iFLYTEK (002230.SHE) Zen Horizon Research Report

Hunting ten-year five-baggers among great growth stocks — pressing the upside question: "Can it get much bigger?"

• QHow large is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?6/10

  A

  Conclusion: iFLYTEK has a very high market ceiling, but it is mainly embedding AI into existing budget pools such as education, government affairs, healthcare, automotive, and office hardware, expanding the pie for “AI applicationization”; truly new markets account for only a small part, mainly large-model API/MaaS, AI Agent, and the fully domestic computing-power ecosystem. So this dimension is positive, but it should not be read as “creating a new continent no one has ever seen.”

  The report defines the company as China’s integrated leader in AI speech + large models + vertical applications. 2025 revenue had reached RMB 27.105 billion, including RMB 8.967 billion from smart education, RMB 6.088 billion from the open platform/large models, RMB 2.183 billion from smart hardware, RMB 1.558 billion from digital government, RMB 1.240 billion from automotive, and RMB 858 million from healthcare; these segment figures are broadly consistent with the company’s disclosed 2025 revenue, segment growth, and R&D investment disclosures. The external industry backdrop also supports a “large ceiling”: under the Ministry of Industry and Information Technology’s framing, China’s core AI industry exceeded RMB 1.2 trillion in 2025, with more than 6200 companies, and the State Council’s “AI+” policy also sets directional goals such as deep integration of AI with six key fields by 2027 and penetration of intelligent terminals and agent applications exceeding 70%. iFLYTEK sits exactly at the intersection of “AI + education / public services / governance / industrial applications.”

  But the distinction has to be made honestly: smart education, government affairs, healthcare, and automotive voice are not essentially new markets. They are AI-upgraded versions of existing education informatization, government IT, hospital information systems, in-car interaction, and consumer electronics. Its opportunity comes from replacement and expansion: schools upgrade from software platforms to AI learning tablets/teacher assistants, governments upgrade from system integration to government large models, and automakers upgrade from voice control to multimodal cockpits. The areas that look more like “new markets” are large-model calls, MaaS, Agent platforms, and the domestic computing-power adaptation ecosystem, but under the briefing’s framing, 2025 large-model API/MaaS revenue was only RMB 385 million, still far from replacing the RMB 8.967 billion education main business as core cash flow.

  So the ceiling can be high, even high enough to support a several-ten-billion-yuan revenue AI application company as it continues to grow; but from a Baillie LTGG perspective, the issue is not TAM, but “how much iFLYTEK can capture, how much profit it can retain, and whether today’s price has already reflected that in advance.” The current price anchor is about RMB 44-45, with market cap around RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x; yet 2026Q1 was still at revenue of RMB 5.274 billion, attributable net profit of RMB -170 million, ex-nonrecurring net profit of RMB -430 million, and operating cash flow of RMB -1.069 billion. The company’s Q1 report also explained that ex-nonrecurring pressure was related to increased R&D and sales investment. In other words, the market space is large, but what it has proved so far is more “there is a large pool to swim in,” not yet “it can convert the large pool into shareholder profit at high returns.”

  Jun 9, 2026
• QCan its revenue at least double over the next five years? Will growth mainly be driven by volume, price, or new businesses?5/10

  A

  Conclusion: It has a chance, but a doubling is not highly certain. Using 2025 revenue of RMB 27.105 billion, up +16.12% year on year as the base, doubling in five years would require reaching about RMB 54.2 billion, implying roughly 14.9% annualized growth. That threshold is not unreasonable, but 2026Q1 revenue of RMB 5.274 billion, up +13.23% year on year was already slightly below the required annualized rate, showing that it needs education, government-enterprise project acceptance, and overseas business to keep ramping in the second half, and cannot rely only on AI-themed valuation support.

  The main drivers are “volume + new businesses,” not price increases. The volume component comes from continued rollout across B/G-end scenarios such as education, government affairs, automotive, and healthcare; the report also emphasizes education as the largest pillar. The annual report disclosed that 2025 smart education was RMB 8.967 billion, +24.04%; digital government was RMB 1.558 billion, +30.35%; smart automotive was RMB 1.240 billion, +25.41%. Meanwhile, large-model API/MaaS was only RMB 385 million, but grew +263% year on year, and overseas business grew +275%. Q1 further disclosed B/C-end business growth of +26.27%, contract value of +28.16%, opportunity reserve of +33%, third-party developer large-model daily average Tokens of +4241%, and overseas revenue of +167%. These look more like revenue growth from “customer count, call volume, installed/deployed volume, and overseas regional expansion” than unit-price increases.

  The real bottleneck is growth quality. 2025 attributable net profit was RMB 839 million and ex-nonrecurring net profit was RMB 264 million, while 2026Q1 attributable net profit was RMB -170 million, ex-nonrecurring net profit was RMB -430 million, and operating cash flow was RMB -1.069 billion; the company explained that the Q1 ex-nonrecurring decline was mainly due to a RMB 349 million increase in R&D and sales expenses. In other words, a revenue-doubling path exists, but the company is still buying growth with RMB 5.364 billion of R&D investment and heavy sales expense, and profit elasticity has not yet been proved. If revenue doubles over the next five years but ex-nonrecurring profit remains thin, it would not qualify as high-quality compounding in the Baillie sense.

  My judgment: this dimension is moderately positive, but certainty is not high. If the education/government-enterprise base can maintain 15%-20% growth, and new businesses such as large-model API, AI Agent, overseas, automotive, and healthcare can keep scaling from small bases, doubling revenue in five years is achievable; if large-model price wars, education policy, project acceptance delays, or U.S. Entity List-related restrictions intensify, growth may fall to 10%-13%. At an anchor of about RMB 44-45, market cap around RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x, the market has already prepaid a large portion of the “five-year doubling” expectation.

  Jun 9, 2026
• QFive years from now, what will take over as the next growth engine? Does this “second curve” exist today?4/10

  A

  Conclusion: the second curve exists today, but it has not yet taken over. Five years from now, the most likely successor is the combination of “Spark large-model platformization + industry Agent + overseas multilingual,” not a single new business. The report lists smart education as the largest pillar in 2025, with revenue of RMB 8.967 billion. The areas that truly look like a second curve are the open platform, large models, enterprise AI, overseas, multilingual licensing, healthcare, government affairs, and automotive, among which the official annual report summary disclosed AI platform and licensing service revenue of RMB 1.252 billion, and large-model API and MaaS platform service revenue of RMB 385 million, up 263% year on year. This shows that demand and commercialization entry points have emerged, but RMB 385 million is still small relative to total 2025 revenue of RMB 27.105 billion, so it cannot be said to have already replaced the education main business.

  The five-year takeover logic has three levels in my view: first, large-model API/MaaS and enterprise Agent move from “call-volume growth” to “chargeable workflows”; second, the Spark education large model, government large model, and healthcare large model embed model capabilities into existing school, government, and hospital customers; third, overseas multilingual products and AI hardware extend Chinese speech capability into Southeast Asia, the Middle East, the U.S., Europe, Japan, and South Korea. 2026Q1 already offered early validation: contract value was up +28.16% year on year, opportunity reserve was up +33%, third-party developer large-model daily average Tokens were up +4241% year on year, and overseas revenue was up +167% year on year. These indicators prove the second curve is not PPT, but they are still more about “orders, calls, and revenue growth,” and have not yet been fully converted into ex-nonrecurring profit.

  The key issue is the quality of the takeover, not whether there is a story. The company’s 2025 R&D investment was RMB 5.364 billion, and 2026Q1 ex-nonrecurring loss was RMB -430 million with operating cash flow of RMB -1.069 billion, showing that the second curve is still burning money to build capabilities and customers. If open-source models and major cloud vendors continue pushing API prices lower, API/MaaS may see high calls but low profit. On top of that, with the current share price around RMB 44-45 and market cap around RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x, the market is already pricing it as if “the second curve will materialize,” rather than valuing an overlooked asset no one has noticed.

  So my answer is: the second curve exists, but it remains in the early stage of moving from “high-growth small base” to “profitable main engine.” The hardest validation points ahead are not new model launches, but whether large-model API/MaaS can scale rapidly from RMB 385 million, whether enterprise/education/healthcare/government Agent can form renewals and gross profit, whether overseas high growth can persist, and whether ex-nonrecurring profit can be pulled back on track while export-control and licensing risks from the Entity List do not worsen further. If it achieves these, it may become the growth engine after education in five years; if not, it is more like an AI narrative supporting valuation than an independent cash-flow curve able to take over.

  Jun 9, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?5/10

  A

  Conclusion: iFLYTEK has a moat, but its core is not that “the Spark large model itself is irreplaceable.” It lies in the combination of “Chinese speech/cognitive intelligence technology + vertical scenario data in education/government/healthcare/automotive + B/G-end customer relationships + compliance around domestic computing power.”The report places this moat in the moderately strong range, and I think that direction is reasonable: over the next three to five years, the vertical-industry moat has a chance to widen, but the general model/API layer will keep being compressed by Baidu, Alibaba, ByteDance, DeepSeek, and open-source models. Overall it is “slightly widening but unstable,” not a top-tier monopoly like NVIDIA CUDA or the Apple ecosystem.

  The first layer of advantage is technology and domestic-replacement credibility. The company’s 2025 annual report summary disclosed that it undertook construction of the National Key Laboratory of Cognitive Intelligence and the National Engineering Research Center for Speech and Language Information Processing, and stated that iFLYTEK Spark is the only general-purpose large model among China’s mainstream large models trained on fully domestic computing power; public reporting also said Spark X2, released in February 2026, completed training based on fully domestic computing power. After iFLYTEK was placed on the U.S. Commerce Department BIS Entity List in 2019, this has instead become a compliance selling point when domestic government, central/state-owned enterprise, education, and healthcare customers make purchases.

  The second layer of advantage is scenario data and customer stickiness. The report notes that iFLYTEK is not just selling a chatbot, but embedding AI into schools, government service halls, primary-level hospitals, in-car cockpits, and learning hardware. The corresponding numbers in the annual report also support this judgment: 2025 smart education revenue was RMB 8.967 billion, smart healthcare was RMB 858 million, digital government was RMB 1.558 billion, smart automotive was RMB 1.240 billion, and the open platform provided 920 AI capabilities to more than 10 million developers. Once these B/G-end systems are deployed, replacement costs, data migration, compliance acceptance, and local service are all relatively heavy, making the moat more tangible than a pure C-end App.

  But the weakness is also hard-edged: this moat has not yet been fully converted into profit. The company’s 2025 revenue was RMB 27.105 billion, attributable net profit was RMB 839 million, ex-nonrecurring net profit was only RMB 264 million, and R&D investment was RMB 5.364 billion; by 2026Q1, revenue was RMB 5.274 billion, but attributable net profit was RMB -170 million, ex-nonrecurring net profit was RMB -430 million, and operating cash flow was RMB -1.069 billion. In other words, the barriers are real, but they still depend on high R&D and high sales investment to maintain, and have not yet proved they can produce stable, high-quality ex-nonrecurring profit.

  So my judgment is: over three to five years, the moats in education, healthcare, government affairs, and automotive, where “industry applications + private data + local delivery” matter, are likely to widen; but the general large-model, API, and C-end user-mindshare layer will narrow. At the current share price of about RMB 44-45, market cap around RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x, the market has already priced in “the moat will widen meaningfully.” Unless we later see major X2 commercialization orders, ex-nonrecurring profit turning positive, and sustained cash-flow improvement, this looks more like a “moderately strong moat + high-valuation overdraw,” not yet a high-conviction ten-year five-bagger quality stock.

  Jun 9, 2026
• QIf its core business is disrupted, does it have the DNA to reinvent itself? How does it handle mistakes and bad news?5/10

  A

  Conclusion: It has the DNA for self-reinvention, but it looks more like “high-R&D-driven continuous migration” than a low-cost, self-disruptive organization that has already been proved. The most positive evidence in the report is the company’s path: it started with Chinese speech recognition, later embedded the technology into education, government affairs, healthcare, automotive, and hardware, then shifted to the Spark large model and domestic computing-power base. External financial reports also confirm that it is willing to sacrifice current profit for the future: 2025 revenue was RMB 27.105 billion, attributable net profit was RMB 839 million, ex-nonrecurring profit was RMB 264 million, and R&D investment was RMB 5.364 billion, equal to 19.79% of revenue. After being constrained by the U.S. Entity List, the company did not simply wait for external supply to recover, but migrated toward Huawei Ascend, domestic computing power, and the Spark X series; iFLYTEK’s inclusion on the BIS Entity List in 2019 itself also explains why “independent and controllable” is not a slogan for it, but a survival constraint.

  But points have to be deducted: its reinvention capability is currently “proved by investment,” not “proved by profit.” If the core business is disrupted, for example speech capability becomes commoditized by free general large models, education AI hardware is constrained by policy, or DeepSeek/open-source models drive API prices down, iFLYTEK is likely capable of changing product forms and computing-power routes; the problem is that every reinvention is expensive. 2026Q1 revenue was RMB 5.274 billion, attributable net profit was RMB -170 million, ex-nonrecurring profit was RMB -430 million, and operating cash flow was RMB -1.069 billion, and the company again explained the ex-nonrecurring decline as a RMB 349 million increase in R&D and sales expenses. This shows it can keep investing through a bad cycle, but it also shows the business model has not yet formed a flywheel where “the more it transforms, the more money it makes.”

  Its handling of mistakes and bad news is “above passing, but not extremely transparent.” On the positive side, the financial reports at least did not conceal the widening Q1 ex-nonrecurring loss; on the product side, after the 2023 AI learning tablet content-safety incident, the company publicly responded by taking the content offline immediately, penalizing partners, and introducing the large-model content review mechanism into the learning tablet. These actions show it does not completely resist bad news and can make emergency corrections and process patches.

  The negative is that its public narrative still leans toward “the strategy is correct, investment is necessary, and domestic substitution is leading,” with insufficient granularity in postmortems on errors. A truly excellent self-reinventing company needs to be able to coldly admit the decline of old businesses, cut wrong projects quickly, and turn bad news into organizational learning; iFLYTEK currently looks more like a reinvestment machine driven by “strong technology conviction + policy/customer scenarios.” At a share price of about RMB 44-45, market cap around RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x, the market is already paying a high price for this reinvention DNA; if ex-nonrecurring profit cannot remain positive, this DNA can only prove resilience, not that it is a ten-year five-bagger stock.

  Jun 9, 2026
• QDoes management, especially the founder, have a long-term view and deep alignment with the company? Is it willing to sacrifice current profit for five to ten years out?6/10

  A

  Conclusion: management’s long-term view and founder alignment are relative positives for iFLYTEK, but this is not a top-tier owner-operator structure. Liu Qingfeng is the founder and long-serving chairman/CEO-level core figure. The report says he has remained in office since founding the company in 1999, and under the “direct shareholding + employee/voting-rights platforms” framing, control is about 20.82%, closer to the founder-led type favored by Baillie than an ordinary professional-manager company. The more conservative public disclosure is: Liu Qingfeng directly holds 5.55%, China Mobile holds 10.03%, and Anhui Yanzhi Technology holds 2.48%; according to the prospectus, as of 2025-09-30, Liu Qingfeng controlled 11.37% of voting rights through direct shareholding, Yanzhi Technology, and voting-right entrustment. So the alignment is real, but it is not the kind of company where the founder directly holds twenty to thirty percent and has absolute control.

  There is also evidence of long-term orientation. In 2025, the company had revenue of RMB 27.105 billion, attributable net profit of RMB 839 million, and ex-nonrecurring profit of RMB 264 million, yet R&D investment still reached RMB 5.364 billion; in 2026Q1, revenue was RMB 5.274 billion, attributable net profit was RMB -170 million, ex-nonrecurring profit was RMB -430 million, and operating cash flow was RMB -1.069 billion, while management still explained this as increased investment in large-model foundation R&D and sales layout, with the ex-nonrecurring decline mainly due to a RMB 349 million increase in R&D and sales expenses. This shows the company is indeed willing to sacrifice current profit for five-to-ten-year variables such as the Spark large model, domestic computing power, and education AI.

  There is also a small but meaningful governance signal: in 2025, Liu Qingfeng and Wu Xiaoru voluntarily lowered their planned compensation, Liu Qingfeng planned to actually receive RMB 4.10 million, and the company launched its second employee stock ownership plan, covering no more than 2000 core employees at a transfer price of RMB 24.59 per share. This is positive evidence for “whether the organization is willing to invest long term and whether employees share in upside.”

  But discounts are also necessary. Direct ownership of about 5.55% is not especially heavy, and there are external structures including China Mobile, state-owned/USTC background, employee platforms, and private placements, so governance objectives cannot revolve only around maximizing the founder’s personal wealth. Meanwhile, the current price is about RMB 44-45 and market cap is around RMB 102-103 billion, corresponding to TTM PE around 120x and Forward PE around 105x, while ex-nonrecurring profit remains thin and Q1 main-business losses widened. In other words, management long-termism is an advantage for iFLYTEK, but the market has already paid a high price for this advantage, and policy, sanctions, and commercialization pressure will force management to produce profit evidence faster. Overall, this dimension is stronger than an ordinary professional-manager company, but below top-tier growth stocks with high founder ownership, very strong cash flow, and long-proven capital allocation.

  Jun 9, 2026
• QIf it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on harming society or regulation?5/10

  A

  Conclusion: if iFLYTEK disappeared tomorrow, the customers who would miss it most would be schools, education bureaus, government hotlines, primary-level healthcare, and some automakers that have already embedded iFLYTEK into daily workflows; C-end hardware and general large-model users would miss it much less. Under the report’s framing, 2025 smart education was RMB 8.967 billion, healthcare was RMB 858 million, digital government was RMB 1.558 billion, and automotive was RMB 1.240 billion. The annual report summary also disclosed that large-model API/MaaS platform service revenue was only RMB 385 million. So dependence mainly comes from vertical delivery, process embedding, data migration, and local service, not from the irreplaceability of the general model. Customers would feel clear migration pain, but Baidu, Alibaba, Huawei, ByteDance, local integrators, and privatized deployment of open-source models could all replace some functions. It is not yet at the level of “if it stops, operations cannot continue.”

  Education is the stickiest piece, but also the one with the highest social risk. In 2025, smart education accounted for about one third of revenue, and the official summary disclosed smart education revenue of RMB 8.967 billion, up 24.04% year on year, showing it is already the company’s largest scenario. Once products such as Zhixue.com, smart classrooms, English speaking/listening, essay grading, and learning tablets enter schools/families, they create switching costs around procurement, training, data, and usage habits; but student data and learning profiles are not assets that can be mined without limit. Personal information of minors under the age of fourteen and medical-health information are both sensitive personal information, and processing them requires a specific purpose, sufficient necessity, and strict protection measures. If growth relies on expanding data collection, manufacturing learning anxiety, or inducing students to over-rely on AI tutoring, sustainability will be discounted; it stands on firmer ground only if positioned as teacher assistance, parent-auditable, and based on data-minimization collection.

  Government affairs and healthcare also have high customer stickiness, but stickiness should not be overstated as monopoly. Systems such as 12345 intelligent customer service, digital civil servants, and primary-level smart medical assistants are usually not replaced frequently once they enter public workflows; domestic substitution, local deployment, and long-term operations and maintenance also favor iFLYTEK. But government and healthcare data are not ordinary SaaS data. The Data Security Law makes clear that competent authorities for sectors such as education and health are responsible for data security supervision in their industries, and important/core data require stricter management. A truly sustainable growth model should improve assisted diagnosis, quality control, primary-level efficiency, and public-service efficiency, rather than black-box decision-making or turning public data into a private closed loop.

  Content safety and the Entity List are two other hard constraints. iFLYTEK Spark, education assistants, office assistants, and government agents all fall under the generative AI regulatory framework. The Interim Measures for the Management of Generative Artificial Intelligence Services explicitly focus on false information, personal-information rights, data security, bias and discrimination, and set systems such as security assessment, algorithm filing, and complaint reporting. Overseas and computing-power risks are not risk-free either: in 2019, the Federal Register confirmed that IFLYTEK was added to the U.S. Entity List, and BIS rules explain that related exports, reexports, and in-country transfers to listed entities generally require a license. The domestic computing-power route can buffer supply-chain risk, but it cannot eliminate compliance discounts from overseas customers, partners, and suppliers.

  Therefore, this dimension is “moderately strong but conditional”: iFLYTEK has real customer missability, especially in localized, long-contract, high-migration-cost scenarios such as education, government affairs, and healthcare; but it is neither a global standard layer nor irreplaceable underlying infrastructure. At the current pricing of about RMB 44-45 per share, market cap around RMB 102-103 billion, and TTM PE around 120x, this stickiness cannot be inflated into a risk-free moat. Its growth can avoid dependence on harming society or regulation, but only if it continues proving protection of minors, clear responsibility boundaries in healthcare/government affairs, content safety, data minimization, and export compliance; otherwise the “data moat” will quickly become a regulatory liability.

  Jun 9, 2026
• QWhat are the unit economics of this business, including gross margin and incremental returns? Do they improve or worsen as scale increases? Where does the money earned go?4/10

  A

  Conclusion first:iFLYTEK is not yet a top-tier platform business whose unit economics automatically improve as scale grows. It is a business with “acceptable gross margin, but a large portion of incremental gross profit is still being consumed by R&D, sales, computing power, and project delivery.” The report says the company’s gross margin is broadly in the 41%-42% range; recalculating from announcements, 2025 revenue was RMB 27.105 billion and operating cost was RMB 15.624 billion, implying gross profit of about RMB 11.481 billion and gross margin of about 42.4%; but 2026Q1 revenue was RMB 5.274 billion and operating cost was RMB 3.219 billion, so gross margin fell to about 39.0%. This shows that scale expansion has not yet brought clear gross-margin uplift.

  The key issue is the expense ratio. In 2025, sales expenses were RMB 5.191 billion, R&D expenses were RMB 4.439 billion, total R&D investment was RMB 5.364 billion and accounted for 19.79% of revenue. Sales expenses plus R&D expenses totaled about RMB 9.630 billion, or 35.5% of revenue, equivalent to consuming about 84% of gross profit. The result was a 2025 attributable net margin of about 3.1%, ex-nonrecurring net margin of only about 1.0%, and ex-nonrecurring ROE of about 1.44%. By 2026Q1, sales expenses were RMB 1.048 billion and R&D expenses were RMB 1.179 billion, totaling RMB 2.227 billion, already exceeding quarterly gross profit of about RMB 2.055 billion; the company also explained in its first-quarter report that the decline in ex-nonrecurring net profit was mainly due to a RMB 349 million increase in R&D and sales expenses.

  Incremental returns have also not yet formed a high-return flywheel. 2025 revenue increased by about RMB 3.762 billion versus 2024, and incremental gross profit was about RMB 1.529 billion, implying an incremental gross margin of about 40.6%; but sales expenses and R&D expenses increased by about RMB 1.655 billion over the same period, exceeding the incremental gross profit. Ex-nonrecurring net profit increased by only about RMB 76 million, implying an incremental ex-nonrecurring net margin of about 2%. Operating cash flow was a bright spot: in 2025, operating cash flow was RMB 3.208 billion and sales cash collection exceeded RMB 27.4 billion, showing improved collections and project acceptance; but 2026Q1 operating cash flow was again RMB -1.069 billion, so seasonality and project-based attributes remain heavy.

  The money earned mainly keeps going into three areas: first, large-model foundation, algorithms, and domestic computing power; second, brand, channel, and overseas sales for C-end products such as AI learning tablets; third, delivery and ecosystem building for B/G-end scenarios such as education, government affairs, healthcare, and automotive. The areas that could genuinely improve unit economics are AI platform licensing, API/MaaS, standardized education products, and replicable C/B-end products, but in 2025, AI platform and licensing revenue was RMB 1.252 billion, of which large-model API/MaaS was only RMB 385 million, still too small relative to total revenue of RMB 27.105 billion. Therefore, at the current share price of about RMB 44-45 and market cap near RMB 100 billion, the honest judgment is: unit economics have paths to improve, but the company has not yet proved a strong flywheel where “the larger the scale, the higher the returns”; later, we must see gross margin recover, sales plus R&D expense ratio decline, and ex-nonrecurring profit and operating cash flow improve together.

  Jun 9, 2026
• QWhat conditions need to hold simultaneously for it to rise fivefold in ten years? Are those conditions realistic? What expectations are embedded in today’s share price?2/10

  A

  Conclusion first: based on the 2026-06-09 Sina quote of about RMB 44-45 and market cap around RMB 102-103 billion, a fivefold increase over ten years would imply a share price around RMB 222-225 and market cap around RMB 510-515 billion; the report’s own Bull range is only RMB 65-95, so fivefold is not an ordinary optimistic case, but an extreme victory case where “iFLYTEK upgrades from a project-based AI application company into a high-margin AI platform company.”

  Reverse-engineering profit is the most intuitive way to see it. If the market still gives 50x PE ten years later, it would need about RMB 10.0-10.5 billion of net profit; at 40x PE, it would need about RMB 12.8-12.9 billion; if it returns to a more mature 30x PE, it would need about RMB 17.0-17.2 billion. Compared with the 2025 annual report summary disclosures of revenue of RMB 27.105 billion, attributable net profit of RMB 839 million, ex-nonrecurring net profit of RMB 264 million, R&D investment of RMB 5.364 billion, and large-model API/MaaS revenue of RMB 385 million, this means attributable net profit CAGR over ten years would need to be about 28%-35%, and if starting from ex-nonrecurring profit, about 44%-52%. The revenue side also cannot merely be low double digit: 15% CAGR would put ten-year revenue at about RMB 110.0 billion, and 18% CAGR at about RMB 140.0 billion; by then, ex-nonrecurring net margin would also need to rise from about 1% in 2025 to roughly 8%-16% to support a fivefold market cap under 40-50x PE.

  The conditions that need to hold simultaneously are: smart education continues steady growth from RMB 8.967 billion and is not suppressed by education policy; healthcare, government affairs, and automotive stop being merely project-based growth and can accumulate reusable products; Spark, AI Agent, the open platform, and API/MaaS grow from the small base of RMB 385 million into several-billion-yuan or even ten-billion-yuan high-margin revenue; R&D investment intensity at 19.79% of revenue gradually dilutes, and sales expenses also show operating leverage; meanwhile, 2026Q1 revenue of RMB 5.274 billion but attributable net profit of RMB -170 million, ex-nonrecurring profit of RMB -430 million, and operating cash flow of RMB -1.069 billion cannot become normal, only an investment-period and seasonal trough. One external condition must also be added: iFLYTEK’s inclusion on the U.S. Entity List in 2019 must not worsen further, and the domestic computing-power route must be both usable and cheap enough, otherwise margin assumptions will be eaten away by computing-power costs and supply constraints.

  These conditions are not completely unrealistic, but the realism of the combination is low. iFLYTEK has real assets: education data, Chinese speech, government/healthcare/in-car scenarios, a long-serving founder, and the domestic computing-power narrative; in 2025, education, healthcare, government affairs, and automotive were all growing in the 20%-30% range, and overseas and token calls also showed signs of high growth. The problem is that fivefold requires not “revenue doubling,” but “sustained high revenue growth + a tenfold-level improvement in ex-nonrecurring margin + no obvious valuation derating” to happen at the same time. Judging from current ex-nonrecurring profit and Q1 losses, the evidence has not reached that level.

  Today’s RMB 44-45 share price already embeds very high expectations: market cap is around RMB 102-103 billion, equal to about 120x PE on 2025 attributable net profit and several hundred times ex-nonrecurring profit; based on Forward PE around 105x, the market is already looking for roughly RMB 1.0 billion of 2026 attributable profit and assuming X2 commercialization, education/government orders, overseas growth, and returns on R&D investment will gradually materialize. In other words, the market has not missed the AI story; it has already paid a very expensive ticket for the story. The only real variant perception can be around “whether iFLYTEK can ultimately turn AI applications into a high-margin platform profit pool”; if it remains mainly an education, government, and industry project-integration company, today’s price embeds rather full expectations, and the odds of a ten-year fivefold return are not high.

  Jun 9, 2026
• QWhy has the market not realized all this yet? Is it unable to understand, looking down on it, or not looking far enough? What will become the “narrative inflection point”?3/10

  A

  Conclusion: the market has not failed to notice iFLYTEK’s AI narrative. It has already seen the theme, but has not yet seen proof at the level of ex-nonrecurring profit and cash flow.Based on the 2026-06-09 Sina quote of about RMB 44-45, market cap of RMB 102-103 billion, TTM PE around 120x, and Forward PE around 105x, the market is not “looking down on” it. On the contrary, it has already assigned a high premium to “domestic large model, education data, government-enterprise scenarios, X2, and overseas expansion.” The real variant perception is that bulls are buying AI application platformization in years 3-10, while bears are focused on 2025 ex-nonrecurring profit of only RMB 264 million, 2026Q1 ex-nonrecurring loss of RMB -430 million, and operating cash flow of RMB -1.069 billion.

  So the issue is not “not understanding,” but “not looking far enough, while also not daring to look too far.” The company’s 2025 revenue of RMB 27.105 billion, attributable net profit of RMB 839 million, and R&D investment of RMB 5.364 billion are all clear in the 2025 annual report summary; but the same annual report also shows that large-model API/MaaS revenue was only RMB 385 million. Although it grew +263% year on year, it remained very small relative to total revenue of RMB 27.1 billion. In other words, the market already knows the “AI narrative,” but has not yet seen AI revenue that is “high-margin, replicable, and sustainably paid” become large enough to support a RMB 100 billion market cap.

  The narrative inflection point will first come from profit quality, not product launches. The strongest positive inflection point would be ex-nonrecurring profit turning positive and continuing to improve in 2026H1 or full-year 2026, especially if R&D and sales investment no longer consumes incremental revenue; next would be operating cash flow turning positive again from Q1’s RMB -1.069 billion, because although the 2026 first-quarter report disclosed revenue growth of +13.23%, contract value of +28.16%, opportunity reserve of +33%, and developer large-model daily average Tokens of +4241%, it also disclosed wider ex-nonrecurring losses and cash-flow pressure. The third inflection point would be X2 commercialization moving beyond the strategic label of “trained on fully domestic computing power” and producing large, renewable orders from central/state-owned enterprises, education, healthcare, and automotive, proving that model capability can convert into revenue and profit.

  Four secondary inflection points can also be watched: first, API/MaaS scaling from a small base of RMB 385 million to the billion-yuan level while resisting price pressure from open-source models such as DeepSeek and Qwen; second, overseas business continuing to scale after high growth in 2025, rather than being a one-off channel burst; third, education business continuing to grow from a base of RMB 8.967 billion without triggering stronger education regulation; fourth, Entity List risk showing marginal improvement, or at least not worsening. The fact that iFLYTEK was placed on the U.S. BIS Entity List in 2019 can be verified in the U.S. Federal Register rule, and this risk does not automatically disappear because of the domestic computing-power narrative.

  Conversely, if X2 has no major orders, API/MaaS remains small and beautiful, and ex-nonrecurring profit continues to be supported by thin profit outside government subsidies and nonrecurring items, the variant perception will converge negatively: it would not be that the market undervalued a great growth stock, but that the market prematurely capitalized the blue-sky scenario of a “ten-year five-bagger.” The most honest judgment now is: the AI banner has already been fully seen by the market; what the market has not fully believed is commercialization quality. Only when ex-nonrecurring profit, cash flow, X2 orders, and MaaS volume ramp all materialize together can a real narrative inflection point form.

  Jun 9, 2026