Is MiniMax Profitable? Revenue Growth, R&D Spending, and China’s Public AI Test
MiniMax's public financials show explosive revenue growth and improving gross profit, but the company remains deeply loss-making as R&D spending expands.
MiniMax gives investors something rare in frontier AI: an income statement
MiniMax is unusually valuable to the AI-profitability debate because it publishes financial results as a Hong Kong-listed company. That replaces speculation with hard numbers. For the first half of 2026, MiniMax reported $116.6 million in revenue, $20.8 million in gross profit and a $358 million loss for the period.[1] The answer to the headline question is therefore clear: MiniMax was not profitable. What makes the company interesting is how quickly revenue is growing and how directly investors can see the cost of maintaining frontier-model ambition.
Public disclosure turns AI economics from narrative into arithmetic
Revenue growth can be compared with cost of sales, R&D and net losses instead of relying on private-company ARR claims.
Revenue growth is extraordinary but starts from a comparatively small base
MiniMax said first-half 2026 revenue increased 283 percent from the prior-year period and already exceeded its entire 2025 revenue.[2] That is genuine commercial acceleration. Enterprise and developer services became the majority of revenue, which is strategically important because business API usage can be more durable than consumer experimentation. Yet the scale still remains modest relative to the hundreds of millions spent on research.
Gross margin improved but remains far below classic SaaS levels
MiniMax reported a first-half gross margin of 17.9 percent.[2] Traditional software investors accustomed to 70 or 80 percent gross margins would view that as infrastructure-like rather than software-like economics. The figure shows how much model inference and service delivery consume before the company even pays for research, sales and administration. Improving gross margin is therefore essential: without it, revenue growth alone can expand workload without generating enough dollars to finance the rest of the company.
Every revenue dollar currently carries substantial serving cost
The company must improve inference efficiency, pricing or product mix before scale can create strong operating leverage.
Research spending explains why fast top-line growth has not produced profit
MiniMax disclosed roughly $297 million of R&D expense in the first half of 2026, vastly larger than gross profit.[2] This is the central frontier-lab problem in one line item. Commercial products may produce positive gross profit, yet the company reinvests several times that amount in building new models. Investors are effectively financing future capabilities with today’s losses.
The enterprise mix is improving the quality of MiniMax’s revenue
Open Platform and other enterprise services reached $73.9 million in first-half revenue, more than 63 percent of the total.[2] This is financially important because API and enterprise workloads can recur and scale with customer production use. Consumer AI products can generate enormous engagement but may require free tiers and high marketing expenditure. A growing enterprise mix gives MiniMax a clearer route toward contracts that can be priced against business value.
Revenue composition can matter more than user count
Hundreds of millions of users create attention; paying enterprise workloads create the cash needed to support research.
MiniMax demonstrates that cheaper Chinese AI does not automatically mean profitable Chinese AI
Reuters grouped MiniMax with Z.AI and DeepSeek as examples of Chinese firms emphasizing lower prices and efficiency while still burning cash.[3] MiniMax’s own public results validate that distinction. The company can be more capital-efficient than a U.S. lab and still lose hundreds of millions because R&D expenditure remains enormous relative to revenue.
Its cash balance buys time for the margin transition
MiniMax reported more than $1.3 billion in cash at June 2026.[1] That provides runway for model development and international expansion even while operating losses continue. Capital availability is therefore part of the strategy, but the public market eventually imposes a stronger discipline than private funding. Shareholders can watch whether losses shrink relative to revenue and whether gross margin improves each reporting period.
The key signal is whether loss intensity declines
A company can remain unprofitable while making real progress if every dollar of revenue requires progressively less operating loss.
So is MiniMax profitable in 2026?
No. Its published results show a substantial net loss and an adjusted net loss even as revenue and gross profit grow rapidly.[1][4] The useful question is whether the trajectory is improving. Revenue is accelerating, enterprise mix is strengthening and gross profit has expanded, but R&D remains far too large for current gross profit to absorb.
MiniMax may become one of the most important case studies in frontier AI finance precisely because the numbers are public. Instead of debating hypothetical unit economics, investors can watch a real model company attempt to move from venture-style growth toward public-company profitability quarter by quarter.[5]
The public results also reveal why efficiency gains must be interpreted carefully. MiniMax’s gross profit grew much faster than revenue, yet absolute R&D spending also surged. That means better serving economics are being reinvested into capability rather than dropping directly to the bottom line. For a frontier lab this may be strategically rational, but investors will eventually require evidence that research spending can grow more slowly than gross profit without causing the model portfolio to fall behind.
Works Cited
- 01MiniMax — First Half 2026 Financial Results minimax.io
- 02MiniMax — 2026 Interim Financial Results Detail prnewswire.com
- 03
- 04MiniMax — Full Year 2025 Financial Results minimax.io
- 05HKEX — MiniMax Listed Company Disclosures www1.hkexnews.hk
CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.
Submit a research lead