2533.HK

A large and fast-growing pie, but Black Sesame is taking a thin slice of an existing market, not creating a new one. The opportunity is genuine. The broader automotive-SoC market was about USD 28.6 billion in 2025 and is forecast to compound near 10% a year, while the narrower autonomous-driving-chip segment is projected to grow from roughly USD 30 billion in 2025 to about USD 191 billion by 2034, a ~23% CAGR. China sits at the center of that demand, expected to account for more than a third of global autonomy-domain-controller and front-view-camera SoC demand by 2030, and smart-driving features are spreading from luxury trims into mass-market cars.

But the honest framing is that this is an existing pie being carved up, not a new market being invented. Black Sesame's 2025 revenue of RMB822.3 million is a sub-1% sliver of the autonomous-driving-chip TAM. More soberingly, it is not among the names the industry expects to lead: forecasters see Horizon Robotics and Huawei taking over half of China's domestic share by 2035, with a global top five of Mobileye, NVIDIA, Qualcomm, Horizon and Huawei reaching roughly 78% share, and Black Sesame absent from that list. Its ceiling is therefore set less by the size of the pie than by how much share a challenger second-source can win and hold. The market is huge and real; the company's claim on it is small and contested.

Yes, doubling revenue within five years is highly likely, driven mainly by volume and a new platform, but profitable doubling is the real test. Black Sesame grew revenue more than thirteenfold from RMB60.5 million in 2021 to RMB822.3 million in 2025, with 2025 alone up 73.4%. The report's own scenarios put 2026 revenue at RMB1.00–1.05 billion (conservative) to RMB1.35–1.45 billion (optimistic); even the low end keeps the company on a path to double by roughly 2027–2028.

The doubling would be volume-led. The A1000 family is already designed into Geely, BYD, Dongfeng and FAW models, and the next-generation A2000 just scored its first publicly disclosed mass-production designation in February 2026 through an AICC partnership with a leading domestic automaker, with first vehicles expected in 2026. The tailwind is structural: China's NOA front-installation reached about 22% in 2025 and L2/L2+ adoption is heading above 95% in 2026–2027. New business helps too, with embodied AI contributing RMB96.3 million in its first meaningful year.

The caveat is price. As smart driving moves downmarket, per-unit economics tighten, and rivals are explicitly cutting costs, with Horizon's "Xingkong" fusion chip designed to reduce per-vehicle cost by RMB1,500–4,000. So volume should comfortably double revenue; whether it doubles without margin erosion and further dilution is the open question.

Two candidate second curves exist today, but neither is yet a proven engine: A2000 (the core's next generation) is the bankable one, while embodied AI/SesameX is genuine optionality. The first, A2000, is really the next chapter of the core automotive business rather than a separate market: a 7nm L2+/L3 platform that cleared U.S. export review in January 2026 and landed its first mass-production designation in February 2026 via an AICC partnership. It exists, it ships soon, and analysts note that winning a handful of A2000 clients would lift volume, ASP and margin, but it inherits the same price-war exposure as the core.

The second, embodied AI, is the more genuinely new curve. The SesameX platform, built on modules Kalos, Aura and Liora, targets logistics, manufacturing and service robots, and the company has signed partners including COSCO Shipping for vessel-inspection robots plus SIIC Tech Capital, Lenovo and Joyson; a Baidu Apollo Go tie-up adds robotaxi and L4 reach.

Honestly, though, embodied AI is only about 12% of 2025 revenue (RMB96.3 million) and unproven at scale; the report rightly calls it optionality and flags distraction risk. So the second curve exists as early revenue and partnerships, not as a self-standing growth engine that could carry the company if the core stalled.

The moat is real but shallow, and more likely to narrow than widen over three to five years. Black Sesame's genuine advantages are automotive qualification and design-in stickiness: A1000 is embedded in Geely, BYD, Dongfeng and FAW models with a stated five-year-plus lifecycle, alongside an integrated chip-plus-software stack and a rare founder pairing spanning imaging semiconductors and auto industrialization.

But the deficits are larger. There is no scale moat against Horizon, which earned RMB3.76 billion of 2025 revenue at a 64.5% gross margin with nearly 400 model designations, versus Black Sesame's RMB822 million at 41.0%. There is no capital moat, as serial dilution makes plain. And there is no data moat of the Mobileye kind, built on 230 million-plus deployed vehicles.

The trajectory looks adverse. Forecasters expect Horizon and Huawei to take over half of China's domestic SoC share by 2035, with Black Sesame absent from the projected global top five. The field is crowded with NVIDIA's Thor, Qualcomm, Huawei and SemiDrive, and a price war is intensifying, with Horizon's "Xingkong" chip targeting per-vehicle cost cuts of RMB1,500–4,000. Most threatening, a key customer is becoming a competitor: BYD is staking its self-driving future on in-house silicon. The moat amounts to "qualified second source." It widens only if OEMs deliberately want dual sourcing, and narrows if they consolidate on one domestic stack plus their own chips.

Black Sesame shows some reinvention DNA, but it is short-tested, and its handling of bad news is reasonable on disclosure yet flattering on financial optics. The company has already remade itself once, from a pure ADAS-chip designer into a bundled compute-plus-solutions supplier, and now into embodied AI, launching the SesameX robot platform and signing partners such as COSCO Shipping and SIIC Tech Capital. The founder mix, pairing Shan Jizhang's OmniVision imaging pedigree with Liu Weihong's Bosch and GM auto-industrial background, gives it adaptive range that pure-AI start-ups lack.

On bad news, the record is mixed-positive. The company issued a profit warning ahead of its 2025 results rather than burying them, kept PwC as auditor with no churn, and disclosed the ugly H1 2025 gross-margin collapse to 24.8% plainly. But the optics cut the other way too: 2024's headline RMB313 million profit was really a non-cash RMB2.047 billion fair-value gain on investor instruments, and management consistently leads with ambitious A2000 and embodied-AI framing while the business stays loss-making.

The deeper truth is that the DNA is untested by real adversity. Its reinvention so far is broadening into adjacencies, not surviving a core disruption or owning a strategic mistake. A RMB1.531 billion cash cushion and repeated equity access give it room to pivot, but that adaptability has been financed by shareholders rather than earned, so the resilience is real but borrowed.

The founders are genuinely long-term and willing to sacrifice present profit, but their alignment with minority shareholders is weakened by amplified control and serial dilution. Commitment to the future is not in doubt: the company spends RMB1.417 billion on R&D against RMB822 million of revenue, about 1.7x sales, deliberately running deep losses to build platforms. Chairman and CEO Shan Jizhang and co-founder Liu Weihong have run the business since 2016 and through listing, and PwC has remained auditor with no change.

The alignment problem is structural. A voting-trust arrangement amplifies Shan's influence well beyond his economic ownership, a sizeable pre-IPO option pool remains outstanding, and share-based compensation is ongoing, so founders steer with outsized votes while economic risk sits more heavily with outside holders. Dilution has become a recurring feature: a February 2025 placing (HK$1.237 billion net), a January 2026 subscription (HK$538 million net), and a second 2026 subscription completed on 8 May at HK$18.88 per share (HK$632 million net).

The timing makes the point sharply. New equity was taken in at HK$18.88 in early May, and by 26 June the shares had fallen to HK$10.38, a roughly 45% drop, so recent outside capital was diluted hard within weeks. Management is clearly sacrificing the present for a five-to-ten-year vision, but it is partly sacrificing shareholders' present, diluting them to fund a journey whose cash payoff is still years away. Long-term in horizon: yes. Deeply aligned with minorities: only partly.

Customers would miss Black Sesame moderately, not desperately. It is a valued second source rather than indispensable, and its growth is socially benign and not dependent on harmful regulation. Within existing programs the switching friction is real: automotive qualification and a five-year-plus design-in lifecycle make a shipped chip costly to rip out. Customer concentration is manageable too, with the top five customers at 38.1% of 2025 revenue and the largest just 9.1%, so no single relationship is existential.

But if the company vanished tomorrow, OEMs have ready alternatives in Horizon, NVIDIA, Qualcomm, Huawei and increasingly their own silicon, since BYD is building self-driving chips in-house. The report itself notes that Horizon, not Black Sesame, is the default ecosystem choice. So the honest answer is "missed but replaceable," not "irreplaceable."

On sustainability, growth rests on a durable, benign tailwind rather than a loophole. China's NOA front-installation hit about 22% in 2025, with L2/L2+ penetration heading above 95% by 2026–2027 as smart driving spreads into mass-market vehicles. The product makes driving safer and more automated; it does not rely on harming society, and export exposure is being managed, with A2000 having cleared U.S. review. The genuine threat to sustainability is competitive and financial, namely commoditization and a price war, not ethical or regulatory. Industry growth is sustainable in principle; whether Black Sesame's share of it is sustainable is the harder question.

Unit economics are improving but the enterprise is still deeply unprofitable: healthy gross margins are overwhelmed by R&D, owner earnings are firmly negative, and earned cash effectively goes to R&D funded by share issuance. Gross margins are not the problem. The group held 41.0% in 2025 (driving-assistance 37.4%, imaging 84.7%, embodied AI 48.7%), and H2 2025 gross margin rebounded to roughly 48% after H1's 24.8% trough.

The problem is everything below gross profit. 2025 R&D was RMB1.417 billion, about 1.7x revenue, producing an operating loss of RMB1.448 billion and an adjusted net loss of RMB1.076 billion that did at least narrow 17.5% year on year. Crucially, operating cash outflow of RMB985.4 million was still larger than the entire RMB337.1 million of gross profit. Owner earnings are negative, and the report rightly warns that earnings-based valuation is "actively misleading" here.

Are economics better at scale? Directionally yes: selling and G&A expenses fell 27.2% and 19.1% while R&D held roughly flat in absolute terms, so losses narrow as revenue scales. But the company is not near the inflection where incremental revenue funds itself. Where does cash go? Into R&D, with the gap plugged by external capital, a February 2025 placing plus two 2026 subscriptions, leaving RMB1.531 billion of cash and FVPL assets at year-end. In short, this is still a capital consumer betting that scale eventually flips burn into self-funding, a flip it has not yet demonstrated.

A ten-year 5x is possible but sits in the optimistic tail: it needs a long chain of conditions to hold at once, and even the bulls are not underwriting it. A 5x from HK$10.38 means roughly HK$52 a share and a market cap near HK$37 billion, versus about HK$7.4 billion today. For that, essentially all of the following must hold simultaneously: A2000 ramps from its first mass-production designation to multi-OEM volume; revenue compounds around 18–20% annually for a decade, from RMB822 million toward RMB4–5 billion, beyond Horizon's current RMB3.76 billion scale; gross margin holds near 40%-plus despite a worsening price war; the roughly RMB1 billion cash burn turns positive so dilution stops; embodied AI becomes a real profit leg; and the company survives BYD's in-house silicon push and the projected Horizon and Huawei consolidation.

What does today's price imply? At HK$10.38 and about HK$7.4 billion market cap, the stock trades on roughly 7.8x trailing sales, between Ambarella (6.9x) and Horizon (12.2x) and above Mobileye (~3.4x). That is no longer scarcity-bubble pricing, but it still embeds substantial scaling. Tellingly, the report's own bull case tops out at HK$16.8–18.7 (under +80%), and even sell-side 12-month targets of HK$23–28, many set before the mid-2026 slide, imply at most about 2.5–2.7x. So a ten-year 5x is not impossible, but it requires near-flawless execution that no one currently prices as the base case.

This is mostly "won't-respect (yet)," not "can't-understand" or "can't-see-far": the market understands the story and has actively de-rated it, demanding cash-conversion proof. Black Sesame is heavily covered, with sell-side 12-month targets clustered at HK$23–28 and mostly buy ratings, so the bull thesis is well known, not overlooked. Nor is the TAM unseen: everyone knows autonomous-driving chips are projected to grow from roughly USD 30 billion to about USD 191 billion by 2034 and that China's NOA adoption is ramping fast. The report itself concedes the gap "is no longer about whether autonomous driving is a large market."

What the market will not yet respect is the financial model. It has re-rated the whole China ADAS-SoC complex from "scarcity growth" to "show me self-funding in a price war." The catalyst was visible: peer Horizon swung to a RMB10.469 billion 2025 net loss despite revenue up nearly 60%, and the group sold off hard. Black Sesame fell from its early-May subscription level near HK$18.88 to HK$10.38 by 26 June, with Horizon down to HK$3.61 near its 52-week low. Serial dilution and roughly RMB1 billion of annual burn reinforce the skepticism.

There may be a thin "can't-see-far" sliver: if A2000 volumes and embodied AI inflect faster than margins erode, today's burn-fixated pessimism could prove short-sighted. The narrative inflection point is concrete, namely clear A2000 mass-production revenue at stable 40%-plus margins with no new discounted raise. Until that arrives, the market respects the technology but not the price.