Biren Technology Deep-Dive Research

In its first two years, Biren was building capability rather than selling products. The prospectus disclosed that R&D expense had already reached RMB 1.0179 billion in 2022, far above revenue for the period. Revenue in 2022 was only RMB 499,000, mainly from other miscellaneous income that was "not representative". In other words, before 2022, the company's core task was to build a high-end GPGPU route that had almost no ready-made domestic template in China, layer by layer from architecture, tape-out, packaging, and software to systems. In that phase, it burned cash on people, EDA, IP, samples, testing, NRE, server adaptation, and round after round of engineering iteration, rather than marketing expenses. It chose this path instead of first making low-end graphics cards or specific ASICs partly because the team's background was better suited to general-purpose GPUs, and partly because a successful general-purpose GPU could theoretically address a larger domestic substitution market.

What first brought Biren into public view was the BR100 launch in August 2022. Xinhua's report at the time showed that the company officially released its first general-purpose GPU chip, BR100, in Shanghai, while also releasing its self-developed architecture, OAM servers, OAM modules, PCIe boards, and the BIRENSUPA software platform. This was a full platform debut, not a product event that only displayed a chip. The market was willing to believe it because Biren offered a full cabinet-ready computing product form, not a PowerPoint chip. The market also could not fully buy in because Chinese GPU startups usually face an even harder test after product launches: can they truly mass-produce, deliver, adapt, and collect cash?

The turning point came quickly in 2023. On 2023-10-19, the U.S. Department of Commerce issued rules adding Shanghai Biren Intelligent Technology Co., Ltd. and several related entities to the Entity List, applying license requirements for all items subject to the EAR and a license review policy of "presumption of denial". For a GPU company still in the ramp-up phase and highly dependent on external toolchains and advanced-process coordination, this risk was not abstract. It struck directly at the center of R&D and supply chain. The prospectus explicitly states that the company recognized RMB 108.7 million of special losses on certain assets in 2023, and those losses "directly resulted from the BIS Listing Incident". The company also disclosed that some EDA tools were "no longer eligible" after listing and that it subsequently procured new EDA tools from mainland Chinese suppliers. Many people understand the Entity List as a short-term disruption, but for Biren it changed R&D organization and supply-chain routing. All later commercialization had to occur within this narrower constraint set.

2024 was the first year in which Biren truly proved it could sell products into customer data rooms. Revenue reached RMB 336.8 million in 2024, up materially from RMB 62.0 million in 2023. Professional technology product customers numbered 14 that year. Revenue mainly came from intelligent computing solutions, and customers shifted from early trial users to leading players in selected industries. Gross margin also fell from 76.4% in 2023 to 53.2%. On the surface this was deterioration; in substance it looked more like a revenue mix shift from early projects and high-margin trial orders toward larger-scale delivery and more complex solution structures. For a company like Biren, a gross-margin decline is not necessarily bad news. The real issue is whether the decline is accompanied by revenue expansion, customer expansion, and cash collection. By the end of 2024, that answer had not fully emerged.

What truly changed capital-market perception was 2025. The prospectus showed that revenue in 2025H1 was only RMB 58.9 million and gross margin was 31.9%. At that time, the market could easily draw a pessimistic conclusion: products had not truly scaled, while inventory had already been stocked. But in the post-listing 2025 annual report disclosure, full-year revenue jumped to RMB 1.035 billion, up 207.2% year on year. The prospectus also disclosed that as of the latest practicable date before listing, the company had 24 outstanding binding orders worth about RMB 821.8 million, plus 5 framework sales agreements and 24 sales contracts totaling about RMB 1.2407 billion. In other words, 2025 was a year when almost no turning point was visible in the first half, while revenue was heavily realized in the second half, not a year of linear growth. Many new hardware companies in business history go through this moment of stocking inventory first and recognizing revenue later. The difference is that some companies then enter a repeat-purchase cycle, while others only complete a one-off shipment. It is still too early to conclude which category Biren belongs to.

The listing path itself is also representative. After updating its prospectus in December 2025, Biren advanced its Hong Kong IPO. The final offer price was HKD 19.60, with 284.8466 million shares offered and gross proceeds of about HKD 5.583 billion. Before the stabilization period ended on 2026-01-28, the over-allotment option was fully exercised, adding 42.7268 million shares, about 15% of the original offer shares. Before listing, the company had completed a long sequence of private-market financings. After listing, it shifted from "telling the long-term domestic GPU story in the private market" to becoming an asset that must prove orders, margins, cash flow, and share-supply absorption capacity every quarter in the secondary market. In June 2026, the near-simultaneous appearance of Southbound Stock Connect inclusion and an H-share full circulation application was the clearest sign of this transition: the funding side is widening, and the supply side may also widen quickly.

Financial Longitudinal Review

Biren's revenue trajectory is very steep, but not smooth. Under the prospectus basis, revenue was only RMB 499,000 in 2022, RMB 62.0 million in 2023, and RMB 336.8 million in 2024. Revenue in 2025H1 was only RMB 58.9 million, but full-year 2025 revenue reached RMB 1.035 billion. This means more than 94% of 2025 revenue was recognized in the second half. For a mature company, this recognition pattern would usually be viewed as a risk. For Biren, it is both a risk and a dividing line. The optimistic interpretation is that GPU projects have long delivery cycles, slow acceptance, and common Q4 concentration. The pessimistic interpretation is that revenue is highly dependent on a few projects and a few customers, and any acceptance delay would cause the next reporting period to slow sharply.

The easiest item in the income statement to distort in external data is the loss basis. In 2025, the company's reported "loss for the year" was RMB 16.493 billion, but about RMB 15.471 billion of that came from changes in the carrying amount of redemption liabilities. The annual report also disclosed an adjusted loss for the year of RMB 873.8 million. The former reflects accounting revaluation of pre-IPO preferred shares and related instruments, and is non-cash. The latter is closer to the operating loss itself. Neither number can be omitted. Reporting only RMB 16.493 billion would mistake accounting noise for operating collapse. Reporting only RMB 873.8 million would miss the financial-reading threshold created by the complex pre-IPO capital structure. For researchers, the correct approach is to read reported loss, adjusted loss, and operating cash flow together.

From gross margin, Biren's commercialization quality has not stabilized. The prospectus shows that gross margin fell from 100% in 2022 to 76.4% in 2023 and 53.2% in 2024, then further to 31.9% in 2025H1. The prospectus attributes this change to product mix and differences in customer demand, which is logically sound: the more the company moves into large projects and solution delivery, the more gross margin usually steps down. But it also points to another fact: Biren has not yet formed a standardized high-margin product curve that the market can accept without hesitation. Current gross margin looks more like a function of project mix and delivery timing.

Cash flow is more severe than the income statement. In 2025, operating cash flow was a net outflow of RMB 2.137 billion, further worse than the RMB 1.009 billion outflow in 2024. Capital expenditure and purchases of intangible assets were about RMB 351 million over the same period. In other words, even after excluding the RMB 15.471 billion non-cash gain or loss from redemption liabilities, Biren's 2025 "owner earnings" were clearly negative. The reasons mainly have three layers. First, R&D remained high, with 2025 R&D spending of RMB 1.4761 billion. Second, the company built inventory and raw-material reserves for BR166 commercialization. Third, the project-based delivery model naturally lengthens cash conversion from shipment to collection. The company's prospectus is very direct: the year-on-year expansion in operating cash outflow in 2025H1 was mainly due to increased raw-material prepayments and inventory stocking.

The balance sheet also has two sides. The risk side is that year-end inventory had risen to RMB 948.6 million, receivables, other receivables, and prepayments occupied meaningful capital, and contract liabilities were not large. The safety side is that the company had liquidity resources of RMB 2.8227 billion at the end of 2025, and the annual report disclosed approximately RMB 5.6314 billion of net IPO proceeds in early 2026. Redemption liabilities formed before listing are also no longer a continuing cash burden after the company entered public trading. In the short term, Biren's biggest financial question is whether money can be efficiently converted into sustainable revenue, rather than being consumed by larger inventory and longer collection cycles. Whether the company will run out of cash immediately is less central.

Share-Price and Valuation History

Biren's Hong Kong share-price history is short, but it has already shown the three classic trading phases of a newly listed hard-tech stock. The first phase was the scarcity premium on the first trading day. The final offer price was HKD 19.60, and the offering size was increased to 284.8466 million shares, corresponding to gross proceeds of HKD 5.583 billion. This was the first time the public market priced "domestic high-end GPU" as a standalone Hong Kong-listed asset. The first-day surge was essentially Hong Kong stocks giving a one-time re-rating to "scarcity narrative + domestic substitution + new-share liquidity", not the market suddenly seeing the profit path clearly.

The second phase was holding at a high level after listing. By intraday 2026-06-11, the share price was HKD 48.52, about 147.6% above the IPO price, corresponding to a market cap of HKD 118.333 billion. For a company whose 2025 revenue had only just passed RMB 1 billion, was still loss-making on an adjusted basis, and still had a large net operating cash outflow, such a market cap shows that the market is pricing the probability of commercialization success over the next several years, not the price of current financial results. Strictly speaking, Biren is already trading on 2027 or even more distant revenue imagination, not on 2025 results.

The third phase is the share-supply re-pricing that began to appear in June 2026. Yahoo historical prices show that the closing price on June 10 was still HKD 55.40, while by intraday June 11 it had fallen to HKD 48.52, down 12.42% from the prior close. In timing, this pullback was highly close to the company's June 5 announcement that it had submitted an H-share full circulation application. The announcement shows that the company proposed to convert 867.5 million domestic unlisted shares into H shares. Meanwhile, cornerstone investors' 6-month lock-up expires in early July, and Pathfinder SIIs under Chapter 18C rules also enter the unlocking window around the six-month anniversary of listing. For a newly listed high-valuation stock, changes in expected supply often move the share price before the financial statements do. The most important point here is that the market has started putting a second issue on the table, beyond how many GPUs can be sold in the future: how many more shares may become tradable.

Business Model, Industry, and Moat

Business Model and Cost Structure

Biren's revenue structure is simple, but the realization path is long. The company's core products are professional technology products, meaning GPU chips and related solutions. The prospectus disclosed that in 2024 the company had 14 such customers, contributing revenue of RMB 336.8 million. In 2025H1, it had 12 customers, contributing revenue of RMB 58.9 million. The company does not sell only bare chips. More commonly, it sells intelligent computing solutions that bundle chips, boards, complete systems, clusters, deployment, and services. The benefit is that a single order can be larger, and a successful deployment may lead to later capacity expansion. The drawback is that the business is more project-heavy, revenue and collections depend more on acceptance, and quarterly volatility is greater.

The second feature of this model is that customer concentration tends to be high in the early commercialization phase. The prospectus disclosed that from 2023 onward, revenue from the largest customer in each year or period was RMB 53.2 million, RMB 183.4 million, and RMB 19.6 million, accounting for 85.7%, 54.5%, and 33.3% of total revenue for the respective periods. This shows that Biren is still relying on a few large customers and a few large projects to cross the early commercialization phase, and has not yet built a broad, diversified, standardized customer pool. For bulls, this means volume can ramp quickly once the company enters leading customers. For bears, it means that if several key projects are delayed, scaled down, or switched to other suppliers, the period's results will look bad.

The cost structure typically reflects the dual features of "fabless + platform R&D". Fixed costs are R&D teams, design and development, software platforms, IDC hosting, equipment leasing, and ecosystem adaptation, not factory depreciation. Variable costs come from raw-material procurement, foundry and packaging, memory, server peripherals, and delivery services. In 2025H1, R&D expense was RMB 571.6 million, already close to 10 times revenue for the same period. Full-year R&D expense further reached RMB 1.4761 billion. Biren's operating leverage direction is clear: if revenue scales sufficiently, R&D as a percentage of revenue will fall quickly and profit elasticity will be strong. But before revenue stabilizes, R&D and supply-chain stocking will press profit and cash flow into a very unattractive shape.

The company also has some supplier concentration. The prospectus shows that during the reporting periods, purchases from the top five suppliers accounted for 56.1% to 64.1% of total purchases, while the largest supplier accounted for 19.6% to 34.9%. For mature chip companies, this level of concentration is not unusual. For Biren, the issue is whether, after Entity List constraints, it can continue obtaining qualified alternatives and ensuring delivery timing and yield, not the percentage itself. In other words, the hardest-to-compress costs in Biren's cost structure are the entire engineering expenses needed for continuing R&D and maintaining delivery under a constrained supply chain, not office rent.

Moat and Governance

Biren has three things closest to a "real moat". The first is team capability. The founder and core executives do not only have financial-capital backgrounds; they also understand AI scenarios, system sales, GPU architecture, and server implementation. This has been partly verified along the 2019-2025 path: the company built the BR100 platform products, rebuilt its supply chain after the Entity List shock, and ultimately grew revenue to the RMB 1 billion level, rather than remaining a financing story. The second is engineering know-how. Xinhua's report on the 2022 BR100 launch showed that the company introduced chips, servers, modules, boards, and software platforms at the same time. This system-level delivery experience is closer to real customer procurement logic than single-chip specifications. The third is timing. U.S. export controls have created a domestic substitution window in China's high-end training and inference chip market, and Biren has at least obtained a seat at that table.

But Biren is still far from proving the kind of "ecosystem moat" the market likes to talk about. The truly hard-to-replace part of Nvidia is CUDA, developer habits, software compatibility, toolchains, mature cases, and path dependence inside customer organizations, not a single chip. Biren is of course promoting its own software platform, but public disclosures do not show third-party large-scale benchmark data, broad developer-community migration evidence, or several consecutive years proving that customers are locked in because of the software ecosystem. Put differently, one can say today that Biren has technical barriers and engineering accumulation, but not that it already has a strong ecosystem moat in a 3–5 year sense.

Governance advantages and risks also coexist. The advantage is that the company does not have the WVR structure common in Hong Kong stocks, and post-listing proceeds have significantly strengthened the balance sheet. The risk is that the shareholder structure is very dispersed, with many private-market investors, and many are financial investors rather than long-term industrial shareholders. The prospectus disclosed that locked-up parties under Chapter 18C rules include the founder, employee shareholding platforms, and five Pathfinder SIIs. Cornerstone investors also have a separate 6-month lock-up. For a company whose market cap has already approached HKD 100 billion and whose history is very short, the governance discount is mainly reflected in how large future supply may be, who may sell first, and at what layer selling may interrupt the market's growth narrative, rather than voting rights.

Industry Structure, Cycle, and Policy

Biren sits in the GPGPU/intelligent-computing chip track driven by policy, capital expenditure, and technology iteration at the same time, not an ordinary semiconductor track. The prospectus cited CIC data saying that in 2024, China's intelligent computing chip market was highly concentrated. The top-ranked U.S. GPU vendor had a revenue share of 76.2%. In China's GPGPU market, the top two players together had a 98.0% share, with the U.S. top player accounting for 97.6%, while Biren's revenue share in China's intelligent computing chip market was only about 0.16%. These data show precisely that Biren's business story is still built on "the market space is large, but the company is still at a very early stage of share ramp", far from "already strong".

This track is highly cyclical. It belongs simultaneously to the technology iteration cycle, the capital-expenditure cycle, and the policy cycle. The most beneficial variables in an upcycle are downstream computing-capacity construction demand, intelligent computing center tenders, domestic substitution willingness, and funding costs. The most vulnerable variables in a downcycle are customer acceptance timing, single-card or single-cluster ASP, server system solution margins, and ecosystem migration speed. Biren has not yet gone through a full public-market cycle from upcycle to downcycle, so it is too early to discuss its ability to cross cycles. What the market can truly see is only that it survived under sanctions and achieved its first revenue jump.

Policy and regulation affect Biren more directly than they affect most Hong Kong technology companies. The first layer is the listing regime. HKEX implemented Main Board Chapter 18C in March 2023, providing a listing route for companies that cannot meet traditional profit tests but satisfy specialist technology conditions. One core threshold for a "Commercial Company" is revenue of at least HKD 250 million in the most recent audited financial year. Biren entered Hong Kong stocks under this framework. The second layer is export controls. After BIS added Biren and related entities to the Entity List, the company became subject to "no license exceptions" and a "presumption of denial" policy, forcing it to shift foundry, EDA, tools, and certain overseas collaboration links as far as possible toward alternatives. The third layer is capital-market accessibility. In June 2026, the company was included in Southbound Stock Connect, meaning mainland investors can directly allocate to it, and both liquidity and investor structure may change. For Biren, policy is both a demand-side tailwind and a supply-side constraint.

Horizontal Competitor Analysis

If one compares only parameter tables, Biren can easily look like "a domestic Nvidia candidate". But once the company is placed in its real peer group, the position becomes much clearer. Nvidia earns money from a global standard platform. Huawei Ascend competes for domestic systematic ecosystem money. Hygon earns money from CPU + DCU combined with server ecosystems. Cambricon has moved from being an A-share "loss-making narrative stock" to a high-beta A-share company with both rising revenue and profit. Biren stands among these large players and remains a challenger that has just pushed commercialization from 0 to 1, but has not yet pushed 1 to 10.

The difference between Nvidia and Biren is not merely a scale gap of hundreds of times. Nvidia's fiscal 2026 revenue reached USD 215.9 billion, with a GAAP gross margin of 71.1%. AMD's fiscal 2025 revenue was USD 34.6 billion, with a gross margin of 50%. These two companies are selling not only chips, but platforms, software, and long-term customer path dependence. Biren is not currently qualified to compete with them globally. Its real participation is in the temporary window opened by the Chinese market for high-end GPU alternatives under export restrictions. The window can be large, but a window is not a moat.

Cambricon is more comparable in capital-market terms. In A-shares, it has already gone through a phase of high R&D, high losses, and high valuation, as well as the market's extreme optimism and extreme skepticism toward domestic AI chips. By 2025, Cambricon's annual report showed revenue of RMB 6.497 billion, up 453.21% year on year, net profit attributable to the parent of RMB 2.059 billion, and a consolidated gross margin of 55.15%, meaning it had turned profitable. In other words, Cambricon's high valuation is no longer only because of the "domestic substitution story"; it has already converted the story into a stretch of financial results. The most fundamental difference between Biren and Cambricon is that Biren still stands today at the early validation point where Cambricon once stood.

Hygon's ecological position is different. It centers on high-end processors and also has a coprocessor/DCU route, supported by server systems, complete-system vendors, and a more mature commercial customer network, rather than being a pure GPU company. Market research summaries citing Hygon's 2025 financial report state that the company had 2025 revenue of RMB 14.377 billion and net profit attributable to the parent of RMB 2.545 billion. Hygon's valuation is also not cheap, but its pricing includes higher cash-flow visibility, more mature server customer relationships, and a clearer profit model. If Biren wants to receive a "high-quality growth" valuation label close to Hygon's, it must first prove that its revenue is replicable continuous expansion, not an occasional jump.

The table below compresses the differences into the most essential items.

Note†: Biren's market cap uses the intraday basis on 2026-06-11. Revenue is converted in RMB while market-cap bases are used separately to calculate price-to-sales ratios. Cambricon's revenue and profit come from its 2025 annual report summary, and market cap comes from Yahoo Finance quote summaries near 2026-06-11. Hygon Information's revenue and profit come from market research summaries of its 2025 financial report, and market cap comes from Yahoo Finance quote summaries near 2026-06-11. Calculations are for horizontal comparison only and do not represent buy or sell advice.

The key conclusion from this table is that the certainty implied in the market's pricing of the three companies is completely different. The focus is neither "Cambricon is more expensive than Biren" nor "Hygon is cheaper than Biren". Cambricon's high price-to-sales ratio is already supported by profitability. Hygon's lower price-to-sales ratio is backed by larger scale and a more mature customer structure. Biren's near-100x price-to-sales ratio is mainly built on scarcity, domestic substitution, and expectations for high revenue growth over the next two to three years. In other words, the market has already placed Biren in the basket of "a future potential leading domestic GPU asset", not the basket of "a hardware startup that has just begun selling products". The price tag rose quickly, but company fundamentals have not fully caught up.

From an ecological-position perspective, Biren is currently a challenger, and one still competing for "market whitespace" rather than stable share. It is most directly trying to capture the profit pool left by restricted overseas high-end GPUs in China, while also competing with Huawei Ascend, Hygon DCU, Cambricon, and a group of unlisted domestic GPU vendors for the same intelligent computing centers, key industries, and system-integration customers. If the industry later enters a price war, supply improves, or policy support further concentrates toward leaders, Biren's position may not naturally strengthen; it may instead be hurt more, because what challengers fear most is not the absence of a story, but major customers deciding that there are more alternatives and no need to bet only on one supplier.

Current Fundamentals, Valuation, and Risks

Current Fundamentals and Long-Short Divide

Biren's hardest current fundamental facts fall into only three categories. The first is the 2025 annual report: full-year revenue of RMB 1.035 billion, reported loss of RMB 16.493 billion, adjusted loss of RMB 874 million, R&D expense of RMB 1.476 billion, net operating cash outflow of RMB 2.137 billion, and year-end inventory of RMB 949 million. The second is the trajectory that can be pieced together from the prospectus and annual report: 2025H1 revenue was only RMB 58.9 million and gross margin was 31.9%, indicating that full-year revenue recognition was highly concentrated in the second half. The third is the capital-market events in June 2026: on June 5, the stock was included in Southbound Stock Connect and became effective, and around the same time the H-share full circulation application was disclosed. New-stock liquidity, incremental capital, and potential incremental supply appeared almost at the same time.

Bulls' core argument today has three layers. The first is industry-level: U.S. restrictions have not disappeared, and domestic demand for high-end computing substitution has not disappeared either. Biren's ability to grow revenue from RMB 337 million in 2024 to RMB 1.035 billion in 2025 under Entity List pressure itself shows it is not a pure concept company. The second is operating-level: before listing, the company still had about RMB 821.8 million of outstanding binding orders and about RMB 1.2407 billion of framework sales agreements and sales contracts, showing that the revenue jump did not appear from nowhere. The third is funding-level: IPO proceeds and Southbound Stock Connect inclusion improved the balance sheet and secondary-market accessibility, respectively. What bulls are really betting on is Biren moving from one-off large-project delivery to continuous large projects, repeat purchases, and new product cycles.

Bears focus on three other groups of evidence. The first is cash flow and inventory: in 2025, operating cash flow was a net outflow of RMB 2.137 billion, far worse than the adjusted loss; year-end inventory was RMB 949 million, close to full-year revenue. This shows commercialization has started, but efficiency has not been proved. The second is recognition timing: 2025H1 revenue was only RMB 58.9 million, while full-year revenue reached RMB 1.035 billion. This extreme back-loading naturally makes the market question sustainability. The third is share supply: as of the end of May 2026, the company had 1.2008 billion H shares and 1.2380 billion domestic shares. On June 5, it applied to convert 867.5 million of those domestic unlisted shares into H shares. At the same time, cornerstone investors held 147.9 million H shares with a 6-month lock-up expiring in early July. Five Pathfinder SIIs held a total of 299.5 million shares, of which about 136.4 million were already H shares, also near the 6-month unlocking window. What bears truly worry about is that valuation remains expensive while share supply may suddenly increase.

This is also the dividing line between "real fundamentals" and "market narrative". Real fundamentals have already proved Biren has products, customers, and revenue, and that the company can survive under strong constraints. The market narrative further assumes that Biren will smoothly cross from RMB 1 billion to several billion RMB of revenue and gain higher share in the domestic GPGPU market. The former already has evidence today; the latter remains an odds-weighted inference. The problem is precisely that the share price looks as if it is pricing the latter.

Valuation Analysis

The least suitable valuation method for Biren today is PE. The most suitable is a forward price-to-sales ratio anchored on 2027. The reason is direct: the 2025 reported loss is heavily distorted by redemption-liability revaluation, while adjusted loss cannot capture cash tied up by inventory, prepayments, and project delivery. Meanwhile, the 2025 net operating cash outflow of RMB 2.137 billion and capital expenditure of RMB 351 million tell us that on an "owner earnings" basis, the company currently has no positive denominator for PE or FCF yield. For this type of company, penetrating cash flow first and then discussing revenue multiples is less likely to be misled by accounting bases.

Start with cash-flow penetration. In 2025, net operating cash outflow was RMB 2.137 billion, while capital expenditure and additions to intangible assets were about RMB 351 million, implying "owner earnings" of roughly negative RMB 2.49 billion. Even looking only at adjusted loss of RMB 874 million, cash flow was materially worse, indicating that inventory stocking, prepayments, and project-delivery timing are amplifying cash pressure outside the income statement. The prospectus already gave the reason: raw-material prepayments and inventory stocking increased in 2025H1 to prepare for BR166 commercialization. For investors, this means looking only at adjusted loss is not enough. The real focus is whether inventory, operating cash flow, and contract liabilities can improve in tandem.

Now look at current valuation. Based on the intraday price on 2026-06-11 and the corresponding market cap, Biren's current market cap was about HKD 118.333 billion, equivalent to about RMB 102.110 billion. Against 2025 revenue of RMB 1.035 billion, the trailing price-to-sales ratio was about 98.7x. This multiple is roughly on the same order as Cambricon and far above Hygon Information. The issue is that Cambricon turned profitable in 2025, while Hygon is larger and clearly profitable. Biren is still at the stage of revenue just passing RMB 1 billion and cash flow remaining deeply negative. So for the same high PS, Biren has weaker certainty than the two mainstream A-share comparables.

My valuation approach is to use three 2027 revenue scenarios with different forward price-to-sales ratios. This has two benefits. First, it acknowledges that the current revenue base is too small, so valuing the company directly on 2025 is not meaningful. Second, it pulls the core debate back to whether Biren can turn billion-RMB revenue into several billion RMB of revenue over the next two years. For multiples, I use roughly 15–18x, 20–22x, and 28–30x 2027 forward PS. This range already looks high, but considering the current valuation environment of the sector where Cambricon and Hygon sit, plus Biren's scarcity premium, it is still a relatively cautious cross-market compromise. The premise supporting these multiples is that the company must continue delivering projects and narrowing losses. If it cannot even do that, even 15x forward PS may not hold.

Note‡: All scenarios in the table are research-framework estimates and do not constitute investment advice. Calculations are based on publicly disclosed 2025 revenue, market cap, current share capital, and the 2026-06-11 exchange rate, using a two-year forward price-to-sales method.

The margin-of-safety conclusion is clear: there is none. Under the three scenarios above, the current share price of HKD 48.52 is already materially above the upper end of my base valuation range of HKD 42, and far above the ideal buy zone of HKD 21–25. More importantly, if revenue shows zero growth over the next three years and remains at the 2025 level of RMB 1.035 billion, even if the market gives an extremely generous 30x price-to-sales ratio, the corresponding share price would be only about HKD 14.7, with a three-year annualized return of about negative 32.8%. Even if the multiple were lifted to 40x, the reasonable share price would be only about HKD 19.7. For a newly listed stock with deeply negative operating cash flow and expanding share supply, such a purchase price has no margin of safety.

So my judgment on the current price is this: the company deserves continued tracking, but the price does not offer enough margin of safety. A good company and a bad price can coexist; Biren is closer to that combination now. Is it worth waiting for a better price? My answer is yes. The opportunity cost of waiting is real, since the domestic GPU theme may continue to be pushed up by Southbound Stock Connect and industry sentiment. But the cost of not waiting is larger, because the buyer is paying a near-100x trailing price-to-sales ratio for a commercialization curve that has not yet stabilized.

Risk Analysis

The first risk that could cause permanent capital loss is discontinuous commercialization validation. I assign it a medium-high probability and high impact. Biren's 2025 revenue delivery was highly back-loaded, showing that revenue depends heavily on a few projects being recognized at a few points in time. If several key customers slow capacity expansion in 2026, acceptance is delayed, or downstream intelligent computing center budgets shift more slowly, revenue growth will drop sharply, and the market will immediately switch from "future high growth" back to "project-based volatile stock". The indicators to watch most closely are customer count, contract liabilities, receivables, and inventory linkages in interim and annual reports. If revenue does not rise while inventory keeps rising, this risk has already started to materialize.

The second risk is long-term supply-chain and technology-route constraints. Probability is medium, impact is high. The Entity List is a long-term structural constraint, not a one-time event. The company has already recognized special asset losses in 2023 because of the BIS incident and shifted to procuring mainland Chinese EDA tools. The problem is that the GPU chain cannot be fully opened by replacing only one tool. It has very high requirements for foundry, packaging, interconnect, memory, and system-level coordination. If external restrictions tighten again, or alternative solutions cannot keep up in cost, yield, and performance, Biren's new product advancement and existing product delivery may both be affected. The observation points are new-product launch timing, delivery rhythm, and whether gross margin deteriorates at the same time, not news headlines.

The third risk is a double hit from valuation and share supply. Probability is high, impact is high. Cornerstone investors hold 147.9 million H shares, with lock-up expiring in early July 2026. Five Pathfinder SIIs hold a total of 299.5 million shares, of which about 136.4 million are already H shares, also near the six-month post-listing unlocking window. The larger variable is the H-share full circulation application submitted on June 5, which proposes to convert 867.5 million domestic unlisted shares. For Biren at a current market cap of HKD 118.3 billion, the potential new H-share supply is worth more than HKD 42.0 billion at the current price. Even if not all shares enter circulation at once, once the market starts expecting future supply to increase materially, the valuation center may step down first.

The fourth risk is customer concentration and price war. Probability is medium, impact is medium to high. The prospectus already shows that the top five customers were highly important during the reporting periods, and the largest customer once contributed more than half of revenue. When GPU supply is tight, the industry talks about performance and delivery timing. When supply improves, it often talks about price. If Biren faces competition from Huawei's ecosystem, Hygon DCU, Cambricon, or other domestic GPU vendors and needs to maintain orders, ASP and project gross margin are the most likely sacrifices. Because the company has not yet reached the scale-profit zone, any price concession would amplify the shock to both the income statement and cash flow. Observable signals are gross margin staying below 35%, and the largest-customer share falling without a significant rise in total customer count.

The fifth risk is that management execution credibility is tested by new-product timing. Probability is medium, impact is medium. Biren's biggest imagination today comes from "BR166 has entered commercialization, and the next-generation BR20X is planned for 2026". In high-end GPUs, the roadmap itself is part of valuation. If BR20X is delayed, performance misses expectations, or software compatibility progress falls below customer requirements, the market will quickly reclassify Biren from "future platform company" to "beneficiary of a temporary supply gap". For growth stocks, reclassification often hurts valuation more than a single quarterly miss.

Catalysts and Tracking Indicators

There are four main positive catalysts. First, if the 2026 interim report can prove that revenue is still growing strongly, rather than being a one-off concentrated recognition in 2025Q4, while inventory turnover improves, the market will be more willing to accept that commercialization has crossed from 0 to 1. Second, sustained southbound net inflows after Southbound Stock Connect inclusion would improve share absorption. Third, BR166 project replication and BR20X roadmap delivery would strengthen the judgment that the next growth round is not merely the tail of old projects. Fourth, if H-share full circulation is advanced in phases and no obvious selling pressure appears after lock-up expiry, the valuation center may instead become more stable.

Negative catalysts are more direct. First, if the 2026 interim report still shows highly uneven revenue recognition, no increase in contract liabilities, and continuing inventory growth, the market will suspect that 2025 was only a one-off conversion. Second, if cornerstone and early investors show obvious selling after unlocking around July, sentiment will weaken quickly. Third, if H-share full circulation is approved and advances faster than the market expects, supply pressure will compress valuation multiples. Fourth, any new round of tightening in U.S. export controls or any delay in key new products could affect delivery and future orders.

Note§: The "normal range" and "warning threshold" in the table are research tracking disciplines, not management guidance. Gross margin, inventory, R&D, operating cash flow, and other bases should follow periodic reports. Southbound Stock Connect and full circulation progress should follow exchange and company announcements.

In this dashboard, I actually care most about only three indicators: whether revenue is sustained, whether inventory falls, and whether cash flow improves. The biggest misjudgment for many hard-tech companies early after listing is that the market tracks only whether orders exist, and not whether orders can become cash. Biren is exactly a company for which these three things must be watched together.

Horizontal-Longitudinal Synthesis

From a longitudinal perspective, what Biren has truly proved along the way is three more basic and harder capabilities, not that it has already produced a mature business model. First, it proved it can build the extremely difficult high-end general-purpose GPU product line in China. Second, after the 2023 Entity List shock, it absorbed asset impairments and toolchain migration costs, then continued advancing without stopping. Third, it took a startup chip company long dependent on private-market funding onto the Hong Kong Main Board, and achieved a secondary-market re-pricing materially above the offer price. Many companies can do only one of these. Biren did all three. That is what deserves respect most.

But these three capabilities are not yet equal to a fourth: making money continuously and replicably. Biren's past stage success came half from team capability and engineering execution, and half from the window opened by the era for domestic computing power. The window is still there today, and the team still appears to be there. But what will decide the next three to five years is who can build repeat purchases, ecosystems, and cash flow inside that window, no longer whether the window exists. From a horizontal comparison, Biren's real advantage over competitors is that it has entered the public market and obtained financing channels, liquidity, and higher brand visibility. Its real weakness is that revenue scale is too small, cash flow is too poor, the customer structure is still in the early validation phase, and its competitors either have more mature ecosystems or more mature profit models. Some of this weakness is temporary, such as a small revenue base. Some may be structural, such as software ecosystem and customer migration costs.

Today's valuation looks more like a reward for the success Biren may achieve in the future than for the report card it has already delivered. The current market cap implies a near-100x trailing price-to-sales ratio, meaning the market is willing to pay in advance for 2027 and even more distant revenue. The market's most likely misjudgment is also here: it may overestimate the continuity of order conversion speed, customer repeat-purchase probability, and new-product iteration, while underestimating the damage that changes in share supply and cash-flow pressure can do to valuation multiples. High valuation is not scary by itself. High valuation meeting continuity that has not yet been proved is the real risk. Biren now stands at that intersection.

The most important variable over the next year is whether the 2026 interim report can upgrade the extremely back-loaded 2025 revenue recognition into a smoother and more credible growth curve. The most important variable over the next three years is whether BR166 commercialization can turn into broader customer repeat purchases, and whether BR20X can advance on schedule. The most important variable over the next five years is whether Biren can grow from a "scarce domestic GPU narrative asset" into a "domestic GPU platform company". If the company can keep growing revenue in 2026-2027, stabilize gross margin above 45%, and materially improve operating cash flow, I would be willing to revise the valuation center upward. If instead revenue growth slows, inventory continues rising, and full circulation plus lock-up expiry bring supply that fundamentals cannot absorb, today's high valuation will reverse quickly.

Bull and Bear Reasons

Bull reasons:

• The company still achieved a commercialization leap after the Entity List shock, showing that its product and organizational execution are not paper capabilities.

• Before listing, it had 24 outstanding binding orders plus additional framework and sales contracts, so demand is not idle.

• Southbound Stock Connect inclusion increases mainland capital accessibility and strengthens the liquidity foundation for a scarce asset.

• Listing proceeds materially strengthened cash resources, giving the company time to continue advancing products and customer expansion.

Bear reasons:

• Net operating cash outflow of RMB 2.137 billion in 2025 already shows commercialization is still consuming large amounts of cash.

• Revenue was only RMB 58.9 million in 2025H1, while full-year revenue reached RMB 1.035 billion, making recognition timing excessively back-loaded.

• Ending inventory was RMB 949 million, close to full-year revenue, so inventory digestion and cash collection still need proof.

• The combination of 6-month unlocking and the H-share full circulation application will directly pressure a high valuation through expanded share supply.

• The current near-100x trailing price-to-sales ratio already embeds very strong future growth assumptions, leaving a very thin margin of safety.

Pre-mortem

The first scenario most likely to make this investment lose 50% after three years is "revenue does not become continuous, and valuation compresses first". The path could be: the 2026 interim report shows revenue growth below market expectations, inventory stays high, BR20X timing slips, and by late 2026 to 2027 the market reclassifies Biren from "future platform company" to "project-based volatile company", compressing the forward price-to-sales ratio from the currently implied high level to around 20x. If revenue at that time reaches only RMB 2.0–2.5 billion, the corresponding share-price range would fall back to about HKD 18–25, more than 50% below the current price. This scenario does not require the company to have a "major problem"; growth only needs to be insufficiently fast to digest valuation.

The second scenario is "sudden supply expansion plus worsening competition". If H-share full circulation advances faster than expected in 2026H2, obvious selling appears after the 6-month unlock, and Huawei Ascend, Hygon DCU, or other domestic GPU players cut prices in key projects, Biren may have to sacrifice ASP to keep customers. If gross margin falls to around 30%, the market would cut profit expectations and valuation multiples at the same time. For a newly listed high-valuation stock, once fundamentals and share supply deteriorate in the same direction, the share price can reset quickly over one or two reporting cycles rather than decline slowly.

Final Research Conclusion

Biren Technology is a company worth serious research, but not worth chasing at any price. It has already proved it is not an empty concept: the team can make products, continue advancing under sanctions, produce the first meaningful revenue slope, and enter the public capital market. Many domestic hard-tech companies fail before product commercialization. Biren has at least reached the commercialization validation phase, which is difficult in itself. The problem is that the price capital markets give it today has already treated the "validation phase" as an "established growth phase".

What deserves the most caution today is the mismatch between pricing and timing, not the technology route. The company really needs to prove that the 2025 revenue jump is the prelude to sustainable repeat purchases and expansion over the next two to three years, rather than one-off project settlement. The share price has already prepaid for that path. If the next several reporting periods go well, today's concerns will be partly digested by growth. If they do not, the near-100x trailing price-to-sales ratio, full circulation, and unlocking window will make the drawdown much faster than many investors expect.

If I were to change my view, there would be only two types of triggers. The first is fundamental evidence: from the 2026 interim report through the 2027 annual report, the company keeps proving high revenue growth, gross-margin recovery, inventory reduction, and cash-flow improvement. The second is price evidence: the share price returns to a range that can absorb uncertainty, allowing investors to buy not only a story but also a margin of safety. Until at least one of these two types of evidence is satisfied, I prefer to view Biren as a high-volatility core watchlist stock, not an immediate long-term core holding candidate.

【Company Profile Score】

• Fundamental quality: Medium

• Growth: High

• Moat: Medium

• Financial resilience: Medium

• Management credibility: Medium

• Valuation attractiveness: Low

• Risk level: High

• Suitable investor type: High-risk speculation

【Investment Rating】

• Rating: Watch

• One-sentence investment thesis: Domestic GPGPU scarcity is real, but the current price already discounts several years of high-growth delivery.

• Three price signals: Ideal buy price: HKD 21–25

• Holdable price: HKD 33–42

• Clearly overvalued price: Above HKD 60

• Current price classification: Outside the three ranges

• Is it worth waiting for a better price: Yes. Better conditions for a buy would be the share price entering around HKD 25 or below, while the interim report also proves revenue was not one-off recognition and inventory no longer continues rising. The opportunity cost of waiting is that if southbound capital keeps flowing in and the theme keeps heating up, the share price may continue rising before fundamentals are fully verified.

• Target holding period: 1–3 years

• Expected annualized return: Bear about -20% to -25%; base about -5% to -10%; bull about +8% to +14%

• Maximum loss risk: 50% to 65%. Triggers include a material slowdown in 2026-2027 revenue growth, gross margin falling below 35%, full circulation advancing too quickly, and insufficient market absorption.

• Signals triggering reassessment: Gross margin below 35% for two consecutive reporting periods

• Inventory keeps rising while revenue does not scale in tandem

• Net operating cash outflow still shows no improvement after revenue expansion

• H-share full circulation is materially faster than market expectations, while the company does not provide matching order conversion

• BR20X launch timing or performance delivery materially lags management's roadmap

【Ideal/Fair Buy Price】HKD 21–25 Rationale: This range corresponds to the conversion of 2027 revenue and valuation multiples under the bear scenario, and additionally requires about a 20% margin of safety to hedge three uncertainties: discontinuous order conversion, unlocking and full-circulation supply, and cash flow weaker than the income statement.

【Valuation Range】

• current: 48.52 (intraday on 2026-06-11)

• bear (conservative · ideal buy zone): [21, 25]

• base (reasonable · acceptable holding zone): [33, 42]

• bull (optimistic · above clearly overvalued line): [60, 72]

Key Data Tables

Note†: Revenue, gross margin, R&D, and operating cash flow for 2022-2025H1 mainly come from the 2025 prospectus. 2025 full-year revenue comes from post-listing media summaries. Adjusted loss, R&D expense, operating cash flow, and ending inventory come from the 2025 annual report. Because publicly retrieved annual report summaries did not fully show 2025 full-year gross margin, this table retains "insufficient information" for that item.

Note‡: The Pathfinder SII total in the table comes from the prospectus lock-up table. The H-share portion is calculated by splitting the prospectus capitalization table. As of 2026-06-05, the H-share full circulation application had only been submitted, the filing had not been completed, and implementation details had not been finalized.

Research Uncertainties

• Since listing, the company has not yet gone through a full interim report and multiple performance verification cycles, so the publicly verifiable financial sample is very short.

• Public disclosure still provides limited third-party independent evidence on the performance, power consumption, compatibility, and customer repeat purchases for BR166 and later BR20X.

• Public materials can confirm order size, but cannot fully confirm customer structure, delivery timing, acceptance conditions, or collection terms for those orders.

• Although the H-share full circulation application has been submitted, the final approval timing, phased arrangements, and actual details of tradable shareholders remain unclear.

• Beyond Hygon and Cambricon, several of the most direct domestic GPU competitors are not yet listed, so peer comparison inevitably involves information asymmetry.

Reference Sources

The high-weight primary or authoritative sources for this report mainly include: Biren Technology's 2025 prospectus, 2025 annual report, IPO allotment results announcement, full exercise of over-allotment option announcement, May 2026 monthly return, and June 2026 H-share full circulation application announcement; HKEX Chapter 18C rules and institutional explanations; the official U.S. BIS announcement on the Entity List; the SSE official notice on Southbound Stock Connect stock adjustments; and annual reports or investor relations materials from Nvidia, AMD, Cambricon, and Hygon Information. Relevant core disclosures have been cited paragraph by paragraph in the main text.

Other Stocks Mentioned in the Report

• 688256.SHG — Core A-share domestic AI chip comparable, used to compare commercialization maturity and tolerance for high valuations.

• 688041.SHG — Leader in server processors and coprocessors, used to compare scale, profitability, and valuation center.

• NVDA.US — Global GPU platform leader, used to compare ecosystem moat, software stack, and industry profit pool.

• AMD.US — Global CPU/GPU vendor, used to compare platform-level product lines and mature-business gross-margin structure.

• 00981.HK — SMIC, an important reference for China's semiconductor manufacturing capability and domestic supply-chain constraints.

• 02007.HK — Country Garden, not a business comparable, but its venture-capital platform is one of Biren's Pathfinder SIIs, used to explain shareholder and unlocking structure.

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