Enflame IPO Date Set: Tencent-Backed AI Chipmaker Opens Subscriptions on September 2

 

The last of China's four homegrown AI chip contenders finally has a timetable. On the evening of August 24, the Shanghai Stock Exchange published Shanghai Enflame Technology's issuance arrangements and preliminary book-building announcement for its STAR Market listing. According to reporting from JRJ, the company will offer 43,035,173 shares, representing 10% of post-offering share capital, with online and offline subscription set for September 2.

 

For investors who could previously only track Enflame as a pending deal, the filing supplies the three variables that were missing: preliminary book-building on August 28, subscription on September 2, and an online subscription code of 787801, with the listed ticker set as 688801. The offer price has not been set. It will emerge from the August 28 book-building process, which means every discussion of market capitalisation and valuation currently rests on assumptions.

 

What deserves attention is not the calendar but the financial structure this company brings to public markets: cumulative non-recurring net losses above RMB 40 billion over three years, more than eighty percent of 2025 revenue coming from a single customer, and that customer being Tencent, which also holds over 20% of the equity. This is a deal being priced simultaneously against a domestic substitution narrative, a concentration risk, and a mechanism built for unprofitable issuers.

 

 

 

On structure, the offering combines strategic placement, offline institutional allocation and online retail distribution, with an initial strategic placement of 8,607,034 shares, an initial offline tranche of 27,542,639 shares and an initial online tranche of 6,885,500 shares. CITIC Securities is sponsor, with Guotai Haitong Securities and GF Securities acting as joint lead underwriters. The company states plainly that it is not yet profitable and will be classified into the STAR Market growth tier if that remains true at listing. Proceeds are targeted at RMB 6 billion, roughly $880 million at recent exchange rates.

 

On fundamentals, revenue ran RMB 301 million, RMB 722 million and RMB 990 million across 2023 to 2025, while net loss attributable to shareholders excluding non-recurring items came in at RMB 1.567 billion, RMB 1.503 billion and RMB 1.197 billion over the same period. For the first half of 2026 the company guides revenue growth of 259% to 289% and an adjusted net loss between RMB 614 million and RMB 582 million. Tencent holds roughly 20% and contributed 83.8% of 2025 sales revenue, while Enflame's share of China's AI accelerator card market sits near 1.7%.

 

 

 

The path has not been slow. Per Ths Finance carrying Dahe Cube's account, the application was accepted on January 22, cleared the listing committee on June 15, and received CSRC registration approval on July 9. Roughly six weeks separated registration from the prospectus filing, which is normal for recent hard-technology deals on this board. That gap says the window was chosen for market conditions rather than forced by review obstacles.

 

Worth noting is that the company re-filed tutoring registration with the Shanghai securities regulator in November 2025, describing the move at the time as a response to changing capital market conditions and strategic needs. In hindsight, that moment sat immediately after Moore Threads and MetaX completed their A-share listings, and the pricing precedent set by peers plainly shaped the timing of the last one to go.

 

 

The online tranche of 6,885,500 shares accounts for roughly 16% of the total offering. Strategic placement takes 8,607,034 shares and the offline institutional tranche 27,542,639. The clawback mechanism can shift the split when online subscription multiples cross thresholds, but with total issuance at just over 43 million shares, the online allocation rate will almost certainly stay very low.

 

The absence of a price is the point that bears repeating. Book-building on August 28 will produce a price range from institutional bids, and only then will the final offer price appear. Because the company is unprofitable, conventional price-to-earnings anchoring does not function here. Offline institutions are more likely to lean on price-to-sales multiples, compute-based comparables and the post-listing trading levels of peers.

 

 

 

According to Sina Finance's review of the prospectus materials, proceeds are allocated as roughly RMB 1.503 billion to fifth-generation AI chip development and industrialisation, RMB 1.197 billion to the sixth-generation programme, and about RMB 3.3 billion to advanced AI hardware and software collaborative innovation.

 

That split is revealing. The software and hardware integration project takes more than half the money, exceeding the two chip programmes combined. For a domestic AI chip vendor, hardware gaps can be narrowed through process nodes and architecture iteration, but software stacks, operator libraries and framework compatibility are usually where switching costs actually live. Putting the largest allocation into ecosystem rather than tape-out says management sees the binding constraint on the software side.

 

Enflame's product line currently centres on its Suisi chip series, CloudBlazer accelerator cards and cluster-level compute systems, spanning both training and inference. The fifth and sixth generation products are scheduled well into the future, which means this raise funds a multi-year research cycle rather than revenue that lands within a quarter or two.

 

 

The company itself flags that it will enter the STAR Market growth tier if still unprofitable at listing. That framework is the precondition for this entire cohort. Rules accommodating unprofitable hard-technology issuers are why all four of these companies reached public markets during early commercialisation. Caixin reported that prospectus materials place the expected turn to profitability somewhere between 2026 and 2027.

 

That is a company expectation, not a commitment or guidance. For a business whose research spending has persistently exceeded revenue, the profitability inflection depends on shipment ramp and per-card margin, neither of which has yet been tested by quarterly disclosure under public market scrutiny.

 

 

 

Tencent occupies an unusual position here. Caixin reported that Tencent and affiliates hold roughly 20% after investing RMB 5.1 billion, and that the company generated 83.8% of 2025 sales revenue from Tencent, up from around 38% in 2024. Put plainly, most of the revenue growth of the past three years has been driven by purchases from a single shareholder-customer.

 

Bloomberg noted in June that Tencent remains a dominant customer alongside its equity position. The benefit is direct: a startup holding roughly 1.7% of the market is able to secure volume orders and validate multiple product generations at scale precisely because of this relationship. The company has said Tencent's demand for accelerator cards exceeds what it can supply.

 

The risk is equally clear. When customer and shareholder overlap, transfer pricing draws sustained scrutiny and the supplier's negotiating room is structurally compressed. More importantly, the exposure is one-sided. If Tencent shifts its compute procurement strategy toward more in-house silicon or alternative vendors, Enflame has no second growth engine of comparable size to absorb the loss.

 

 

In book-building, customer concentration typically shows up as a valuation discount. Institutions pricing an unprofitable chip company weigh revenue growth against revenue quality. Enflame scores strongly on the first and carries a visible discount on the second. How those two forces net out is something only observable after August 28.

 

Over a longer horizon, the measure of whether this listing succeeds should not be first-day performance but whether the company can lift the non-Tencent share of revenue over the following several quarters. Whether the internet, intelligent computing centre and financial verticals described in the prospectus generate durable orders will say more than any single session of price action.

 

For investors looking to observe expectations before the A-share price is fixed, platforms including MEXC have listed Enflame-linked equity contracts whose quotes can serve as a sentiment reference. One caveat is essential: those quotes carry no direct relationship to the offer price produced by STAR Market book-building, since the participant base and trading mechanisms differ entirely.

 

 

 

 

Three years of data form a clean line. Revenue rose from RMB 301 million in 2023 to RMB 722 million in 2024 and RMB 990 million in 2025. Adjusted net loss narrowed from RMB 1.567 billion to RMB 1.503 billion and then to RMB 1.197 billion. Cumulative adjusted losses of roughly RMB 4.27 billion over that stretch exceed total revenue for the same period by more than double.

 

The first-half 2026 guidance is the number to watch. Revenue growth of 259% to 289% represents a clear step change in magnitude, while the adjusted loss of RMB 582 million to RMB 614 million did not expand proportionally, indicating scale effects have begun to appear. Still, burning close to RMB 600 million in half a year means cash consumption remains rapid absent the raise, which explains why RMB 6 billion is the target rather than something smaller.

 

 

Caixin's figures place Enflame's share of China's AI accelerator card market near 1.7%, in a market still dominated by Nvidia. On broader industry measures, individual domestic GPU vendors have generally hovered around 1%, with Bernstein Research data showing Cambricon, Moore Threads and MetaX all clustered at that level in 2024.

 

Those numbers exist to calibrate expectations. Domestic substitution is a large enough story, but the slice each company currently holds remains small. Pricing the offering entirely off the ceiling of the substitution opportunity leaves a wide gap against present shipment reality.

 

 

 

Three listed peers provide direct reference points, though the sample itself is highly distorted. Per Sina's tally, Moore Threads and MetaX rose 425.46% and 692.95% respectively on their A-share debuts. Biren Technology listed in Hong Kong on January 2, 2026 at HK$19.60, opened 82.14% higher and closed at HK$34.46 for a 75.82% gain, valuing it at HK$82.6 billion.

 

The gap between the two markets is the most informative part of that data. Same sector, comparable commercialisation stage, radically different pricing outcomes. That tells you the A-share debut moves contained substantial components unrelated to fundamentals: new-issue supply cadence, the structure of subscription capital, and the pricing conventions applied to unprofitable STAR Market issuers.

 

 

The follow-through matters more. Moore Threads climbed as high as RMB 941.08 before retreating for several sessions, down roughly a quarter from that peak by the December 16 close. By early January, both Moore Threads and MetaX were trading well below their highs, and Biren had also pulled back after listing.

 

For anyone participating in book-building or subscription here, that observation is worth more than the debut percentages. Peers went through a transition from narrative pricing to earnings verification, and Enflame arrives last into a market that has already partially digested the substitution premium. Scarcity value applied to the fourth entrant is usually smaller than for the first.

 

 

Price is the first variable. Only after August 28 will the final offer price and implied post-offering valuation become known. A price materially above comparable peer levels would compress the margin of safety after listing.

 

Customer mix is the second. The first periodic report after listing will be the first disclosure of Tencent's related-party sales share under public company reporting standards. Whether that ratio keeps climbing or starts to fall determines how the market prices revenue quality.

 

Product cadence is the third. The fifth and sixth generation programmes funded by this raise are multi-year projects, and interim iteration and customer validation progress will need to be tracked through announcements and results calls.

 

Two macro factors also deserve weight. Access to advanced process nodes remains constrained by external restrictions and feeds directly into product competitiveness and delivery schedules. And the broader capital expenditure cycle matters: if procurement plans at China's large internet platforms decelerate, every downstream chip vendor feels it at once, and a company with concentrated customer exposure absorbs a magnified share of that shock.

 

 

What deserves recording about this deal is not the September 2 date. It is that once all four of these companies trade publicly, the pricing method for domestic AI silicon has to change. When only one name in a sector is buyable, the market is buying scarcity. When four have continuous quotes, the market is buying relative valuation. Enflame is the last one in, and its book-building is the first to run with a complete set of comparables on screen.

 

The most likely misreading is treating the 400% to 700% debut moves of the first two A-share listings as a baseline expectation. Those numbers reflect a specific supply and demand condition in the new-issue market far more than any judgment about enterprise value. Biren, from the same cohort, gained 75.82% on its Hong Kong debut. When the same category of asset produces answers an order of magnitude apart under different pricing mechanisms, the disciplined response is to treat both as range endpoints rather than selecting whichever is more convenient as an anchor.

 

The second underappreciated point concerns the weighting of revenue quality against revenue growth. Tencent accounted for 83.8% of revenue in 2025, up from roughly 38% in 2024. The slope of that line is itself the signal: revenue growth and customer concentration rose together. That means growth cannot be read in isolation at this stage. The way to assess genuine competitiveness is to strip out Tencent-related revenue and examine the growth rate of what remains, and that series will not be continuously disclosed until periodic reports arrive post-listing.

 

From a risk management standpoint, three information sets matter most between the close of book-building and the first trading day. The first is the price-to-sales multiple implied by the final offer price, compared laterally against where Moore Threads and MetaX currently trade, which is the most direct test for a pricing gap. The second is the final strategic placement roster and the distribution of valid offline bids, since the dispersion of quotes usually reveals more about institutional disagreement than the median does. The third is early turnover and the lock-up structure, because the peer experience suggests sentiment tends to fade two to four weeks after debut rather than on day one.

 

There is a cross-asset dimension as well. Public market pricing of domestic AI compute is, at bottom, a vote on the durability of China's AI capital expenditure cycle. That cycle drives A-share semiconductors, Hong Kong technology names, and at the far end connects to globally correlated high-beta assets. Once all four are listed, the market gains a set of continuously observable price signals for judging whether this compute investment wave sits early or mid-cycle. The value of those signals lies in their continuity, not in any single day's move. Turning them into a reliable sector indicator will take at least several quarters of reported results.

 

 

 

Online and offline subscription is set for September 2, 2026, with an online subscription code of 787801. Preliminary book-building takes place on August 28, and the final offer price will be determined after that process concludes. The offering covers 43,035,173 shares, or 10% of post-offering share capital of roughly 430 million shares, with an initial online tranche of 6,885,500 shares. Given the modest online allocation, the retail allotment rate is likely to be very low.

 

 

The target is RMB 6 billion, roughly $880 million at recent exchange rates. Proceeds break down into about RMB 1.503 billion for fifth-generation AI chip development and industrialisation, RMB 1.197 billion for the sixth-generation programme, and around RMB 3.3 billion for advanced AI hardware and software collaborative innovation. The software project receives more than the two chip programmes combined, indicating management views ecosystem maturity as the current bottleneck.

 

 

Not yet. The price emerges from preliminary book-building on August 28, formed from bids submitted by qualified offline institutional investors. Until then, any discussion of offering market capitalisation, oversubscription or post-listing valuation rests on assumptions. Because the company is unprofitable, price-to-earnings anchoring has limited relevance, and institutional bids will more likely reference price-to-sales multiples and peer trading levels.

 

 

Tencent is simultaneously shareholder and largest customer. Caixin reported that Tencent and affiliates hold roughly 20% after investing RMB 5.1 billion, and that the relationship generated 83.8% of Enflame's 2025 sales revenue, up from around 38% in 2024. That structure secured stable orders during early commercialisation while creating clear concentration risk and ongoing scrutiny of related-party transaction pricing.

 

 

No. Net loss attributable to shareholders excluding non-recurring items came to RMB 1.567 billion in 2023, RMB 1.503 billion in 2024 and RMB 1.197 billion in 2025, totalling roughly RMB 4.27 billion over three years. First-half 2026 guidance points to an adjusted net loss between RMB 582 million and RMB 614 million. The company explicitly flags that it will be placed in the STAR Market growth tier if still unprofitable at listing.

 

 

With caution. Moore Threads and MetaX gained 425.46% and 692.95% on their A-share debuts, while Biren rose just 75.82% in Hong Kong. The same sector produced pricing outcomes an order of magnitude apart across the two markets, which shows how heavily debut moves depend on supply structure and market mechanics. All three also retraced meaningfully after their peaks, with the sentiment fade typically arriving within weeks of listing.

 

 

Caixin's figures put its share of China's AI accelerator card market near 1.7%, a market still led by Nvidia. On broader industry measures, individual domestic GPU vendors have generally held around 1%. Domestic substitution is a substantial long-term direction, but the share each company presently commands remains limited, so pricing directly off the ceiling of that opportunity is not well supported by shipment data.

 

 

Three things. The price-to-sales multiple implied by the final offer price relative to current peer trading levels, which determines the early margin of safety. The direction of Tencent's related-party revenue share in the first periodic report, which reflects whether revenue quality is improving. And progress on fifth and sixth generation development and customer validation, since the raise funds a multi-year product cycle whose delivery has to be verified through continued disclosure.

 

 

This article is provided for information and market analysis purposes only and does not constitute investment advice, financial advice, legal advice, tax advice, or any recommendation to subscribe or transact. New issues carry the risk of trading below the offer price, and prices of equities, crypto assets and other related financial instruments can move sharply. None of the financial data, issuance arrangements, regulatory milestones or peer price performance referenced here can guarantee future outcomes. The offer price, proceeds, profitability expectations and product plans described remain subject to change, and the official announcements of the issuer and the Shanghai Stock Exchange take precedence. Readers should verify the latest disclosure before acting, reach independent conclusions based on their own financial circumstances, investment objectives, experience and risk tolerance, and consult a qualified professional where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect loss arising from the use of, or reliance on, the information contained in this article.

 

 

James Mitchell specializes in technical analysis, market trends, and trading strategies for both Bitcoin and altcoins. Based in London, he has over 10 years of experience in financial markets. Before joining MEXC Learn, James worked as a senior analyst at a leading European investment firm, where he developed expertise in risk management and quantitative trading. His transition to cryptocurrency markets began in 2017, and he has since become recognized for his data-driven approach. He holds a Master's degree in Financial Economics from the London School of Economics. His analytical approach combines traditional technical analysis with on-chain metrics to provide readers with actionable insights.

 

His areas of expertise span technical analysis, market trends and cycles, trading strategies, Bitcoin and altcoin analysis, and risk management.

 

 

 

Want the fastest access to MEXC's latest updates? Join our official Telegram group now!

MEXC Platform: Download App | Zero Fee Trading | Buy USDT Now

Join MEXC Community: X (Twitter) | Telegram | Discord

Get Insights: MEXC Learn | Terms of Service | MEXC Blog | MEXC News

Account Verification: Understand KYC | How to Complete KYC

External Content Platforms: Substack | Medium | Paragraph | LinkedIn | X(News)