Two Chinese AI Unicorns Face Six-Year Road to Profitability
Z.ai and MiniMax may burn cash through the decade's end as compute costs and revenue mismatches define the new AI economics in Asia.
The Profitability Problem
Z.ai and MiniMax, two of China's most prominent artificial intelligence start-ups, may not turn a profit until 2030 or later, even as revenues climb sharply. The projection, delivered by Ellie Jiang, Macquarie Group's head of Asia internet and software research, underscores a structural reality across the region: building and operating frontier large language models requires capital outlays that dwarf near-term revenue potential.
At DailyTechWire, we have tracked the funding rounds across Asia's AI sector for the past eighteen months, and the pattern is consistent. Companies raise hundreds of millions of dollars, deploy those funds into GPU clusters and inference infrastructure, and watch their revenue curves lag years behind their cost curves. The Chinese market intensifies that dynamic.
Compute Scarcity and the China Multiplier
Jiang pointed to the expense of compute as the principal drag. Training and serving state-of-the-art models demand thousands of high-end GPUs running continuously. In China, that challenge is compounded by limited access to the latest Nvidia chips, restricted under US export controls. The resulting scarcity drives costs two to three times higher than in markets with unrestricted access, according to Jiang's remarks.
Chinese firms have pivoted toward domestic alternatives and older-generation hardware, but those substitutes carry trade-offs in efficiency and performance. Inference workloads, which must deliver low-latency responses to millions of users, become more expensive per token when running on less capable silicon. The gap between what a model costs to operate and what customers will pay for its output remains wide, and closing it will take years of engineering optimisation and scale.
Revenue Growth Without Margin Relief
Both Z.ai and MiniMax have demonstrated traction. Z.ai's consumer chatbot applications have attracted tens of millions of monthly active users, while MiniMax has built enterprise partnerships across media, gaming, and customer service verticals. Revenue is growing, but gross margins remain deeply negative because infrastructure costs rise in step with usage.
The monetisation models compound the problem. Consumer AI products in China face intense price competition, with many services offered free or at negligible subscription fees to build user bases. Enterprise contracts often involve pilot pricing or volume discounts that reflect customers' own uncertainty about AI's return on investment. Neither channel generates the unit economics needed to cover compute at current scale.
The path to profitability, in Jiang's assessment, depends on three variables: continued revenue growth, gradual improvement in model efficiency, and eventual moderation in hardware costs as domestic chip supply matures. All three will take years to converge. Until then, these companies will rely on venture and strategic capital to fund operations.
The Broader Pattern Across Asia's AI Layer
The Z.ai and MiniMax outlook mirrors a wider trend among Asia's foundation model builders. At DailyTechWire, we have observed similar dynamics in Seoul, Tokyo, and Singapore, where well-funded AI labs are burning through nine-figure rounds with no clear line of sight to breakeven. The capital intensity of this technology is unprecedented in software, more akin to semiconductor fabs or telecom infrastructure than to the lean SaaS businesses that dominated the previous decade's venture landscape.
Investors continue to fund these companies because the strategic stakes are high. Whoever controls the model layer controls access to the next generation of applications, and no major economy wants to cede that position. But the timeline to returns has lengthened considerably. Limited partners in venture funds will need to accept that AI investments made in 2024 and 2025 may not exit until the early 2030s, if at all.
Implications for the Funding Environment
The extended loss-making period has already begun to reshape capital allocation. Early-stage AI start-ups without a clear path to efficient inference or differentiated datasets are finding it harder to raise follow-on rounds. Investors are concentrating capital in a smaller number of companies with proven model performance and strategic backing from cloud providers or state-linked funds.
In China specifically, the government's willingness to support AI development through subsidised compute, favourable procurement contracts, and patient capital from state investment vehicles will determine which companies survive the profitless stretch. Z.ai and MiniMax both benefit from such support, but the depth of those resources is not unlimited, and the number of companies competing for them is growing.
What the Market Is Pricing In
Macquarie's analysis reflects a sobering recalibration. A year ago, many investors assumed that Chinese AI companies would reach profitability by 2027 or 2028, riding a wave of enterprise adoption and improving hardware access. That timeline has now been pushed back by at least two years, and even the revised forecast carries significant uncertainty.
The delay does not necessarily mean these companies will fail. It does mean that their valuations must be justified by revenue potential in the 2030s, not the 2020s, and that the risk of dilution for early shareholders has increased. For employees holding equity, the calculus has shifted: stock options granted in 2024 may not be liquid until well into the next decade.
At DailyTechWire, we see this as a defining moment for Asia's AI sector. The region has built genuine technical capability and attracted world-class talent. But the economics of frontier AI remain brutal, and the companies that emerge as long-term winners will be those that can endure years of losses while continuing to improve their models and expand their user bases. Z.ai and MiniMax are test cases for that endurance.
