VeriSilicon(688521) · Chip IP

VeriSilicon From a Long-Term Owner's Perspective

Other languages
Quick ReadPlain-language overview · read this first

VeriSilicon is a company built on its own semiconductor IP, providing customers with one-stop chip customization and IP licensing. Up front, it collects design and licensing fees by project; after customers enter mass production, it receives royalties based on shipments. In essence, this is a hybrid business of “project work + platform reuse + a small stream of ongoing royalties,” with customers spanning chip design houses, IDMs, system vendors, internet and cloud companies, and automakers.

The rating is Avoid: not because the company is poor, but because this is a strong track, a scarce company, and a bad price. By 2024 metrics, it ranked first in China’s domestic IP licensing market and eighth globally, making it one of the few local players with world-class comparability. The problem is that it has not yet proven itself to be a machine capable of consistently producing real free cash flow, while the current share price has already pulled forward and priced in a very long-dated, very optimistic earnings scenario.

The facts supporting this judgment are hard to ignore: large losses for three consecutive years from 2023–2025; 2025 net profit attributable to the parent of RMB -528 million; and persistently weak operating cash flow and free cash flow. Based on roughly RMB 240 per share and 525.9 million shares, the market capitalization is about RMB 126.4 billion, implying a price-to-sales ratio of about 40x, while the upper end of the optimistic intrinsic value range offered by the analyst is only RMB 150. There is almost no margin of safety, and a permanent drawdown of 50%–80% in the worst case would not be an exaggeration. The ideal buying range is RMB 40–80; before real cash flow is proven, investors should not pay dream prices.

Lead

A good market, a scarce company, and a bad price. VeriSilicon signed RMB 5.960 billion of new orders in 2025, more than doubling year on year, but it has lost money for three straight years and has long had negative free cash flow. Rating Avoid: at roughly RMB 240 per share, a 40x P/S ratio and 37-40x P/B ratio already discount the most optimistic scenario for the next decade, leaving almost no margin of safety.

Full report

Prices in the article are as of publication; see the valuation band above for the live price.

Conclusion First

As of the intraday public market data I found for 2026-06-04, VeriSilicon trades on the STAR Market of the Shanghai Stock Exchange, with public quote pages showing a price range of roughly RMB 235.40 to RMB 245.01 per share. To reduce intraday noise, the valuation below uses RMB 240 per share as an approximation. Based on the roughly 525.9 million shares disclosed in the Q1 2026 report, this implies a market capitalization of about RMB 126.0 billion to RMB 129.0 billion.

My preliminary conclusion is: for a long-term value investor with a horizon of more than 10 years and a balanced but conservative stance, VeriSilicon is better treated as an "avoid" at the current price, rather than a buy. The reason is not that the company lacks strategic value. It operates in a market with sound long-term demand, and it is scarce in China's domestic IP licensing and ASIC customization fields. The real problem is that the company still has not proven that it is a machine capable of consistently producing real free cash flow, while the current share price has already pulled very distant and very optimistic profit scenarios into today's valuation. In other words, what you are buying now looks more like a long-dated option on high growth in AI ASIC orders than a validated cash-flow asset.

Laying out the judgments: the investment rating is Avoid, and the current price offers no margin of safety. I score business understandability at 4/5, industry attractiveness at 3/5, moat strength at 2/5, and management and capital allocation at 2/5. In terms of investor fit, it suits sector/growth/theme investors more than traditional conservative value investors. The largest uncertainties are concentrated in three areas: whether orders can smoothly convert into high-quality revenue and cash flow; whether margins can improve meaningfully after volume-production business scales; and whether subsequent financing and M&A will continue to dilute shareholder value.

Sources: the company's listing venue and market quote pages, 2025 annual report summary, Q1 2026 report, and public financing and H-share listing announcements.

If the question is reframed as "if the stock market closed for 5 years today, would I be willing to buy this business at the current price," my answer is: I would be willing to study the business, but I would not be willing to buy it at the current price. As a business, it has technology, customers, and strategic position. As a stock, its price has run far ahead of operating cash flow.

Business and Industry

How This Company Actually Makes Money

VeriSilicon's core business model is not complicated: it is essentially a company that relies on its proprietary semiconductor IP to provide customers with one-stop chip customization services and semiconductor IP licensing services. In its 2025 annual report summary and prior annual reports, the company divides its business into two major segments. The first is one-stop chip customization, internally divided into chip design services and chip volume-production services. The second is semiconductor IP licensing, including upfront IP license fees and royalties after customers enter volume production.

The charging model is also clear. IP licensing usually collects a front-end license fee during the customer design phase, then collects royalties according to the contract after the customer ultimately mass-produces and sells chips. Chip customization provides all or part of the services from specification definition, IP selection, design, and verification to wafer manufacturing and packaging/testing management, and charges fees by project stage and volume-production business. In other words, this is neither a typical SaaS business nor a stable utility. It is a hybrid model of "project-based work + platform reuse + a small amount of recurring royalties."

Customer types are broad. According to the company's disclosures over the years, major customers include chip design companies, IDMs, system vendors, large internet companies, cloud service providers, and automakers. By 2025, roughly 40% of the company's revenue came from system vendors, internet companies, cloud service providers, and automaker customers. In Q1 2026, revenue from customer groups such as system vendors, large internet companies, cloud service providers, and automakers accounted for 33.75% of revenue, up 94.10% year on year. This indicates that the company is penetrating from traditional chip-company customers into higher-value "system-level customers."

Revenue has average repeatability, medium stability, and visibly improved predictability over the past two years, but it is still far from excellent. The reason is that IP licensing revenue is naturally project-based, and order signing and delivery cadence is uneven. Royalties have some continuity, but their share is limited. Volume-production business depends on customers' chip shipments; once it scales, it resembles a manufacturing-chain service, which can bring higher revenue but does not naturally bring high gross margins. In 2025, the company signed RMB 5.960 billion of new orders, up 103.41% year on year, and ended the year with RMB 5.075 billion of backlog. By the end of Q1 2026, backlog had further increased to RMB 5.133 billion, and new orders signed year to date as of 2026-04-29 had risen to RMB 8.240 billion. This materially improves short- to medium-term revenue visibility, but "high orders" do not automatically mean "high free cash flow."

The cost structure is the key to understanding VeriSilicon. On one hand, it is an IP/design platform company with very high R&D spending. R&D investment was 42.78% of revenue in 2025. In Q1 2026, R&D investment was RMB 485 million, or 58.04% of revenue. On the other hand, volume-production business brings more outsourced wafer, packaging/testing, and supply-chain management costs. The result is that when revenue grows, gross margin does not necessarily rise in tandem; profit improvement depends heavily on scale effects and expense absorption. In Q1 2026, the company's consolidated gross margin was 32.29%, down 6.76 percentage points year on year, which the company itself mainly attributed to changes in revenue mix.

This business is not free of dependencies. First, customer concentration is relatively high. Sales to the top five customers in 2025 were RMB 1,434.341 million, accounting for 45.51% of annual sales; in 2023, the top five customers also accounted for 46.48% of sales. Second, volume-production business naturally depends on foundries, packaging/testing houses, and availability of advanced process nodes. Third, the business is highly tied to advanced processes, IP roadmaps, and technology evolution such as RISC-V and AI, and is exposed to geopolitics and supply-chain volatility. Supplier concentration was not fully extracted with official page references from the materials reviewed this time, so I leave it as an item requiring supplementary information.

If the only question is "is this a business I can understand," my answer is yes. It is essentially a company that "builds chips for others, sells reusable design modules to others, and tries to turn design experience into a platform." The business logic is clear, and the revenue drivers are understandable. The issue is not that it is incomprehensible. The issue is that once understood, it is clear that the company is not yet the kind of excellent business whose price can be ignored.

Industry and Competitive Landscape

At the industry level, VeriSilicon operates in the semiconductor design service and IP licensing market, which has long-term growth but still carries short- to medium-term cyclicality. In its 2025 annual report summary, the company cites IBS and industry association data showing that the number of Chinese chip design companies increased from 736 in 2015 to 3901 in 2025. China's semiconductor market was about US$274.4 billion in 2025 and is expected to reach about US$373.7 billion by 2030. This shows that the industry's broad direction remains upward, especially driven by applications such as AI, big data, cloud computing, 5G, IoT, smart vehicles, and edge computing.

But this is not an easy good industry. First, it is vulnerable to technology substitution, process iteration, geopolitical restrictions, and changes in customers' in-house development strategies. Second, the industry's profit pool is unevenly distributed. The truly fat profits often stay with architecture-monopoly IP vendors, key EDA vendors, advanced-process leaders, and end-product definers. Design services and project-based customization have growth, but turning revenue into high-quality profit is not easy. VeriSilicon's Q1 2026 report illustrates this directly: revenue doubled, yet net loss still widened to RMB 341 million.

In terms of industry position, VeriSilicon is indeed a relatively scarce domestic listed company. In public investor communications and annual reports, the company has repeatedly cited IPnest statistics: based on 2024 data, VeriSilicon's semiconductor IP licensing business ranked No. 1 in mainland China and No. 8 globally; its IP license-fee revenue ranked No. 6 globally; and by IP category breadth, it ranked in the top two among the global top ten IP companies. This shows it is not an ordinary follower. It is one of the few domestic players with a degree of world-class comparability.

Competitors should be viewed in two groups. In IP licensing, comparable companies disclosed in IPO feedback materials include ARM, CEVA, Faraday, Global Unichip, and Alchip; research reports also list Synopsys, Cadence, and Imagination as key comparables. In design services, more direct comparables are Global Unichip, Faraday, Alchip, and Brite Semiconductor. The most important conclusion here is not that "VeriSilicon has no competitors." It is that VeriSilicon faces rivals with greater scale, more profit, stronger ecosystems, or sharper niche focus.

Therefore, my assessment of industry attractiveness is: the market is good, but profit distribution is not gentle; VeriSilicon is a scarce company in a good market, yet it is not yet a proven high-return company. If classified, it is more like a "company in a good industry whose economics have not yet been validated" than a "good company in a bad industry." I score industry attractiveness at 3/5.

Moat and Management

Moat Assessment

VeriSilicon has a moat, but at present it looks more like a narrow moat than a wide moat.

From the perspective of brand and technical credibility, the company has six categories of controllable proprietary processor IP: GPU, NPU, VPU, DSP, ISP, and display processors. Its website also discloses more than 1,700 mixed-signal, RF, and interface IP blocks. Its design capability spans traditional CMOS, FinFET, and FD-SOI, covering successful tape-out experience in 14nm/10nm/7nm/6nm/5nm/4nm FinFET and 28nm/22nm FD-SOI. For large customers, this kind of proven, reusable, cross-process-node engineering capability is itself a reputational asset.

From the perspective of scale advantages and switching costs, the company has some advantages, but they are far from insurmountable. Once a customer builds a chip based on VeriSilicon's IP and platform and enters volume production, switching costs rise for subsequent cooperation, especially in highly complex projects such as advanced processes, automotive-grade chips, and AI ASICs. This is also why the company can form a business chain of front-end licensing + back-end royalties + customization extensions. Meanwhile, system-level customers accounted for roughly 40% of 2025 revenue and have stayed above 30% for five consecutive years, showing that the company is becoming more deeply embedded in complex customer needs.

From the perspective of network effects, channel advantages, and data advantages, VeriSilicon is not strong. It is not an operating system, not a developer-platform monopoly, and not a company controlling massive proprietary data. RISC-V ecosystem building, such as the company's role in leading the establishment of the China RISC-V Industry Alliance and the alliance's 204 members as of the end of 2025, can strengthen its industry influence and ecosystem position. But this is closer to "soft influence" and "participation rights in a technology ecosystem," and is not enough to constitute a strong network effect like Arm's.

From the perspective of cost advantages and pricing power, I consider it weak. Scaling volume-production business can lift revenue, but it can also drag down consolidated gross margin. The year-on-year decline in Q1 2026 consolidated gross margin already shows that the company does not yet have pricing power strong enough to offset business-mix changes. In other words, it can win orders and deliver projects, but it may not be able to raise prices comfortably.

From the perspective of regulatory barriers, patents, and certifications, VeriSilicon has certain thresholds. Its website discloses that its NPU IP and display processing IP have passed ISO 26262 certification, and it continues to invest in automotive-grade and advanced packaging directions. Achieving this requires years of R&D accumulation, customer validation, and engineering organization capability. For new domestic entrants, replicating "a smaller VeriSilicon" may require more than 3 to 5 years and several hundred million to more than RMB 1 billion of R&D investment. Replicating "a globalized VeriSilicon with IP breadth and the ability to serve top customers" is harder. I emphasize that this is an inference based on IP categories, R&D intensity, process-node coverage, and customer validation thresholds, not a figure explicitly stated by the company.

Overall, VeriSilicon's moat is closer to stable with some improvement than clearly widening. The reason is that AI ASICs, RISC-V, and penetration into system-level customers are indeed strengthening its strategic position. But global competitors are too strong, customer bargaining power is not weak, and VeriSilicon's own cash flow and profitability have not yet "crystallized" these technical advantages into a thick moat. I score moat strength at 2/5.

Management and Capital Allocation

Management's strengths are that it understands technology, understands customers, and invests in R&D for the long term. Management's problem is that it looks too much like growth-oriented entrepreneurs and too little like capital allocators who are extremely strict about per-share intrinsic value.

The company's chairman and president is Dai Weimin. Public information shows the company has no actual controller. As of the end of 2025, the largest shareholder, VeriSilicon Limited, held 11.39%, and Dai Weimin personally held 1.66%; the two are parties acting in concert. According to an earlier equity-change announcement, the combined shareholding ratio after acting in concert was about 13.058%. This means the founder does have some "skin in the game," but equity control is not strong, and this is not a highly owner-operator structure.

On information disclosure and financial audit, I have not seen obvious red flags. The 2021 annual report, 2024 annual report summary, and 2025 annual report summary all show that Deloitte Touche Tohmatsu CPA issued a standard unqualified audit opinion. This does not prove the company is excellent, but it at least indicates that no major audit-level abnormality is currently visible.

The truly debatable area is the capital allocation record. The company raised net proceeds of about RMB 1.678 billion in its 2020 IPO, completed a private placement raising RMB 1.807 billion in July 2025, and advanced an H-share listing application in 2026. The use of proceeds includes R&D for key technologies and main services, global marketing ecosystem, strategic investments or acquisitions, and working-capital replenishment. For a company with long-term negative free cash flow, this logic is fully reasonable in the language of growth companies. But from a long-term owner's perspective, it means that the company's own cash "blood-making" is still insufficient to support its technology ambitions, and it needs to keep relying on external capital.

On buybacks, the company repurchased 287,000 shares through centralized bidding in 2025, for a total of RMB 24.8319 million, at a buyback price range of RMB 85.80-87.00 per share, for employee stock ownership plans or equity incentives. The buyback itself was rational because the buyback price was materially lower than today's market price. But in scale, it was too small, far from enough to offset share expansion from prior and potentially future financing.

The change in share count also illustrates this. Share capital was about 500.4 million shares at the end of 2024 and increased to 525.9 million shares by the end of 2025, up about 5.1% in one year. By the end of Q1 2026, it had further increased to more than 525.9 million shares. Comparing total share capital after the IPO with Q1 2026, cumulative expansion was close to 9%. For a company without stable positive free cash flow, this dilution should not be brushed aside.

More caution is needed because the company also disclosed non-GAAP metrics such as "adjusted net profit attributable to the parent" and "adjusted EBITDA" in its Q1 2026 report, and noted that major adjustment items include share-based payment expenses and non-operating or non-cash expenses arising from acquisitions. These metrics have reference value, but for value investors, what truly matters is GAAP profit, operating cash flow, and free cash flow per share, not how smooth the adjusted story looks.

Therefore, my judgment on management is: no obvious integrity issue is visible for now, and strategic and technical judgment is not weak, but capital allocation leans more toward a growth-company paradigm of "keep investing, keep expanding, keep financing" than a cash-flow paradigm of "improve per-share intrinsic value." I score management and capital allocation at 2/5.

Financial Quality

Key Financial Metrics

The table below uses verifiable data from the company's periodic reports and full annual reports as much as possible. Unless otherwise stated, the basis is consolidated, and monetary amounts are in RMB 100 million. Here, FCF = net operating cash flow - cash paid for purchase and construction of fixed assets, intangible assets, and other long-term assets. The 2025 cash outflow item was not fully extracted from the materials reviewed this time, so FCF is temporarily marked "unknown." ROE, ROA, and asset-liability ratio are rough estimates based on disclosed net profit, net assets, and total assets, for directional analysis only and not a substitute for Wind/Choice standard calculations.

Year Revenue Revenue Growth Net Profit Attributable to Parent Net Margin Operating Cash Flow Capex Free Cash Flow Rough ROE Rough ROA Rough Asset-Liability Ratio
2020 15.06 N/A -0.26 -1.7% -1.25 1.41 -2.67 -1.0% -0.8% 17.8%
2021 21.39 42.0% 0.13 0.6% 1.55 1.36 0.19 0.5% 0.3% 29.5%
2022 26.79 25.2% 0.74 2.8% -3.29 8.44 -11.73 2.5% 1.7% 34.3%
2023 23.38 -12.7% -2.96 -12.7% -0.09 1.15 -1.24 -11.0% -6.7% 38.7%
2024 23.22 -0.7% -6.01 -25.9% -3.46 1.44 -4.90 -28.3% -13.0% 54.2%
2025 31.52 35.8% -5.28 -16.7% -2.22 Unknown Unknown -15.4% -6.8% 55.7%

Sources: 2021 annual report, 2023/2024/2025 annual report summaries, and cash-flow statements in the 2023/2024 full annual reports. Growth rates, ROE, ROA, and asset-liability ratios in the table are calculated from disclosed data.

This table has three very important implications.

First, revenue growth has not steadily converted into profit and cash flow. From 2020 to 2025, the company's revenue grew from RMB 1.506 billion to RMB 3.152 billion, a five-year compound growth rate of about 15.9%. But during the same period, it posted large losses for three consecutive years from 2023 to 2025, and cumulative net profit attributable to the parent for 2020-2025 was about RMB -1.364 billion. This model of "high growth but no cash output" is not rare in early-stage technology platform companies, but value investors must apply a discount to it, not a premium.

Second, operating cash-flow quality is unstable, and at times even weaker than accounting profit. In 2021 and 2022, the company reported small profits, but 2022 net operating cash flow was RMB -329 million. In 2023-2025, it was RMB -9 million, RMB -346 million, and RMB -222 million, respectively. This shows that profit does not always turn into cash in a timely way, and working capital and project cadence have a large impact on cash flow.

Third, free cash flow is even weaker. 2021 was one of the few recent years that can barely be called "positive FCF," with FCF of only RMB 19 million. In 2022, due to a sharp rise in long-term asset investment, FCF was about RMB -1.173 billion. In 2023 and 2024, it was about RMB -124 million and RMB -490 million, respectively. Even if 2022 is treated as an abnormal year, the overall conclusion is clear: VeriSilicon has not yet become a machine that can steadily generate cash returns for shareholders.

Cash Flow, Working Capital, and Balance Sheet

Looking more closely at the balance sheet: at the end of 2024, the company had RMB 747 million of cash and cash equivalents, RMB 112 million of trading financial assets, RMB 943 million of accounts receivable, RMB 396 million of inventory, RMB 245 million of contract assets, and RMB 607 million of contract liabilities. Accounts payable were RMB 163 million, long-term borrowings were RMB 833 million, and non-current liabilities due within one year were RMB 434 million. By the end of 2025, based on the opening figures in the Q1 2026 report, cash and cash equivalents had increased to RMB 2.008 billion, trading financial assets to RMB 924 million, accounts receivable to RMB 1.203 billion, inventory to RMB 498 million, prepayments to RMB 467 million, and contract liabilities to RMB 1.258 billion. By the end of Q1 2026, contract liabilities had risen further to RMB 1.760 billion, but cash and cash equivalents had fallen back to RMB 1.242 billion, while prepayments rose further to RMB 885 million.

This set of data tells me two things.

First, the rise in orders and contract liabilities does improve business visibility. This is also one of the core reasons the market is willing to assign a high valuation. In particular, the Q1 2026 report explicitly says that more than 90% of backlog is expected to convert within one year, and volume-production business orders exceed RMB 3.0 billion. If these orders convert smoothly, revenue and profit may both step up.

Second, working capital still consumes a heavy amount of cash. Accounts receivable, inventory, contract assets, and prepayments all tie up funds when viewed together. The surge in prepayments from the end of 2025 to Q1 2026 is especially conspicuous; this usually means volume-production business expansion, front-loaded supply-chain investment, or acquisition factors are increasing cash pressure. Therefore, looking only at revenue and orders can easily lead to optimism. But when cash flow and working capital are viewed together, this remains a "long-cycle, heavy upfront-investment" business.

On leverage, the company's short-term survival ability is not the main contradiction for now, but asset quality and financial flexibility are declining. At the end of 2024, the company had no short-term borrowings. In Q1 2026, short-term borrowings appeared at RMB 26 million, long-term borrowings were RMB 1.164 billion, and non-current liabilities due within one year were RMB 589 million. Because Q1 2026 EBIT was negative, interest coverage and net debt/EBITDA metrics are not meaningful. At the current stage, the real issue to watch is whether "cash + trading financial assets" can cover debt and M&A payment pressure, and whether contract liabilities/backlog can continue to convert.

On accounting quality, I have not seen obvious signs of fraud, but two points require conservative treatment. First, the Q1 2026 report clearly notes that the fair-value assessment of assets and liabilities acquired through business combinations not under common control has not yet been completed, so future purchase-price allocation may be adjusted. Second, in Q1, goodwill rose to RMB 629 million, intangible assets rose to RMB 898 million, and development expenditure was RMB 82 million, creating future amortization and impairment risk. For value investors, this means: do not treat book net assets as a solid safety cushion.

Therefore, my overall judgment on financial quality is: there is no obvious sign of financial fraud, but the operating reality shown in the financial statements is not attractive: high revenue growth, weak cash flow, heavy working capital, non-light capital expenditure, and no closed loop of shareholder returns yet.

Owner Earnings and Valuation

Owner Earnings Analysis

Viewed through owner earnings, VeriSilicon looks colder than it does through revenue.

In 2025, net profit attributable to the parent was RMB -528 million, and net operating cash flow was RMB -222 million. If we follow a Buffett-style approach and avoid overcomplicating depreciation and amortization, share-based payments, and acquisition fair-value adjustments, instead using operating cash flow minus maintenance capital expenditure as a conservative version, then 2025 owner earnings would at least not be higher than operating cash flow. Considering that the company's cash paid for long-term assets in 2021, 2023, and 2024 was mostly in the RMB 115 million to RMB 144 million range, a conservative maintenance capex estimate for 2025 can start at RMB 120 million to RMB 150 million. This implies conservative 2025 Owner Earnings of roughly RMB -340 million to RMB -370 million.

This means two things. First, the current valuation cannot be normally explained by "how many times owner earnings," because Owner Earnings under a conservative basis are still negative. Second, buying VeriSilicon now is not essentially buying "today's real cash-earning power," but buying the probability that several years later orders convert, expenses are absorbed, margins turn positive, and margins expand. Growth investors can accept this. For conservative value investors, it is not an ideal target.

Three Valuation Methods

Owner Earnings Discount Method

I will give the most important conclusion first: a DCF based purely on Owner Earnings shows no margin of safety for VeriSilicon at the current price.

I used three scenarios, aiming not for numbers that "look beautiful," but for laying out optimistic assumptions. The common assumptions are: do not linearly extrapolate the explosive order growth in Q1 2026; use a 10%-12% discount-rate range; use a 2.5%-3.5% terminal growth rate; and assume the company does indeed transition from current negative owner earnings to positive owner earnings within 10 years. The scenario differences mainly lie in revenue growth speed, the final level of Owner Earnings margin, and the degree to which scale effects materialize. These assumptions are all based on 2025-2026 orders, backlog, volume-production business mix, and R&D intensity.

Scenario Core Assumptions DCF Point Estimate Wider Intrinsic Value Range I Assign
Conservative Orders convert well, but long-term margins only recover to low- to mid-single digits; R&D remains high; cash-flow improvement is slow About RMB 6/share RMB 10-25/share
Base Revenue grows quickly from 2026 to 2030, and Owner Earnings margin gradually rises to high single digits About RMB 18/share RMB 25-60/share
Bull AI ASIC scales smoothly, volume-production business forms scale effects, and long-term Owner Earnings margin reaches double digits About RMB 50/share RMB 90-150/share

The "wider intrinsic value range" is higher than the mechanical DCF point values because I add some upward adjustment for industry scarcity, the AI order option, and the value of a domestic IP platform. Even so, at around RMB 240/share, the current share price remains clearly above the upper end of my optimistic range. In other words, the market is already paying for outcomes far beyond the bull scenario.

Relative Valuation Method

Using approximately RMB 240 per share and total share capital of about 525.9 million shares, VeriSilicon's current market capitalization is about RMB 126.4 billion. Relative to 2025 revenue of RMB 3.152 billion, the P/S is about 40x. Relative to net assets attributable to the parent of RMB 3.418 billion at the end of 2025, P/B is about 37x. If based on parent-company equity of RMB 3.147 billion at the end of Q1 2026, P/B is even close to 40x. Because the company is still loss-making, PE, EV/EBITDA, P/FCF, and similar metrics are either negative or have no practical explanatory power.

Compared with peers, this valuation looks even more aggressive. Public pages show that Brite Semiconductor's current P/B is about 9.76x. Global Unichip's current market value is about NT$623.8 billion, with a trailing P/E of about 166x; it is an ASIC service company with stronger profitability, larger scale, and a more mature business model. Arm's current market value is roughly in the US$367.0 billion to US$377.0 billion range, and its CPU architecture IP moat is materially wider than VeriSilicon's. In other words, VeriSilicon's valuation is no longer merely benchmarking domestic peers; it is benchmarking the imagination attached to the world's highest-quality IP/design platform companies.

Therefore, the conclusion from relative valuation is not "peers are all expensive, so it is cheap." It is the opposite: even within the high-valuation semiconductor growth-stock framework, VeriSilicon currently belongs to the group whose future profit-realization requirements have been set very high.

Asset or Liquidation Value Method

The asset method provides almost no positive support for VeriSilicon. Based on net assets attributable to the parent of RMB 3.418 billion at the end of 2025, book value per share was about RMB 6.50. Based on parent-company equity of RMB 3.147 billion at the end of Q1 2026, book value per share was about RMB 5.98. If goodwill, intangible assets, and development expenditure in the Q1 report are further deducted, tangible net assets would be even lower. This shows that the relationship between current market capitalization and book asset value is not "asset-supported," but "priced on long-term future earning power."

So, combining the three methods, my range assessment is as follows:

Judgment Range
Conservative intrinsic value RMB 10-25/share
Reasonable intrinsic value RMB 25-60/share
Optimistic intrinsic value RMB 90-150/share
Ideal buy price RMB 40-80/share
Acceptable hold price RMB 80-120/share
Clearly overvalued price Above RMB 180/share

At RMB 240 per share, the current price still represents roughly a 60% premium to the upper end of my optimistic intrinsic value range; relative to the midpoint of the reasonable range, the premium is even larger. For a conservative investor, this almost automatically means no margin of safety.

Margin of Safety

On margin of safety, the conclusion is very direct: insufficient, and close to none.

The current valuation rests on three especially fragile assumptions. First, the market assumes orders will convert into revenue at a high ratio and on schedule, and revenue will further convert into profit and cash flow. But VeriSilicon's history has repeatedly shown that this chain is not automatic. Second, the market assumes the scale effects of volume-production business will ultimately cover high R&D expenses and project volatility. But as of Q1 2026, the company still had not proven this with GAAP profit. Third, the market assumes subsequent financing, M&A, and incentives will not materially erode per-share value. But over the past few years, the company has already gone through an IPO, a private placement, and is advancing an H-share listing.

If future growth falls short of expectations, the current balance sheet can hardly provide valuation support. If margins fall or remain negative for too long, the "long-term owner earnings multiple" implied by the current price will look very difficult. If valuation multiples contract, then even if revenue continues to grow, the stock may still suffer serious permanent capital loss. This is the classic case of a good story, good market, but bad price.

Risks, Counterarguments, and Alternatives

The most important risks are not short-term volatility, but the following factors that can create permanent capital loss.

On competitive risk, VeriSilicon is not growing in a vacuum. On the IP side, competitors include Arm, CEVA, Synopsys, Cadence, and Imagination. On the design-service side, competitors include Global Unichip, Faraday, Alchip, Brite Semiconductor, and others. VeriSilicon is scarce, but it is certainly not irreplaceable.

On technology substitution risk, the value of semiconductor IP and ASIC platforms depends heavily on processes, architectures, software ecosystems, and the direction of end demand. Today the market favors AI compute and custom chips. If customers' in-house development capabilities improve tomorrow, standardized solutions mature, or a certain architecture becomes locked in by a stronger ecosystem, VeriSilicon's pricing power may be compressed.

On regulatory and supply-chain risk, the company's business is naturally exposed to advanced-process supply, packaging/testing capacity, cross-border technology restrictions, and changes in the geographic distribution of downstream customers. The company itself also includes factors such as geopolitics, international conditions, and industrial transfer in its industry description in annual report summaries.

On financial and capital allocation risk, the company posted losses for three consecutive years from 2023 to 2025, cumulative operating cash flow was negative from 2020 to 2025, and it has already gone through a large private placement and is advancing an H-share listing. If it continues to rely on external financing, even a strong growth story may be offset by per-share dilution.

On customer concentration risk, the top five customers accounted for 45.51% of revenue in 2025. This means the delay, cancellation, or below-expectation volume production of a single large project can materially disturb annual profit.

On M&A and accounting risk, the Q1 2026 report already showed a visible rise in intangible assets and goodwill, and noted that purchase-price allocation had not yet been completed. If acquisition integration goes poorly, future risks include amortization, impairment, and even strategic drift.

The strongest counterargument is simple: this may not be a "value investment" at all, but a "high-valuation sector bet." Bulls see "China's scarce domestic IP + ASIC platform," "explosive AI orders," "upgrading system-level customers," and "RISC-V ecosystem position." Bears see "multiple years of negative free cash flow," "an unproven profit model," "continued financing and potential dilution," and "a current price that has already discounted a decade of good news." Neither group is seeing false facts. The difference is whether you are willing to bear huge execution risk at today's price.

What facts would make me admit that my judgment needs to be raised, or even that I was wrong? Mainly three categories. First, operating cash flow and free cash flow turn sustainably positive in 2026-2027 without reliance on continued large-scale financing. Second, after volume-production business scales, net profit and ROIC improve meaningfully under GAAP, rather than only "adjusted profit" looking better. Third, order growth is not a flash in the pan, but can steadily convert into higher earnings per share and cash flow per share. Conversely, if order growth slows, contract liabilities fall, gross margin and expense ratios fail to improve, refinancing continues to increase, or goodwill impairment appears, I would be more convinced that the current price is unacceptable.

On alternatives, VeriSilicon is not clearly superior to indices for conservative capital. The CSI 300 was around 4,844 points near 2026-06-02, representing a basket of large-cap Chinese earning assets. The S&P 500 closed at 7,553.68 points on 2026-06-03. China's 10-year government bond yield was about 1.71% on 2026-06-03. Compared with these alternatives, VeriSilicon must prove that it deserves your capital by delivering a long-term return far above the 1.71% risk-free yield and materially outperforming broad indices. At the current price, I do not think it offers that level of compensation.

If I could hold only 5 assets, VeriSilicon currently would not qualify for the portfolio. Not because it lacks potential, but because at the current price, the requirements it places on profit realization, capital expenditure, free cash flow, and valuation maintenance are too high, leaving too little room for error.

Checklist and Final Investment Conclusion

Investment Checklist

Checklist Item Conclusion Brief Judgment
Can I understand this business? Pass Clear business model: IP licensing + chip customization
Does it have long-term stable demand? Pass Long-term demand exists in AI, data processing, automobiles, and IoT
Does it have a durable moat? Uncertain It has technical and platform barriers, but the moat is not wide
Does it have pricing power? Fail Changes in revenue mix directly pressure gross margin
Can it generate stable free cash flow? Fail Historically, it mostly has not
Are its returns on capital excellent? Fail ROE/ROA have been materially negative in recent years
Is management trustworthy? Uncertain No obvious integrity issue for now, but capital allocation is financing-driven
Is capital allocation rational? Uncertain to failing High R&D is reasonable, but dilution and financing are excessive
Is the balance sheet robust? Uncertain Liquidity is acceptable, but leverage and intangible assets are rising
Is valuation below intrinsic value? Fail Clearly above my reasonable range
Is the margin of safety sufficient? Fail Almost none at the current price
Would I feel comfortable holding it long term? Fail Not comfortable for conservative investors
What key facts would make me sell? N/A Failed order conversion, FCF not turning positive, continued major dilution, goodwill impairment
Am I only interested because of emotion and sector hype? High vigilance This is an easy "good company, bad price" trap

Sources: the full text is organized based on the aforementioned public annual reports, quarterly reports, announcements, public market data, and valuation calculations.

Final Rating

Avoid

One-Sentence Investment Thesis

VeriSilicon is a domestic IP+ASIC platform company that I can understand and whose strategic scarcity I acknowledge, but it still has not proven that it can produce high-quality free cash flow over the long term, while the current share price already reflects too much of a very distant and very optimistic profit-realization expectation.

Core Bull Points

  • In semiconductor IP licensing, the company ranked No. 1 in mainland China and No. 8 globally on a 2024 basis, giving it scarcity in the domestic market.

  • New orders signed in 2025 were RMB 5.960 billion, up 103.41% year on year. Backlog at the end of Q1 2026 was RMB 5.133 billion, and new orders signed year to date as of 2026-04-29 had reached RMB 8.240 billion, giving strong short- to medium-term revenue visibility.

  • Penetration into system-level customers is improving. In 2025, revenue from system vendors, internet companies, cloud service providers, automakers, and similar customers accounted for roughly 40%, showing deeper embedding in higher-value projects.

  • The company has a relatively complete processor IP portfolio, 1700+ mixed-signal/interface/RF IP blocks, and cross-process-node and RISC-V ecosystem layouts, all of which form engineering and platform barriers.

Core Bear Points

  • The company posted large losses for three consecutive years from 2023 to 2025. Net profit attributable to the parent was RMB -528 million in 2025, and it still lost RMB 341 million in Q1 2026. The profit model has not yet been validated under GAAP.

  • Operating cash flow and free cash flow have been weak for a long time, and revenue growth has not steadily converted into distributable cash.

  • The current price implies about 40x P/S and about 37-40x P/B. For a company that is still loss-making and has unstable cash flow, the margin of safety is very poor.

  • Over the past few years, the company has gone through an IPO, a private placement, limited-scale buybacks, and is advancing an H-share listing, showing that its capital needs remain heavy and that future per-share value faces dilution risk.

Key Assumptions

  • AI ASIC and data-processing-related orders can continue and convert at a high ratio into revenue and cash flow over the next two to three years.

  • Scaling of volume-production business can ultimately bring scale effects, so GAAP margins and free-cash-flow margins keep improving, rather than only "adjusted profit" improving.

  • Subsequent financing, M&A, and equity incentives will not continue to erode ordinary shareholders' per-share interests.

Fair Buy Price

For conservative long-term investors, I think a more reasonable buy range should be RMB 40-80/share. This is the result after reserving a certain margin of safety on top of my reasonable intrinsic value range. If the company proves itself with real free cash flow over the next two years, the range can be revised upward. Until then, I do not recommend relaxing discipline because of sector enthusiasm.

Target Holding Period

If bought at a more reasonable future price, the observation and holding period should be at least 5-10 years. But the precondition is that the company must move out of the "high orders, low cash flow" phase and enter a phase of "high orders, positive cash flow, positive ROIC."

Expected Annualized Return

Based on the current price of about RMB 240 per share and my scenario valuation, I assign the following conservative range for expected annualized return over the next 10 years:

  • Conservative scenario: -18% to -12%

  • Base scenario: -10% to -5%

  • Bull scenario: 0% to 5%

These numbers are low not because I deny the company's growth, but because the current entry price is already too high. If the starting price is wrong, even many later operational successes may fail to deliver good returns to shareholders.

Maximum Loss Risk

In the worst case, if AI order enthusiasm fades, order conversion falls short of expectations, the company continues financing and diluting shareholders, and the market's valuation of the company reverts from a "long-dated sector premium" to a "realization-based technology-stock valuation," a 50%-80% permanent drawdown in the share price would not be exaggerated. Asset-based valuation provides very weak support for the current price, so book value can hardly provide a floor in a downturn.

Tracking Indicators

In the future, I will keep watching the following indicators, rather than short-term share price moves:

  • Year-on-year and sequential changes in new orders and backlog.

  • One-year conversion ratio in backlog and the share of volume-production business orders.

  • When net operating cash flow and free cash flow turn sustainably positive.

  • Linked changes among contract liabilities, contract assets, accounts receivable, and prepayments.

  • Whether consolidated gross margin and R&D spending ratio show the combination of "revenue growth, expense-ratio decline, and profit turning positive."

  • Share capital expansion, equity incentives, refinancing, and H-share progress.

  • Subsequent amortization and impairment risk for goodwill, intangible assets, and development expenditure.

  • Whether the share of system-level customers and revenue from data-processing fields continues to rise.

Signals That Would Trigger Reassessment

If the following situations occur, I would immediately re-examine the investment logic:

  • The company achieves sustained positive free cash flow in 2026-2027 without relying on continued large-scale equity financing.

  • Net profit and ROIC under GAAP improve meaningfully, proving that scale effects have truly appeared.

  • High order growth suddenly reverses, or contract liabilities and backlog clearly turn down.

  • H-share listing, acquisition integration, and equity incentives further and materially dilute ordinary shareholders' value.

  • Goodwill and intangible assets suffer major impairment, or purchase-price allocation from acquisitions creates worse-than-expected consequences.

Final Recommendation

If you are a long-term, disciplined, conservative investor, I do not recommend buying VeriSilicon at the current level as a "value stock." It may of course continue to perform strongly because of AI orders, market preferences, and sector enthusiasm. But that is not the kind of investment I recognize as being based on real business cash flow and a reasonable price. The best attitude toward this company is not emotional bearishness, but acknowledging that it is a scarce enterprise worth tracking while insisting that before real cash flow is proven, one should not pay an excessive price for dreams.

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

ARMSNPSCDNSCEVA

Semiconductor IPChip CustomizationASICRISC-VSTAR MarketValue InvestingOwner Earnings
Reader Q&A10

Baillie Framework · Ten Questions for Growth Investing

10

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

Baillie Framework · Ten Questions for Growth Investing — score profile: 41/100 total Ceiling 6/10 · Revenue 2x 6/10 · Next engine 4/10 · Moat 4/10 · Reinvention 5/10 · Management 3/10 · Customer need 5/10 · Unit economics 3/10 · 5x path 2/10 · Blind spot 3/10 0510 How large is its market ceiling? Is it expanding an existing pie, or creating an entirely new market? — 6/10 Ceiling 6 Can its revenue at least double over the next five years? Will growth be driven mainly by volume, price, or new businesses? — 6/10 Revenue 2x 6 After five years, what will take over as the next growth engine? Does this "second curve" exist today? — 4/10 Next engine 4 What is its core competitive advantage? Will this moat widen or narrow over the next three to five years? — 4/10 Moat 4 If its core business is disrupted, does it have the genes to reinvent itself? How does it treat mistakes and bad news? — 5/10 Reinvention 5 Does management, especially the founder, have a long-term view and deep alignment with the company? Is it willing to sacrifice current profit for five to ten years out? — 3/10 Management 3 If it disappeared tomorrow, how much would customers miss it? Is its growth model sustainable and not dependent on harming society or regulation? — 5/10 Customer need 5 What are the unit economics of this business, including gross margin and incremental returns? Do they improve or worsen as scale increases? Where does the money it earns go? — 3/10 Unit economics 3 What conditions must hold at the same time for it to rise fivefold over ten years? Are these conditions realistic? What expectations are embedded in today's share price? — 2/10 5x path 2 Why has the market not realized all this yet? Is it because investors do not understand, look down on it, or cannot look far enough? What will become the "narrative inflection point"? — 3/10 Blind spot 3
  • How large is its market ceiling? Is it expanding an existing pie, or creating an entirely new market?6/10

    Conclusion First

    VeriSilicon is both expanding an existing pie and riding the tailwind of a new market now being created, but the latter has the higher ceiling and is not exclusively owned by VeriSilicon. Objectively, the ceiling is high enough that LTGG does not need to worry about the track topping out. The real constraint is not TAM, but the share VeriSilicon can capture and its ability to turn that share into cash flow. That only becomes clear in Questions 4 and 8.

    Existing Pie: IP Licensing, Scarce but Not a Large Pool

    Semiconductor IP licensing itself is a slow-growing installed-base pie: under third-party estimates, the global IP licensing market was about USD 7.0 billion in 2024 and about USD 9.0-11.0 billion in 2030, with a CAGR of only 6-8%. Under IPnest's 2024 ranking, VeriSilicon was No. 1 in mainland China and No. 8 globally, making it a scarce asset. Still, even if its share doubled, IP licensing alone could not support a RMB 120 billion market cap. This is the "moat," not the "ceiling."

    New Pie: AI ASIC Customization, the Real Blue-Sky Case

    The new market actually being created is AI compute custom silicon / ASIC: based on Broadcom and similar estimates, the custom AI accelerator market may reach about USD 60.0-90.0 billion by 2027, while more optimistic views put the full market above USD 100.0 billion by 2030. The report cites industry data showing China's semiconductor market rising from about USD 274.4 billion in 2025 to about USD 373.7 billion in 2030. VeriSilicon's newly signed orders in 2025 were RMB 5.960 billion, up +103% YoY, with AI compute-related orders accounting for more than 60% under one disclosure basis and even higher under the annual-report basis. It is stepping onto this curve. This part is "creating a new market."

    The Honest Boundary of the Ceiling

    But not all of the enlarged pie belongs to VeriSilicon: in the custom silicon pool, leading giants such as Broadcom control most of the market. VeriSilicon is benchmarked against design-service firms such as Global Unichip and Alchip, not against market definers. The highest-margin links, including architecture and end-customer definition, are not in its hands. The RISC-V alliance, with 204 members by the end of 2025, is a potential "new ecosystem" option, but today it remains soft influence. Conclusion: the ceiling is genuinely high, combining existing-pie expansion and new-market creation. The answer to this question is that the track is not too small. The bottlenecks come in later questions: share, pricing power, and cash-flow realization.

    Jun 5, 2026
  • Can its revenue at least double over the next five years? Will growth be driven mainly by volume, price, or new businesses?6/10

    Conclusion first: revenue doubling over the next five years, about 15% annualized, is likely achievable and may even be conservative. But this is order-driven visibility, not cash already realized. Growth is driven mainly by volume, meaning production ramp and customer upgrading. Price, through AI ASIC and advanced-node project size, is an accelerator. The real upside comes from new businesses, namely cloud-side AI ASICs and large system-level customers. Keep this in mind: high orders ≠ high revenue ≠ high profit ≠ high cash flow.

    Start with how close the doubling threshold already is. Revenue rose from RMB 1.506 billion in 2020 to RMB 3.152 billion in 2025, a five-year CAGR of about 15.9%, already on the "doubling in five years" line. Revenue in Q1 2026 roughly doubled YoY, about +114%, marking the starting point of another doubling. Visibility comes from orders: newly signed orders in 2025 were RMB 5.960 billion, up +103.41% YoY, with RMB 5.075 billion in backlog at year-end. As of 2026-04-29, newly signed orders for the year had reached RMB 8.240 billion, with backlog of RMB 5.133 billion. More than 90% of backlog is expected to convert within one year, and production orders exceed RMB 3.0 billion. If backlog plus new orders are realized at around 80-90%, doubling revenue in five years to about RMB 6.3 billion is not difficult and could happen earlier.

    Breaking down the drivers: new business is the main engine. Under the company's disclosure, AI compute-related orders account for a very high share of newly signed orders, more than 90%, mainly from cloud-side AI ASICs and supporting IP. In Q1 2026, revenue from system-level customers such as system vendors, internet platforms, cloud companies, and automakers accounted for 33.75%, up +94.10% YoY, versus about 40% for full-year 2025. Volume means production ramp plus a customer-base shift from chip design companies toward higher-ticket system-level customers. Price means larger single-project amounts for advanced nodes and AI ASICs, but pricing power is weak. The Q1 2026 blended gross margin was 32.29%, down 6.76pct YoY, showing that ramp-up relies on scale rather than price increases.

    The conversion risk must be pinned down: VeriSilicon's history repeatedly shows that "revenue growth does not automatically turn into monetization." It was loss-making throughout 2023-2025, with 2025 net profit attributable to shareholders of RMB -528 million and operating cash flow of RMB -222 million, while surging prepayments consumed cash. Revenue doubling is very likely to happen, but turning profit and free cash flow positive remains the next unproven gate.

    Jun 5, 2026
  • After five years, what will take over as the next growth engine? Does this "second curve" exist today?4/10

    Conclusion first: VeriSilicon's true "second curve" has not yet been established. It looks more like the same main curve, IP plus customization, being thickened by the AI wave rather than a heterogeneous new engine. Five years from now, the handoff will most likely still come from "AI ASIC / system-level customization" itself, only converted from today's orders into back-end production royalties. This curve has real orders and revenue, so it is not pure narrative. But it shares the same roots as the core business, so from a Baillie Gifford lens the quality of the "second curve" deserves a discount.

    Screen the candidates one by one. If it exists, say it exists:

    • AI ASIC / system-level customization, already landed but an extension of the main curve: this is the hardest part. VeriSilicon has entered mass production for leading customers' 5nm media / AI acceleration ASICs. AI compute-related orders account for a very high share of newly signed orders, more than 90% under the company's disclosure. Production-business backlog exceeds RMB 3.0 billion, system-level customers account for about 40%, and Q1 2026 reached 33.75%, +94% YoY. This is genuine incremental growth, but in substance it is the main curve scaling up, not a separate second curve.

    • RISC-V compute platform, germinating and acquired through M&A: in September 2025, VeriSilicon disclosed a proposed acquisition of about 97% of China-based RISC-V CPU IP company Nuclei System Technology through share issuance and cash payment, filling its missing high-performance CPU core and combining it with existing GPU/NPU assets into a full-stack compute IP offering. This is the closest step toward a true "second curve." After the deal closes, VeriSilicon may become one of the few domestic RISC-V suppliers with full-stack CPU + non-CPU IP capabilities. But Nuclei is still small relative to RMB 3.152 billion of revenue, and integration plus goodwill risks have already shown up in the statements, with goodwill rising to RMB 629 million and intangible assets to RMB 898 million in Q1 2026.

    • Chiplet / automotive-grade, still more platform narrative than scale revenue: the three-step direction of "IP chipletization - chip platformization - platform ecosystemization" is directionally right and well positioned. NPU/display IP has also passed ISO 26262 automotive-grade certification. But at present this is more platform layout than a separately identifiable scaled revenue stream.

    The honest line: does the second curve "exist today"? The AI ASIC engine already exists and is running, but it is the same engine as the main curve. The truly heterogeneous RISC-V compute platform has only just been seeded through M&A and has not yet sprouted. With a price of about 40x P/S, this long-dated option has already been fully priced. The original report's Avoid rating still holds on this question.

    Jun 5, 2026
  • What is its core competitive advantage? Will this moat widen or narrow over the next three to five years?4/10

    Conclusion first: VeriSilicon has a moat, but it is a "narrow moat"; the report assessed moat strength at 2/5. It is more likely to remain narrow, caught between a stronger strategic position and pressure from powerful rivals plus customers, rather than widen meaningfully over time. Baillie Gifford cares about whether a moat can compound and widen. On this point, VeriSilicon honestly fails.

    Sources of advantage, where they exist: first, IP breadth and engineering credibility: proprietary GPU/NPU/VPU/DSP/ISP/display processor IP across 6 categories, plus 1700+ mixed-signal, RF, and interface IP blocks, with successful tape-outs across 14nm to 4nm FinFET and FD-SOI. This "verified, reusable, cross-node" capability creates real switching costs for large customers. Second, domestic scarcity: under IPnest's 2024 ranking, VeriSilicon was No. 1 in mainland China and No. 8 globally in semiconductor IP, and No. 6 globally by licensing revenue. Third, deep embedding with system-level customers, which contribute about 40% of revenue and have stayed above 30% for five consecutive years. Fourth, RISC-V ecosystem position: it leads the China RISC-V Industry Consortium, with 204 members by the end of 2025, and plans to acquire about 97% of Nuclei System Technology to complete CPU/full-stack heterogeneous IP, alongside automotive-grade ISO 26262 certified assets.

    Weaknesses, where they do not exist: it is not an Arm-style architecture monopoly, has no strong network effects, and has no proprietary data barrier. The RISC-V alliance is closer to "soft influence." Weak bargaining power is hard evidence: Q1 2026 blended gross margin was 32.29%, down 6.76pct YoY, meaning production ramp actually pressured gross margin. Rivals have wider moats: Global Unichip is tied to TSMC's most advanced processes and is far stronger than VeriSilicon in profitability and scale, while Arm/Synopsys/Cadence have thicker positions in architecture and EDA.

    Three-to-five-year judgment: the widening and narrowing forces are both running. AI ASIC strength, RISC-V full-stack positioning, and automotive-grade qualification are reinforcing the strategic position, while global rivals' scale advantage, customer in-house development, and customer bargaining pressure are squeezing it, and gross margin has not crystallized into thick profits. Before VeriSilicon proves scale effects through real free cash flow, the more likely outcome is "better strategic position, still narrow moat," rather than the compounding moat widening Baillie Gifford wants.

    Jun 5, 2026
  • If its core business is disrupted, does it have the genes to reinvent itself? How does it treat mistakes and bad news?5/10

    Conclusion first: VeriSilicon does have the genes for "self-reinvention." It has transformed itself from a pure IP seller into a one-stop customization + production + Chiplet + RISC-V + AI ASIC platform, so its ability to pivot proactively passes the bar. But when facing bad news, its posture is "active response + order narrative as offset." This is not concealment, but it clearly leans toward reporting good news through non-GAAP metrics and order stories, with limited self-reflection on hard problems such as free cash flow not turning positive and gross margin decline.

    Start with the reinvention gene. VeriSilicon's evolution is itself a history of "shifting gears before the core business becomes marginalized": it began with IP licensing, while one-stop chip customization has now become the revenue mainstay, with process coverage across 14/7/5/4nm FinFET and 28/22nm FD-SOI. More importantly, when IP licensing growth hit a ceiling and AI compute became the main line, it did not cling to the old base. The AI compute-related share of chip-design revenue rose quickly. In September 2025, it also announced a proposed acquisition of about 97% of RISC-V CPU IP company Nuclei System Technology, filling the most missing CPU piece in its proprietary IP matrix, while promoting a Chiplet roadmap of "IP chipletization → chip platformization → platform ecosystemization." This move to leave its comfort zone and migrate toward harder but more valuable links is exactly the self-disruption Baillie Gifford values.

    But the handling of bad news needs an honest two-sided split. Positive side: in April 2026, market rumors about "competitors including Broadcom and ASR taking orders, and performance missing expectations" triggered an intraday flash drop. Chairman Wayne Dai personally denied order losses and released backlog data in response, showing a non-evasive attitude. Deloitte Touche Tohmatsu Certified Public Accountants has also issued standard unqualified audit opinions consecutively. The sign that must be called out: for the real bad news, including losses throughout 2023-2025 totaling about RMB 1.4 billion, Q1 2026 net profit attributable to shareholders of RMB -341 million, and gross margin of 32.29%, down -6.76pct YoY, VeriSilicon tends to cover it with the "record-high orders" narrative. In the Q1 report it highlighted non-GAAP metrics such as "adjusted net profit attributable to shareholders / adjusted EBITDA," excluding share-based payments and non-cash M&A expenses, making the losses sound smoother. This is a typical preference for reporting the good over the bad.

    Summary: reinvention capability is strong, meaning yes; candor has a baseline but carries a beautifying bias, meaning half yes and half no. It is willing to refute rumors and disclose losses, but is better at softening bad news with adjusted profit and order stories. It is still some distance from Baillie Gifford's expected willingness to face mistakes nakedly.

    Jun 5, 2026
  • Does management, especially the founder, have a long-term view and deep alignment with the company? Is it willing to sacrifice current profit for five to ten years out?3/10

    One-line conclusion: the long-term technical vision is real, and sacrificing current profit for the future is also real. But "skin in the game" is thin, and per-share value discipline is missing. Under Baillie Gifford's owner-operator yardstick, this is a "half-qualified" management team. The original report gave management and capital allocation only 2/5, and I agree.

    Long-term vision and "sacrificing current profit": full marks. Chairman and president Wayne Dai founded VeriSilicon in 2001 and has led it for more than two decades. He is a technical founder with a PhD in computer engineering. The hardest evidence is R&D intensity: R&D investment was 42.78% of revenue in 2025 and surged further to 58.04% in Q1 2026, with R&D of RMB 485 million. Even under years of GAAP losses, including 2025 net profit attributable to shareholders of RMB -528 million, the company still put more than half of revenue into R&D. This is exactly the "sacrificing current profit for five to ten years out" that Baillie Gifford appreciates.

    Interest alignment: an obvious weakness. The company has no actual controller. The largest shareholder, VeriSilicon Limited, holds about 11.39%, while Wayne Dai personally holds only about 1.66%. Persons acting in concert hold about 13.058% in total, and had already reduced from a higher proportion through inquiry transfer before 2025. This is far from the high-ownership owner-operator Baillie Gifford prefers. The founder has limited "skin in the game," and alignment with minority shareholders is not deep.

    Capital-allocation discipline: closer to a "financing + dilution" paradigm than a "per-share intrinsic value" paradigm. The company has raised about RMB 3.67 billion cumulatively in 6 years since its A-share listing and paid zero dividends, including IPO net proceeds of about RMB 1.678 billion and private-placement net proceeds of about RMB 1.780 billion: it raised about RMB 1.8 billion through private placement in 2025, and in 2026 it again filed with the Hong Kong Stock Exchange to pursue H shares. Share count expanded from 500.4 million to 525.9 million in one year, +5.1%, and has expanded nearly 9% since the IPO. The 2025 buyback was only 287,000 shares, RMB 24.83 million, for equity incentives; a drop in the bucket. There is no audit red flag, with Deloitte Touche Tohmatsu Certified Public Accountants issuing standard unqualified opinions consecutively.

    Baillie Gifford judgment: the founder has genuine long-termism, genuine technical conviction, and genuine willingness to sacrifice current profit. That side honestly deserves credit. But low ownership + no actual controller + continuous equity financing dilution + zero cash return mean there is a gap between "investing for the future" and "creating value for shareholders on a per-share basis." In a good track, he is a qualified technical leader, but he is not yet the kind of capital allocator Baillie Gifford has in mind, one who treats per-share intrinsic value as a lifeline.

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

    Conclusion first: VeriSilicon does create switching costs for customers whose production has already been validated on its IP/platform, and it is a scarce domestic supplier. But it is absolutely not irreplaceable. The growth model itself is healthy and causes no social harm. The true sustainability variables are external risks such as geopolitics and export controls, not moral or regulatory red lines. It belongs in the "missed, but replaceable" bucket.

    ① Indispensability: sticky, but not a "power-outage level" dependency. VeriSilicon is an engineering-service provider spanning "one-stop customization + IP licensing + production royalties." Once customers complete tape-out and enter mass production on its GPU/NPU/VPU IP and cross-node platform, the cost of reselecting, revalidating, and replacing automotive-grade certification is high. The more complex the project, such as AI ASICs and automotive-grade chips, the stronger the stickiness. This deep embedding shows up in system-level customers, including system vendors, internet platforms, cloud companies, and automakers, which have contributed about 40% of revenue for many years, with Q1 2026 at 33.75% and up 94.10% YoY. It is also a scarce domestic supplier, ranking No. 1 in mainland China and No. 8 globally in IP licensing under the 2024 ranking. But stickiness stops at the "project level," not the "architecture level": on the IP side, there are ecosystem/architecture-stronger rivals such as ARM, Synopsys, Cadence, CEVA, and Imagination. On the design-service side, Global Unichip, Alchip, Faraday, and Brite Semiconductor are on the same stage. Customers can completely switch suppliers at the starting point of new projects. With the top 5 customers contributing about 45.51%, the "missing it" is more from individual large customers than from the whole market having no substitute.

    ② Social and regulatory sustainability: clean, but externally fragile. This is a pure toB chip-design/IP engineering business. Growth is driven by real demand from AI compute, automotive, and IoT. It does not rely on harvesting users or crossing data/compliance red lines, and there is no social-harm issue. This dimension is healthy. The real sustainability risks are exogenous: exposure to advanced-node supply, packaging and testing capacity, cross-border technology restrictions, downstream customers' geographic distribution, and changes in in-house development strategy. If export controls tighten or major customers shift to in-house development, stickiness will be compressed by external factors.

    One line: customers would miss it, especially for in-production projects, and it is a scarce, compliant, healthy business. But "indispensable" needs a discount, and the critical sustainability risk lies in geopolitics rather than morality. VeriSilicon has achieved only half of the "irreplaceability" Baillie Gifford wants.

    Jun 5, 2026
  • What are the unit economics of this business, including gross margin and incremental returns? Do they improve or worsen as scale increases? Where does the money it earns go?3/10

    Conclusion first: VeriSilicon is a counterexample of "revenue doubling while unit economics deteriorate." As scale increases, unit economics have not improved; instead, blended gross margin has fallen. The money earned has long been negative, the company has relied on financing to survive, and capital has mainly gone into R&D and M&A expansion rather than shareholder returns. Baillie Gifford prefers businesses where greater scale produces higher incremental returns. VeriSilicon is the opposite. This question is its hardest wound.

    Start with gross margin and incremental returns. Q1 2026 revenue rose +114% YoY, almost doubling, yet blended gross margin was only 32.29%, down 6.76 percentage points YoY. The company attributed this to "changes in revenue structure." In substance, the high-growth one-stop production business, with heavy outsourced wafer and packaging/testing costs and inherently lower gross margin, took a larger share and dragged down blended gross margin. This means its incremental revenue is "dilutive," not "accretive": each additional unit of production business pulls the average margin lower, the opposite of Baillie Gifford's desired "scale effects amplifying unit economics."

    Then look at "where the money earned goes." The cold answer: there is no distributable money earned at all. In 2025, revenue was RMB 3.152 billion, net profit attributable to shareholders was RMB -528 million, operating cash flow was RMB -222 million, and Q1 2026 was still a loss of RMB 341 million. Free cash flow has been negative for years. There are 2 destinations for the money: first, R&D, with R&D investment equal to 42.78% of revenue in 2025 and surging to 58.04% in Q1 2026, with R&D of RMB 485 million; second, M&A expansion, with goodwill and intangible assets rising to RMB 629 million and RMB 898 million respectively. And this money is not self-generated; it comes from the capital market. The company has raised about RMB 3.67 billion cumulatively in 6 years since its A-share listing and paid zero dividends, raised about RMB 1.8 billion through private placement in 2025, and in 2026 again filed with the Hong Kong Stock Exchange for H shares. The shareholder-return portion is tiny: the 2025 buyback was only RMB 24.83 million and 287,000 shares, nearly negligible versus continued dilution.

    One line: revenue growth ≠ improvement in unit economics. Until the cycle of production ramp dragging down gross margin, R&D and M&A swallowing cash, and financing diluting shareholders is broken, meaning GAAP profit and free cash flow sustainably turn positive, this business has not proven the high incremental-return characteristics Baillie Gifford wants.

    Jun 5, 2026
  • What conditions must hold at the same time for it to rise fivefold over ten years? Are these conditions realistic? What expectations are embedded in today's share price?2/10

    Conclusion first: for VeriSilicon to rise fivefold over ten years, from about RMB 240 to RMB 1,200 and to a market cap of about RMB 630 billion, at least 5 conditions must hold simultaneously. Any one of them failing would be enough to break the path, so the joint probability is low. More importantly, today's price of about RMB 240, corresponding to about 40x P/S on 2025 revenue of RMB 3.152 billion, has already priced in expectations far beyond the report's optimistic scenario.

    A ten-year fivefold outcome requires this chain of conditions to materialize together:

    • Revenue growth stays high without stalling: with current total share capital of about 526 million shares, if the valuation multiple remains unchanged, revenue needs to rise roughly in line with market cap by about 5x. That means about 16% CAGR over ten years with no major-year reversal, even though 2023-2024 saw 2 consecutive years of zero growth or even negative growth.
    • Gross margin and net margin turn meaningfully positive and expand: the company still lost RMB 528 million in 2025, while Q1 gross margin was still down 6.76 percentage points YoY. The production business naturally dilutes gross margin as it scales, so scale effects must pull GAAP net margin from negative to double digits.
    • Free cash flow stays positive: operating cash flow was negative in most years from 2020-2025, and the report's conservative owner-earnings estimate was still between RMB -340 million and RMB -370 million. This machine first needs to "generate its own blood."
    • No more major dilution: IPO, 2025 private placement, a buyback of only RMB 24.83 million, and an H-share push. For per-share value to rise fivefold, share count cannot keep expanding.
    • The valuation multiple does not contract: even if all the above are achieved, if P/S falls from 40x back toward mature ASIC peer levels, the stock price could still fall rather than rise.

    Are these conditions realistic? Each individual condition is possible, but achieving them all at the same time is optimism stacked on optimism. The report's optimistic intrinsic-value upper band is only RMB 90-150, and the optimistic scenario's ten-year annualized return is only 0%-5%. A fivefold outcome, by contrast, requires about 17% annualized. In other words, the fivefold path requires a step beyond even the report's most optimistic scenario.

    What does today's share price imply? The current price is still at about a 60% premium to the report's optimistic upper band of RMB 150. Given the recent share price once surged to about RMB 278 and the violent volatility of a 52-week range of about RMB 81-297, the market is already pricing in "perfect AI ASIC order realization + margin turning positive and expanding + no valuation contraction" all happening together. In other words, what you are buying today is not a "ten-year fivefold option." It is closer to a position where the optimistic scenario has already been charged upfront, and any miss versus expectations risks permanent drawdown. The report's downside case is a 50%-80% drawdown. Honestly: good track, scarce company, but the current price makes a "ten-year fivefold" outcome both unrealistic and extremely low in margin of safety.

    Jun 5, 2026
  • Why has the market not realized all this yet? Is it because investors do not understand, look down on it, or cannot look far enough? What will become the "narrative inflection point"?3/10

    Conclusion first: for VeriSilicon, Baillie Gifford's classic premise that "the market has not realized it yet" most likely does not hold. It is not being "misunderstood, looked down on, or underappreciated for the long term." It has already been looked at too far out and too optimistically. About RMB 240 corresponds to about 40x P/S and about 37-40x P/B, already pricing in long-term AI ASIC profitability. It is still at about a 60% premium to the report's optimistic intrinsic-value upper band of RMB 90-150. This looks more like a story that has been fully, even excessively, priced than an undiscovered cheap compounder.

    Break down the 3 forms of "not realized" one by one: not understood? Not valid. The narrative of "China's scarce domestic IP+ASIC platform and picks-and-shovels seller" has been told repeatedly. AI compute-related orders have risen all the way to more than 90% of newly signed orders, and both sell-side analysts and retail investors are chasing this line. Looked down on? The reverse is true. The market is not dismissive; it is enthusiastic, with the 52-week price surging from about RMB 81 to about RMB 297. Cannot look far enough? The opposite. The market is looking too far, linearly extrapolating 2025 newly signed orders, already RMB 8.240 billion as of 4-29, into future cash flow, while selectively ignoring 3 consecutive years of losses, Q1 2026 gross margin of 32.29%, -6.76pct, goodwill of RMB 629 million, and continued dilution.

    Narrative inflection points differ for bulls and bears. The bullish inflection point is orders truly converting into GAAP profit and sustained positive free cash flow, without relying on refinancing, plus ROIC turning positive and gross margin recovering with scale. Once that is proven, the current high valuation would have support. The bearish inflection point deserves more attention: in April 2026, rumors of competitors taking orders once caused a flash drop and forced the company to release order data to refute the rumor, already exposing the fragility of the narrative. If order growth reverses, contract liabilities turn down, gross margin fails to improve, H-share/private-placement dilution continues, or goodwill impairment lands, the valuation could move from "track premium" back to "execution-based technology stock." A 50%-80% permanent drawdown would not be exaggerated. For VeriSilicon, the real risk is not underestimation. It is that the narrative has been priced before cash flow.

    Jun 5, 2026
Ask about this report

Members can ask about this report; once answered it appears under "Reader Q&A" on this page. You can also highlight a passage in the text to ask about it directly.