| Meta Info | Content |
|---|---|
| Rating | Neutral |
| Target Price | HKD 2.0–2.45 (base-case fair range, 12 months) |
| Current Price | HKD 2.65 (2026-11-20 close, frozen anchor) |
| Probability-weighted fair value | approx. HKD 2.39 (expected return approx. −10%) |
| Margin of safety | approx. −24.5% ((base-case fair lower bound 2.0 − current price 2.65) / current price; no margin of safety) |
| Market cap / total shares | approx. HKD 4.585 billion / approx. 1.730 billion shares |
| Time horizon | 12 months |
| Key catalysts | 2026-12 KoolPAD Conference; 2027-03 annual report and HSCI review; 2027-10-16 founder lock-up expiry |
Data integrity statement: During this documentation cycle, the company's locally stored financial report files could not be directly read due to a system environment failure (PDF parsing suspended). Financial data were cross-verified via the following channels: ① the mirrored full text of the 2026 interim results announcement (balance sheet line items verified verbatim for valuation); ② structured statements from Eastmoney/financials (compared against media-reported figures for consistency); ③ media reposts of the prospectus and results announcements. All figures not verified against original documents are flagged in the text. This is the main downgrade factor behind the report's evidence grade (Grade B).
Manycore Tech is the #1 cloud-based design SaaS company in China's spatial design software market by share (23.2%, Frost & Sullivan, 2024 revenue basis). After its April 2026 listing, it went through a complete bubble deflation of "HKD 48.5 peak → lock-up expiry → HKD 2.65" (a drawdown of approx. 94.5%). The current price shows no systemic undervaluation: the current price trades at an approx. 8% premium to the upper bound of this report's base-case fair range and an approx. 11% premium to the probability-weighted fair value; the valuation is assessed as mildly overvalued. In terms of value structure, under a zero-growth conservative framework, approx. 53% of the current price (HKD 1.41 per share) represents option pricing on the AI/spatial intelligence second growth curve — yet that business contributed only 7.7% of revenue in 2026H1, its new-order momentum is weaker than headline growth, and historical clients exhibit one-off large-order patterns. Under the growth assumptions of J.P. Morgan, the sole external deep-cover analyst, the option premium narrows to approx. 7% — whether the current price is "expensive" or not hinges almost entirely on one's judgment of the operating value of the core business, which is the single most critical point of divergence for this stock. On fundamentals, the SaaS core business has seen three consecutive quarters of decelerating growth (+13.7% → +8.6% → +1.5%) and net dollar retention has fallen below 100%; the transmission of the property-chain contraction is materializing. Net cash of approx. RMB 1.55 billion (39.8% of market cap) and an 82–83% gross margin provide a buffer, but Hong Kong precedents (Mingyuan Cloud's market cap persistently below net cash) suggest the market may discount that buffer. Strategically, we do not enter at current levels (wait for price): the entry precondition is a pullback to the HKD 2.0–2.45 fair range without the core-business or AI theses being falsified. Key signals to monitor: AI revenue share and repeat-purchase verification, the pace of absorption of unlocked shares, and the revenue direction in the FY2026 annual report.
Opportunity characterization: A recently listed stock down approx. 94.5% from its peak in 7 months post-listing (still −65.2% vs. the IPO price of HKD 7.62). The 18-month payoff is determined by two variables — whether the AI/spatial intelligence option, worth approx. 53% of the current price, can deliver, and whether the 64.5% of unlocked shares without lock-ups can be absorbed. Evidence grade B (original filings unreadable; reliance on mirrored announcements and multi-source cross-checks).
Changes most likely to reprice the stock over the next 6–12 months:
Ranking of primary price drivers: ① slope of AI option realization (more than half the current price is priced on it) → ② core-business retention (NDR/RPO) → ③ digestion of share supply. Liquidity is the amplifier: post-unlock single-day swings of ±10%–25% have become the norm.
Expectation gap: Negative. Nominal sell-side target prices (HKD 15.1–30, moomoo consensus approx. HKD 20) were all set around the post-unlock crash and never updated — diverging from the market price by 5.7–11.3x, they no longer constitute valid market expectations. Working backward from price, the current level implies roughly 40% probability of the bull case, while verifiable reality is core growth of +1.5% and NDR of 94.4%. Action gate: wait_price — wait for price to enter the fair range with R2/R5 falsification conditions not triggered.
Under the zero-growth conservative framework (SBC fully expensed, normalized FCFF ≈ HKD 0.02/share), the current HKD 2.65 = net cash 1.05 (40%) + zero-growth operating value approx. 0.18 (7%) + AI option approx. 1.41 (53%). The current EV implies a five-year revenue CAGR of approx. 3%–13% (AI revenue would need to scale 6–11x from an annualized approx. HKD 73 million), while core-business reality shows 2026H1 revenue +1.5% and RPO over one year at −43.8% YoY — nearly all implied growth must be supplied by the AI business whose slope remains unproven. But this conclusion is highly sensitive to the operating-value assumption: under J.P. Morgan's assumptions (initiated 2026-05-28 — a neutral rating from an IPO joint sponsor) of a 2025–2028 revenue CAGR of 11% and a 2028 adjusted net margin of 17%, operating value rises to approx. HKD 1.42/share, the option premium narrows to approx. 7%, and the current price carries only a 7.3% premium to EPV. This report's base-case fair value is set at HKD 2.0–2.45 (EV/S 1.8–2.4x × FY2027 revenue + net cash, between Mingyuan Cloud's real-world floor and the quality ceiling of global leaders); the current price sits approx. 8.2% above the range's upper bound and approx. 11% above the probability-weighted fair value — assessed as mildly overvalued.
Reverse case: If FY2026 revenue growth returns above +5% and NDR recovers past 100%, the JPM-style growth assumptions are validated and the valuation framework moves up across the board; a breakout above HKD 4.6 with AI share exceeding 15% would invalidate the "expensive" judgment.
Revenue growth has decelerated three times in a row: FY2024 +13.7% → FY2025 +8.6% → 2026H1 +1.5%. Net dollar retention (NDR) on an annual basis: 106.1% (2023) → 98.6% (2025), and on the 2026H1 interim basis further down to 94.4% (100.3% a year earlier; a 5.9pp decline in one year) — enterprise clients 96.4%, individual clients 83.3%, largest client cohort 101.8%. The drop below 100% is mainly dragged by the low-ticket, high-churn individual/small-B structure, but the overall installed base is contracting. The share of new-client contracts with terms over one year collapsed from 48.6% in 2023 to 9.8% in 2025 (single media outlet citing the prospectus; basis not verified against originals). Downstream transmission is materializing: listed custom-furniture companies all posted revenue declines in the first three quarters of 2026 (Oppein −29.6%, Sofia −24.1%, cumulative basis), while KoolHome subscription revenue accounts for 96.9%–97.1% of total revenue, directly tied to home-furnishing industry budgets. This report forecasts FY2026 full-year revenue of RMB 825–845 million (+0.6% to +3.0%, implying 2026H2 YoY of −0.2% to +4.5%; H2 has possible support from a low base, continued national subsidies, and hardware launches).
Reverse case: Any quarter with revenue growth back above +5% and NDR recovering above 100% would shift the "deterioration" judgment to "stabilization"; downstream declines narrowing to single digits would also be a repair signal.
As of 2026-06-30, net cash was approx. RMB 1.550 billion (cash and equivalents 1.1149 billion + time deposits 461.5 million + restricted cash 1.4 million + FVTPL financial assets 70.2 million − short-term borrowings 90.0 million − lease liabilities 7.9 million), approx. HKD 1.825 billion, or 39.8% of market cap; IPO net proceeds of HKD 1.335 billion (official figure, including over-allotment) — only HKD 156 million deployed through H1. Gross margin 82.2% (FY2025) → approx. 83% (2026H1); adjusted net profit consecutively positive, reaching RMB 55.42 million in 2026H1 (+210.9%). But three constraints soften the floor: ① Hong Kong precedent — Mingyuan Cloud, also a property-chain SaaS, holds nearly RMB 4.0 billion in cash against a market cap of approx. HKD 980 million; its net cash has been discounted by the market to approx. 0.3x for nearly three years, showing Hong Kong can refuse full credit for such companies' cash; ② approx. 90% of IPO net proceeds are earmarked for growth uses (international expansion 30%, sales & marketing 20%, core technology infrastructure 20%, product enhancement 20%; only 10% for working capital) — the cash is ammunition slated for deployment, not a static floor; ③ the improvement in adjusted net profit relies on SBC add-backs and cost compression (R&D RMB 390 → 290 million; sales RMB 356 → 274 million), while operating cash flow was negative for 2023–2025 — media figures diverge by 10x between −RMB 19 million and −RMB 192 million (2025); with originals unreadable, no adjudication is possible and both figures are retained as reported.
Reverse case: Year-end cash and cash-like assets below RMB 1.0 billion, a large equity financing launch, or consecutive deterioration in operating cash flow would falsify the "buffer" narrative; conversely, operating cash flow turning positive would materially reinforce the floor.
Revenue from new AI applications and products reached RMB 31 million in 2026H1 (+177%, 7.7% of total revenue); LuxReal monthly revenue grew +470% MoM (tiny base); SpatialVerse (synthetic training data for embodied AI) booked RMB 6.8 million in new orders in 2026H1, with clients including Zhiyuan Robotics, Galbot, Noetix, and Hesai. But slope evidence is weaker than the highlighted figures: ① SpatialVerse's H1 new orders of RMB 6.8 million < the period's recognized revenue of RMB 11.81 million (professional services revenue +29.3%, mainly attributed to it); annualized new orders ≈ 0.58x recognized revenue — "orders exceed all of last year" mismatches order flow against revenue stock; ② the prospectus's top-five client list shows that in 2020, Company G (speculated to be Meta) placed a RMB 31.346 million order with no subsequent renewal; the business historically shows a "one-off, intermittent" pattern with poor repeat purchasing; ③ daily average token consumption of approx. 2.4 billion (2026-07) covers only approx. 19 days (July 13–31) and is a company self-reported figure, with no token unit-price-to-revenue mapping; ④ two co-branded capture hardware products with Hesai (11-09) and CHCNAV (11-16) are launched but not yet on sale, with zero revenue. The verification points for option realization lie in the two reporting periods of 2027: whether AI revenue share can reach 15%, whether top clients show repeat purchases/multi-year framework procurement, and whether hardware and token billing show up as revenue line items.
Reverse case: AI share below 9% for two consecutive periods or growth below 50% would falsify the option value; conversely, a single tens-of-millions repeat-purchase contract or multi-client framework procurement in 2027 would turn the option from narrative into asset.
On 2026-10-19, 1.599 billion shares (92.64% of total, 35 shareholders) were unlocked; the stock fell over 35% cumulatively in the first two trading days (−25.37% on 10-20 alone). The four founders and CFO voluntarily extended the lock-up on 483 million shares (27.99%) to 2027-10-16; cornerstone investors (Linear Capital, Yunqi Capital, Hesai, Goldenintek, Wusong, Huaying Construction, etc.) stated they would continue holding. Excluding the founders' extension, approx. 1.116 billion shares (approx. 64.5% of share capital) carry no lock-up; Pre-IPO cost is extremely low (media back-calculated IDG's cost at approx. HKD 0.14/share from "53x gain vs. IPO price," leaving approx. 18x paper gains at the current price; that cost is a single-media estimate). But absorption evidence is equally notable: within one month of unlock, zero disposal filings by ≥5% shareholders on HKEXnews; southbound funds bought on 16 of the last 20 trading days, with cumulative net buying of approx. 176 million shares (approx. 10.2% of total shares, equivalent to 15.8% of the unlocked free float), raising holdings to 192 million shares (11.09%) — turnover from old holders to southbound money is happening in an orderly manner, not disorderly dumping. Hard node: 2027-10-16 founder lock-up expiry; expectations of a second supply wave may begin trading into the stock from mid-2027.
Reverse case: Any ≥5% shareholder/cornerstone disposal filing, or two consecutive weeks of net southbound outflows, would restart disposal-pressure pricing; a break below HKD 1.5 would confirm the bear case (supply + fundamentals double hit).
| Metric (RMB) | FY2023 | FY2024 | FY2025 | 2026H1 (Latest Period) |
|---|---|---|---|---|
| Revenue (RMB 100M) | 6.64 | 7.55 | 8.20 | 4.05 |
| Revenue YoY | — | +13.7% | +8.6% | +1.5% |
| Net profit attributable to shareholders (RMB 100M) | −6.46 | −5.13 | −4.28 | −1.46 |
| Adjusted net profit (RMB 100M) | — (negative) | — (negative) | 0.571 (first positive) | 0.554 (+210.9%) |
| Gross margin | — | — | 82.2% | ~83% |
| Operating cash flow (RMB 100M)* | −0.63 | −1.08 | −0.19* | Not obtained (source unreadable) |
| Free cash flow (RMB 100M, OCF−CapEx) | ≈−0.94 | ≈−1.39 | ≈−0.54 | — |
| Cash and cash equivalents (period-end, RMB 100M) | — | — | 3.57 | 16.48 (cash 11.149 + time deposits 4.615 + restricted 0.014 + FVTPL 0.702) |
| Net cash (RMB 100M) | — | — | — | 15.50 (net of short-term borrowings 0.900 and leases 0.079) |
| Interest-bearing debt (RMB 100M) | — | — | — | 0.98 (2026-06-30) |
| Net assets (RMB 100M) | — | — | — (preferred shares carried as liability) | 10.90 (2026-06-30, preferred shares converted to equity post-IPO) |
Data sources: interim results announcement mirror (item-by-item verification of the 2026-06-30 balance sheet), National Business Daily 2026-02-24 (FY2025), Eastmoney/financials structured data and Lanjinger (FY2023-25 profit and cash flow series, cross-consistent). "—" indicates not obtained; FY2023/24 gross margins not disclosed in accessible channels.
Reasons for metric changes (items with YoY ≥ ±20%):
Revenue of RMB 405M, +1.5% YoY—the slowest growth since listing—of which subscription revenue was RMB 393M (97.1% of total), up only +0.8% YoY; new AI applications and products at RMB 31M (+177%, 7.7% of total) were the only significant growth engine, but their scale is insufficient to change the slope of total revenue. Net loss attributable to shareholders was RMB 146M (loss was larger in the prior-year period, narrowed); adjusted net profit of RMB 55.42M nearly matched FY2025 full-year (RMB 57.1M)—but the improvement came mainly from expense compression and removal of one-off items (see above), not from revenue. Quality-side signals were weak: overall NDR 94.4% (−5.9pct YoY), enterprise customer NDR 96.4%, RPO over one year −43.8% YoY—deteriorating forward revenue visibility. The day after the interim report (08-28), CCB International maintained "Outperform" with a target price of HKD 30—in sharp contrast to the subsequent October lock-up expiration crash, showing sell-side anchors lagging significantly. At the market level, the interim report itself did not trigger a crash (the stock traded in the HKD 10-13 range during Aug-Sep); the real repricing was triggered by the 10-19 lock-up expiration, indicating the market priced the "+1.5% growth" gradually but the shareholding structure change abruptly.
Business model sketch: Cloud-native vertical SaaS—Kujiale (domestic home/public-space design + rendering) + Coohom (overseas, 18 languages covering 200+ countries) + SpatialVerse/Aholo (spatial intelligence as the second curve). Subscription-based (subscription revenue at 96.9%-97.1%), freemium funnel for customer acquisition (65M registered users → MAU ~2.5M → paying customers: 48,617 enterprise + ~434K individual), enterprise customers pay over RMB 10K annually while individual customers average only ~RMB 200/year—extremely tiered ARPU. In-house Qizhen rendering engine + Matrix CAD engine, paired with a self-built GPU cluster ("consumer-grade low-cost GPUs" strategy) to dilute rendering costs—this is the foundation of the 82-83% gross margin. Revenue is highly recurring, but pricing power is weakening (NDR <100%, long-term contracts with new customers at 9.8%).
Cash content of earnings: book losses far exceed cash burn—FY2023-25 cumulative attributable loss of ~RMB 1.59B vs cumulative operating cash outflow of ~RMB 190M over the same period (−0.63/−1.08/−0.19 RMB 100M series), with the gap mainly non-cash items such as SBC and preferred share fair value changes; even under the −1.92 RMB 100M disputed figure, cash burn remains far below book losses. FCF was negative for three years (~−0.94/−1.39/−0.54 RMB 100M) but narrowed significantly in FY2025. The looseness of the adjusted metric must be called out: in 2026H1 adjusted net profit of RMB 55.42M, SBC of RMB 34.33M (8.5% of revenue) was fully added back—SBC is a recurring annual expense, and adding it all back is an aggressive treatment; the economic operating profit after removing listing expenses and preferred share items is ~RMB 21.1M per half-year (5.2% margin), so the "profitability inflection" deserves a discount. Long-term, the traditional OCF/net profit and FCF/net profit tests fail for loss-making companies; the more relevant framing is: cash burn rate (~RMB 50-200M/year) vs cash reserves (RMB 1.55B)—self-sustaining cash generation is not yet achieved, but the runway is long (5+ years).
Return on capital: under persistent GAAP losses, ROIC is negative and a positive return on invested capital cannot be measured; using adjusted-basis profit (~RMB 100M/year) against net cash + net assets (~RMB 2.6B combined), the implied return is ~4%—below any reasonable WACC (this report uses 11%), so "growth creates value" does not yet hold.
CapEx quality: fixed asset purchases of 0.31/0.31/0.35 RMB 100M vs depreciation & amortization of 0.57/0.46/0.41 RMB 100M, CapEx/D&A <1 for three years—asset-light holds; but from 2026H1, GPU compute investment is accelerating (annualized CapEx ~RMB 55M, +124% YoY), plus 20% of IPO proceeds earmarked for "core technologies and infrastructure"—future capital intensity trends upward; the company does not disclose a maintenance/growth split, so this cannot be determined.
Red flags: ① Rising customer concentration—411 largest customers contribute 46% of revenue (prospectus basis); ② NDR falling below 100% + RPO over one year at −43.8%, deteriorating forward visibility; ③ The SaaS-specific leading indicator "contract liabilities/deferred revenue" was not obtained due to unreadable source, to be verified in the annual report.
Consistency of words and deeds: only 7 months post-listing, insufficient guidance track record—only event-based characterization is possible: the four founders + CFO voluntarily extended lock-ups by 12 months during the post-expiration crash (a skin-in-the-game commitment, positive); the tension on the other side: R&D expenses compressed 25% over three years (3.9 → 2.9 RMB 100M) coexists with management's public statements about "continuously increasing investment in world models/compute"—the direction of expense discipline vs narrative expansion awaits annual report verification. Verdict: existing materials insufficient for a full assessment; leaning toward a hybrid style of "pragmatic retrenchment + optionality-preserving narrative". Shareholder friendliness: no dividends, no buybacks (reasonable during the loss-making period); no insider selling filings or placement dilution seen post-IPO; ongoing SBC accrual (~RMB 70-100M/year scale) constitutes chronic dilution. Verdict: neutral. Risk signals: October 2026 renaming to "Manycore Tech" (brand unification move, neutral); no abnormal related-party transaction disclosures found (not obtained); two of the covering sell-side firms are joint sponsors (CCB at HKD 30/J.P. Morgan HKD 19 neutral)—sponsor target price independence is questionable, and their views were already discounted when adopted.
The company discloses by brand line rather than strict segment reporting; the available structural breakdown is as follows:
| Business Line | 2026H1 Revenue | Share | YoY | One-line Business Logic |
|---|---|---|---|---|
| Subscription revenue (Kujiale + Coohom core) | RMB 393M | 97.1% | +0.8% | Design/rendering tool subscriptions, efficiency-driven demand from real estate chain customers |
| New AI applications and products | RMB 31M | 7.7% | +177% | AI design tools + LuxReal + SpatialVerse, second curve (may overlap with subscription basis) |
| Professional services revenue | RMB 11.8M | 2.9% | +29.3% | Mainly SpatialVerse synthetic data project-based revenue |
| Of which: overseas revenue (Coohom) | ~RMB 37M in 2025 | ~4.5% (FY2025) | +15-20% (FY2027E) | Low cloud design penetration overseas, long-term upside but small base |
Profit driver: subscription is the absolute mainstay (96.9%-97.1% of revenue + the only scaled profit source); SpatialVerse/hardware are still in investment phase. Gross margin structural differences: the company does not disclose gross margin by business line (total 82-83%); AI/professional services project-based revenue gross margin is likely below pure subscription, and its rising share could dilute total gross margin by 1-2 percentage points—needs annual report verification. Under a single-core-business structure, mix shifts matter more than totals: AI share climbing from 0.6% (2024) → 7.7% (2026H1) is the core variable in the valuation narrative, but note possible overlap with the subscription basis (due to company disclosure).
Accounting red flags: ① Large gap between adjusted and GAAP net profit (~RMB 200M in 2026H1)—besides preferred share fair value changes (disappears post-IPO, reasonable) and listing expenses (one-off, reasonable), full SBC add-back is a loose treatment (see Business Model and Earnings Quality chapter); ② Pre-IPO statements showed an apparent insolvency—"total liabilities of RMB 4.735B vs negative net assets"—due to fair-value measurement of convertible redeemable preferred shares; this is standard Hong Kong IPO accounting, not real solvency risk; post-IPO preferred shares converted to equity, with net assets of RMB 1.090B at 2026-06-30; ③ The two-figure OCF discrepancy (−0.19 vs −1.92 RMB 100M, see financial table) is itself a disclosure verification gap. Cross-period consistency: in verifiable series, revenue growth, gross margin (82% → 83%), and expense compression direction are consistent across periods; the NDR series stitches an annual basis (106.1% → 98.6%) with an interim basis (100.3% → 94.4%)—each is valid on a like-for-like YoY basis (annual −7.5pct, interim −5.9pct), consistent trends, no contradiction. Insufficient disclosure history: only 7 months post-listing, fewer than 3 complete reporting periods available; cross-period consistency is overall "partially assessable"; note-level checks (receivables, contract liabilities, capitalization policy) are missing due to unreadable source. Conclusion: based on available materials, no obvious accounting tricks or cross-period manipulation found; but the depth of this check was limited by the unreadable PDF—this is "checked but not fully verified"; once the annual report is readable, the cash flow statement and SBC details should be re-examined first.
Closing price HKD 2.65 (2026-11-20), market cap ~HKD 4.585B (total shares ~1.730B), EV ≈ HKD 2.760B (net cash HKD 1.825B, 39.8% of market cap). PS(TTM) ≈ 4.72x (TTM revenue RMB 826M ≈ HKD 972M); EV/S(TTM) ≈ 2.84x; PB ≈ 3.65x—with only 7 months since listing, the historical percentile window covers only a one-way decline, multiple percentiles are not meaningful, and this report does not use percentiles as cheap/expensive signals (PE is also inapplicable for a loss-making company; FY2026E adjusted PE ≈ 35x, clearly rich for a company growing +1.5%).
| Peer | PS(TTM) | Revenue Growth | Gross Margin | Notes |
|---|---|---|---|---|
| Mingyang Cloud (0909.HK) | ~0.85x | FY2025 −10.5%, 2026H1 −13.3% | ~80% | China real estate chain SaaS "reality multiple anchor," market cap below net cash |
| Autodesk (ADSK) | ~6.6x | +16%~18% | ~91% | Global general-purpose design software leader |
| Bentley (BSY) | ~7.9x | +10%~14% | ~82% | Infrastructure AEC vertical SaaS |
| Manycore Tech (00068.HK) | ~4.72x | +1.5% (2026H1) | 82-83% | Multiple between Mingyang Cloud and global leaders, growth and multiple mismatched |
(All peer multiples use 2026-11-20 market cap ÷ respective TTM revenue; cross-market currencies handled separately.)
Core mismatch: Manycore's PS is ~4.5x higher than Mingyang Cloud and ~28% below Autodesk, but its growth (+1.5%) belongs to the same "stall camp" as Mingyang Cloud (−13%) and is far below global leaders (+10-18%)—the premium the market retains for it can only be explained by the AI/spatial intelligence narrative.
Reverse DCF (EV ≈ RMB 2.345B, 11% discount, interim economic FCF ≈ −RMB 50M/year): at terminal EV/S of 2.75x (a successful transformation into a Chinese SaaS mainstay), the current price implies a five-year revenue CAGR of ~13%; at terminal 4.5x (AI premium realized), ~2.8%—i.e., the current price requires revenue to compound at roughly 3%-13% over five years, with almost all incremental growth supplied by AI (AI revenue expanding 6-11x from annualized ~HKD 73M, CAGR ~63%), while reality is core business +1.5%, NDR 94.4%, RPO −43.8%. Another extreme test: with a terminal anchor at Autodesk's 6x PS, the implied CAGR is negative—the current price is self-consistent even without crediting any growth; this shows the "implied growth range" is highly sensitive to terminal multiples, so no one-directional "expensive/cheap" conclusion can be drawn; it must be combined with three-layer value and scenario distribution.
| Scenario | Probability | Fair Range (HKD) | Decisive Move |
|---|---|---|---|
| Bear | 28% | 0.9–1.5 (midpoint 1.2, −54.7% vs current) | Core business turns negative growth + AI share <9% + lock-up selling pressure materializes; priced at EV/S 0.5-1.0x × FY2027 revenue + cash counted at 65% (Mingyang Cloud-style discount) |
| Base | 48% | 2.0–2.45 (midpoint 2.225, −16.0% vs current) | Core business stabilizes at low single digits + adjusted net profit expands (FY2027E RMB 120-150M) + AI share climbs to 15-18%; priced at EV/S 1.8-2.4x × FY2027 revenue + net cash—anchored at the "Chinese SaaS mainstay" tier between Mingyang Cloud (0.7-0.8x, the reality floor for shrinking revenue) and ADSK/BSY (6-8x, the quality ceiling), not this company's current multiple |
| Bull | 24% | 3.6–4.6 (midpoint 4.1, +54.7% vs current) | AI second curve realized: AI revenue FY2027 RMB 150-170M → FY2028 RMB 300-350M, share 25-30% (software-hardware integration + robotics synthetic data volume); FY2028 EV/S 3.5-4.5x (discounted to global leaders, heavier China revenue structure and capital intensity) |
Probability-weighted fair value ~HKD 2.39 (expected return ~−10%); current price sits between base and bull, toward the upper end—already pricing in roughly 40% of the bull case probability. Odds structure: to base midpoint −16%, to bear midpoint −55%, to bull midpoint +55%; nominal upside/downside symmetry but probability distribution skewed downward (base + bear = 76% of scenarios imply fair value below current price). External anchor comparison: the more cautious J.P. Morgan target of HKD 19 (2026-05-28, when the stock traded ~HKD 19.5, ~26x PS) is obsolete in absolute terms—its value lies in the growth assumptions (11% CAGR), not the price; the more optimistic Soochow/Guosheng HKD 15.1 (2026-09-08)/CCB HKD 30 (2026-08-28) were all formed before the lock-up expiration crash and not updated—coverage is thin and two are sponsors; they constitute no valid expectations and provide no downside protection.
| Period | Revenue | Net Profit (Attributable) | Adjusted Net Profit | Driving Assumptions |
|---|---|---|---|---|
| FY2026E | RMB 825-845M (+0.6%~+3.0%) | −200 to −180M | RMB 105-115M | Subscriptions 0~1%; AI +170%~180% to ~RMB 80M; gross margin ~83%; implied H2 YoY −0.2%~+4.5% (low base/trade-in subsidies/hardware launch support) |
| FY2027E | RMB 850-880M (+2.4%~+6.0%) | −160 to −120M | RMB 120-150M | AI doubles to RMB 150-170M (~18% share); overseas +15-20%; top-customer NDR 101.8% supports structure; SBC gradually declining |
The company has not disclosed revenue/profit guidance; no reliable sell-side consensus earnings forecasts are available for comparison (existing target prices lack verifiable forecast figures).
Mildly overvalued: the current price is about 8% above the baseline fair upper bound, about 11% above the probability-weighted fair value, with no margin of safety (−24.5% on the baseline lower bound). Quality and price should be viewed separately: the company is a qualified China SaaS asset (net cash at 40% of market cap, 83% gross margin, #1 market share, profitable on an adjusted basis), but the current price has already priced in that quality along with roughly 40% bull-case probability—at the current level, the return source for buyers is AI option realization rather than value reversion. Note on baseline exit multiple anchoring: EV/S 1.8–2.4x reflects the reasonable tier for a China vertical SaaS with "83% gross margin + 15% adjusted net margin + low single-digit growth"; a comparison to the 5-year median is unavailable (listed for 7 months), and it represents a 60–70% discount to the global peer median (6–8x), justified by decelerating growth and China property-chain exposure. If one deems this discount excessive, the appropriate action is still to wait for a better price rather than buy immediately.
The niche Manycore operates in is "space design software" (cloud design SaaS), a subset of the broader "general design and visualization software market." Per the primary institutional source for this research (Frost & Sullivan, exclusive industry consultant for the company's IPO):
Note the industry's absolute size is small (RMB 3.3 billion in China); even doubling by 2029 yields only RMB 6.6 billion—Manycore's FY2025 revenue of RMB 820 million already represents ~25% of the Chinese market; the industry ceiling means its growth must come from share capture, ARPU expansion, or breaking beyond space design (AI/spatial intelligence/overseas). This is the scale context needed to understand its valuation narrative.
Upstream are GPU cards/cloud compute and IDC (rendering is the core compute consumer)—Manycore dilutes unit rendering costs with a "consumer-grade low-cost GPUs + self-built GPU cluster" strategy, but has weak bargaining power over the NVIDIA ecosystem and IDC resources, and compute dominates operating costs; Frost & Sullivan lists "capital investment barriers" (GPU infrastructure) as one of the industry's four major entry barriers—upstream dependence, while a cost disadvantage, also raises the hurdle for latecomers. The asset/model ecosystem (~150 million design assets accumulated by Coohom, 300 million+ design plans, covering ~90% of Chinese floor plans) is a quasi-upstream asset with clear data network effects. The midstream is the cloud design SaaS platforms (Coohom/Kujiale, Coohom international, 3vjia, Easyhome Design), where value is retained: Manycore's gross margin is 82.2% and Autodesk's ~90%, far above downstream custom furniture (~30% and broadly declining sharply in 2026). Downstream are furniture customization manufacturers (Oppein, Sofia, etc.), home/commercial renovation firms, real estate and commercial space clients, and individual designers—while gross margin sits in the midstream, bargaining power over downstream is deteriorating: property-chain client budget cuts have pushed Manycore's NDR down from 106.1% to 94.4%, and the share of new-client contracts over one year from 48.6% to 9.8%—the traditional SaaS logic of high switching costs is being suppressed by industry-wide budget cuts; this is the real-world version of "gross margin in the midstream, pricing power downstream."
Demand side: volume drivers are digitalization penetration and cloud migration in home renovation/commercial decoration/real estate/e-commerce display scenarios (commercial & industrial space design, 2024–2029 CAGR 18.1%, the fastest sub-segment); price drivers are subscription add-ons (accelerated rendering, AI features, enterprise editions). But total demand is suppressed by the deep property correction: nationwide new commercial housing sales area was 974 million ㎡ in 2024 (−12.9%) → 881 million ㎡ in 2025 (−8.7%) → 450 million ㎡ in Jan–Jul 2026 (−11.8%), contracting for four consecutive years; all nine listed custom-furniture companies saw revenue declines in the first three quarters of 2026 (Oppein −29.6%, Sofia −24.1%). The software layer keeps growing on penetration + cloud migration + AI offsets (Manycore +8.6% in 2025), but the growth center has systematically shifted down (+1.5% in 2026H1)—the industry shows a pattern of "shrinking demand, fragmented supply, head-tier capture." Supply side: China has over 50 active participants but the market is concentrated at the top; Manycore leads with a 23.2% share (2024 revenue basis, Frost & Sullivan)—beyond CR1, 3vjia and Easyhome Design (formerly Tangping Design, Alibaba-affiliated) have undisclosed financials; CR3/CR5 unavailable (Frost & Sullivan only provided CR1). Some media claim Manycore held a 70%+ share of domestic home design software during 2018–2021 (single source, unverified)—even on that basis, the share trend is declining. Entry barriers (Frost & Sullivan): technology (high-performance 3D rendering parallel computing + AI integration), capital (GPU infrastructure), talent (scarce graphics + AI hybrid skill sets), brand (copyright + trust); plus Manycore-specific data network effects from its asset library and its design-to-production closed loop. Competitive intensity and substitution threats: moderately high but no vicious price war observed; substitution threats are layered—AI text-to-image tools are diverting basic rendering demand at the low end (the most direct threat to individual subscription ARPU and retention; individual NDR of 83.3% may partly reflect this); high-end rendering is held by SketchUp+D5 and the 3ds Max professional pipeline (interviews with frontline designers: interior design favors SketchUp+D5, with Coohom seen as "lacking realism quality"); Autodesk Revit dominates commercial decoration/BIM. Coohom's advantage zone is the vertical workflow of "fast rendering + marketing display + front-to-back production integration"—fast and comprehensive, but not superior in realism or professional depth; this positioning faces pressure from both ends in the AI rendering era.
The industry is in a mismatched combination of "technology growth phase × deep downstream downturn." Leading indicators: ① nationwide commercial housing sales/completion area (monthly, National Bureau of Statistics); ② quarterly revenue of listed custom-furniture companies (leads Manycore revenue by ~1–2 quarters); ③ Manycore NDR and new-client long-contract ratio (annual/interim reports); ④ MAU (2.7 million average in 2024 → ~2.5 million in 2025); ⑤ share of revenue from new AI businesses (0.6% → 7.7%). Regulation is neutral to mildly restrictive: the Interim Measures for the Administration of Generative AI Services (effective 2023-08) require algorithm filing, security assessments, and content labeling for AI generation services—Coohom's AI rendering is directly subject to these, raising compliance costs for feature launches; the Data Export Security Assessment Measures create compliance friction for Coohom's cross-border cloud services; the property "trade-in" national subsidy (first tranche of RMB 62.5 billion allocated in January 2026) provides interim support for downstream demand. The industry's true headwind is the property cycle, not regulation.
| Company | Revenue Scale | Growth | Gross Margin | ROE | Positioning and Differentiation |
|---|---|---|---|---|---|
| Autodesk | FY2026 (to 2026-01) USD 7.21 billion | +17.5% | ~90.6% | N/A (buybacks drove equity negative) | Absolute global leader in general CAD/BIM; Manycore shares its high-margin subscription model but trails by an order of magnitude in market depth and contract size |
| Bentley | FY2025 USD 1.50 billion | +11.0% (ARR +11.5%) | ~81.5% | ~6–8% | Infrastructure AEC vertical SaaS with utility-like demand resilience; Manycore's downstream is a strong property cycle—both are verticals, but their demand sources differ fundamentally |
| Mingyuan Cloud | FY2025 RMB 1.284 billion | −10.5% (2026H1 −13.3%) | ~80% | ~1.5% | "Late-property-cycle mirror control": deeply tied to developers' IT budgets, shrinking at double digits; Manycore maintains positive growth via multi-scenario and AI extensions, demonstrating relative resilience |
| Manycore Tech | FY2025 RMB 820 million | +8.6% (2026H1 +1.5%) | 82.2% | Negative (loss-making) | #1 share in China space design software; vertical-scenario diversification + AI second curve, but growth is decelerating |
Positioning: absolute niche leader + pressured share trend + unverified AI narrative. The moat is real (rendering cost advantage, asset-library data network effects, ecosystem of 40,000+ enterprise clients, 23.2% #1 share), but proxy indicators of the share trend (growth, NDR, long-contract ratio) are weakening across the board—the durability of leadership and the speed of AI realization are the two key variables to track. Note on market-position figures: the 23.2% is single-sourced from Frost & Sullivan (2024 revenue basis); the media-reported historical share (70%+) is questionable; both figures are disclosed in parallel, with the former taking precedence. For this industry, AI is "augmentation-dominant, partially substitutive, and still small in magnitude" rather than disruptive (this research's industry-line judgment): the quantified chain—Manycore's new AI application revenue of RMB 5.2 million (2024, 0.6% of revenue) → RMB 31 million (2026H1, 7.7%, +177% YoY); the company claims its AI platform generates full-home plans in 5 minutes (marketing claim, unaudited); external benchmark—Figma's 2025 survey found only 31% of designers directly use AI to generate design assets—AI has entered workflows but has not eliminated roles. For Manycore, AI cuts both ways: it is both the growth option most priced in by the market (about half the current price) and potentially the substitution source eroding its low-end rendering base (individual-client NDR of 83.3% is a warning)—the same technological shift is being booked on both ends of the income statement simultaneously; the two reporting periods of 2027 will decide the outcome.
The valuation of cloud design SaaS rests not on current profit but on stock metrics of subscription quality; this section presents them together (all figures on the latest available basis):
| Metric | Value | Interpretation |
|---|---|---|
| Subscription revenue share | 96.9% (FY2025) → 97.1% (2026H1) | Highly pure recurring revenue structure; genuine SaaS attributes |
| Overall NDR | 106.1% (FY2023) → 98.6% (FY2025) → 94.4% (2026H1, −5.9pct YoY) | Below 100% = stock base contracting; growth can only come from new-client acquisition, while acquisition spend is being cut |
| Tiered NDR (2026H1) | Enterprise 96.4% / Individual 83.3% / Largest clients 101.8% | Large clients still net-expanding while individual churn accelerates—client mix concentrating at the top; the higher-contract-value strategy faces a "move upmarket vs lose the base" balance |
| Client structure | 48,617 enterprise clients + ~434,000 individual clients; 411 largest clients contribute 46% of revenue | Concentration is high and trending up, amplifying exposure to top-client budget swings |
| Contract structure | Share of new-client contracts over one year: 48.6% (2023) → 9.8% (2025, single source) | Forward visibility collapsed, corroborated by >1-year RPO of −43.8% YoY (the two may be two facets of the same underlying fact) |
| User funnel | 65 million registered → MAU ~2.5 million → paid conversion | Broad freemium funnel but shallow conversion; AI features are the lever for funnel stickiness |
| Rule of 40 | Growth 1.5% + adjusted net margin ~13.7% ≈ 15 | Well below the 40 threshold; even on FY2026E adjusted net margin of 13% + 2% growth it is still ~15—by the SaaS quality framework, it merits only a "low-multiple tier"; the premium of 4.72x PS can only be explained by the AI option |
| Cash conversion | SBC at 8.5% of revenue (2026H1), ~RMB 70–100 million per year | Source of the "water" in adjusted profit; GAAP breakeven requires revenue growth back above 15% or a substantial decay in SBC |
The overall picture from these metrics: a vertical SaaS with a pure subscription structure but whose growth engine has degraded from "stock expansion (NDR>100%)" to "new-client replenishment"—valuation elasticity for such companies is extremely sensitive to the direction of NDR; a return of NDR to 100% is a more leading bullish signal than revenue growth, while NDR below 90% would confirm base erosion. Tracking priority: tiered NDR > AI revenue share and repurchase > new-client long-contract ratio > revenue growth.
Overall rating: Neutral (confidence 0.55, 12 months). Manycore Tech is a combination of "good industry position (niche leader) + bad industry cycle (property-chain downturn) + real cash buffer (which can be discounted) + unverified option (worth half the current price)." The bubble has deflated 94.5%, but the endpoint of deflation is not automatically a buy point: the current price remains ~8%–11% above baseline fair value, expected return ~−10%, and the odds distribution is skewed downward (fair value below the current price in 76%-probability scenarios). Strategy: do not enter now; wait for one of two conditions—① price falls back to the fair range of HKD 2.0–2.45 (begin entry evaluation below 2.45; if it breaks 2.0 without fundamental deterioration, evaluate for deep margin of safety); ② fundamental evidence reversal (AI share above 15% + revenue growth back above +5% + NDR back to 100%), at which point the valuation framework would be revised up wholesale. For existing holders' discipline levels: below HKD 1.5 (bear case realized) or upon disclosure of major-shareholder selling, re-examine rather than add.
Key risks (downside): second wave of lockup-expiry selling pressure (founder lockup expires 2027-10-16, with anticipatory trading possibly starting mid-2027); core revenue turning negative (momentum of NDR 94.4%, RPO −43.8%); AI option falsified (share <9% or growth <50%); deepening market discount on net cash (Mingyuan Cloud-style 0.3x) and removal from Stock Connect (FTSE/Hang Seng review 2027-03). Key catalysts (upside): new product launches at the Coohom+ Conference in December 2026; Aholo's two co-branded hardware products' pricing/launch/orders; SpatialVerse million-level repeat contracts; disclosure of token-billing revenue for the world model/LuxReal; confirmed southbound accumulation signaling chip clearing.
Tracking checklist (machine monitoring of invalidation conditions and catalysts in watchlist): FY2026 annual report (revenue direction/adjusted net profit/cash/AI share/tiered NDR/cash flow—and adjudicating the OCS-basis discrepancy); HKEx disclosure of interests ≥5% shareholders and weekly southbound holding changes; Aholo hardware launch evidence and SpatialVerse client repurchases; price crossings of the four levels 2.45/2.0/1.5/4.6.
This report is based on cross-verified public information and structured data; the original financial statements could not be read directly due to a system environment failure (detailed in the data integrity statement); all judgments follow the evidence and do not constitute investment advice.
<!-- publication:1977952ad28df5dcd2161915 -->