Enflame vs Moore Threads vs MetaX vs Biren Compared as China Builds Its AI GPU Challengers

 

With Enflame Technology completing its STAR Market registration, the capital-markets map for China's second tier of AI chip designers is essentially complete. Moore Threads and MetaX are already listed in Shanghai, Biren trades in Hong Kong, and for the first time all four can be placed on the same table.

 

The comparison does not fully match the narrative of four champions arriving together. Per statistics carried by Xincai, 2025 revenue was highest at MetaX with RMB 1.644 billion, followed by Moore Threads at RMB 1.506 billion, Biren at RMB 1.035 billion and Iluvatar at RMB 1.034 billion. Per Ifeng's summary, Enflame's 2025 revenue was RMB 990 million with growth slowing to 37%. Over the same period Moore Threads and Biren grew 243.37% and 207.2% respectively, with MetaX at 121.26%.

 

In other words, the earliest-founded company among the four currently has the smallest revenue and the slowest growth. Understanding that contrast requires looking at architecture, customer structure, gross margin and listing sequence together rather than comparing revenue alone.

 

 

 

2025 revenue: MetaX RMB 1.644 billion, Moore Threads RMB 1.506 billion, Biren RMB 1.035 billion, Enflame RMB 990 million.

 

2025 revenue growth: Moore Threads 243.37%, Biren 207.2%, MetaX 121.26%, Enflame roughly 37%.

 

2025 losses: Moore Threads net loss attributable of RMB 1.024 billion (narrowing 36.70%), Biren adjusted annual loss of RMB 874 million (widening 13.9%), MetaX net loss attributable of RMB 789 million (narrowing 43.97%), Enflame net loss attributable of RMB 1.164 billion.

 

Gross margin comparison: Moore Threads above 70% in 2024, MetaX between 53% and 55% across 2023 to 2024, Enflame roughly 36%.

 

Customer structure: Enflame derived 83.79% of 2025 revenue from Tencent with top five customers at 96.89%; Moore Threads' top five customers exceeded 80% in every reporting period; MetaX's single Xiyun C500 product series accounted for 97.28% and 97.87% of main business revenue in 2024 and Q1 2025; Biren had 14 customers contributing RMB 337 million in 2024.

 

Listing venues and status: Moore Threads (688795) and MetaX (688802) trade on the STAR Market, Biren is listed in Hong Kong, and Enflame has completed STAR Market registration seeking RMB 6.0 billion.

 

MetaX achieved a profitable quarter in the second quarter of 2025, with revenue of RMB 595 million, up 237.87%, and net profit attributable of RMB 46.62 million.

 

Per IDC data, 2024 China GPU shipments were 1.9 million units for NVIDIA, 640,000 for Huawei Ascend, 69,000 for Kunlunxin, 38,000 for Iluvatar, 26,000 for Cambricon, 24,000 for MetaX and 13,000 for Enflame.

 

 

 

Placing the disclosed data side by side makes the positional differences immediately visible.

 

 

These figures come from different sources with different disclosure conventions, some are adjusted measures, and the gross margin years are not aligned, so the table illustrates orders of magnitude rather than precise comparison and each company's official filings are authoritative. No comparable public gross margin disclosure for Biren was obtained, so that cell is left unfilled.

 

 

An easily overlooked detail is that the label itself lacks a settled definition. Per The Paper, Moore Threads, MetaX, Biren and Iluvatar are called the four GPU champions because their revenue scale lags the first tier and none has turned profitable. Per Securities Times, the grouping is Moore Threads, MetaX, Biren and Enflame.

 

The disagreement is informative: Iluvatar and Enflame are close in scale, with 2025 revenue of RMB 1.034 billion and RMB 990 million respectively, while IDC data puts Iluvatar's 2024 China GPU shipments at 38,000 units against Enflame's 13,000. Ordering within this tier is not settled.

 

 

 

Founding team backgrounds directly shaped each company's architecture. Per QbitAI's summary, Moore Threads founder Zhang Jianzhong served as NVIDIA's global vice president and China general manager over more than 15 years at the company, while MetaX's founding team carries AMD DNA, with founder Chen Weiliang formerly senior director of AMD GPU design, CTO and chief hardware architect Peng Li being AMD's first ethnic Chinese female Fellow, and CTO and chief software architect Yang Jian having spent 14 years at AMD and served as GPU chief architect at Huawei HiSilicon.

 

Per Time Weekly coverage carried by Investing.com, Biren was founded in Shanghai in September 2019 by Zhang Wen, former president of SenseTime, whose founder has said China's domestic high-end GPU talent pool likely numbers under 500 people and that Biren's density of such talent should be the highest in the Chinese market.

 

Enflame's path is comparatively independent. Per Eastmoney's summary, the company has self-developed four generations of architecture across five cloud AI chips and built a full product system spanning AI chips, accelerator cards and modules, intelligent computing systems and clusters, and an AI computing and programming software platform named TopsRider.

 

 

Product positioning is among the four's most substantive differences. QbitAI's summary notes Moore Threads focuses on full-function GPU chips and computing platforms covering AI training and inference, graphics rendering and intelligent SoC, with a self-developed MUSA unified architecture supporting fusion of AI, graphics and scientific computing while remaining compatible with mainstream ecosystems. MetaX focuses on high-performance general-purpose GPU design and computing platforms covering AI training and inference, general computing and graphics rendering, with full-stack self-developed GPU IP and instruction set architecture, a software ecosystem compatible with CUDA, and support for kilo-card cluster deployment.

 

Moore Threads also addresses both data centre and consumer gaming graphics markets, the only one of the four to cover the consumer side.

 

Enflame currently concentrates on inference. Eastmoney's summary states plainly that its products are mainly used in inference scenarios, that its fourth generation and first combined training and inference product is about to enter mass production, and that a higher-performance fifth generation is in development.

 

MetaX started earlier on combined training and inference. Time Weekly's coverage shows MetaX launching its first such GPU, the Xiyun C500, in 2023, followed by the C550 and C588, with the C600 built on domestic supply chains having taped out and lit up.

 

For investors seeking detail on any single company's business model and how it trades, this explainer on Enflame Technology and how to trade it offers an entry point into this tier's product structure.

 

 

 

 

All four are concentrated, but the object of concentration differs, and that matters more than the percentage itself.

 

Enflame is customer-concentrated. Per Beijing News, top five customers accounted for 96.50%, 92.60% and 96.89% of revenue across 2023 to 2025, with sales to Tencent Technology (Shenzhen) at 33.34%, 37.77% and 83.79%, and Tencent Technology being a related party whose sales constitute connected transactions.

 

MetaX is product-concentrated. QbitAI's summary notes that in 2024 and the first quarter of 2025, MetaX's main revenue came from its core combined training and inference chip series Xiyun C500, accounting for 97.28% and 97.87% of main business revenue respectively.

 

Moore Threads has a broader product line, but the same summary notes its top five customers exceeded 80% of revenue in every reporting period.

 

Biren discloses customer counts most explicitly. Time Weekly's coverage shows 14 customers contributing RMB 337 million of revenue in 2024.

 

 

A distinction is warranted here. Product concentration represents a bet on a technical path and can be eased by launching new products. Customer concentration represents demand-side dependence and can only be eased by winning new customers, which is harder. And when the single customer is also the largest shareholder, the question of related-party pricing is layered on top.

 

Beijing News notes the Shanghai Stock Exchange has asked whether existing AI chips are highly adapted to Tencent and whether the company can develop and adapt to other customers, alongside questioning the reasonableness of the sharp increase in Tencent-related transactions. No equivalent item appears in the public review information for the other three.

 

 

 

Backlog is the key indicator for revenue over the next 12 to 18 months, but the four disclose it very differently, so direct comparison requires care.

 

The Paper reports MetaX's backlog reaching RMB 1.43 billion as of September 2025, close to twice its 2024 full-year revenue, and Moore Threads disclosing approximately RMB 2.004 billion of expected orders under negotiation for which customers have indicated purchase intent, including more than RMB 1.7 billion in AI computing cluster project contracts where customers have completed product testing or staged evaluation. Time Weekly's coverage shows Biren holding five framework sales agreements and 24 sales contracts totalling RMB 1.241 billion as of December 15.

 

Enflame discloses in a different unit. Ifeng's summary notes the company stating in its inquiry response that an order backlog exceeding 100,000 chips, mass production of the third-generation S60 inference accelerator card and contributions from kilo-card intelligent computing centre projects support breakeven as early as 2026. Disclosing in units rather than value makes it non-convertible against the other three.

 

Worth emphasising: Moore Threads' RMB 2.004 billion reflects purchase intent rather than signed contracts, MetaX's RMB 1.43 billion is stated excluding tax, and Biren's RMB 1.241 billion includes framework agreements. The three carry different degrees of commitment.

 

 

A first watershed has appeared in commercialisation. Time Weekly's coverage shows that from April to June 2025, MetaX generated revenue of RMB 595 million, up 237.87% year over year, with net profit attributable of RMB 46.62 million, achieving a profitable quarter.

 

That is the first verifiable quarterly profit among the four. It establishes that under current price and cost structures, profitability becomes mathematically achievable past a certain shipment scale, leaving the height of that threshold as the open question.

 

Enflame's timeline sits later. Per Sina Finance, the company projects consolidated profitability in 2026 or 2027, with first quarter 2026 revenue of RMB 287 million, up 1,474.85% year over year, and a net loss attributable of RMB 444 million.

 

 

 

Before debating who leads, a larger coordinate is necessary. Per The Paper, MetaX's prospectus states that the AI chip market has essentially formed an oligopoly of NVIDIA and AMD with NVIDIA holding above 80% global share, while Yicai calculations put MetaX's 2024 China AI chip market share at roughly 1%. Biren states that by 2024 China-market revenue the top two participants together held 94.4% of the market, with the remainder relatively dispersed across more than 15 scaled participants none of which exceeded 1%, and that in China's general-purpose GPU market two US vendors hold 98%.

 

The same report's R&D data shows the intensity of investment: across 2022 to 2024, R&D expenses were RMB 3.81 billion at Moore Threads, RMB 2.247 billion at MetaX and RMB 2.73 billion at Biren. Enflame's 2023 to 2025 R&D totalled RMB 3.676 billion, or 1.83 times cumulative revenue over the same period.

 

 

The Paper cites Omdia chief semiconductor analyst He Hui observing that domestic GPU companies still lack real earnings support, that most remain lossmaking with small revenue scale, and that the market assigns high valuations based on expectations of future growth.

 

That observation explains an apparent anomaly: the four generated roughly RMB 5.2 billion of combined 2025 revenue while market capitalisations run into the hundreds of billions of renminbi. Valuations anchor on the long-run domestic substitution opportunity rather than current revenue or profit, which means any information about the pace of substitution moves prices far more than ordinary results data does.

 

 

The first risk is homogeneous competition. All four compete for the residual space in a market where two US vendors hold 98% of general-purpose GPU share, and the technical differences among them have not yet produced clear stratification. That is tolerable while demand grows rapidly, but once growth slows, price competition would compress already modest gross margins.

 

The second is the software ecosystem. Hardware performance can be closed through tape-out iteration, while ecosystems require accumulated developer time. MetaX chose CUDA compatibility, Moore Threads built the MUSA unified architecture with mainstream compatibility, and Enflame has its own TopsRider platform, and the migration cost of each path differs for customers.

 

The third is customer and product concentration. Enflame's single customer at 83.79% involves related-party transactions, MetaX's single product series exceeds 97%, and Moore Threads' top five customers exceed 80%, all of which amplify the transmission of demand fluctuations.

 

The fourth is the distance between valuation and results. With combined 2025 revenue near RMB 5.2 billion and none profitable on an annual basis, market capitalisations are already high, so any delivery data below expectation gets amplified.

 

On scenarios, the base case is the industry maintaining high growth with all four scaling shipments at their own pace, MetaX's quarterly profit gradually replicated by the others, and attention shifting from growth rate to gross margin and cash flow. A second is a clear winner emerging, with one company establishing a generational advantage in combined training and inference or cluster capability, accelerating the concentration of resources and orders and widening valuation dispersion. A third is demand slowing, where given industry-wide R&D intensity and loss levels, financing capability becomes the primary determinant of survival order.

 

Status distinctions: financial data, customer structures and backlog figures come from prospectuses, inquiry responses and credible reporting with differing conventions and cut-off dates. Enflame's pricing and listing date had not been formally announced as of writing. Market share figures come from third-party research estimates. Analyst views are those institutions' judgements rather than statements of fact.

 

 

Placing the four side by side, the most counterintuitive finding is that the earliest-founded company, Enflame, currently ranks last on several dimensions: 2025 revenue of RMB 990 million is the lowest, growth of roughly 37% sits far below the 121.26% to 243.37% range of the other three, gross margin near 36% trails MetaX's 53% to 55% and Moore Threads' above 70% in 2024, and single-customer concentration at 83.79% is the highest of the four. That combination sits in visible tension with the optimistic framing of four champions arriving together, and that tension is exactly what investors need to verify for themselves.

 

Three misreadings look likely. The first is treating revenue scale as a proxy for technical leadership. More than 97% of MetaX's RMB 1.644 billion came from the single Xiyun C500 series, and 83.79% of Enflame's RMB 990 million came from a single customer. The two forms of concentration differ in nature but both indicate that the durability of current revenue is unproven, so ranking by scale easily produces the wrong technical conclusion. The second is ignoring the gross margin gap. Moore Threads exceeded 70% in 2024 against Enflame's roughly 36%, a difference approaching twofold that cannot plausibly be explained by scale effects alone within the same sector. It more likely reflects differing product form, meaning the revenue mix between accelerator cards and full system deployments, which directly determines future earnings leverage. The third is treating narrowing losses as a uniform trend. MetaX narrowed 43.97% and Moore Threads narrowed 36.70%, while Biren's adjusted annual loss widened 13.9%. The direction is not consistent.

 

Three verifiable metrics deserve tracking, and none of them is revenue. First, each company's gross margin trajectory, the most direct quantification of product competitiveness and the leading signal for whether price competition has begun. Second, how quickly quarterly profitability spreads. MetaX achieved it in the second quarter of 2025, and if the other three follow within several quarters the scale threshold has been crossed, whereas if only one sustains it the lead is structural. Third, the absolute level of non-Tencent revenue in Enflame's 2026 interim report, given first-half revenue guidance of RMB 1.06 billion to RMB 1.15 billion. Whether that increment comes from Tencent or from other customers determines its true position within this tier.

 

For cross-asset investors, the lesson is that companies within a single sector are frequently priced as one exposure while their risk structures differ entirely. These four diverge on architecture, customer structure, product form and listing venue, meaning identical sector-level good news transmits to them at different magnitudes and identical bad news hits them asymmetrically. Crypto knows the pattern well: protocols on a single chain tend to rise together in a bull market, while genuine differences only surface under stress, at which point outcomes are determined by each protocol's revenue sources and liquidity structure rather than by sector membership. When evaluating names like this, establish each company's revenue composition and form of concentration first, then assess how sector variables transmit. That sequence controls risk better than allocating to the sector as a whole. From a risk management standpoint, while the entire sector remains unprofitable on an annual basis and valuations anchor on long-run potential, diversification fits the available information better than concentration does.

 

This analysis rests on published prospectus disclosures, regulatory inquiry records and credible reporting available now. Subsequent financial data, listing progress and industry structure could each change the conclusion, and no single scenario should be treated as a fixed expectation.

 

 

 

The list is not settled. Some reporting names Moore Threads, MetaX, Biren and Iluvatar, while other reporting names Moore Threads, MetaX, Biren and Enflame. That disagreement reflects unsettled ordering within the tier: 2025 revenue was RMB 1.034 billion at Iluvatar against RMB 990 million at Enflame, while IDC data puts 2024 China GPU shipments at 38,000 units for Iluvatar against 13,000 for Enflame.

 

 

On 2025 data, MetaX leads at RMB 1.644 billion, followed by Moore Threads at RMB 1.506 billion, Biren at RMB 1.035 billion and Enflame at RMB 990 million. The growth ranking differs: Moore Threads 243.37%, Biren 207.2%, MetaX 121.26% and Enflame roughly 37%. Note that revenue scale does not directly equate to technical leadership, since revenue composition and product form differ across the four.

 

 

Not on an annual basis. None was profitable in 2025: Moore Threads posted a net loss attributable of RMB 1.024 billion, Enflame RMB 1.164 billion, MetaX RMB 789 million, and Biren an adjusted annual loss of RMB 874 million. MetaX did achieve a profitable quarter in the second quarter of 2025 with revenue of RMB 595 million, up 237.87%, and net profit attributable of RMB 46.62 million, the first verifiable quarterly profit among the four.

 

 

Founding backgrounds shaped the paths. Moore Threads' founder came from NVIDIA, focusing on full-function GPUs with a self-developed MUSA unified architecture while addressing both data centre and consumer graphics. MetaX's founding team came from AMD, with full-stack self-developed GPU IP and instruction set, CUDA-compatible software and kilo-card cluster support. Biren was founded by a former SenseTime president emphasising high-end talent density. Enflame has developed four architecture generations across five cloud AI chips with the TopsRider software platform, currently focused on inference.

 

 

All four are concentrated but differently. Enflame is customer-concentrated, with Tencent at 83.79% of 2025 revenue, top five customers at 96.89%, and Tencent as a related party. MetaX is product-concentrated, with the Xiyun C500 series at 97.28% and 97.87% of main business revenue in 2024 and Q1 2025. Moore Threads' top five customers exceeded 80% in every reporting period. Biren had 14 customers contributing RMB 337 million in 2024.

 

 

Conventions differ, so direct comparison requires care. MetaX reported RMB 1.43 billion excluding tax as of September 2025, close to twice its 2024 full-year revenue. Moore Threads disclosed approximately RMB 2.004 billion of expected orders under negotiation with indicated purchase intent, including over RMB 1.7 billion in computing cluster projects. Biren held five framework agreements and 24 sales contracts totalling RMB 1.241 billion as of December 15. Enflame discloses in units, citing backlog exceeding 100,000 chips.

 

 

Very little. Per prospectus disclosures and third-party estimates, NVIDIA holds above 80% of the global AI chip market, MetaX's 2024 China AI chip share was roughly 1%, Biren states the top two participants held 94.4% of the China market by 2024 revenue with more than 15 other scaled participants none exceeding 1%, and two US vendors hold 98% of China's general-purpose GPU market.

 

 

Valuations anchor on long-run substitution potential rather than current results. A semiconductor analyst has noted that domestic GPU companies lack real earnings support, that most remain lossmaking with small revenue scale, and that the market assigns high valuations based on future growth expectations. The four generated roughly RMB 5.2 billion of combined 2025 revenue while none was annually profitable, and that gap means information about substitution progress moves prices far more than ordinary results data.

 

 

This article is provided for informational and research purposes only and does not constitute investment advice, financial advice, legal advice, tax advice or any recommendation to transact. The financial data, customer structures, order backlogs and market share figures referenced here come from publicly disclosed prospectuses, inquiry responses, third-party research estimates and credible reporting; sources differ in accounting convention, disclosure cut-off date and adjustment methodology, the comparison table illustrates orders of magnitude rather than precise comparison, and each company's official filings are authoritative; comparable data for certain companies could not be obtained and is marked as such in the text; as of writing, no formal announcement of Enflame's pricing or listing date has been identified; and analyst views and third-party estimates are those institutions' judgements rather than statements of fact. Prices of crypto assets, equities and other related financial instruments can move sharply over short periods, particularly for unprofitable companies whose valuations anchor on long-run expectations, and newly listed securities are especially prone to severe volatility in early trading, with the possibility of total loss of principal. Futures contracts referencing a stock carry leverage risk that magnifies losses proportionally, holding such contracts is not equity ownership and confers no shareholder rights, and the risk profile differs fundamentally from holding shares directly. Historical performance, technical indicators and on-chain data cannot guarantee future outcomes. Readers should conduct their own independent research, verify official information directly, and evaluate any decision against their own financial circumstances, investment objectives, experience and risk tolerance, consulting a qualified professional adviser where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect loss arising from use of or reliance on the information 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.

 

Areas of Expertise:

 

 

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)