarios. The transmission path is simple: lower gross margin, slower improvement in losses, greater need for financing, lower multiple.
The second risk is financial. 2025 operating cash burn of RMB985.4 million was still larger than 2025 gross profit. Repeated external fundraisings in 2025 and 2026 reduced immediate liquidity stress, but they also taught the market to expect dilution when runway tightens. Probability is medium; impact is high; observable indicator is net operating cash outflow remaining above roughly RMB800 million annualized without a matching rise in cash-generative revenue.
The third risk is technological credibility at volume. Management’s language around A2000 is ambitious and strategically sensible, but the decisive test is no longer review completion or designated projects. It is shipped units, program starts, and customer expansion without destroying margin. Probability is medium; impact is high; observable indicators are explicit mass-production references, rising automotive revenue share from next-generation platforms, and the absence of another H1-style cost shock.
The fourth risk is governance-by-dilution rather than governance-by-fraud. I found no evidence here of auditor churn or accounting scandal. The risk is that ordinary shareholders are diluted while the company remains years away from self-funding. Because a large pre-IPO option plan remains outstanding and founder influence is elevated through voting-trust arrangements, public investors should assume capital allocation discipline will matter more than formal governance optics. Probability is medium; impact is medium to high.
The main positive catalysts are all measurable. The strongest would be a clear statement that A2000 has moved into meaningful mass production, backed by revenue acceleration rather than just project counts. The second would be gross margin holding above 40% while driving-assistance revenue continues to scale. The third would be another year in which adjusted loss narrows and operating cash outflow shrinks materially. The fourth would be evidence that embodied AI adds profitable revenue without demanding another heavy round of cash burn.
| Indicator | Normal range | Alert threshold |
|---|---|---|
| Group revenue growth | above 30% year on year | below 20% for a full year |
| Group gross margin | around 40% or better | below 35% |
| Driving-assistance gross margin | high-30s | below 30% |
| Annual operating cash outflow | falling from 2025 level | above RMB1.0bn again |
| New equity financing | infrequent and strategic | recurring discounted issuance |
| A2000 newsflow | sample, validation, mass production | repeated delay without revenue proof |
| Top-customer concentration | largest customer below 15% | largest customer above 20% |
| Peer premium to Mobileye | justified by growth | stays high while cash conversion does not improve |
These indicators matter because they separate real progress from thematic noise. Revenue growth alone can look good while value is destroyed through low-margin shipments. Gross margin alone can look fine while revenue stalls. Cash burn alone can improve simply because R&D is cut too hard. The right way to watch Black Sesame is as a three-part test: commercial scale, unit economics, and financing dependence. If two improve and one worsens, the story is still unresolved.
Cross-synthesis summary
Looking vertically across its whole journey, Black Sesame has already proven one thing that deserves real credit: it can build automotive-grade compute products that reach production vehicles. That sounds basic, but it is not. Many chip start-ups never cross that threshold. Black Sesame did. The founders’ backgrounds explain why. Shan gave the company technical depth in imaging and semiconductors; Liu gave it the automotive-industrial instinct that pure AI start-ups often lack. The company’s history since 2016 shows a real progression from product ambition to automotive shipment, then from a single-curve car-compute story to a broader on-device AI aspiration. That is a meaningful achievement.
Past success, though, came from a blend of management capability and era tailwind rather than from a finished moat. The era tailwind was China’s rush toward intelligent vehicles and the strategic preference for domestic core technologies. Management capability showed up in surviving the pre-IPO years, bringing A1000 into real vehicles, getting through listing, and broadening the product stack without abandoning the core business. What has not yet been proven is the next-order question: whether those strengths can become an enduring financial structure rather than a technologically respectable but permanently cash-hungry one. That is the whole argument of the stock now.
Horizontally, Black Sesame’s advantage over competitors is not dominance. It is relevance. It is one of the few listed domestic names in a field that Chinese OEMs increasingly care about. Its products are real, its customers are real, and its technology stack is broad enough to support an open-platform positioning. Against Horizon, it lacks ecosystem scale. Against Mobileye, it lacks global installed-base credibility and cash generation. Against NVIDIA and Qualcomm, it lacks capital firepower. But it does have a believable niche where local supply preference, automotive design-in stickiness, and acceptable performance can matter. That is enough for a business. It is not yet enough for a premium equity unless execution keeps compounding.
The current valuation is rewarding some future success, but not all of it. The stock has already been derated dramatically from IPO levels. That means investors are no longer paying absurd scarcity prices. They are still paying for a future in which A2000 becomes important, embodied AI becomes more than optionality, and financing dependence eases. At 7.8 times trailing sales and with no margin of safety against a conservative scenario, the price is not reckless. It is simply not generous. That distinction matters. This is a stock where upside exists if execution goes right, but downside still exists if the company needs to prove the same things for longer than expected.
What the market is most likely misjudging now is not the industry’s size. It is the timing mismatch between commercial proof and financial proof. H2 2025 showed that the business can improve faster than H1 implied. Investors who focus only on the ugly H1 margin print can miss that. But the opposite mistake is also easy: extrapolating one stronger half into a clean path to scale. The next year is about A2000 evidence and burn-rate discipline. The next three years are about whether Black Sesame remains a narrow domestic challenger or becomes a durable second platform in Chinese intelligent-vehicle compute. The next five years are about whether it can reach self-funding and avoid becoming one more interesting semiconductor name that repeatedly diluted before the payoff arrived.
Bull and bear reasons
The core bull case is that 2025 finally looked like commercialization rather than just validation: revenue rose 73.4%, the automotive business reached RMB686.9 million, and full-year gross margin held at 41.0% despite a weak first half.
A1000 has already crossed the most valuable threshold for a start-up automotive chip company, with integration into vehicle models associated with Geely, Dongfeng, BYD, and FAW and a stated lifecycle of more than five years.
A2000 gives Black Sesame a plausible next platform at the part of the market where domestic OEM demand is rising, and management says it has passed relevant reviews, secured designated projects, and is expected to start mass production.
The liquidity picture is materially better than it was pre-IPO because the 2025 placing plus the two 2026 subscriptions extended runway and reduced near-term insolvency risk.
The core bear case is that Black Sesame still spends like a platform leader without yet earning like one: 2025 R&D was RMB1.417 billion on RMB822 million of revenue, and operating cash burn was RMB985 million.
The company has taught the market to expect dilution whenever runway tightens, through a placing in 2025 and two new-share subscriptions in early 2026.
Its closest domestic listed peer, Horizon, is already much larger and deeper in ecosystem scale, while global players such as Mobileye still dominate on installed-base trust and cash conversion.
The embodied-AI second-curve story is commercially interesting but too early to deserve major valuation weight, and it could distract from the still-unfinished core automotive transition.
Pre-mortem
A plausible 50% down case three years from now looks like this: A2000 wins designated projects but ramps more slowly than management expects through 2027, while Chinese OEMs force down system costs as smart-driving functions spread to lower-priced vehicles. Black Sesame then keeps growing revenue, but driving-assistance gross margin stays closer to the H1 2025 pressure zone than the H2 2025 recovery. Operating cash burn remains near the 2025 level, the company raises more equity at a discount, and the sales multiple compresses from roughly 7–8 times trailing sales today toward 3–4 times. In that script, the stock could drift into the HK$4–6 range.
A second bad script is more strategic. Embodied AI attracts market attention and encourages management to widen the product agenda before the car business has consolidated. The result is higher R&D demand across too many fronts, slower progress toward self-funding, and a market conclusion that Black Sesame is becoming a “good technology portfolio” rather than a focused automotive platform. The revenue story survives, but the stock loses its strategic clarity and trades down on fatigue.
Final research conclusion
Black Sesame is worth following because it has already proved more than many automotive-chip start-ups ever do. It is worth owning only with more caution than the industry narrative invites. The company has built a real position in Chinese intelligent-driving compute, and 2025 showed meaningful commercial progress. But the stock today is still pricing a future that requires several things to go right in sequence: A2000 must ramp, margins must hold better than they did in H1 2025, embodied AI must stay additive rather than distracting, and the next funding need must be pushed far enough away that dilution ceases to dominate the narrative. If those things happen, the upside is real. If they do not, the market will not forgive the company simply because autonomous driving remains a large theme.
What keeps me from a more positive judgment is not disbelief in the technology. It is distrust of the current margin of safety. Black Sesame has crossed the credibility threshold, but not the cash-conversion threshold. At the current price, an investor is not buying distress. An investor is buying a still-loss-making challenger with enough liquidity to keep going and enough technical proof to matter, but without enough financial proof to call the price comfortably cheap. The stock therefore deserves patience more than enthusiasm. I would change that view if A2000 begins to contribute visible mass-production revenue while operating cash burn drops sharply and no new discounted raise appears.
【Company-profile scores】
- Fundamental quality: medium
- Growth: high
- Moat: medium
- Financial soundness: medium
- Management credibility: medium
- Valuation attractiveness: low
- Risk level: high
- Suitable investor type: high-risk speculation
【Investment rating】
- Rating: Hold
- One-line thesis: Real commercialization exists, but recurring dilution risk and still-negative owner earnings limit valuation support at the current price.
- Three price signals:
- Acceptable hold price: HK$9.5-HK$12.5
- Clearly overvalued price: above HK$16.8
- Current-price classification: acceptable hold
- Whether to wait for a better price: yes; below roughly HK$6 would offer a true margin-of-safety entry, and the opportunity cost of waiting is missing an early A2000 rerating if execution improves faster than expected.
- Target holding horizon: 3–5 years for investors already involved; 12 months matters mainly for execution checkpoints.
- Expected annualized return:
- Conservative scenario: about -35% to -37%
- Base scenario: about +2% to +20%
- Optimistic scenario: about +47% to +80%
- Max-loss risk: roughly 50% or a bit more, if A2000 ramps slowly, margins stay compressed, and another discounted financing resets the multiple lower.
- Reassessment-trigger signals:
- Group gross margin below 35% for two consecutive reporting periods
- Annualized operating cash burn staying above RMB1.0 billion after 2026 revenue growth
- Another major discounted equity raise before clear A2000 revenue proof
- Evidence that embodied AI revenue grows while core automotive margin deteriorates
- Largest customer concentration rising above 20% without offsetting diversification
【Ideal Buy Price】HK$5.2-HK$5.8 Basis: at least a 20% margin of safety below the conservative scenario value derived from roughly 4.0x forward-sales logic on about RMB1.00–1.05 billion of 2026 revenue, converted into HKD.
【Valuation Range】
- current: 10.38 (close as of 2026-06-26)
- bear (conservative · ideal buy zone): [5.2, 5.8]
- base (fair · acceptable hold zone): [9.5, 12.5]
- bull (optimistic · above the clearly-overvalued line): [16.8, 18.7]
Sources and research uncertainties
The highest-confidence sources for this report were the company’s 2024 interim report, 2024 annual report, 2025 interim report, 2025 annual report, prospectus-related HKEX materials, and 2025–2026 capital-markets announcements; peer references came primarily from official company disclosures for Horizon Robotics, Mobileye, and Ambarella, supplemented by market-data pages for current prices and market caps.
A few blind spots remain. First, HKEX monthly-return mechanics and market-data services did not line up perfectly on shares outstanding and market cap after the 2026 subscriptions and treasury-share activity, so I used the latest reliable closing price and a market-cap approximation rather than claiming an exact official base-date share count. Second, I could verify the existence of prospectus lock-up schedules and cornerstone participation, but I could not verify every cornerstone-investor name cleanly from the accessible pages gathered here. Third, Black Sesame does not provide full quarterly disclosure in the way U.S. investors may expect, so the “last four quarters” analysis had to be reconstructed from interim and annual data. Fourth, industry penetration numbers for NOA and L2+ adoption are widely cited but not always published in equally accessible primary form; where I used them, I kept them directional rather than thesis-critical.
Other tickers mentioned
- 09660.HK: closest listed domestic peer in intelligent-driving SoCs and solutions
- MBLY.US: global ADAS reference for installed base, pipeline depth, and cash conversion
- AMBA.US: adjacent edge-AI semiconductor reference with automotive and robotics exposure
- NVDA.US: global high-end autonomous-driving compute reference and competitive ceiling
- QCOM.US: diversified automotive compute platform reference
- 01211.HK: OEM example of smart-driving features moving into lower-priced vehicles
- 09888.HK: mentioned through Apollo Go as a robotaxi ecosystem partner
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Full report
Sign in to read the full report
Sign up free to unlock the full text, the Baillie growth scorecard, and full-text search.
Log in / Sign up free