In-Depth Value Investment Analysis of Empyrean Technology

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

• QHow high is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?3/10

  A

  Bottom line: the “nominal ceiling” of Empyrean Technology’s market is not high. The entire global EDA market is only around $20 billion in scale (about $19.2 billion in 2024 and about $20.8 billion in 2026), far below the TAM of most companies Baillie Gifford would classify as “great growth stocks.” At its core, this is mainly a story of taking share through domestic substitution inside an existing pie dominated by the three giants, rather than creating a large new market from scratch. It does have incremental themes such as 3DIC, advanced packaging, and AI+EDA, but their scale is limited and cannot change the basic constraint that the pie is small. The real ceiling bottleneck is therefore not “how much share Empyrean Technology can take,” but “the pie itself is not large.” Under Baillie Gifford’s LTGG standard, the nominal height is constrained, and the effectively reachable opportunity must be discounted again for dependence on domestic-substitution policy and the technology gap versus the three giants. This is the first hard flaw exposed in the ten questions, making Question 1 hard to score highly.

  1. Nominal ceiling: global EDA is a small but difficult market

  EDA is one of the three strategic foundations of the chip industry. It has a long runway, deep technical snow, and a “picks and shovels” profile, but a deep slope does not mean a large market. The current global EDA market is only about $19.2 billion (2024) and $20.8 billion (2026, EDA tools basis). Even if one extrapolates various institutions’ 8%–10% CAGR to around 2032, it is only around $35–38 billion. Compared with Baillie Gifford’s past major holdings such as Tesla (autos + energy worth trillions of dollars), Amazon (retail + cloud), and ASML (unique lithography position tied to the entire advanced-node capex cycle), the whole EDA market sits in the smaller bucket on the LTGG TAM yardstick.

  The first honest point is this: the ceiling Empyrean Technology can nominally reach is capped by the size of the global EDA pie itself. This is not a business in a market that is still undefined with unlimited imagination. It competes in a mature industrial software market with clear boundaries, clear players, and a limited buyer base of global chip design and manufacturing companies.

  2. Nature of the opportunity: overwhelmingly expanding an existing pie through domestic substitution, with incremental creation as a secondary layer

  Empyrean Technology’s growth can be split into two parts:

  1. Replacement of existing spend (the main driver and most of the logic). This is a story of taking share from the existing pie of the three giants. The global EDA profit pool is highly concentrated: Synopsys has about 31%–32%, Cadence about 29%–30%, Siemens EDA about 13%, and the three together exceed 70%. After Synopsys completed its $35 billion acquisition of Ansys in July 2025, the combined share of Synopsys and Cadence in EDA software moved closer to about 85%. The giants’ moats are not loosening; they are being reinforced through full-stack integration across multiphysics and advanced packaging. Empyrean Technology’s growth narrative is mainly about taking the share in China that originally belonged to the three giants. The report notes that 90.09% of its revenue comes from mainland China and that its global share remains in the single digits. That is the signature of replacement of existing spend, not incremental market creation.

  2. Incremental creation (secondary and limited in scale). Empyrean Technology does have several moves that create new markets or position early for new demand: according to the report, it is the only domestic vendor providing full-flow EDA for 3DIC design verification, and it has launched the PyAether platform with AI assistance, covering RF, flat-panel display, wafer manufacturing, and advanced packaging. In the post-Moore era, Chiplet, heterogeneous integration, and AI+EDA are indeed new directions that raise EDA tool complexity and pricing, making the pie somewhat thicker. But the honest point is that these increments still occur within the same $20 billion global EDA pie. They make the same pie higher value; they do not open a separate new continent alongside EDA. Moreover, 3DIC and advanced packaging are precisely the multi-die direction that Synopsys is prioritizing after integrating Ansys. Empyrean Technology is domestically leading here, not globally exclusive.

  Conclusion: expanding the existing pie through domestic substitution is the main source of its ceiling; creating new markets is only a marginal thickening and cannot sustain an independent growth story.

  3. Effective reach: two further discounts apply on top of the nominal ceiling

  The real Baillie Gifford question is not “how high is the ceiling,” but “how much can Empyrean Technology effectively convert.” Two discounts must be applied to the nominal height:

  • The addressable space is really a subset of “China market × domestic penetration.” China’s EDA market in 2025 is about RMB 14.95–18.49 billion (about 18% of the global market), and the domestic penetration rate rose from less than 5% in 2020 to about 11.5%–18% in 2025. The direction of penetration improvement is real, and this is Empyrean Technology’s strongest bullish anchor. But Empyrean Technology’s current share of China’s EDA market is only about 5.9%, ranking fourth domestically. In other words, even if the domestic-substitution thesis fully plays out and the domestic penetration rate reaches 30%, Empyrean Technology would still receive only the narrowed subset of “China’s pie × domestic penetration × its own share among domestic vendors.” Between the nominal ceiling (global $20 billion) and the effectively reachable opportunity sit two orders of discounting.

  • Discount 1: policy dependence. The rise in domestic penetration depends heavily on national major projects, stronger SASAC support for integrated circuits, and the demand for “independent and controllable” supply chains forced by supply cutoffs, rather than purely market-based product superiority. This kind of ceiling is space granted by policy. Its realization pace depends on trade friction, subsidy intensity, and customer validation cycles, making it less certain than a TAM driven by rigid demand.

  • Discount 2: the technology gap versus the three giants. The report and external data are consistent: the domestic penetration rate in advanced nodes below 5nm is even below 5%; digital back-end tools are still dominated by international giants, with a replacement rate below 20%. Empyrean Technology’s strength is concentrated in full-flow analog, where domestic penetration has exceeded 40%, but it remains a follower in the largest and most profitable full-flow digital SoC area. The highest and thickest part of the ceiling, advanced digital design, is precisely the part it has the hardest time reaching.

  4. Overall judgment on Question 1 under the Baillie Gifford LTGG lens

  • Nominal height: low. Global EDA is about $20 billion, with growth from single digits to 10%. It is not the “huge and fast” category LTGG prefers, and Empyrean Technology is nominally capped by that pie.
  • Market nature: mainly expanding an existing pie through domestic substitution, with new-market creation as a secondary layer through limited 3DIC/AI+EDA increments. Its growth ceiling depends on how much it can take from the three giants, and the ceiling for share-taking is far lower than the ceiling for defining a new category.
  • Effective reach: on top of the nominal ceiling, one must further discount for policy dependence and the technology gap. The true addressable opportunity is the triple subset of “China market × domestic penetration × own share.” The direction, higher domestic penetration, is real, but the scale, pace, and certainty of realization are insufficient to support the “ten years of high growth” implied by the current RMB 57.5 billion market capitalization. This already aligns with the report’s “good company, bad price” conclusion at Question 1: on the ceiling question, Empyrean Technology’s answer is “the narrative is real, but the space is constrained,” and it cannot explain why the market has not realized this could rise tenfold.

  Jun 5, 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

  Bottom line: the certainty of a five-year revenue doubling is relatively high. The historical six-year compound rate is about 31%, and broker consensus still expects 28%–33% growth in 2026–2028 (Sina Finance commentary on the 2025 annual report and 2026 Q1 report, forecasting 2026–2028 revenue of RMB 1.765/2.300/2.954 billion). A five-year doubling only requires about 15% compound growth, a low bar. It will be driven mainly by “volume” (domestic-substitution penetration plus more licenses and customers), supported by “new businesses” (platformization, 3DIC, AI+EDA), while “price” contributes almost nothing because pricing power is weak. But the honest point is that this is a low-bar, low-quality doubling: revenue doubling does not equal per-share value doubling. Net margin has slid from about 19.9% in 2023 to about 4.6% in 2025, 2026Q1 turned to a loss of about RMB 73 million, growth abruptly slowed to +9.65%, and share-based payment amounted to 237.98% of net profit, creating dilution. Revenue doubling may well come with stagnant or even declining profit. Under Baillie Gifford’s standard of finding great growth stocks that can rise fivefold in ten years, Question 2 gets a mixed judgment: it passes on revenue, but fails on quality.

  1. Feasibility of a five-year doubling: history suggests a high probability, but 2026Q1 is a deceleration warning

  A five-year revenue doubling implies about 15% compound growth. Empyrean Technology went from about RMB 257 million in 2019 to RMB 1.325 billion in 2025 (2025 revenue of RMB 1.325 billion, up +8.40% YoY), a six-year CAGR of about 31%, more than twice the 15% threshold. The historical cushion is ample. Sell-side consensus is also not pessimistic: the commentary cited above gives 2026–2028 revenue of RMB 1.765/2.300/2.954 billion, with growth of about 33%/30%/28%. If realized, revenue would more than double from 2025 by 2028, in only three years. On this line alone, a five-year doubling is almost the built-in result of industry beta plus domestic-substitution policy, with relatively high certainty.

  But one point cannot be covered by the narrative: the near-term growth inflection. Reported 2025 growth had already fallen to only +8.40%, and 2026Q1 fell further to +9.65% (revenue of RMB 257 million), down from the frequent 20%–30% growth of prior years to single digits or low double digits. The report also records that “in the first quarter of 2026, the company still recorded a RMB 73 million loss despite 9.65% YoY revenue growth.” The judgment here is whether this is quarterly noise or a trend slowdown. My view is that it is both, but more seasonal noise plus a high base: Empyrean Technology has a very strong Q4 revenue-recognition concentration (the report records 2025Q4 single-quarter revenue of RMB 520 million, far above the first three quarters), and Q1 is naturally a low season with a low full-year weight, so one quarter’s growth is less representative. At the same time, the high base from 2023–2025 naturally suppresses YoY readings. Therefore, the five-year doubling should not be scared away by Q1’s single-digit-like growth, but an optimistic return to 30% compound growth cannot be assumed. A more realistic anchor is 15%–25% compound growth: a doubling is achievable, but it may not be a doubling in the sense of high growth.

  2. Growth breakdown: volume is primary, new businesses support it, and price is almost zero

  • Volume (main driver): This is the core engine for Empyrean Technology’s five-year doubling. Domestic EDA penetration remains low (the report says the company is first domestically, but international giants still dominate the domestic market and local vendors’ overall share remains small), the customer base has exceeded 700, and mainland China accounts for 90.09% of revenue. Growth comes from two “volume” lines: first, new customers and new design teams adopting the tools, increasing license seats and modules; second, existing customers expanding procurement from point tools to full flows. This line is supported by domestic-substitution policy and the capex cycle of foundries and design houses. It is the most certain part, but it is also most exposed to downstream capex and policy rhythm, making it industry beta rather than company alpha.

  • New businesses (key source of optionality): These include platformization (the PyAether unified database framework), AI+EDA (Aether Coder intelligent code generation, Hima EMIR, and others mentioned in the commentary above), and advanced packaging/3DIC (the report says the company is “the only domestic provider of full-flow EDA for 3DIC design verification,” while the commentary mentions the Argus 3DIC verification platform and 4 products already launched). These are the sources of optionality that could pull revenue doubling toward a higher compound rate, but the pace of realization is uncertain. In the short term, they are more about positioning and filling capability gaps than scaled monetization. Overseas expansion, with overseas revenue at about 9.91%, is theoretically incremental, but trade friction constrains it and makes it hard to be the main force.

  • Price (basically no contribution): Weak pricing power is a hard constraint. The report judges that the company has local bargaining power but lacks strong overall pricing power. Licenses are priced by duration, quantity, tool mix, and technical difficulty, and for major customers pricing is more project-based negotiation. As a follower with still-small share, the company has little general price-raising power across the market. Therefore, the contribution of “price” to a five-year doubling is close to zero; the doubling must come from volume expansion and category expansion.

  3. Honest point: the doubling bar is low, but this is low-quality growth

  This is the layer that matters most from a Baillie Gifford perspective: revenue doubling does not equal per-share value doubling, for three reasons:

  1. Margins are thinning in the opposite direction, and revenue and profit are badly decoupled. From 2019 to 2025, revenue rose about 4 times, while net profit attributable to shareholders barely moved (RMB 57 million to RMB 61 million). Net margin went from about 19.9% in 2023 to about 9.0% in 2024 and about 4.6% in 2025, with 2026Q1 directly turning loss-making. R&D was 64.84% of revenue (RMB 859 million), continuing to consume profit. In other words, even if revenue rises to about RMB 2.6 billion over the next five years, if R&D intensity and upfront expenses do not improve, profit could remain stagnant or even negative. The “doubling” is a doubling of scale, not distributable profit to shareholders.

  2. Heavy dilution from share-based payment. Share-based payment was RMB 145 million in 2025, equal to 237.98% of net profit attributable to shareholders. Per-share value is continuously diluted. A meaningful portion of the fruits of revenue doubling flows to employee incentives rather than outside shareholders, making growth on a per-share basis materially weaker than growth in aggregate figures.

  3. Strong cycle and policy dependence: beta rather than alpha. Growth is highly tied to the capex cycles of foundries and design houses and to the strength of domestic-substitution policy. If downstream conditions or financing weaken, or if policy implementation falls short, the volume engine will stall. The 2026Q1 growth plunge was already a rehearsal.

  Final judgment on Question 2: five-year revenue doubling has relatively high certainty, leaning toward “pass”, driven mainly by volume from domestic-substitution penetration, supported by new businesses, with no price contribution. But high-quality growth does not pass. This is a low-bar, lower-quality doubling: margins are thinning, Q1 turned loss-making, near-term growth slowed, and dilution is heavy. Revenue growth has not yet translated into growth in per-share value and sustainable free cash flow. In the context of the Baillie Gifford ten questions, this exactly confirms the report’s “good company, bad price” undertone: there is evidence that the company will become larger, but not yet evidence that it will become more profitable as it grows, and the latter is the real precondition for a fivefold return over ten years.

  Jun 5, 2026
• QFive years from now, what will take over as the next growth engine? Does this “second curve” exist today?4/10

  A

  Bottom line: the second curve has visible outlines, but almost none of it has taken over today. Empyrean Technology has many second-curve candidates: full-flow digital platformization (PyAether / AI-native EDA), 3DIC / advanced-packaging EDA, wafer-manufacturing tools (DTCO / memory / advanced processes), and overseas globalization. Most are no longer pure PPT; the company has named products that can be checked (Argus 3DIC physical verification platform, AndesAMS, HimaSim/HimaTime, PyAether agent Aether Coder). But measured honestly against the Baillie Gifford LTGG ruler of “visible today and capable of taking over in the future”: the two most imaginative lines, AI-native EDA and the full-flow digital platform, contribute close to 0 revenue today and are not even separately disclosed in the annual report, remaining in the technology breakthrough/adoption phase; 3DIC is “domestically unique” but the market is still early and no visible revenue has been split out; what actually converted into revenue growth in 2025 was instead technical services, overseas sales, and hardware agency, these auxiliary revenues rather than any disruptive new platform. The current revenue base still depends mainly on established strengths such as analog and flat-panel display. In other words, the outlines do exist, but the handoff has not happened today.

  Below is a line-by-line check of each candidate curve: whether it exists today, whether it already contributes visible revenue, or whether it is still a vision. This is exactly the test Baillie Gifford cares about.

  ① Full-flow digital platformization (PyAether + digital EDA): the outline exists, but it is “getting smaller” and is far from taking over. This should be Empyrean Technology’s core battlefield for catching up with the three giants and opening a second growth curve. On capability, there has been progress: in 2025 the company launched 4 digital products including HimaSim and HimaTime, and its digital EDA tools already cover nearly 80% of the main tools for digital circuit design, with HimaTime timing accuracy reaching industry benchmarks. It also launched the PyAether platform on a unified database (more than 12,000 Python APIs). But looking honestly at revenue: digital circuit design EDA as a share of EDA software revenue has fallen year by year, from 36.32% in 2019 to 17.96% in 2021, while the revenue mix has become increasingly reliant on analog. This means the digital full flow has not only failed to take over today; its share is being diluted by existing strengths. The chairman himself also admitted that “ecosystem is the current main shortcoming,” and that there is still a significant gap versus the three giants in sub-3nm advanced processes and multiphysics simulation. Judgment: the outline exists and capabilities are being filled in, but revenue contribution is small and the share is falling. This is the most promising line, but also the one least appropriate to include in valuation today.

  ② AI-native EDA (PyAether agent / Aether Coder / AndesAMS): products exist, but revenue is close to 0; today it is a narrative, not a curve. This is the story the market most likes to use when assigning imagination to Empyrean Technology. In 2025, the company did launch PyAether agent Aether Coder, ArgusFPD Triage AI for false-error filtering in flat-panel display, and the intelligent platform AndesAMS, said to improve efficiency by 50%+. But the key point is: the annual report only describes AI+EDA as having “achieved major breakthroughs,” without any separate revenue disclosure or independent revenue line. It is still in the technology breakthrough/adoption stage rather than visible cash flow. For Baillie Gifford, an AI efficiency tool looks more like a moat investment that strengthens existing tool competitiveness, with commercialization model, willingness to pay, and scale curve still standing between it and becoming a new engine five years from now. Judgment: today it remains at the vision layer and should not be inflated to fit a growth narrative.

  ③ 3DIC / advanced-packaging EDA: the scarcity is the strongest, but the market is early and independent revenue realization is not visible. This is Empyrean Technology’s most differentiated card: it is the only domestic provider of full-flow EDA for 3DIC design verification, launched its first Argus 3DIC physical verification platform in 2025, supports 2.5D/3D heterogeneous integration, and shortens the manual design cycle by about 60%. Under the Chiplet/post-Moore narrative, the direction is almost unimpeachable. But two honest points matter: first, advanced packaging/3DIC is still in early adoption in China, and the annual report does not split out independent revenue; being “domestically unique” is more positioning value than a realized profit pool. Second, its actual contribution to total revenue of RMB 1.325 billion cannot be verified as meaningful today. Judgment: this has the clearest outline and the highest strategic value, but “the market is early + revenue is not split out” means it is only a visible seed today, not an engine that has taken over.

  ④ Wafer-manufacturing EDA (DTCO / memory / advanced processes): capability milestones are frequent, but revenue remains invisible. In 2025 the company continued to advance manufacturing tools: ALPS has 4nm/5nm process verification capability, and the company launched the only domestic full-flow EDA for memory chips that can support ultra-large-scale Flash/DRAM mass production; its EDA tools cover 70%+ of domestic foundry process nodes. These are solid capability fills and ecosystem investments that bind it more deeply with foundries. Again, however, manufacturing EDA has no separate, meaningful revenue disclosure and is more of a long-term foundation for entering customer workflows. Judgment: capability exists, revenue has not emerged, so this is under construction and not yet realized.

  ⑤ Overseas / globalization: the smallest base and fastest growth; the only second curve that has “moved in the numbers,” but still far from being an engine. Interestingly, among the five candidates, the one that most resembles visible incremental contribution today is this least glamorous one: overseas revenue was RMB 131 million in 2025, up +127.45% YoY, representing about 9.9% of revenue (mainland China RMB 1.194 billion / 90.09%). Growth is eye-catching, but the absolute amount is still small and rests on high uncertainty from geopolitics and trade friction, which the report already lists as a source of permanent capital-loss risk. Judgment: it exists and contributes visible revenue, but its scale and certainty are insufficient to carry the handoff five years from now.

  Putting the account together: an honest score for whether the second curve exists today. This is the key honest test: inside 2025 revenue of RMB 1.325 billion, EDA software sales actually declined slightly by 1.63% YoY to RMB 1.075 billion, and the full-year increment mainly came from technical services +74.93% (RMB 201 million), hardware agency/software/other +228.7% (RMB 49 million), and overseas +127.45% (RMB 131 million). In other words, what actually converted into revenue growth in 2025 was auxiliary expansion such as services, overseas, and agency, while the most hoped-for disruptive new platforms, such as full-flow digital, AI-native EDA, and 3DIC, are still almost absent from software revenue growth today. Given Empyrean Technology’s only about 6–9% share in domestic EDA and the fact that the three giants still hold about 68–78% of the China market, even if all these second curves are directionally right, they remain in the early stage of positioning, customer adoption, and waiting for volume.

  Conclusion for Baillie Gifford: the second curve’s “degree of existence” is roughly “outline visible, handoff not yet arrived.” It meets Baillie Gifford’s low bar of being visible today (the products, platforms, and positions are real, not pure PPT), but it is far from meeting the higher bar of visible handoff signals and revenue migration toward new engines. The most imaginative AI+EDA / full-flow digital lines are too small to be separately disclosed today; 3DIC is scarce but the market is still early; existing cash flow remains highly dependent on established strengths such as analog and display. For a company with a market value of about RMB 57.5 billion and a PS ratio of about 42.8 times, these second curves carry most of the “prepayment” in the valuation, while their true contribution today is, honestly, still close to 0.

  Jun 5, 2026
• QWhat is its core competitive advantage? Will this moat widen or narrow over the next three to five years?3/10

  A

  The core competitive advantage is a “composite moat” rather than a single overwhelming barrier: more than ten years of technical accumulation (402 authorized patents and 186 software copyrights), scarce EDA talent with a training cycle of about ten years, medium-to-strong switching costs from tools deeply embedded in customer design/verification/process flows, local adaptation and service response, the policy window for domestic independent controllability, and the domestic-only full-flow position in 3DIC design verification. The true nature of this moat is a local relative advantage under the domestic-substitution window, not a global monopoly. Empyrean Technology is first domestically, but has only about 6% share of China’s EDA market and single-digit global share, with limited pricing power over leading customers. Under Baillie Gifford’s LTGG standard that a moat should be able to widen over the next three to five years, my net judgment is that domestically it will probably be stable to slightly wider, while globally it is hard to call it meaningfully wider. The moat direction is mildly positive, but its width and pricing power are far from a rent-collecting position, consistent with the report’s 3/5 score.

  Breakdown of moat sources and real strength

  1) Technology and product accumulation: medium-to-strong, and widening. The company’s products already cover analog, memory, RF, digital, flat-panel display, wafer manufacturing, advanced packaging, and 3DIC, with full-flow analog as its clear strength. The most persuasive incremental signal is that the Argus 3DIC physical verification platform disclosed on May 25, 2026 made it the “only domestic provider with full-flow 3DIC design verification EDA capability,” with verification efficiency said to reach 5 times that of traditional overseas tools and to compress high-end 3D stacked-chip full-chain verification from 2 weeks to 1-2 days. This directly positions it in the incremental post-Moore track of Chiplet/heterogeneous integration and is the strongest argument on the “moat widening” side. But the shortcomings must be stated honestly: its full-flow capabilities in digital circuits, wafer manufacturing, and advanced packaging remain weaker than the three giants. The company’s own wording that these areas are “expected to be completed as soon as possible” admits they are not yet complete. The planned acquisition of Xpeedic to fill system-level EDA gaps was also terminated in July 2025 because the parties failed to agree on core terms, so filling gaps will rely more on internal R&D and move more slowly.

  2) Talent barrier: real but an industry-wide barrier. The fact that EDA R&D talent needs about ten years from university research to practical industry competence is a high threshold that blocks most new entrants. But it does not form a moat against the three giants, whose talent pools are deeper. Against domestic peers such as Primarius, Semitronix, and X-Epic, it is relative rather than exclusive leadership. This is more an industry entry barrier that protects the entire domestic EDA sector, not something exclusive to Empyrean Technology.

  3) Workflow embedding / switching costs: medium-to-strong and the core source of stickiness. Once a customer has used an EDA toolchain to complete chip design, verification, and process collaboration, models, PDK adaptation, methodology, and team habits all become embedded. Switching requires re-validation and carries high risk. The company’s 700+ customers and mainland revenue of about 90.09% confirm that stickiness has formed. But the same sword cuts both ways for the three giants: because switching costs are high, leading customers already using full-flow Synopsys/Cadence tools are actually harder for Empyrean Technology to dislodge. Switching cost is both Empyrean Technology’s shield for retaining existing business and a wall that blocks its attack on leading incremental accounts.

  4) Local adaptation and service response: positive, but hard to convert into pricing power. Being close to local customers, iterating quickly, and responding to service needs promptly are real advantages in the domestic-substitution context, but they are service quality rather than structural barriers. Peers can copy them in similar ways, and they are hard to translate into broad price-raising power.

  5) Policy moat from domestic independent controllability: the strongest current tailwind, but external and reversible. This is the biggest variable in the moat narrative over the past year. The repeated U.S. EDA export controls directly catalyzed faster domestic substitution. But two points cannot be ignored. First, it is external to the company and set by geopolitics; when 2025 controls were once lifted, the three giants’ share prices surged, showing that the window can tighten or loosen. Second, it protects the whole domestic EDA sector. Empyrean Technology still needs to win within the sector through its own hard strengths in items 1-4. Treating the policy window as a company-exclusive moat is a typical overstatement in a growth narrative.

  6) The only domestic full-flow 3DIC capability: the sharpest differentiated position. As in item 1, this is the genuinely exclusive segment relative to domestic peers (Primarius leans toward device modeling/yield, Semitronix toward finished-product testing and yield improvement), and one of the few areas where it can claim local leadership versus the three giants. It contributes the most certain widening to the moat over the next three to five years.

  Compared with the three giants: the gap is clear and should not be glossed over

  Global EDA is dominated by Synopsys (about 31-32%), Cadence (about 29-30%), and Siemens EDA (about 13%); the three together exceed 70% globally and about 78-80% in China. The gap is not just share, but four structural dimensions: ecosystem completeness (the three giants have end-to-end loops from front-end synthesis and place-and-route to signoff, plus deeply certified PDK/IP libraries with foundries), IP library depth (Synopsys/Cadence’s semiconductor IP businesses are themselves multibillion-dollar moats, while Empyrean Technology has almost no comparable scale), full-flow digital maturity (advanced-node digital signoff remains a clear weakness for Empyrean Technology), and global customer base (overseas revenue is only about 9.91%). Conclusion: Empyrean Technology’s moat is a relative advantage “in China, in segments such as analog/3DIC, and under the control window.” All three qualifiers matter. It is not a global price-setting oligopolist, has limited pricing power over leading customers, and lacks broad general pricing power.

  Next three to five years: net judgment on widening versus narrowing

  Forces widening it: ① deeper localization, as local customers are forced to adopt domestic tools under control catalysts and domestic EDA is considered to have “opened a breach”; ② platformization, as the PyAether unified database plus AI assistance advances toward full-flow toolchains and strengthens positive feedback; ③ 3DIC positioning, as Argus occupies a domestic blank space in the incremental heterogeneous-integration track.

  Forces narrowing it: ① counterattack from the three giants, which can rapidly recover through ecosystem/IP/price bundles once controls loosen and are also increasing investment in AI+EDA; ② customer path dependence, as leading design companies’ key flows still depend heavily on international tools and switching costs work against Empyrean Technology’s assault; ③ domestic peer competition, as Primarius has completed six acquisitions, including Boda Micro, Magwel, and Aragio, to fill out its process, while Semitronix and X-Epic approach in their own segments. The sector’s dividend is split among multiple players and price-war risk rises.

  Net judgment under the Baillie Gifford LTGG lens: The moat direction is slightly toward “widening,” but the positive magnitude is limited and highly dependent on the external policy window. Domestically, it will probably be stable to slightly wider; globally, it is unlikely to materially narrow the generation gap versus the three giants over the next three to five years. For the standard of finding a great growth stock that can rise fivefold in ten years, this moat can support “surviving and defending domestic share,” but it is not yet enough to support “expanding pricing power plus a step-change in margin and free cash flow.” The latter is what turns a moat into shareholder compounding. This qualitative point is confirmed by the fundamentals: 2025 revenue of RMB 1.325 billion, net profit attributable to shareholders of only RMB 61 million, R&D at 64.84% of revenue, and gross margin of about 89.25% that does not reach the bottom line. This shows the moat is still at the stage of a defensive technical barrier, not an offensive pricing moat. Therefore, the moat item gets 3/5: it genuinely exists and is slightly widening, but its width, exclusivity, and monetization are insufficient to justify the current extreme valuation of about RMB 105.8 per share and PS of about 42.8 times.

  Jun 5, 2026
• QIf its core business is disrupted, does it have the genes for self-reinvention? How does it treat mistakes and bad news?5/10

  A

  Bottom line: Empyrean Technology’s “self-reinvention gene” is reflected more in proactive technical-route iteration than in surviving a near-death crisis and being reborn. The evidence is neutral to slightly positive, but not enough to form the vivid conviction anchor Baillie Gifford looks for. There are two verifiable positive lines: first, sustained high R&D without cuts (2025 R&D investment of RMB 859 million, 64.84% of revenue, maintained even as profit fell sharply); second, proactive transition from point tools toward a full-flow digital platform (PyAether) + AI (the large-model-based Aether Coder agent, used for API retrieval and natural-language code generation), while positioning in the new 3DIC / advanced-packaging paradigm (the report says it is the only domestic full-flow provider for 3DIC design verification). But on the character dimension Baillie Gifford cares about most, “how it treats mistakes and bad news,” it lacks a vivid, verifiable example of founders publicly reviewing a major failure. Together with one governance fact that must be honestly updated, I can only give this question a roughly neutral judgment after offsetting positives and negatives.

  1. Self-reinvention gene: there is evidence of proactive iteration, but no real-world test of rebirth after disruption.

  Baillie Gifford asks this question to understand whether, if AI-native EDA reshapes the traditional point-tool paradigm, or if the three giants open-source or attack from above, or if the domestic-substitution policy window narrows, the company has the organizational genes to change its way of living. Empyrean Technology offers signals of proactive iteration at the technical-route level, not evidence of organizational rebirth:

  • Willingness to spend for the long term and not cut R&D when profit collapses. In 2025, net margin had already fallen from about 19.9% in 2023 to 4.6%, with net profit attributable to shareholders only RMB 61 million, yet R&D still reached 64.84% of revenue (RMB 859 million). Maintaining high R&D when “larger scale means less profitability” is a positive sign of willingness to invest for the long term and avoid being bound by short-term profit. This aligns with Baillie Gifford’s preference for tolerating long periods of low profitability in exchange for disruptive capability.
  • Proactive migration to new paradigms, rather than passively waiting to be disrupted. The company is building the PyAether platform around a unified database and making AI a main line of transformation (API retrieval, natural-language automatic code generation, process automation), while also laying out 3DIC / advanced packaging. Notably, the Xpeedic acquisition it terminated in 2025 targeted a scarce company described as “globally leading in 3DIC design platforms.” The deal failed, but Empyrean Technology is still developing internally on this frontier that is most likely to be disrupted by AI + heterogeneous packaging. This shows it senses that the paradigm is shifting and is actively betting on it.

  But honestly, these are proactive iterations, not rebirth after disruption. The company is relatively young and has not been tested through a full semiconductor cycle. Although it is first domestically, it remains a follower globally under the dominance of the three giants, in the report’s framing. It has never proven, in a real near-death crisis where one leg of the core business is cut away, that it can change its bones and be reborn. The AI transition currently remains “embedding AI into tools for efficiency,” not “reconstructing the whole product paradigm.” Whether it can withstand an AI-native EDA wave led by stronger players remains an unproven assumption rather than a verified gene.

  2. How it treats mistakes and bad news: one disciplined M&A example is notable, but there is no vivid failure-review case, and bad-news disclosure is still only at the level of honest accounting.

  • Positive: rationally cutting losses and not forcing M&A. In July 2025, the company announced termination of the acquisition of 100% of Xpeedic, officially because “the parties did not reach agreement on core terms”; market analysis pointed to differences over valuation, control integration, and performance commitments. For a high-valuation technology company, refusing to force through a deal that cannot be agreed just to “increase scale” or tell an acquisition story is a positive signal of capital discipline. It shows management can say no when facing a temptation that does not pay.
  • Neutral to slightly positive: bad-news disclosure is transparent, with no obvious major fraud signs. The company has honestly disclosed slowing growth, year-by-year margin thinning, and even the 2026 Q1 loss (net profit attributable to shareholders of RMB -73 million while revenue still grew +9.65% YoY). The report also judges that “no obvious signs of fraud were seen.” But this is only honest bookkeeping, still some distance from what Baillie Gifford truly wants to see: management actively reviewing its own mistakes and clearly explaining lessons learned. Terminating Xpeedic is more like avoiding a trap beforehand than admitting an error afterward. It is not essentially a sample of “we messed up, then corrected it this way.”
  • Honest point: one governance fact must be updated. The report and common descriptions say Empyrean Technology has “no controlling shareholder / no actual controller, with CEC and concert parties holding about 33.9%.” But public announcements show this is already outdated: on December 2024, the company formally changed from having “no actual controller” to having China Electronics Corporation (CEC) as the actual controller. CEC holds about 34.06% through China Electronics Co., Ltd. (21.22%) + CEC Jintou Holdings (12.84%), obtained 6 of 11 board seats, and consolidated the company. This is a double-edged fact for this question: on one hand, it is no longer a governance vacuum with nobody responsible; central SOE control strengthens strategic support and long-term investment willingness if the domestic-substitution policy window narrows. On the other hand, from the Baillie Gifford paradigm, control resting with a central SOE system rather than a founder/manager with a large equity stake weakens the case that, if the core business is disrupted, a high-conviction owner will drive painful self-reinvention. The motivation for reinvention is more likely to come from industrial policy and system-level considerations than from the entrepreneurial instinct of having one’s life tied to the company.

  3. Net conclusion after positives and negatives.

  • Positive items (real and verifiable): long-termism shown by not cutting high R&D, proactive technical iteration in AI + platformization, internal efforts to position in the new 3DIC paradigm, M&A discipline shown by terminating Xpeedic, and transparent disclosure of bad news.
  • Negative items (also real and unavoidable): no vivid, verifiable example of founders publicly reviewing a major failure; self-reinvention to date is only technical-route iteration and has not been tested by rebirth after disruption; governance has shifted to CEC as actual controller, not a founder structure with strong ownership alignment, making it less like Baillie Gifford’s preferred image of self-revolution driven by high-conviction owners; the company is young and has not gone through a full semiconductor cycle.

  Overall judgment: neutral, slightly positive. Empyrean Technology has a growth base of willingness to spend for the long term plus active technical-route iteration, and it has one disciplined example in the terminated Xpeedic deal, making it clearly stronger than a typical theme stock. But it has not yet offered the vivid evidence Baillie Gifford truly wants: self-reinvention in bad news and near-death moments. With governance now under a central SOE actual controller and no founder-like strong ownership conviction anchor, its “self-reinvention gene” should be understood as the potential to proactively change technical routes, not the proven fact of having survived disruption and been reborn. For an LTGG investor seeking a great growth stock that can rise fivefold in ten years, this is worth continued observation, but the conviction evidence is not yet sufficient. This dimension should not be forced into a high score.

  Jun 5, 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 from now?3/10

  A

  Bottom line: this is a split case: willingness to sacrifice current profit for the long term is present, but deep alignment between management/founder and outside shareholders’ per-share returns is absent. Under Baillie Gifford’s LTGG lens, it can only be neutral to weak. Empyrean Technology offers strong evidence on the dimension Baillie Gifford likes most, willingness to sacrifice current profits for five to ten years from now: 2025 R&D expense was RMB 859 million, equal to 64.84% of revenue. Multi-year high investment compressed current net margin to about 4.6% (net profit attributable to shareholders only RMB 61 million), and the company did not simply cut R&D when profit fell. In 2025 it also proactively terminated the Xpeedic acquisition because core terms could not be agreed, showing discipline in not forcing a transaction just to tell an acquisition story. But on Baillie Gifford’s most preferred item, a founder/owner-operator with a large stake and interests aligned with outside shareholders, it is the opposite: there is no typical founder with a large shareholding, and its incentive mechanism is far from the private owner-operator paradigm of personally owning the stock and sitting in the same boat as minority shareholders. Together, it is a company with a strong willingness to invest for the long term, but a weak structure that hard-binds that willingness to outside shareholders’ per-share returns.

  1. Long-term view and “sacrificing the present for the long term”: strong (a Baillie Gifford positive). Continuously giving current profit to R&D is the part of this company that looks most like a Baillie Gifford candidate. R&D as a share of revenue rose from 44%–53% during the prospectus period to about 67.77%, about 71%, and 64.84% in 2023–2025, while net margin slipped from about 19.9% in 2023 to about 4.6% in 2025. This pattern of “revenue grows, profit is sacrificed” is directionally exactly the LTGG preference of betting for five to ten years out. One honest addendum: R&D spending in EDA tools is itself life-support spending required to maintain product competitiveness, not purely optional expansionary investment, so the quality of the “sacrifice” deserves a small discount. But it does not change the judgment that the company is not dressing up statements by squeezing short-term profit.

  2. Skin in the game: weak and special in nature (a Baillie Gifford negative). This is the core contradiction of the question. The report says the company has “no controlling shareholder” and no actual controller, with the top ten shareholders mainly state-capital systems, industry funds, and financial investors, and no large founder-manager shareholding structure visible. A latest fact outside the report’s framing, and crucial to this question, must be added: in December 2024 the company announced the actual controller would change from “no actual controller” to China Electronics Corporation. China Electronics and concert parties together held about 33.9% (under the announcement-date basis, 34.06%, consisting of China Electronics Co., Ltd. at 21.22% plus CEC Jintou and others, with nomination of a majority of 11 board seats and consolidation into its accounts). In April–May 2026, CEC Jintou further planned to increase holdings by no less than 1% and no more than 2% of total share capital, and had already injected RMB 250 million of state capital operating budget funds through an entrusted loan in December 2025. This means Baillie Gifford should view it in two layers:

  • “Is there a long-term, well-funded major shareholder willing to keep investing?” Yes. The state-controlled shareholder is willing to increase holdings, inject funds, integrate the industry chain, and coordinate policy. This provides a genuine long-term backing, which is positive compared with short-term speculative capital.
  • “Is management/founders’ personal wealth tightly bound to outside minority shareholders’ per-share returns?” No. Incentives and checks under state-capital governance operate by a different logic from Baillie Gifford’s preferred founder who has personal wealth tied to the company and fully shares per-share gains and losses with outside shareholders. The controlling party’s objectives include industrial strategy and domestic substitution, not purely per-share returns; managers’ compensation incentives are not the same as large personal shareholdings. For an LTGG investor looking for “a tenfold-minded management team in the same boat as me,” this alignment is clearly insufficient.

  3. Capital allocation quality: medium, with clear negatives. Three pieces of evidence pull against each other: ① negative: share-based payment is heavy, with about RMB 145 million accrued in 2025, equal to 237.98% of that year’s net profit attributable to shareholders, creating real dilution for existing shareholders and transferring part of shareholder value to employee/management incentives; ② negative: high cash and high borrowing coexist, with RMB 891 million in cash and another about RMB 1.178 billion in term deposits/certificates, while about RMB 250 million of short-term borrowing was added and there were large transactions with the group finance company. The clarity and efficiency of capital flows are not friendly to outside shareholders; ③ positive: M&A discipline shown by terminating the Xpeedic acquisition, and no buyback at a high valuation despite dividends (RMB 1.50 per 10 shares in 2025, dividends of about RMB 81.82 million, no buyback). Not repurchasing when the stock is expensive is rational. Overall, the direction is not chaotic, but each RMB 1 of retained earnings has not yet proven it can create more than RMB 1 of intrinsic value. That falls short of the high capital return required by the “sacrifice for the long term” narrative.

  4. Balanced honest judgment. The whole question should not be scored high simply because “R&D is 64.84% of revenue.” High R&D only satisfies half of the issue, willingness to sacrifice the present; the other half, founder/management alignment with outside shareholders’ per-share returns, is missing. Nor should the state-capital background and absence of founder ownership produce a complete veto: it brings a controlling shareholder willing to invest long term, increase holdings, and inject capital, which is a real positive for long-term view. Net conclusion: long-term view (strong ✓) + willingness to sacrifice current profit (strong ✓) + skin in the game (weak ✗) + capital allocation (medium, with heavy dilution) offset one another. Under the Baillie Gifford LTGG lens, this “management and founder alignment” question is neutral, slightly weak, broadly consistent with the report’s “management and capital allocation 3/5.” It is a long-term investor worth respecting, but not the owner-operator type in the Baillie Gifford mold where the founder has tied their fate tightly to yours.

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

  A

  Bottom line: EDA as a tool category is absolutely indispensable to chip design, but Empyrean Technology as a company is only moderately indispensable. In most links, customers can switch to the three giants, Synopsys, Cadence, and Siemens EDA, which together account for more than 70% of global EDA revenue and about 75% CR3 in China. The truly scarce parts are only the domestic-only full-flow 3DIC capability, local strengths in analog/flat-panel display, and strategic indispensability under the “independent and controllable” context. Its growth model does not cross ethical red lines and has no obvious negative social externalities, but it is highly tied to domestic-substitution policy/subsidies, the capex cycles of foundries and design houses, and geopolitics, where export controls are both a tailwind and an uncertainty. Overall judgment: some segments are scarce and the category is essential, but the company’s business is relatively substitutable and strongly dependent on policy and cycles. It does not meet Baillie Gifford’s threshold of “if it disappeared tomorrow, the whole world could not live without it.”

  1. Indispensability: separate the “category” from “this company”

  (1) Category level: EDA is indispensable to chip design (high). Without EDA, modern chips cannot be designed. The report positions EDA as one of the three strategic foundations of the integrated-circuit industry and notes that the post-Moore era, 3DIC, Chiplet, cloudification, and design-methodology innovation will continue to increase EDA complexity and usage. This is a classic picks-and-shovels position: as long as chips continue to evolve, EDA demand will not disappear. From this angle, if the EDA category disappeared tomorrow, the entire semiconductor design chain would immediately stall. It would be missed extremely badly.

  (2) Company level: Empyrean Technology can be replaced, and its scarcity is materially lower than that of an industrial monopoly leader (medium). The key is that “EDA is indispensable” does not mean “Empyrean Technology is indispensable.” Global EDA is dominated by the three giants: Synopsys about 31%, Cadence about 30%, Siemens about 13%, with the three together exceeding 70% and even higher by revenue. The report also clearly admits that the domestic market is still mainly led by the international three giants and that local suppliers’ overall share remains small. Empyrean Technology’s mainland China revenue accounted for 90.09% in 2025, and full-year revenue was only RMB 1.325 billion, a different scale from the three giants. This means that for most customers in most design links, Empyrean Technology is one replaceable option, not the only node without which work cannot continue. If it disappeared tomorrow, most customers would feel switching costs and local capability gaps, not the shutdown of an entire production line. This is fundamentally different from a true industrial monopoly leader whose supply cutoff means production stops.

  (3) The three truly scarce points. Empyrean Technology’s indispensability is not uniformly high; it concentrates in three places: ① full-flow 3DIC, where the report says it is the only domestic EDA provider for the full flow of 3DIC design verification, the closest it comes to being “irreplaceable in the short term”; ② local strengths such as full-flow analog + RF + flat-panel display, where it competes more directly against the three giants; ③ strategic indispensability for “independent and controllable” supply. In the context of China-U.S. technology rivalry and the possibility that chip design software becomes a chokepoint, the very existence of local EDA has strategic value. This was validated in 2025: when the U.S. BIS sent letters at the end of May restricting EDA exports to China by the three giants, then revoked the controls in early July, the narrative that “domestic substitution is indispensable” was repeatedly amplified. But this also exposes that its scarcity is more strategic scarcity granted by policy and geopolitics, not a commercial necessity formed by customers voting with their feet.

  In one sentence: category indispensability is high, company indispensability is medium. Empyrean Technology is the most reliable one within domestic independent controllability, but not the one global customers cannot live without.

  2. Growth sustainability + society/regulation: the bottleneck is policy and cycle dependence, not negative social externalities

  (1) The growth model does not cross ethical red lines (positive). Empyrean Technology sells industrial software licenses and technical services. Independent controllability itself is a positive mission: it serves local chip-industry supply-chain security and does not have the addiction, data abuse, regulatory arbitrage, or social negative externalities often seen in platform companies. On Baillie Gifford’s red line of whether growth depends on harming society or regulators, it is clean. Growth comes from legitimate R&D and product replacement, not from harming users or exploiting regulatory loopholes. This is better than many high-growth companies Baillie Gifford would exclude.

  (2) But the real sustainability bottleneck is triple dependence on policy, cycles, and geopolitics. This is where it fails to reach Baillie Gifford’s bar:

  • Policy/subsidy dependence: the report repeatedly notes that the company benefits from domestic substitution and support from national major projects. Domestic EDA penetration has risen from “supporting 30%-40% of domestic demand” in 2019 to 70%-80% according to Liu Weiping, with about 20% domestic market share (note: this is the company’s product coverage of different demand scenarios, not the share of domestic tools in total market sales, which is still around single digits to 20%). But this climb is largely policy-driven rather than purely a market choice. If subsidies fade or localization priority declines, growth momentum would weaken visibly.
  • Dependence on customer capex cycles: revenue has a pronounced Q4 concentration and budget-driven pattern (2025 fourth-quarter revenue was RMB 520 million, far above the first three quarters). It is essentially tied to the budget and business cycles of foundries and design companies. The report’s 2026 Q1 data, revenue still up 9.65% YoY but a RMB 73 million loss and operating cash flow of RMB -172 million, shows that even in a good track, single-quarter cash generation is unstable. Growth quality is strongly constrained by customer capex timing.
  • Geopolitics as a double-edged sword: export controls are both a tailwind (when supply is cut, the domestic-substitution narrative explodes and customers are forced to adopt domestic tools) and an uncertainty (the 2025 sequence of “May restriction → July lifting” shows policy can reverse overnight, and the urgency of local substitution falls once the three giants resume supply). Building growth on “competitors being sanctioned” is inherently fragile, not the endogenous certainty Baillie Gifford prefers, where customers worldwide choose your product because it is better.

  3. Overall judgment against Baillie Gifford’s “how much would the world miss it tomorrow” bar

  Combining the two lines: the EDA category is highly indispensable, but that is an industry attribute rather than company-specific; Empyrean Technology as a company is moderately indispensable, with most links replaceable by the three giants and only 3DIC/analog/strategic independent controllability locally scarce; the ethical/social side of the growth model is clean and does not cross red lines; but the sustainability and endogeneity of growth are weak because it depends heavily on policy, cycles, and geopolitics, with external drivers stronger than internal cash generation.

  Final judgment: Empyrean Technology has real scarcity in three areas: full-flow 3DIC, analog/display niches, and the strategic value of domestic independent controllability. Its growth model is legitimate and has no negative social externalities. But for most customers in most links, the business is relatively substitutable, and it is strongly dependent on domestic-substitution policy/subsidies, customer capex cycles, and geopolitics. Its indispensability is more “strategic and policy-granted” than a commercial necessity that customers cannot live without. If it disappeared tomorrow, global chip design would not stop; the three giants would fill the gap. The main impact would be on China’s progress toward independent controllability, a national-strategy loss rather than the Baillie Gifford standard that “the whole world would deeply miss it.” Therefore, on Question 7, it is “category essential, company medium, growth sustainability weak”, and it does not meet the Baillie Gifford LTGG standard of being something the world cannot live without if it vanished tomorrow.

  Jun 5, 2026
• QWhat are the unit economics of this business (gross margin, incremental returns)? Do they improve or deteriorate as scale grows? Where does the money it earns go?3/10

  A

  Unit economics: this is one of Empyrean Technology’s weakest dimensions. It has the “good foundation” of a top software business, with gross margin of about 89.25%, but this high gross margin has not yet translated into net profit or stable free cash flow. More importantly, over the past few years its scale has kept growing while its unit economics have kept worsening, the opposite trajectory from the self-funding compounding machine Baillie Gifford prefers. The conclusion: the unit-economics potential is good, but it is far from proven.

  ① Unit economics themselves: the good foundation is consumed by R&D. EDA is a classic software-license business with very low marginal cost, and Empyrean Technology’s 2025 gross margin reached about 89.25%. This is exactly the high-gross-margin base Baillie Gifford values because it can support compounding. But above gross profit sits a giant R&D machine: 2025 R&D expense was RMB 859 million, 64.84% of revenue, almost consuming all gross profit. Final net margin was only about 4.6%, and net profit attributable to shareholders only RMB 61 million. A business with 89% gross margin but only 4.6% net margin, and then a direct 2026Q1 loss (net profit attributable to shareholders of RMB -73 million, with quarterly gross margin falling back to 78.8%), means that all current unit-economics quality has been offset by “investment,” and no meaningful money has yet been left for shareholders.

  ② Incremental returns / scale-effect direction: the core test, and the answer is “worse.” The mark of a Baillie Gifford good business is that as scale grows, unit economics improve. Empyrean Technology is the opposite. Revenue grew from RMB 257 million in 2019 to RMB 1.325 billion in 2025, a six-year compound rate of about 31%, but net margin fell one way from about 19.9% in 2023 to about 9.0% in 2024 and about 4.6% in 2025; weighted ROE over the last three periods collapsed from 4.26% to 2.24% to 1.19%, down to about 1.19%. The most striking year was 2024: revenue rose +20.98% YoY, but net profit attributable to shareholders fell -45.46%, and recurring net profit moved directly from RMB 64 million in 2023 to RMB -57 million. In other words, each additional RMB 1 of revenue has not produced higher marginal profit; it has diluted overall returns. Incremental return, or incremental ROIC, is deteriorating rather than improving. One can charitably explain this as an expansion phase of R&D and product-gap filling, with operating leverage not yet released, but under Baillie Gifford’s lens it must be recorded honestly: to date, scale effects are negative, and operating leverage is a promise, not a fact.

  ③ Free cash flow quality: highly volatile and containing water; “larger revenue → steadily larger FCF” has not been proven. Operating cash flow was RMB -52 million in 2024 and rebounded to RMB +561 million in 2025. Looking only at 2025, it appears “fat,” but about RMB 262 million came from working-capital release through receivables collection and higher payables, not pure endogenous cash generation. Corresponding free cash flow was about RMB -131 million in 2024 and about RMB +375 million in 2025, swinging sharply from negative to positive. The report therefore estimates more conservative owner earnings at RMB 120 million / RMB 200 million / RMB 300 million, far below the surface CFO of RMB 561 million. In 2026Q1, operating cash flow returned to RMB -172 million. Baillie Gifford wants free cash flow that compounds upward and is predictable; Empyrean Technology’s FCF is neither stable nor free from one-off working-capital water. This standard is clearly not met.

  ④ Where the money earned goes: high-intensity reinvestment, with returns not yet realized. The company has net cash and a light-leverage balance sheet, with an asset-liability ratio of about 17%. The money mainly goes into R&D: talent and tool development. Of the RMB 859 million of 2025 R&D expense, employee compensation was RMB 576 million and share-based payment RMB 99 million. From Baillie Gifford’s reinvestment logic, continuously reinvesting profit into a high-potential track is directionally correct (domestic EDA substitution is a long-runway, deep-snow track). The problem is that the reinvestment return has not yet appeared in net profit or FCF. Investment increases each year, while unit economics thin each year. Reinvestment is a compounding engine only when it generates high incremental returns; otherwise it consumes value. Current evidence leans toward the latter. In addition, share-based payment in 2025 reached 237.98% of net profit attributable to shareholders, a real economic cost to shareholders that further depresses the unit economics left for owners.

  ⑤ Honest point: high gross margin is a good foundation, but the current combination is “high gross margin + thin net profit + unstable FCF + declining margins during the investment phase.” Taken together, Empyrean Technology has the embryo of the kind of business Baillie Gifford likes: software-like high gross margin, low capital intensity, and a long track. But it has not yet shown the compounding machine Baillie Gifford actually underwrites: larger scale, better unit economics, and ever-larger FCF. Today it looks more like “unit-economics potential is good but unproven.” The quality of a high-gross-margin foundation is temporarily suppressed by extremely high R&D intensity, negative scale effects, and volatile FCF. To turn this dimension from “weak” to “strong,” clear inflection signals are needed: net margin and operating margin stopping their decline and recovering, ROE climbing again, operating cash flow breaking dependence on working-capital release, and incremental revenue starting to bring higher rather than lower marginal profit. Before such evidence appears, Question 8 (unit economics) should honestly be rated as one of Empyrean Technology’s weakest dimensions.

  Jun 5, 2026
• QFor it to rise fivefold in ten years, what conditions must all hold at the same time? Are these conditions realistic? What expectations does today’s share price imply?2/10

  A

  Bottom line: starting from today’s roughly RMB 105.8 share price and RMB 57.5 billion market value, a fivefold rise in ten years, to about RMB 287.5 billion, requires five things to almost all happen together: revenue must multiply several times, net margin must recover sharply from 4.6%, valuation multiples must not collapse from already extremely overdrawn levels, domestic substitution must fully play out, and competition must not worsen. This is a chain product of low-probability events. Today’s share price, with PS about 42.8 times, static PE about 900 times, above the report’s optimistic intrinsic value of RMB 75, and reverse-implied annual returns of -17%/-6%/+5% across the three scenarios, has already prepaid “high-speed growth + sharp profitability recovery + consolidated leadership” as if they were high-probability events. Two questions must be separated: Empyrean Technology may well grow into an excellent local EDA leader at the company level; whether it can rise fivefold from today’s price over ten years is a different matter. By honest LTGG math, that is a low-probability event.

  1. First make the math concrete: a fivefold rise in ten years = about RMB 287.5 billion market value

  The core of Baillie Gifford LTGG is never simply “is this a good company,” but “from today’s price, is there real fivefold upside over ten years.” The simplest arithmetic breaks it down this way:

  Current market value is about RMB 57.5 billion (share price about RMB 105.8, closing price RMB 105.75 on 2026-06-01). Fivefold = about RMB 287.5 billion. Terminal market value can be split into two multipliers: (terminal revenue or profit) × (terminal valuation multiple). To reach RMB 287.5 billion, the following “condition chain” must basically hold at the same time. Assessing each condition:

  ① Revenue must multiply several times again. 2025 revenue was RMB 1.325 billion. Even assuming the company can sustain the roughly 31% compound growth of the past six years over the next ten years, revenue would be around RMB 20 billion in ten years. But the report has already recorded 2026Q1 growth abruptly slowing to +9.65% and a RMB 73 million loss. The global EDA pie itself is limited (the three giants, Synopsys/Cadence/Siemens EDA, together have only tens of billions of dollars of annual revenue), and China’s local EDA pie is smaller. To grow revenue by an order of magnitude in ten years would require Empyrean Technology not only to fully capture the domestic-substitution dividend, but also to break through overseas, where overseas revenue is currently only 9.91%. That itself means challenging the three giants head-on, which is extremely difficult. Realism: it requires an optimistic scenario and becomes harder to sustain later in the period.

  ② Net margin must recover sharply from 4.6%. This is the hardest gate. 2025 net profit attributable to shareholders was only RMB 61 million, net margin about 4.6%, and it has been falling year by year from 19.9% in 2023 to 9.0% in 2024 and 4.6% in 2025. 2026Q1 directly turned loss-making. The root of margin collapse is R&D at 64.84% of revenue (2025 R&D expense of RMB 859 million) plus share-based payment (RMB 145 million in 2025, equal to 237.98% of net profit attributable to shareholders). To support a fivefold market value, revenue alone is not enough; margins must structurally return to double digits or higher. That requires R&D intensity to be materially spread over higher revenue and commercialization efficiency to improve sharply. But EDA is a talent-barrier, continuously high-R&D business, and R&D cannot be cut deeply. The timing and magnitude of margin recovery are highly uncertain. Realism: the direction is possible, but the size and pace are the biggest question marks.

  ③ Valuation multiples must not collapse from already extremely overdrawn levels. This is the most fatal link that many growth narratives ignore. Today’s PS of about 42.8 times is an extreme value for the global EDA sector. Compared with more mature and more profitable global leaders, Synopsys’ current TTM PE is about 78 times and Cadence’s is about 70–82 times (the report had recorded about 114/95 at its cutoff; current levels had fallen back to the 70–80 range), still far below Empyrean Technology’s roughly 900 times static PE. Even A-share peers Primarius (PS about 30) and Semitronix (PS about 26) are lower than Empyrean Technology. If a ten-year fivefold return is to be driven by fundamentals, valuation multiples must at least not shrink materially. But starting from a historical high PS of 42.8 times, the more realistic long-term path is mean reversion toward peers. Even returning to local peer PS of 26–32 times would imply roughly 25%–40% multiple compression, directly eating a large portion of fundamental growth. Realism: requiring multiples not to collapse is itself a bet on a permanent high premium, contrary to mean reversion.

  ④ Domestic substitution must realize, and ⑤ competition must not worsen. These two are preconditions for ② and ③. Domestic EDA penetration must rise sharply over ten years, and Empyrean Technology must keep its local No. 1 position while upgrading toward full flows (the report notes that China’s market is still mainly led by the international three giants and that local share remains small overall). At the same time, local followers such as Primarius and Semitronix must not erode its share or pricing power. If any one of these falls short, ① and ② cannot be discussed. Realism: industry and policy support the direction, but “full realization + no worse competition” is an ideal script, not a base case.

  Treating ①–⑤ as five independent probabilities, each may not look absurd alone, but requiring them all to land in the ideal range within ten years produces a low joint probability. This is the key to Baillie Gifford’s honest math: fivefold upside is not judged by the most optimistic single assumption, but by whether the conditions can hold together.

  2. What today’s share price implies

  The report’s reverse calculation already puts the answer on the table and is consistent with the math above:

  • Absolute valuation level: PS about 42.8 times, PB about 11.1 times, static PE about 900 times (the 2026Q1 loss distorts/turns negative the TTM basis). This is not paying for existing profit or existing free cash flow. 2025 net profit attributable to shareholders was only RMB 61 million, and owner earnings under the report’s conservative adjustment were roughly RMB 100–200 million, implying the market value is already about 290–580 times.
  • Position relative to intrinsic value: the report’s three intrinsic-value bands are conservative RMB 10–25, reasonable RMB 30–55, and optimistic RMB 55–75. The current RMB 105.8 stands above the upper end of the optimistic band, RMB 75, with about 92%–253% premium to the reasonable range and still about 41% above the optimistic upper end. In other words, today’s price has already overdrawn even the report’s most ideal scenario.
  • Reverse-implied annual return: starting from the current price and using the report’s ten-year scenarios, the conservative case is about -17%, the neutral case about -6%, and even the optimistic case only about +5%. All three scenarios run from negative to +5%.

  These three sets of numbers say one thing together: today’s price has already prepaid “high-speed growth + sharp profitability recovery + consolidated leadership” as a high-probability event, as if it had already happened. The market is giving a strategic premium, not a cash-flow discount. When even the optimistic scenario for a stock implies only a single-digit positive return, the room for “another fivefold” has already been squeezed out mathematically by the price itself, because a fivefold return would need to be added on top of a baseline that is already negative to slightly positive.

  3. Honest point: separate “can it become a good company” from “can it rise fivefold from today’s price over ten years”

  Put plainly, the answers to these two questions are different, and mixing them up is the most common cognitive trap in high-valuation growth stocks.

  • “Can Empyrean Technology become a good company?” Probably yes. It is the local EDA leader, has a complete product line, technical and talent barriers, net cash, and a long-term demand track (post-Moore, Chiplet, advanced packaging, domestic substitution). From a business perspective, Baillie Gifford would be willing to put it on a watchlist and keep tracking it.

  • “From today’s roughly RMB 105.8 share price and RMB 57.5 billion market value, can it rise fivefold in ten years?” Low probability. The reason is not that the company is poor, but that the starting point is already extremely overdrawn. A ten-year fivefold return requires the stock to rise another fivefold from a position already above optimistic intrinsic value, with all reverse-implied returns negative, PS 42.8 and PE 900. That requires valuation and fundamentals to simultaneously realize an extreme scenario: revenue multiplying several times × sharp margin recovery × multiples not collapsing × full domestic-substitution realization × no worsening competition. This is a product of low-probability events. Even if the fundamentals successfully develop into an excellent leader, a “fivefold return from this price” remains unlikely. The likely outcome could be that the company continues to grow while the share price stagnates for a long time or even falls, which the report also identifies as the largest permanent-capital-loss scenario: buying a good company too expensively, with a 40%–60% drawdown not exaggerated.

  Baillie Gifford judgment (Question 9): the ten-year fivefold question fails on price, not on the company. LTGG looks for growth stocks that start from a reasonable entry point and have real asymmetric upside. At today’s price, Empyrean Technology’s asymmetry is reversed: the downside, from valuation mean reversion plus insufficient margin recovery, is thicker than the upside of a fivefold rise. If the price returned to the report’s ideal buy range of RMB 20–40, the ten-year fivefold discussion would re-enter a plausible probability range. At RMB 105.8, it is a low-probability chain product and does not constitute an LTGG buy thesis.

  Jun 5, 2026
• QWhy has the market not realized all this? Does it not understand, look down on it, or fail to look far enough? What will become the “narrative inflection point”?3/10

  A

  Bottom line: this is the opposite of the default premise behind Baillie Gifford’s question. The market is not overlooking Empyrean Technology or looking down on it so that it is cheap; it understands it too clearly and is paying too generously. When domestic EDA is layered with the triple narrative of “independent controllability + semiconductor + AI,” the market has already pushed Empyrean Technology to an extreme valuation of PS about 42.8 times and static PE about 900 times (current price about RMB 105.8, total market value about RMB 57.5 billion). More importantly, the 12-month average sell-side target price is about RMB 122.77, with 5 institutions in the past 90 days showing 3 Buy and 2 Hold, and some targets reach RMB 125, implying about 39% upside, all above the current price. This is not a market-neglected discount; it is a market-prepaid premium. Therefore the honest landing point for this question is: there is no positive perception gap here. The narrative premium is already sufficient or even excessive. The true “narrative inflection point” is more likely to be triggered from the side that disproves the high valuation, not from the side where the market finally discovers value. This is consistent with the report’s “Watch, no current margin of safety” judgment.

  1. Why among “does not understand / looks down / cannot look far enough,” only a two-sided “cannot look far enough” holds

  • It is not “does not understand.” EDA, as one of the three strategic pillars of the chip industry and a key industrial software link in domestic substitution, has already been repeatedly explained by the market. The report also gives “business understandability” 4/5. If the market did not understand it, it would not assign a PS of 42.8 times, which is pricing that shows strong understanding and strong belief.
  • It is not “looks down.” Quite the opposite: the market looks up to it. Brokers maintain “Buy,” consensus expects 2026 revenue growth to accelerate to 42% and net-profit growth to 59.23%, and sees a clear fundamental inflection point. This is being respected to the point of premium, not being dismissed to the point of discount.
  • What really holds is a two-sided “cannot look far enough.” Bulls linearly extrapolate “domestic substitution will inevitably realize, R&D investment will eventually turn into profit, gross and net margins will inevitably recover,” and discount into today’s market value a harvest that might only appear ten years from now. Potential shorts or value investors, meanwhile, focus only on current thin profits and losses: 2026Q1 revenue was up +9.65% YoY, but net profit attributable to shareholders was RMB -73 million, net margin -28.44%, and gross margin fell 12.61 percentage points YoY to 78.80%; 2025 net profit attributable to shareholders was only RMB 61 million and R&D was 64.84% of revenue. Neither side has looked far and clearly enough at the fundamental return question: ten years from now, is this company a global price-setting oligopolist, or a strategically important local follower with ordinary economic returns? The current price implies that the bullish version of “not looking far enough” has prevailed.

  2. Comparables also show the whole sector is generously priced, not that Empyrean Technology alone is undervalued

  The peer comparison in the report also points to premium, not discount: Primarius (688206) PE about 750 / PS about 30, Semitronix (301095) PE about 222–252 / PS about 26, while global leaders Synopsys (current TTM PE about 78 times) and Cadence (about 70–82 times) still trade far below Empyrean Technology. Empyrean Technology is not even cheap inside the already high-premium domestic EDA sector. This means there is no perception gap where the market has temporarily failed to give it its deserved valuation. Instead, there is double overdrawing from sector narrative premium plus a relatively high individual-stock position.

  3. Narrative inflection points (catalysts) in both directions, with weight clearly tilted negative

  Negative catalysts, which are more likely and more damaging once triggered, matching the report’s “largest risk is excessive valuation”:

  • Margins fail to recover / losses continue. If the 2026Q1 pattern of “revenue growth without profit growth” continues for one or two more quarters and the consensus “2026 net profit +59% inflection” is disproved, the foundation of the premium, inevitable profit recovery, will be knocked out.
  • Valuation reverts toward intrinsic value. The report’s reasonable intrinsic value is RMB 30–55, and the optimistic upper end is only RMB 75. The current RMB 105.8 is about 41% above the optimistic upper end. If high valuations in the sector contract overall, a 40%–60% drawdown would not be exaggerated.
  • Domestic-substitution progress / customer capex disappoints, competition from the three giants or geopolitics repeats. Synopsys, Cadence, and Siemens EDA still dominate the domestic market, and local share remains small. Any evidence that replacement is slowing would shake the “inevitable realization” narrative.
  • Unlocking / selling down. Shareholders are mainly China Electronics system entities, industry funds, and financial investors, without a typical founder holding a large locked-up stake. Selling pace can directly hit the premium.

  Positive catalysts, which support the bull narrative but are mostly prepaid at the current price, compressing marginal surprise:

  • Faster domestic-substitution penetration and new benchmark customers.
  • Full-flow digital / AI+EDA (PyAether) / 3DIC converts into volume, with the report calling it the only domestic EDA provider for the full flow of 3DIC design verification.
  • A real profit inflection appears, with R&D investment converting into high gross profit and lower expense ratios.
  • Policy catalysts, such as major projects and stronger independent-controllability support.

  4. Honest landing point

  Translated into Baillie Gifford language: this is not a hidden compounding machine the market has yet to discover; it is a stock where the market has already fully priced, and even prepaid, the optimistic domestic-substitution narrative. The current price has paid too much upfront for the story of “domestic EDA will inevitably succeed + profitability will inevitably recover.” Therefore, the meaning of a “narrative inflection point” is the opposite of a typical Baillie Gifford growth stock. It is not “the inflection arrives, the market wakes up, and the stock is re-rated upward.” It is more likely “profitability or domestic-substitution progress fails to meet already-priced high expectations, the premium is disproved, and valuation reverts toward intrinsic value.” In the LTGG framework, for a company whose business deserves respect but whose price has already paid for the next ten years of good news in one go, the rational stance is to put it on a watchlist, wait for profit and owner earnings to truly materialize, or wait for price to return to a level that provides margin of safety. This is fully consistent with the report’s “Watch, no current margin of safety, ideal buy range RMB 20–40.”

  Jun 5, 2026