중국의 4개 GPU 제조사(Biren Technology, Enflame, MetaX, Moore Threads, Zhuiyuan Technology)가 공개한 반기 보고서에서 뚜렷한 수익성 격차가 드러났으며, Biren은 매출을 약 20배 성장시켰음에도 적자 상태를 유지하고 있어 공공 시장에서의 감시가 강화되고 있다.

Earlier, Biren Technology, Tianday Zhixin, MetaX, and Moore Threads successively released their first or most recent semi-annual reports since going public. In September, Enflame also officially initiated its subscription process, marking an era where domestic GPUs are entering the capital markets with unprecedented density.

But as more domestic GPU companies face public market scrutiny, the most intuitive question—“who profits first”—is losing its explanatory power.

The four semi-annual reports present starkly different income statements: Biren’s revenue grew nearly 20 times year-on-year yet remains unprofitable; Tianday Zhixin and MetaX have achieved accounting profitability, but their profits contain factors that cannot be simply equated with core operational improvement; Moore Threads is one step away from accounting breakeven, yet its net profit after deducting non-recurring gains and losses (non-GAAP) remains negative.

Financial reports are becoming more polished, but what the secondary market is truly asking now is not who crosses the breakeven line first, but whose growth can be sustainably delivered.

After Lei Feng Network spoke with several secondary market research professionals, a consensus emerged despite differing views on valuations and stock prices: For domestic GPU companies still in rapid commercialization phases, current-period net profit is not the primary pricing variable.

Whether revenue can continue growing, the quality of orders, whether supply can support deliveries, and whether these expectations have already been traded into stock prices are becoming far more critical questions.

As investible targets multiply, the market is shifting from “which domestic GPU can we buy?” to “whose growth is worth buying.”

**01**

Turning profitable is not the watershed moment for financial reports. On the surface, the four companies fall into different tiers: some still report losses, others have turned a profit, and one is just steps away from breakeven. But digging into the income statements reveals this ranking is fragile. Biren Technology is the most typical example. In the first half of fiscal year 2026, it reported revenue of 1.236 billion RMB while recording a net loss of 377 million RMB. Compared to a 1.601 billion RMB loss in the same period last year, the loss narrowed by over 70%.

This looks like a strong figure, but restoring the financial changes before and after listing shows the actual operational improvement is not as dramatic. Last year’s loss included approximately 1.011 billion RMB from a financial gain or loss related to pre-IPO financing arrangements, not core operations. With the company’s January listing, related redemption rights terminated, and this financial impact ceased. According to the company’s disclosed non-IFRS figures, the adjusted loss decreased from 552 million RMB to 337 million RMB, a narrowing of about 39%.

In other words, Biren’s losses are indeed improving, but a “76% narrowing in losses” does not equate to a 76% improvement in operational conditions. The structural financial changes around the listing amplified the apparent improvement on the income statement.

“We don’t really look at current-period profit.” Secondary market analyst Li Ran told Lei Feng Network. For domestic GPU companies still in rapid commercialization phases, he focuses more on revenue, gross margins, and balance sheet metrics.

On the other end, Tianday Zhixin has also crossed the accounting profitability line, reporting a net profit of 106 million RMB for the first half, turning a profit year-on-year. However, a closer look at the report reveals a gain far exceeding the net profit itself behind that figure.

During the first half, Tianday Zhixin recorded a 760 million RMB gain from the appreciation of its holdings in listed company equity, which was not cash earned from the GPU business itself. The company explicitly stated that this turnaround was primarily driven by this accounting gain.

The semi-annual report shows Tianday Zhixin’s general-purpose GPU revenue has clearly ramped up. But based solely on the 106 million RMB net profit figure, one cannot directly conclude that the core GPU business has become profitable.

Li Ran’s assessment of Tianday Zhixin’s current stage is: “It stands on the eve of a major volume expansion in its core business.”

MetaX’s situation differs again. Its semi-annual report shows net profit attributable to shareholders reached 612 million RMB, turning from a loss last year to a profit, yet after non-recurring adjustments, the company still posted a loss of 48.86 million RMB. This means there remains a gap between accounting profitability and non-GAAP profitability.

Moore Threads, meanwhile, is approaching accounting breakeven. Its first-half net loss attributable to parents was only 11.56 million RMB, but the non-GAAP loss remained at 151 million RMB. Rather than simply discussing “how far from profitability,” the persistent non-GAAP loss indicates the company’s core operating metrics have not yet turned positive in tandem.

Thus, a notable misalignment emerges among the four companies: Biren’s losses appear to narrow faster on a comparable basis; Tianday Zhixin and MetaX have achieved accounting profitability, but their income statements contain significant non-recurring or non-core factors; Moore Threads is one step from accounting breakeven, yet remains unprofitable on a non-GAAP basis.

Ranking the four companies purely by profitability versus losses easily yields a distorted answer.

Zhou Heng, an A-share institutional researcher, told Lei Feng Network that for domestic GPU companies still in commercial expansion phases, whether core business revenue can continue growing is far more noteworthy than occasional accounting turnarounds.

For instance, regarding MetaX’s accounting profit alongside a non-GAAP loss, he believes this discrepancy won’t change his judgment: “Neither bulls nor bears rely on this metric.”

In Zhou Heng’s view, the underlying reason lies in a shifting valuation logic for domestic GPUs. As investible targets increase, the pricing premium previously driven by scarcity is weakening, and the market is beginning to compare growth trajectories more closely. “At this stage, the sustainable growth of operating revenue is far more important than occasional accounting turnarounds.”

If profitability is no longer the watershed, what kind of growth is credible?

**02**

Can revenue growth be sustained? It depends on orders and supply. “We generally prioritize looking at current-period revenue first, checking whether it holds up,” Li Ran told Lei Feng Network.

In his view, for domestic GPU companies still in commercial expansion, the market focuses more on three dimensions: revenue, demand, and supply. Revenue represents realized performance, orders reflect future demand, and supply determines whether that demand can ultimately be converted into deliveries.

Li Ran emphasizes that what truly needs verification is whether revenue growth can be sustained. Orders are precisely a key indicator for observing future demand. Whether these orders persist, their client profiles, and whether supply can support subsequent deliveries will all impact whether follow-up growth materializes.

In his assessment, securing procurement system access to top-tier internet companies often carries more reference value than clients like telecom operators or local intelligent computing centers. On the supply side, further observation is needed regarding wafer tape-outs, capacity guarantees, HBM procurement, and supply channels.

This means that after revenue begins to surge rapidly, the market’s concern shifts from merely “how much was sold” to what supports this growth and how long it can last.

Taking Tianday Zhixin as an example, in the first half of fiscal year 2026, general-purpose GPU product revenue reached 916 million RMB. Among them, inference series revenue surged 651.8% year-on-year to 654 million RMB, which the company attributed to rising AI inference market demand, along with increased product sales and pricing.

As for the training series, although revenue grew 38.0% compared to the same period last year, its slower growth rate relative to the inference series caused its revenue share to drop by 30.8 percentage points, falling to 27.7%.

This shift in Tianday Zhixin’s revenue structure directly impacted its gross margin performance. The company’s overall general-purpose GPU gross margin fell from 50.3% to 43.9%, which the report explicitly attributes to product mix changes. Meanwhile, the gross margins for both the training and inference product lines actually improved individually.

In other words, the decline in Tianday Zhixin’s overall GPU gross margin does not indicate a simultaneous deterioration in the profitability of its two product lines, but rather stems from a revenue shift toward the lower-margin inference products.

Li Ran judges that Tianday Zhixin’s core business is nearing a larger-scale volume expansion, with the subsequent key lying in whether new products can further enter the procurement systems of top-tier clients.

Biren’s financial report offers another way to observe the future. “When revenue looks solid, we prioritize specific balance sheet indicators that can foreshadow performance over the next six to twelve months,” Li Ran said. These include prepayments, contract liabilities, and inventory.

Biren’s prepayments for inventory and services rose from 465 million RMB at year-end to 1.534 billion RMB in the first half, a 230% year-on-year increase. This at least indicates a significant rise in upfront investments for inventory and related services. However, prepayments and stockpiling do not directly equate to secured orders, nor can they be prematurely counted as future revenue.

Consequently, once domestic GPUs enter a revenue expansion phase, the new test has only just begun: realized revenue requires order sustainability for validation, and orders require supply and delivery capabilities to ultimately materialize.

**03**

Reports are just the outcome; the market trades “expectations.” As noted above, since revenues are improving across the board, why hasn’t the market rewarded all growth?

For brokerages, this question seems traceable. By checking if current-period revenue holds up, assessing order quality, verifying supply alignment, and combining these with balance sheet metrics like prepayments, contract liabilities, and inventory, analysts can attempt to project performance for upcoming quarters. For Tianday Zhixin, Li Ran even states bluntly that the market is currently pricing in 2027 revenue.

But on the actual investment decision-making side, matters are not so straightforward. “That represents the sell-side’s expectation; the buy-side actually has its own expectations,” investment firm researcher Ethan told Lei Feng Network. As buyers, they must determine how much of these growth expectations are already reflected in the current stock price. Financial reports are merely outcomes, while the market trades expectations.

Another public fund researcher shared a similar view: “The market is actually trading subsequent orders and expectations more than current-period revenue itself, which may not be the primary pricing factor.”

Ethan pointed out that while risk appetites vary across institutions and performance expectations differ, even if final revenue and profit exceed sell-side forecasts, it does not necessarily mean they surpassed the expectations the market was truly trading.

Buy-side and sell-side researchers focus on different aspects of financial reports. Brokerage analysts care more about how much future performance a company can achieve; buy-side institutions making actual investment decisions must further assess how many of these growth expectations have already been priced in.

“The market does not follow the sell-side’s expectations,” Ethan said. If a company has already experienced a significant rally, even if the final report beats sell-side forecasts, the stock price may still fall if it fails to exceed the buy-side’s pre-formed expectations.

This means the commonly cited “beat expectations” actually operates under two different reference frameworks: one is beating publicly available brokerage forecasts, and the other is surpassing the expectations investors had already baked into the stock price.

For domestic GPU companies, judging short-term stock price reactions requires looking beyond whether results beat sell-side forecasts, and examining whether final performance exceeded the expectations already traded by the market. (For further insights on buy-side and sell-side perspectives regarding domestic GPUs, readers are welcome to add the author’s WeChat: treelog10 for discussion.)

In other words, financial reports verify whether a company has delivered, while stock prices trade the “expectation gap” between actual delivery and prior expectations. This explains why domestic GPU financial reports are increasingly important, yet rarely dictate short-term stock prices in isolation.

Li Ran attempts to find the future within the reports: whether revenue, orders, and supply can continue to support growth. Ethan cautions that even if all these clues hold true, they must be recalculated against current valuations. The former focuses on “how much can be achieved in the future,” while the latter cares more about “how much of that future has not yet been bought in advance by the price.”

At this point, re-examining this round of semi-annual reports makes it clear that the market no longer asks merely “who profits first,” nor even just “who grows fastest.”

What is truly being questioned now is who can sustainably deliver growth, and how much of that growth remains unpriced.

*Li Ran, Ethan, and Zhou Heng are pseudonyms.