SG Micro: 42.7% H1 Growth at a 51.84% Gross Margin Proves Share and Content Gains, but the A Share at CNY 120.55 Is 105x Trailing Earnings and Above the CNY 89 Base Value

The build-out took revenue to CNY 1.20 billion and attributable profit to CNY 289 million by 2020. The 2021–22 shortage and substitution boom almost tripled revenue to CNY 3.19 billion in 2022, with attributable profit of CNY 874 million; the product gains were genuine, but the 36.9% and 30.0% ROEs also reflected extraordinary pricing and supply conditions.

The 2023 reversal is the most informative episode: revenue fell to CNY 2.62 billion, attributable profit to CNY 281 million and ROE collapsed to 7.63%. SG Micro is a high-quality participant in a cyclical industry, not a cycle-free compounder. Even after the 2024–25 rebound lifted revenue above the 2022 peak, 2025 recurring profit of CNY 428 million was roughly half of 2022 attributable earnings on CNY 710 million more revenue; R&D intensity, price normalization and competition had reset the economics.

The current stage began in late 2025 and accelerated sharply in 2026, with the Jiangyin design and test base operating, automotive-qualified SKUs spreading and optical-module demand turning network/computing into a major revenue engine. After the H-share IPO the constraint is shifting from funding toward whether management can deploy capital far beyond the historic balance sheet without diluting returns.

Financial trajectory

† H1 measure, not comparable with full-year ROE and distorted by the June H-share capital raise. 2020–22 figures are from company annual filings; 2023–25 and H1 2026 are CAS figures.

Over 2021–25, OCF of about CNY 2.90 billion matched roughly CNY 2.90 billion of attributable profit, but the bad years matter: the 2023 destocking trough produced only CNY 171 million of OCF, 2025 OCF was CNY 466 million against CNY 547 million of attributable profit, and H1 2026 conversion improved to 91%. Profit is broadly cash-backed over the cycle, but working capital absorbs cash precisely when the cycle weakens.

The 2023 and 2025 3-for-10 capitalizations break the published EPS series, so I restate on the September 24, 2026 total-share base of 686.091 million shares: roughly CNY 0.409 of attributable earnings per share for 2023, 0.729 for 2024, 0.797 for 2025 and 0.613 for H1 2026, and trailing earnings of about CNY 1.12 a share attributable and 0.95 recurring, before future option dilution.

What changed in the latest two quarters

H1 revenue was CNY 2.595 billion, up 42.7%; attributable profit CNY 420 million, up 109.3%; recurring profit CNY 356 million, up 164.6%. Subtracting Q1 leaves Q2 revenue of about CNY 1.497 billion, 45.5% above Q2 2025 and 36.4% above Q1, and attributable profit of around CNY 297 million. H1 cost of sales of CNY 1.250 billion implies a 51.84% gross margin; the implied Q2 margin of about 52.0% sits slightly above Q1's 51.6%, so the surge was not bought through visibly lower pricing.

At the August 31 investor meeting management said network and computing, about 25% of H1 revenue, had grown nearly 100%, with Q2 up more than 100% year on year and more than 45% sequentially, led by optical modules. That implies about CNY 649 million of H1 revenue; assuming “nearly 100%” means roughly a doubling, it contributed around CNY 323 million, about 42% of SG Micro's CNY 777 million H1 increase. This is an inference, not a disclosed segment bridge. On the same approximation the rest grew around 30%, which the company attributes to continuous product launches, a rising high-end mix, industrial-market accumulation and wider customer coverage; that breadth is why I weight structural share/content gains above channel restocking, while allowing a sizeable cyclical component.

Inventory is the counterweight. Net inventory rose from CNY 1.448 billion at December 2025 to CNY 1.677 billion at June 2026; gross inventory was CNY 2.030 billion and the write-down allowance CNY 353.4 million, up from CNY 310.4 million at year-end. H1 reserve additions of about CNY 108 million were overwhelmingly raw materials and finished goods, and the company attributed CNY 107.8 million of asset-impairment losses to inventory provisioning. Average net inventory equals roughly 226 days of H1 cost of sales, though a reserve ratio of about 17.4% of gross inventory, slightly below year-end, tempers the concern. The blind spot is the channel: management says it is intentionally carrying more stock as breadth and revenue expand, but without disclosed distributor sell-through, investors cannot independently separate healthy SKU proliferation from demand pulled forward into distributors.

Q4 2025 should not be annualized: recurring profit roughly doubled from Q3 on an 11.8% revenue increase, but the annual disclosure lacks the quarter-specific mix, provision and operating-expense detail to isolate a durable incremental-margin change. The H1 evidence, with gross margin and recurring earnings improving across two quarters, is the firmer basis for the earnings thesis.

Business model, moat, governance, industry and policy

Revenue, cost structure and earnings quality

The residual group revenue includes sensors, memory/peripheral products and other categories.

The signal-chain mix is particularly important for valuation: operational and instrumentation amplifiers, ADCs, DACs, analog front ends and interface products need process/device-model expertise and often qualify more slowly than commodity power parts, and at a 57% H1 gross margin each incremental yuan of signal chain earns more gross profit than one of power management, which partly explains why group margin rose as revenue accelerated.

R&D is the company's main operating investment. All expensed, it was CNY 617 million in H1, 23.77% of revenue, against CNY 508 million a year earlier; 1,401 R&D employees made up 72.7% of staff, cumulative patents reached 724 by June and nearly 900 products launched in 2025. Expensing makes current earnings more conservative than a model that capitalizes development spending, but the business must sustain roughly a quarter of sales in R&D to keep the catalog moving.

Operating leverage is beginning to appear: H1 revenue rose 42.7% while R&D rose about 21.6%, selling expense 27.2% and administrative expense 22.0%. That is legitimate while new products keep scaling; a downturn reverses it, because analog engineering teams cannot be cut in line with quarterly sales without damaging the franchise.

Non-recurring income was a major 2025 issue: CNY 119.5 million, including CNY 71.8 million of government subsidies, roughly CNY 54.6 million of financial-asset/fair-value gains and CNY 6.5 million of entrusted-investment income, less taxes, minority interests and other items, against only about CNY 49.1 million in 2024. H1 2026 added CNY 64.6 million, including CNY 24.9 million of subsidies and about CNY 44.9 million of financial-asset gains. These items recur in the loose sense that SG Micro regularly holds investments and receives subsidies, but they are neither customer-generated nor predictable enough to capitalize at an analog-growth multiple.

The H-share raise will sharpen this distinction: with more than CNY 2 billion already in low-risk bank wealth-management products, H2 interest and wealth-management income can rise with no improvement in analog economics, so my operating valuation strips those gains from the earnings base and adds cash separately.

The moat and where it stops

SG Micro's strongest moat is breadth combined with analog process know-how. Thirty-eight categories and more than 7,200 saleable products let a distributor or systems engineer qualify one supplier across amplifiers, converters, references, switches, interface ICs, LDOs, DC/DC converters, drivers and sensors, giving each successful relationship cross-selling value. Switching costs are moderate per product but meaningful once customers have qualified dozens of parts, especially in industrial and automotive designs.

The second moat is the accumulated design/process/test loop: SG Micro stays fabless for normal manufacturing but integrates proprietary process technology into foundry production and now performs some special testing at Jiangyin. TSMC has been the principal foundry since the founding, with SMIC subsidiaries and DB HiTek as additional wafer sources; regular packaging and testing remain with suppliers including JCET, Tongfu Microelectronics, Tianshui Huatian and Carsem. That is a real moat but also a dependency. The exact share of wafers from TSMC, the Taiwan fabs used, domestic-foundry yield comparisons and process-transfer costs are not disclosed in the primary A-share material I could verify. I therefore do not present a made-up localization ratio. Porting an analog design is harder than transferring GDS files, since device models, high-voltage behavior, noise, reliability and qualification all interact with the process; multiple foundries reduce concentration risk only after the relevant products qualify at commercially acceptable yields.

The Jiangyin test base is useful rather than transformative: it gives control over special-test know-how, engineering feedback and selected lead times. Fixed assets rose from CNY 539 million at end-2025 to CNY 641 million at June 2026, with CNY 271 million of contracted long-term-asset commitments outstanding; the likely margin benefit is gradual: fewer outsourced special tests and faster learning cycles, not a sudden conversion into an integrated device manufacturer.

The moat is weakest against global vendors in the highest-reliability sockets: Texas Instruments and Analog Devices have decades of process libraries, internal manufacturing, sales engineering and qualifications, and NXP, Infineon, STMicroelectronics, Renesas and onsemi deep automotive relationships. SG Micro's China analog share of only about 1.8%, according to the issuer-commissioned Frost & Sullivan work, leaves room to gain but also shows how much of the profit pool incumbents still control.

Governance, incentives and capital allocation

Dr. Zhang Shilong remains chairman, executive director and general manager. His wholly owned Hongshun Xiangtai held 17.35% at June 30; with Baoli Hongya, Hongwei International, Zhang Qin, Lin Lin and other acting-in-concert parties the controlling group held about 221.5 million shares, roughly 32.3% of the current 686.1 million share base. Effective control remains considerable despite H-share dilution.

The alignment is imperfect. Director and concert-party member Lin Lin sold 7.887 million A shares, 1.27% of the company at the time, between November 2025 and January 2026 after pre-disclosing the plan in October, at averages of CNY 64.74 in block trades and CNY 75.02 by centralized bidding. The timing, between the HKEX application and the H-share offering, is a legitimate negative governance datapoint rather than something to explain away, even though the sale was pre-announced and was not a founder exit.

Equity incentives are economically large. H1 equity-settled share-payment expense was CNY 91.3 million at company level plus CNY 7.6 million in a controlled subsidiary, about CNY 99 million or roughly 24% of attributable profit, after about CNY 158 million embedded directly in 2025 administrative, selling and R&D expenses before any subsidiary amount; I treat all of it as compensation cost. The primary disclosure adds a nuance: at June 30, residual 2022 stock options were still outstanding with about five months of contractual life, alongside the 2023 plan, both 2025 option plans and the 2021 restricted-share instrument, so a count of “four live plans” understates the residual equity instruments at the H1 balance-sheet date. Exercise prices ranged from CNY 44.26 to CNY 78.09.

Known dilution alone is meaningful. The second 2025 plan's first exercise period allowed 2.853 million options at CNY 55.49, implying CNY 158 million if fully exercised, and the 2023 plan's September window covers about 2.594 million options at CNY 50.34, roughly another CNY 131 million. Those two tranches alone represent 5.45 million potential shares and at least CNY 289 million of proceeds; the actual overhang is higher because future tranches and the other 2025 plan remain. The H1 summary note gives prices but not a clean current option count by every instrument, hence the deliberately conservative share count used in valuation.

The second 2025 plan's first tranche required 2025 revenue of at least CNY 3.709 billion to trigger and CNY 3.80 billion for full company-level vesting, with annual/cumulative targets rising through 2028; actual 2025 revenue of CNY 3.898 billion cleared the first full target. The incentives reward sales growth, but investors should monitor whether growth comes with cash returns, because revenue-only hurdles can encourage inventory, distributor loading or low-return expansion if governance discipline weakens.

Capital allocation is becoming the next governance test. In March 2025 SG Micro bought 67% of 感睿智能科技（常州）有限公司 for CNY 160.8 million; the acquired business contributed CNY 23.4 million of revenue and a CNY 16.1 million loss from acquisition through year-end. In April 2026 it paid CNY 57.66 million for an ADC/DAC business asset group with only CNY 6.53 million of identifiable net assets, booking CNY 51.13 million of goodwill; that business had no revenue and lost CNY 2.38 million through June. The deals are small, but 26% of the H-share proceeds is earmarked for investments and acquisitions, and management has yet to prove it can compound large amounts of acquisition capital at the returns of its organic R&D.

EY Hua Ming became auditor for FY2025, replacing Grant Thornton China/Zhitong, the auditor named in earlier annual filings; the company cited development and overall audit needs, not an accounting disagreement. The timing just before an H-share listing warrants monitoring, but the 2025 CAS opinion was unqualified, and auditor rotation alone is not evidence of a problem.

The CNY 44 million contract dispute has moved in SG Micro's favor. The same CNY 44 million bank deposit remained frozen under property-preservation measures at June 30, but the company had won at first instance; the judgment was not yet effective, and management assessed economic outflow as extremely unlikely. The claim also covers overdue-payment interest and related expenses. The financial note reviewed does not name the counterparty, so I do not invent one.

Industry cycle, tax and geopolitics

SG Micro's own history is the cleanest cycle indicator: 2021–22 combined shortage pricing, rapid electronics demand and domestic substitution; after 2023 destocking and 2024–25 recovery, 2026 adds an AI/optical-content cycle on top of the analog recovery, placing the cycle in early-to-mid recovery, not at a fresh shortage peak. The biggest upside variable is volume/mix in network, industrial and automotive; the most fragile is gross margin, if distributors restock ahead of demand and competitors then cut prices.

Substitution has more room in signal chain and power management than the 1.8% aggregate share implies, but progress will be uneven: consumer and communications designs can change faster, while industrial and automotive customers need long qualification histories, failure-rate data and stable supply, which is where incumbents are hardest to displace. SG Micro's accelerating automotive-qualified portfolio matters more over three to five years than over the next quarter.

China's September 2025 anti-dumping investigation into certain U.S.-origin general-purpose interface ICs and gate-driver ICs on 40nm-and-above processes is directionally relevant because SG Micro sells interface circuits and MOSFET drivers. As of September 28, 2026, MOFCOM's trade-remedy case database still listed the case as under investigation, and I found no preliminary or final determination, duty schedule, extension or termination notice. A November 2025 White House fact sheet said China would terminate its investigations targeting U.S. semiconductor companies, including anti-dumping cases, but MOFCOM has published no corresponding termination. With the original one-year deadline passed and no published outcome, the legal status is an unresolved research item, not evidence that no duties exist. I assign trade protection zero earnings benefit in valuation; SG Micro does not disclose revenue from the covered product definitions, so no credible sensitivity is possible.

Export-control exposure runs the other way. The core analog parts are generally mature-process devices, not the leading-edge AI accelerators at the center of U.S. advanced-computing controls, but reliance on TSMC creates an indirect path: restrictions on foundry customers, technology transfer or Taiwan production can affect availability even when the analog SKU itself is not a controlled advanced chip. Without a disclosed TSMC wafer percentage or fab-by-fab production map, more precision would be false.

Tax benefits are material. The parent qualifies for a 10% income-tax rate in 2026 as a nationally encouraged key integrated-circuit design enterprise; Harbin SG Micro and Jungying Semiconductor Shanghai received high-tech-enterprise certificates in December 2025 carrying a 15% rate for three years; and an IC-industry VAT policy permits a 15% additional input-VAT credit through December 31, 2027 for qualifying design, manufacturing, packaging/testing, equipment and materials companies.

The very low H1 tax expense can be reconciled more specifically than management's description of “lower taxable income.” Pretax profit was CNY 438.55 million and tax expense only CNY 16.81 million, a 3.8% effective rate. The tax note starts with CNY 43.85 million at the 10% rate and subtracts CNY 37.63 million for the R&D additional deduction; current tax was only CNY 0.06 million and deferred tax CNY 16.75 million. The R&D deduction is the principal visible reason taxable income is much lower than accounting profit.

For normalized valuation I do not assume a permanent 3.8% rate: the 10% parent rate depends on keeping qualifying IC-design status, the subsidiary certificates run three years, and the VAT extra credit sunsets in 2027 under the filing's cited policy. A 10–15% normalized cash tax burden would still leave SG Micro profitable but would remove one source of current earnings leverage.

Competitive landscape and niche

What SG Micro became relative to its competitors

Frost & Sullivan, commissioned by SG Micro for the H-share prospectus, estimates the 2025 China analog market at roughly CNY 218 billion. SG Micro's CNY 3.9 billion gave it 1.8% share, eighth overall and first among Chinese vendors; signal chain, at roughly CNY 1.5 billion, ranked sixth overall and first domestically, and power management, at about CNY 2.4 billion, seventh overall and second domestically.

The geography deserves skepticism. H1 CAS reporting books CNY 1.157 billion in Mainland China and CNY 1.274 billion in Hong Kong, together 93.7% of revenue, and the Hong Kong figure largely reflects the distribution and sales structure, not end-customer location. Frost & Sullivan assigns essentially all 2025 consolidated revenue to the “China analog market,” which cannot be reconciled mechanically to the CAS geographic line, so I do not treat the 1.8% figure as a precise auditable market-share statistic.

The commissioned report also anonymizes competitors. Its descriptions fit Company A being Texas Instruments, B Analog Devices, D Infineon, E STMicroelectronics and F NXP; I could not establish C, G and H from primary public disclosure with enough confidence to call the identities confirmed. Monolithic Power Systems, Renesas, onsemi and Microchip are plausible global candidates, but presenting one mapping as fact would defeat the purpose of testing the commissioned report.

Texas Instruments became the scale-and-manufacturing competitor, with enormous catalog breadth, deep industrial penetration and owned mature-node capacity; Analog Devices the high-performance signal-chain specialist with deep instrumentation, industrial and communications content; Monolithic Power Systems a high-density power specialist increasingly exposed to AI infrastructure. Infineon, ST, NXP, Renesas and onsemi combine analog with power, automotive and embedded franchises. Customers choose them when validation history, edge-condition performance and global supply assurance outweigh the savings from domestic substitution.

SG Micro became the broad Chinese challenger. The economic difference is scale: CNY 3.9 billion of 2025 revenue supports a large catalog by Chinese standards but remains much smaller than the multinational incumbents. Its strategic advantage is proximity to Chinese engineering teams, rapid catalog expansion, domestic-supply preference and many “good enough to high end” replacements from one vendor; its weakness is highest-end industrial and automotive qualification history and dependence on external foundries.

Among domestic peers, 3PEAK is the closest signal-chain/power-management comparison; Novosense has a sharper identity in automotive, isolation, sensors and industrial signal paths; Southchip is more concentrated in power management and charging; Awinic is better placed in consumer/audio and mixed-signal niches; Joulwatt and Silergy add power-management comparisons. A September 2026 sector survey called Chinese analog recovering and named SG Micro a principal beneficiary of optical-module, industrial and automotive demand.

The current data support a premium for breadth: the 51.8% H1 group gross margin, 57.1% signal-chain margin and 23.8% R&D intensity show an economic profile closer to a differentiated broad analog vendor than a commodity power-chip house. The premium should have limits: the growth still came with CNY 353 million of inventory reserves and roughly one quarter of sales spent on R&D, and high gross margin does not automatically become high free cash flow.

I do not publish a false-precision current-multiple table for 3PEAK, Novosense and Southchip, because I was unable to cross-verify all three companies' September 24, 2026 prices, fully diluted share counts and H1 2026 accounting bases to the same standard as SG Micro before the research cut-off. The qualitative conclusion holds regardless: the A line, above 100 times trailing earnings, demands a growth premium even against high-growth Chinese analog comparables, while the H line sits much closer to a defensible growth valuation.

Ecological niche

SG Micro is a challenger taking sockets from multinational analog franchises, not a mature leader harvesting an installed base. Its best near-term attack surface is Chinese communications, optical modules and industrial systems where local engineers can qualify domestic parts quickly; the longer-term prize is automotive and high-reliability industrial content. Its profit pool is most vulnerable to 3PEAK and other domestic vendors closing the signal-chain gap, to power-management price competition and to aggressive incumbent responses in China. A pure price war weakens SG Micro; geopolitical localization with stable end demand strengthens it, as, currently, does a technology cycle rewarding broad analog content per optical/AI system.

Valuation, price path and margin of safety

Price path, cross-listing and what the tape is saying

The H-share listing on June 26, at an offer price of HKD 85.20, made price discovery unusually complicated. Exchange data show the H shares opened about 23% above the offer and closed the first day at HKD 125.30, up about 47%, a sign of heavy IPO demand, while the A share fell more than 3% that day as investors gained access to an economically similar claim at a lower price. The H line later traded as high as HKD 142.30 but fell to HKD 94.55 on September 23 and HKD 91.85 on September 24.

The A line closed around CNY 116.58 on August 20 during a broader analog/AI-hardware rerating and finished September 24 at CNY 120.55. Northbound ownership rose from 2.77% at end-2025 to 6.60% at June 30, so foreign/institutional demand through Stock Connect coincided with the fundamental recovery; that flow is supportive evidence for a rerating, not proof that the valuation is correct.

At the same September 24 closes and CNY 0.86054/HKD, HKD 91.85 converts to about CNY 79.04, so the A share at CNY 120.55 carries an A/H premium of roughly 53%. The A line, about 91% of share capital, remains the appropriate valuation anchor, but an investor who can own H shares is paying materially less for the same consolidated earnings.

H liquidity is poor relative to the A line. After the full over-allotment there are 62.101 million H shares; the 26.941 million cornerstone shares, 43.38% of the line, are locked for roughly six months from listing, leaving only about 35.16 million H shares before other sticky holders. The prospectus names commercial counterparties among the cornerstones, including subsidiaries of customer Sungrow Power Supply and packaging-and-test supplier Tongfu Microelectronics. September 24 volume was only about 82,600 shares and HKD 7.64 million of turnover, and when the cornerstone lock-up ends on December 25, 2026, potentially available supply rises by an amount equal to about 77% of today's non-cornerstone H float.

Southbound Stock Connect access is already in place: Shanghai and Shenzhen Stock Exchange records show 3661.HK was added to both Southbound eligible-security lists effective July 24, 2026, two months before the research cut-off. The thin September 24 turnover therefore already reflects post-inclusion liquidity, so I do not build a Southbound-demand catalyst into the valuation.

Balance sheet after the H issue

The June 30 balance sheet already includes the base H offering, to which the company explicitly attributes the cash increase: cash of CNY 4.630 billion, trading financial assets of CNY 2.072 billion, total assets of CNY 11.769 billion and attributable equity of CNY 9.774 billion.

Interest-bearing obligations were modest: CNY 452 million of short-term loans, CNY 49.9 million of current long-term borrowings and CNY 34.9 million of long-term borrowings. Excluding CNY 45.3 million of restricted cash and adding about CNY 236 million of large time deposits gives roughly CNY 6.35 billion of net liquid financial assets at June 30; the July over-allotment added HKD 681.78 million net, about CNY 0.59 billion at the rate used here, taking the pro-forma figure toward CNY 6.94 billion.

The pre-over-allotment prospectus allocated roughly 60% of H proceeds to R&D over five years, 26% to strategic investments/acquisitions, 6% to overseas sales infrastructure and 8% to working capital. At June 30 the money had only just arrived and was overwhelmingly visible as cash and wealth-management assets rather than operating deployment, and I found no subsequent disclosure through the base date that allows a reliable project-by-project deployment bridge, so valuation treats cash as cash, not as future earnings.

Headline ROE is now a poor comparison measure: the H1 weighted-average 7.61% spans an equity base that changed abruptly at the end of June. A rough operating-capital approach is more revealing: excluding financial assets and using inventories, receivables and operating fixed/intangible assets net of non-interest operating liabilities puts deployed operating capital in the mid-CNY 2 billion range; H1 core operating profit after inventory write-downs annualizes to a return on it in roughly the high-20% area. That is an estimate, but it shows the operating franchise remains far more productive than post-IPO headline ROE suggests.

Cash-flow passthrough

The five-year cash record justifies using cash earnings rather than distrusting the whole income statement: aggregate 2021–25 OCF/attributable-net-income was approximately 1.00x. The problem is maintenance versus growth spending. With no formal maintenance-capex split disclosed, I use depreciation and amortization as a rough proxy; the 2025 reconciliation shows roughly CNY 199 million of it (fixed-asset depreciation of about CNY 94 million plus right-of-use depreciation and intangible and long-term prepaid amortization) against CNY 466 million of OCF, leaving about CNY 268 million of crude 2025 “owner earnings” before growth capital expenditure.

That figure is far below accounting profit and the gap exceeds 30%, so the scenario valuation below uses normalized owner earnings rather than headline P/E. Part of the 2025 gap is temporary working capital; part reflects a business that must keep spending on test infrastructure, labs and SKU expansion. H1 2026 cash conversion is improving, but it is too early to declare the working-capital burden gone.

At the class-adjusted CNY 80.1 billion equity value, subtracting about CNY 7 billion of pro-forma net financial assets leaves an enterprise value of roughly CNY 73 billion, still about 113 times trailing recurring earnings: the cash raise does not make the operating business cheap.

Scenario valuation

The scenarios value 2027 normalized owner earnings on an EV/owner-earnings basis, add CNY 7.0 billion of net financial assets and divide by a conservative 710 million fully diluted share assumption, a count that deliberately absorbs part of future option dilution without crediting all option-exercise proceeds.

This is valuation-scenario analysis within a research framework, not investment advice.

The multiples are intentionally generous for a normalized cyclical analog producer, reflecting the domestic-substitution runway, product breadth and high current growth. The conservative 45x is still not a distressed multiple; the optimistic 65x assumes SG Micro is proving itself a long-duration Chinese analog compounder rather than merely enjoying another upcycle.

The current A price requires something close to the optimistic path. At a 55x normalized multiple, a CNY 73 billion operating enterprise value requires about CNY 1.33 billion of sustainable owner earnings, roughly twice trailing recurring earnings and materially above my base 2027 estimate. The market is pre-spending perhaps two years of successful execution.

The H line gives a different entry point. At CNY 0.86054/HKD the same values translate to roughly HKD 70 conservative, HKD 103 base and HKD 155 optimistic, so the September 24 H close of HKD 91.85 lies around the lower part of the base valuation neighborhood while the A close lies well above it. That difference is why the investment conclusion must specify the share class.

Expectation gap and margin of safety

The expectation gap centers on AI durability. The bullish market reading is that optical modules opened a new high-growth profit pool where SG Micro can sell many auxiliary analog components without competing for the main optical-channel silicon. The bearish reading is that a 25%-of-revenue end market growing near 100% has temporarily exaggerated group growth and will normalize before the automotive and industrial ramps are large enough to replace it. The next two quarters should tell the two apart.

A second gap sits in cash deployment: the market can treat the H proceeds as future R&D and M&A value, but today they are largely cash, and paying a growth multiple for both current operations and undeployed cash would double-count optionality. The acquisition record is too short to assume that the 26% strategic-investment bucket earns SG Micro's organic R&D returns.

The current A price has no discount to the conservative scenario; it trades at roughly twice that CNY 60 value, so margin of safety is zero. Cut the most fragile base assumption, normalized 2027 owner earnings of CNY 1.02 billion, to 70% and base value falls to roughly CNY 65 a share even before compressing the 55x multiple.

Flat earnings for three years would also be unattractive. The FY2025 dividend was CNY 0.20 per pre-H-share dividend-base share, only about a 0.17% cash yield at CNY 120.55; with no earnings growth and no multiple expansion that carry does not clear a sovereign-bond hurdle, and any normalization in a triple-digit earnings multiple would make returns negative. This is a good-company/bad-price setup on the A line.

Margin-of-safety sufficiency verdict: none.

The final conclusion's valuation bands follow directly from these scenario values.

Risks, catalysts and tracking

Risks that can cause permanent loss

The first risk is valuation compression: high probability, high impact, with the multiple relative to growth as the indicator. Trailing recurring earnings support a class-adjusted valuation around 124x, and the A line is even more expensive than the H line on identical corporate economics; if revenue growth falls into the teens before recurring earnings reach CNY 1 billion-plus, a move from today's triple-digit multiple toward 45–60x can overwhelm years of profit growth. The path to permanent loss is multiple compression, not insolvency.

The second is an optical-module and distributor inventory reversal: medium probability, high impact. Network/computing supplies about one quarter of H1 sales and roughly 42% of my estimated incremental revenue, and distributor sell-through is undisclosed. The warning sequence would be network growth below 50%, inventory days above roughly 260, accelerating write-down provisions and gross margin below 48%: revenue would slow first, provisions would hit earnings next, then the AI narrative would re-rate.

The third is technology/supply dependence: medium probability, high impact. TSMC remains the principal foundry. A Taiwan disruption, an export-control problem or failure to qualify a core proprietary analog process at another foundry could constrain supply, while rushing localization at worse yields could lift wafer cost and damage margin. Watch new foundry qualifications, wafer-source disclosures, unexplained gross-margin deterioration and any change in lead times.

The fourth, compensation dilution, has high probability and medium impact. H1 SBC equaled about 24% of attributable profit, and multiple option plans have in-the-money exercise prices far below the CNY 120.55 A price. Dilution alone will not destroy the company, but compounding per-share value becomes much harder if 3–5% potential dilution is repeatedly replenished; watch annual SBC as a percentage of recurring profit and the actual share-count increase through self-exercise windows.

The fifth is poor deployment of the H-share cash: medium probability, medium-to-high impact, because that cash represents nearly CNY 7 billion of downside protection. The ADC/DAC deal booked CNY 51 million of goodwill on a CNY 58 million purchase and produced no revenue by June, and the 2025 感睿 acquisition lost money after purchase. If the 26% acquisition pool is spent on similar assets at scale, cash protection can disappear before operating earnings rise; the tells would be acquisition price-to-revenue, goodwill creation and acquired-business cash profit.

Catalysts

Positive catalysts are time-specific. The Q3 2026 report, due by the end of October under the ChiNext timetable, matters most: revenue growth above 35% with gross margin at or above 51% would materially strengthen the structural-share-gain case. Network/computing holding above 50% year-on-year growth in H2 while industrial and automotive also accelerate would be a second; explicit evidence that H-share proceeds are earning measurable revenue and cash returns in R&D or acquisitions, rather than simply increasing investment income, a third.

Negative catalysts are equally observable. Q3 gross margin below 48% or inventory days above 260 would point toward channel digestion or price competition, and a material rise in inventory reserves beyond the current CNY 353 million without comparable revenue growth would weaken cash quality. The six-month cornerstone lock-up expiry around late December 2026 can increase H-share supply sharply; a widening A/H premium alongside it would be evidence that the A line remains insulated from cheaper price discovery, not of higher intrinsic value.

Tracking dashboard

Current operating figures come from the June 2026 CAS report and August 31 investor communication.

Gross margin, network growth and inventory form the most useful three-variable test: buoyant network demand with stable margin and falling inventory days would validate design wins and end demand, while buoyant reported revenue with falling margin and rising inventory days would look like channel loading. SBC belongs beside them, since reported growth is less valuable when a large portion goes to employees through new shares.

Cross-synthesis and research conclusion

What the vertical and horizontal evidence says together

SG Micro has proved beyond reasonable doubt that it can repeatedly broaden an analog catalog and turn breadth into customer coverage: revenue went from CNY 1.2 billion in 2020 to nearly CNY 3.9 billion in 2025 despite a severe 2023 analog downturn. That downturn also marked the boundary: the 30%-plus shortage-era ROEs were partly cyclical, and sustainable returns are lower.

The 2026 improvement looks different from 2021–22 in a constructive way: network/computing nearly doubling, signal chain growing above 50%, Q2 gross margin near 52% and recurring profit growth far ahead of revenue, without reliance on widespread analog shortages or announced broad price increases. With optical modules appearing to be a real new content pool and industrial breadth helping outside it, my central 60–70% structural estimate is more persuasive than a pure restocking explanation.

Horizontally, the advantage is breadth at Chinese-vendor scale, local design support and a growing high-performance catalog; the weakness against TI, ADI and the multinational automotive franchises is structural, not temporary, because manufacturing depth, decades of qualification history and absolute R&D scale take years to reproduce. The 1.8% estimated China analog share makes the opportunity large but shows that domestic substitution has not already been won.

Pro-forma net financial assets approaching CNY 7 billion remove financing as a near-term constraint and make capital allocation harder: SG Micro must turn cash into engineers, product platforms and acquisitions at returns high enough to justify the dilution from 62.1 million new H shares plus ongoing employee options, and the early acquisition evidence is too small and too mixed to award value in advance.

The A-share valuation is pre-spending success: at CNY 120.55 investors pay about 124 times trailing recurring earnings on a class-adjusted basis and more than 110 times after net cash, against my base 2027 owner-earnings value of about CNY 89. The business has earned a higher quality rating than it had in 2023; the A share has already capitalized most of that upgrade.

Over one year the variables are optical/network growth, gross margin, inventory and SBC; over three years, automotive qualification, high-end signal-chain share and efficient use of H proceeds matter more. Over five years the decisive question is whether SG Micro can become a broad global-quality analog franchise with domestic foundry redundancy while sustaining returns after dilution. If it can, today's optimistic valuation becomes defensible; if it remains a cyclical domestic analog challenger, today's A multiple is too high.

Bull and bear reasons

Core bull reasons:

• H1 2026 recurring profit rose 164.6% on 42.7% revenue growth, while group gross margin rose to 51.8%; the earnings acceleration is operating, not merely non-recurring income.
• Network/computing reached roughly 25% of sales and nearly doubled, with optical modules the main driver; it likely supplied about 42% of incremental H1 revenue.
• Signal chain grew 53.1% at a 57.1% gross margin, improving the economic mix as SG Micro moves toward more difficult, higher-value analog sockets.
• Pro-forma net liquid financial assets near CNY 7 billion create substantial R&D capacity and remove refinancing risk.
• At only about 1.8% of the issuer-commissioned China analog market estimate, successful domestic substitution does not require SG Micro to become remotely dominant.

Core bear reasons:

• The A line is roughly 105x trailing attributable earnings and 124x recurring earnings; a normal growth slowdown can create a large price loss without any deterioration in the underlying franchise.
• About 94% of revenue is distributor-led, yet channel inventory is undisclosed; SG Micro itself holds CNY 1.68 billion of net inventory and CNY 353 million of provisions.
• Share-based compensation was roughly CNY 99 million in H1, about 24% of attributable profit, while multiple option tranches are materially in the money.
• The effective H1 tax rate was only 3.8% because the 10% IC rate and an approximately CNY 37.6 million R&D extra-deduction benefit suppress taxable income; tax normalization is a real earnings sensitivity.
• The H share represents the same consolidated business at roughly CNY 79 equivalent versus CNY 120.55 for the A line. A 53% A/H premium is a warning that part of the A valuation is market segmentation rather than corporate economics.

Pre-mortem: where this could go wrong

By 2027, optical-module capex slows after the 800G/1.6T build, while MPS, TI and Chinese analog competitors cut auxiliary power and signal-chain pricing. Network growth falls from nearly 100% to zero, SG Micro's gross margin falls from 52% to 45%, inventory provisions rise, and owner earnings stay below CNY 700 million. A normalization from a triple-digit earnings multiple to 45x would put the A share around CNY 50–60, more than 50% below today's price.

A second script is capital and supply execution. SG Micro spends part of the H-share acquisition pool on low-revenue analog assets while TSMC-linked processes prove slow or expensive to qualify at domestic foundries. SBC keeps the share count rising, the effective tax rate normalizes above the current 3.8%, and cash protection shrinks before acquired earnings arrive. Even with revenue growth, per-share owner earnings disappoint and the premium multiple compresses.

Final research conclusion

SG Micro is worth following because the 2026 acceleration contains evidence of a real franchise upgrade: optical modules have opened a new profit pool, signal-chain growth and margins are strong, and product breadth is becoming harder for smaller domestic competitors to replicate. The balance sheet can fund several years of aggressive R&D without financial stress.

The A-share price offers poor asymmetry. CNY 120.55 already requires execution near my optimistic case, while distributor inventory, option dilution, TSMC concentration and the deployment of CNY 7 billion of net financial assets remain incompletely observable. The H share is materially more sensibly priced, but its short trading history, thin turnover and late-December cornerstone unlock deserve a separate liquidity discount. I would change my A-share view if recurring owner earnings move toward CNY 1.2–1.3 billion without sacrificing gross margin, or if the price falls far enough that the current execution risks are no longer prepaid.

【Investment rating】

• Rating: Watch
• One-line thesis: H1 proves share and content gains, but the A share already discounts near-optimistic owner earnings while channel inventory and dilution remain opaque.
• Ideal buy price:

【Ideal Buy Price】42–48 CNY Basis: a 20–30% margin of safety below the CNY 60 conservative scenario value, provided gross margin remains at least 49% and revenue growth stays above 20%.

• Acceptable hold price: CNY 78–100, within roughly ±15% of the CNY 89 base-case value.
• Clearly overvalued price: CNY 146–160, beginning about 10% above the CNY 133 optimistic scenario value.
• Current-price classification: outside the three bands; above the acceptable-hold range but below the clearly-overvalued range.
• Whether to wait for a better price: yes. For the A line, I would require CNY 48 or below plus gross margin of at least 49%, OCF/net profit above 0.8x and no evidence of network/computing growth collapsing below 25%.
• Target holding horizon and expected annualized return: 3–5 years; conservative about -10% to -15%, base roughly 4–6%, optimistic about 14–17% annualized, assuming eventual convergence toward the valuation regimes used above.
• Max-loss risk: a combination of optical-module normalization, gross margin near 45% and multiple compression toward 45x can plausibly produce a 50–60% A-share loss without threatening solvency.
• Reassessment-trigger signal: group gross margin below 48%.
• Reassessment-trigger signal: network/computing growth below 50% while inventory continues rising.
• Reassessment-trigger signal: inventory days above 260.
• Reassessment-trigger signal: trailing OCF/net profit below 0.75x.
• Reassessment-trigger signal: SBC remains above 25% of attributable profit for another twelve months.

For the H line, the same CNY bands translate at CNY 0.86054/HKD to an ideal-buy area of roughly HKD 49–56, an acceptable-hold area of roughly HKD 91–116 and a clearly-overvalued area beginning around HKD 170. The September 24 H close of HKD 91.85 is therefore already within the acceptable-hold band, near its lower edge, on identical corporate fundamentals; this does not alter the report's A-anchored Watch rating.

【Valuation Range】

• current: 120.55 CNY (A-share close as of 2026-09-24)
• bear (conservative · ideal buy zone): [42, 48]
• base (fair · acceptable hold zone): [78, 100]
• bull (optimistic · above the clearly-overvalued line): [146, 160]

Research uncertainties and key sources

The largest operating blind spot is distributor sell-through. Management comments on inventory and the company details own-stock provisions, but neither distributor inventory nor end-customer sell-through is disclosed; the 60–70% structural versus 30–40% cyclical growth split is therefore my inference, not company guidance.

The second is wafer concentration. Primary disclosures identify TSMC as the principal long-term foundry and name SMIC subsidiaries and DB HiTek as alternatives, but I could not verify an exact TSMC purchase share, specific Taiwan fabs or cross-foundry yield/cost data.

The third is the issuer-commissioned Frost & Sullivan ranking, which anonymizes major competitors; SG Micro's CAS Mainland/Hong Kong split does not cleanly reconcile to F&S's CNY 3.9 billion “China” revenue attribution, so I use the rankings only as evidence of approximate scale.

The fourth is share-count verification. I could not fully reconcile the September 24 A-share total from an exchange monthly return after every self-exercised option, so I use the 623.990 million A-share and 62.101 million H-share base, cross-checked against the company's June share table, July over-allotment and September incentive announcements.

The fifth is peer-market valuation precision. Proprietary peer-comparison reports were not available as source documents in this research session, and I would not substitute stale or mismatched peer multiples, so the competitive conclusions rely more heavily on primary SG Micro disclosures, the H-share prospectus framework and the disclosed current operating economics than on a September 24 peer-multiple league table.

The main primary sources are SG Micro's 2025 CAS annual report, published March 28, 2026; its 2026 H1 CAS report, published August 29; the August 31 investor-relations record; the 2022 annual report for full-cycle history; and the 2026 incentive-plan and over-allotment disclosures. The H1 filing is particularly useful because it reconciles product margins, inventory provisions, tax benefits, share-based compensation, financing proceeds and litigation within one CAS framework.

Other tickers mentioned

• TXN.US: Texas Instruments is the scale, manufacturing and broad-catalog benchmark for analog IC competition.
• ADI.US: Analog Devices is the high-performance signal-chain reference and a key competitor in industrial and communications analog.
• MPWR.US: Monolithic Power Systems is the closest global reference for high-growth fabless power management and AI-infrastructure content.
• 688536.SHG: 3PEAK is a direct Chinese signal-chain and power-management challenger.
• 688052.SHG: Novosense competes in automotive, industrial, sensor and signal-chain applications.
• 688484.SHG: Southchip is a Chinese power-management and charging specialist.
• 688798.SHG: Awinic competes in consumer analog, audio and mixed-signal niches.
• 688141.SHG: Joulwatt is a domestic power-management competitor.
• 6415.TW: Silergy is a major Chinese/Taiwan-listed power-management comparator.
• 2330.TW: TSMC is SG Micro's principal long-standing wafer-foundry partner.
• 0981.HK: SMIC subsidiaries provide alternative domestic wafer-manufacturing capacity.
• 600584.SHG: JCET is a major outsourced packaging and testing supplier.
• 002156.SHE: Tongfu Microelectronics is both a packaging/test supplier and commercially connected cornerstone-investor group.
• 300274.SHE: Sungrow Power is identified in the H-share offering materials as a customer-linked cornerstone investor.
• NXPI.US: NXP is a global analog/automotive competitor used to frame the difficulty of high-reliability automotive substitution.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.