MiniMax reports 159% revenue surge as China's AI challenger eyes global markets
MiniMax posts $79M revenue with 159% year-over-year growth after its $614M Hong Kong IPO, with over 70% of revenue coming from outside China.
Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR — MiniMax, one of China's leading AI startups, reported $79M in revenue for 2025 — up 159% year-over-year — shortly after completing a $614M Hong Kong IPO in January 2026. More than 70% of that revenue came from international markets, making MiniMax a genuinely global AI business, not just a domestic Chinese story. The company posted a net loss of $1.87B, but that figure is dominated by non-cash valuation adjustments, not operational burn. If you think the global AI race is a two-horse contest between OpenAI and Google, MiniMax is here to disrupt that narrative.
In January 2026, MiniMax completed a $614M IPO on the Hong Kong Stock Exchange, making it one of the largest AI listings in Asia in recent memory. The timing was deliberate. Chinese AI companies have watched US IPO windows close and open unpredictably, shaped by regulatory scrutiny and geopolitical tensions. Hong Kong, by contrast, offers access to institutional capital from both Asia and Europe while keeping the company within a regulatory environment that Chinese AI labs can navigate with confidence.
The IPO was not just a liquidity event — it was a statement. MiniMax chose to go public at a moment when DeepSeek had just rattled Western AI markets with its cost-efficient reasoning models, and when China's AI sector broadly was demonstrating that it could compete on model quality, not just price. MiniMax's listing signals that at least some Chinese AI companies are mature enough to bear the scrutiny of public markets: audited financials, disclosed unit economics, and genuine accountability to shareholders.
Founded in 2021 by Yan Junjie, a former SenseTime researcher, MiniMax has taken a product-first approach rather than positioning itself primarily as an API vendor. Its flagship consumer products — the character AI platform Talkie and the video generation tool Hailuo AI — have built genuine user bases outside China, giving it revenue streams that US-China trade tensions cannot easily sever. That matters enormously for a company trying to sustain growth through an IPO and beyond.
The Hong Kong listing also unlocks future capital raises in a market where AI infrastructure spending is accelerating. MiniMax is not done investing in compute and model development, and having public-market credibility makes subsequent equity or debt raises substantially easier.
$79M in annual revenue sounds modest relative to OpenAI's reported $3.7B annualized run rate or Anthropic's multi-billion-dollar ambitions. But context matters enormously here. MiniMax grew from approximately $30.5M in 2024 to $79M in 2025 — a 159% year-over-year increase. That growth rate is not incremental; it is the kind of acceleration that signals genuine product-market fit, not just favorable comparables.
The revenue mix is also meaningful. MiniMax generates income across two primary channels: consumer subscription revenue (driven by Talkie and Hailuo) and enterprise/API revenue (developers and businesses building on the MiniMax foundation models). Neither channel dominates overwhelmingly, which is a healthier revenue base than pure API-dependence or pure consumer dependence.
Consumer subscription revenue is stickier than it looks. Users who pay monthly for character AI companions or video generation credits build habits that are difficult to break, particularly when the product quality continues to improve. MiniMax has benefited from the broader tailwind of AI-native consumer products finding paying audiences — a dynamic that took longer to materialize than many predicted but is now clearly underway.
Enterprise and API revenue, meanwhile, benefits from a different dynamic: developers who integrate MiniMax models into their own products generate recurring usage that scales with their own growth. As those downstream applications grow, MiniMax's revenue grows alongside them without requiring proportional increases in sales effort.
At $79M, MiniMax is not yet at the revenue scale where it can self-fund heavy model training runs. But it is at the scale where it can credibly claim that it has a business, not just a research lab with a consumer wrapper.
The 143.4% growth in consumer subscription revenue is the headline that deserves the most attention from product-focused observers. This is not B2B revenue growing because a handful of enterprise contracts were signed. This is individual users, at scale, choosing to pay for an AI product every month.
Talkie, MiniMax's character AI platform, has been the primary driver. The product allows users to create, customize, and converse with AI personas — a category that Character.AI pioneered in the West and that has proven to have remarkable stickiness with specific demographic groups, particularly younger users seeking social or entertainment experiences. Talkie has been downloaded millions of times outside China and has built a particularly strong presence in Southeast Asia, the Middle East, and parts of Europe.
Hailuo AI, MiniMax's video generation platform, added a second consumer revenue stream. As text-to-video AI moved from novelty to utility through 2025, Hailuo competed directly with tools like Sora, Runway, and Kling. The platform's subscription tiers — which gate access to longer video durations, higher resolutions, and faster generation queues — converted free trial users into paying subscribers at rates that exceeded the company's internal projections.
The consumer subscription model has a structural advantage that pure API revenue lacks: users who subscribe monthly develop loyalty and brand affinity. They follow product updates, share outputs, and advocate for the platform within their networks. MiniMax has leveraged this dynamic aggressively, investing in community features and creator programs that turn power users into unpaid distribution channels.
The 143.4% growth rate also reflects a relatively small base effect — consumer AI subscriptions were an emerging revenue category in 2024, and MiniMax was early enough to capture disproportionate share as the market expanded. Sustaining anywhere near that growth rate through 2026 will require continued product investment and increasingly aggressive international marketing.
The single most important data point in MiniMax's financial disclosure is that more than 70% of its revenue came from outside China. For a Chinese AI startup, this is extraordinary. Most of its domestic peers — even well-capitalized ones — derive the overwhelming majority of their revenue from Chinese enterprise clients or Chinese consumer markets. MiniMax has inverted that dynamic.
The international revenue is not evenly distributed. Southeast Asia has been the largest market outside China, where Talkie's character AI format found early adoption among Gen Z users in Indonesia, the Philippines, Thailand, and Vietnam. These markets have high mobile penetration, younger demographic profiles, and consumer willingness to engage with AI companions that mirrors trends seen in South Korea and Japan years earlier with virtual influencer culture.
The Middle East has emerged as a second significant international market. MiniMax has invested in Arabic-language model capabilities, and enterprise clients in the GCC countries have used the MiniMax API for customer service, content generation, and translation applications. The regional AI investment boom — driven partly by sovereign wealth funds deploying capital into AI infrastructure — created an opening that MiniMax moved into ahead of many Western competitors who were slower to localize.
Europe represents a smaller but growing share. Hailuo AI has found users among creative professionals in Germany, France, and the UK who needed affordable, high-quality video generation without the per-minute costs of premium Western tools. Enterprise API usage in Europe is growing but remains constrained by data residency requirements that MiniMax is still working to fully address.
Notably, the United States is not a major revenue source for MiniMax — a reflection of the political environment rather than product limitations. As US-China AI tensions persist, MiniMax has rationally prioritized markets where its Chinese origin does not create procurement friction.
MiniMax's reported net loss of $1.87B requires careful interpretation, because the raw number is deeply misleading without context. The vast majority of this loss is attributable to fair value changes in preferred share liabilities — a standard accounting treatment for pre-IPO preferred equity that requires mark-to-market adjustments as the company's valuation changes.
When a startup raises preferred equity at progressively higher valuations, accounting standards require it to record the increasing value of those liabilities on the income statement. As MiniMax's valuation climbed through multiple funding rounds — it was valued at approximately $2.5B in its last private round — those non-cash charges accumulated. The IPO converts preferred shares to ordinary shares, eliminating this accounting mechanism going forward.
Strip out the valuation-related charges and MiniMax's operational loss profile looks materially different. The company is investing heavily in compute, model training, and international expansion — all capital-intensive activities that justify operational losses at this stage of growth. But the $1.87B headline number is not indicative of a business hemorrhaging cash on operations.
Investors who have followed Chinese tech IPOs — from Meituan to ByteDance's various subsidiaries — will recognize this pattern. The accounting treatment for preferred share liabilities creates dramatic-looking loss numbers that disappear post-IPO. Analysts covering MiniMax appropriately adjust for these items when assessing the company's underlying unit economics.
The more meaningful metric is gross margin trajectory. MiniMax has not disclosed detailed gross margin data, but the revenue mix suggests margins are improving as higher-margin consumer subscriptions grow faster than lower-margin API calls. A healthier gross margin profile is the foundation of a credible path to profitability.
MiniMax's three-product architecture — Talkie (consumer character AI), Hailuo AI (consumer video generation), and the MiniMax API (developer and enterprise platform) — gives it more diversification than most AI startups at its stage.
Talkie is the oldest and most established product. It operates as a social AI platform where users create named personas with defined personalities, backstories, and conversational styles. The platform supports both text and voice interactions, and MiniMax has invested significantly in voice synthesis quality that makes conversations feel genuinely naturalistic. Talkie competes with Character.AI and Replika internationally, and with similar domestic Chinese products at home. Its international user base — which drives the majority of its consumer subscription revenue — is measured in the tens of millions of monthly active users.
Hailuo AI is the newer and faster-growing product. Launched in earnest in mid-2024, it rode the wave of consumer interest in AI video generation sparked by OpenAI's Sora announcement. MiniMax's video models have been benchmarked favorably against Western competitors, particularly on motion consistency and realistic physics, which are the primary quality complaints users level at first-generation video AI tools. Hailuo's subscription tiers range from free (with watermarks and generation limits) to premium tiers that unlock commercial usage rights — a model that has proven effective at converting professional creators who need clean outputs for client work.
The MiniMax API provides programmatic access to the company's foundation models: MiniMax-Text, MiniMax-Speech, and MiniMax-Video. Enterprise clients use these APIs to power customer service bots, internal knowledge management tools, and content generation workflows. The API platform is priced competitively against OpenAI and Anthropic, and its multi-modal capability — covering text, voice, and video in a single API relationship — is a meaningful differentiation point for developers who need cross-modal workflows without managing multiple vendor relationships.
China's AI sector is not monolithic, and MiniMax occupies a distinct position within it. The most useful peer comparisons are Zhipu AI, Moonshot AI (maker of Kimi), Baichuan AI, and the recently prominent DeepSeek.
DeepSeek has attracted the most Western attention, primarily because of its cost-efficient reasoning models that appeared to match or exceed GPT-4-class performance at dramatically lower training costs. But DeepSeek is primarily a research-forward organization — it has released models publicly but has not built substantial consumer or enterprise revenue streams. MiniMax, by contrast, is a revenue-generating business with a consumer product track record.
Moonshot AI's Kimi has built strong domestic Chinese traction as a long-context reasoning assistant, but its international presence is limited. Kimi is effectively a China-first product competing in a crowded domestic market where ByteDance's Doubao and Baidu's ERNIE Bot have substantial distribution advantages.
Zhipu AI has focused heavily on enterprise B2B contracts within China, making it a fundamentally different business model — less scalable internationally, more dependent on large domestic deals.
MiniMax's distinctiveness is its genuine international revenue. No other Chinese AI startup of comparable scale can credibly claim that more than 70% of its revenue comes from outside China. That international footprint is both a competitive moat and a hedge against domestic Chinese market dynamics, including regulatory shifts and intense local competition.
The conventional Western narrative about AI is that the US leads — in model capability, in funding, in enterprise adoption — and that China is perpetually catching up. MiniMax's financial results complicate that story in important ways.
First, consumer AI product-market fit is not exclusively a US phenomenon. The assumption that OpenAI or Anthropic will inevitably capture the global consumer AI market underestimates the importance of localization, cultural alignment, and pricing. MiniMax's success in Southeast Asia and the Middle East reflects a product that understood local user preferences and pricing sensitivities better than its Western competitors.
Second, the AI capability gap between US and Chinese labs is narrowing faster than US incumbents publicly acknowledge. DeepSeek's models were the most visible demonstration of this, but MiniMax's Hailuo video generation quality — competitive with Sora and Runway on multiple benchmarks — suggests the gap in applied model quality is smaller than the narrative suggests.
Third, the Chinese AI industry is not a monolith, and MiniMax's IPO demonstrates that market-driven, revenue-focused AI companies can emerge from China alongside the government-adjacent labs that dominate Western perception of Chinese AI. MiniMax is not building AI for state infrastructure contracts. It is building consumer products and APIs that compete on merit in open international markets.
None of this means US AI labs are in danger of losing their leading position in the near term. OpenAI's revenue scale, Anthropic's safety-focused enterprise positioning, and Google's infrastructure advantages are real. But the assumption that the AI race is essentially settled in America's favor is increasingly untenable.
MiniMax's last disclosed private valuation was approximately $2.5B, established during its Series B+ round. The Hong Kong IPO at $614M raised implies a public market valuation that will be determined by where the stock trades post-listing — a number that was not yet stable at the time of this writing.
The IPO timing is notable. MiniMax chose to list in Hong Kong rather than wait for a potential US window, a decision that reflects both geopolitical realism and the maturation of Hong Kong's technology listing infrastructure following the HKEX reforms of recent years. Hong Kong has been actively courting technology listings, and MiniMax's IPO is a validation of that effort.
For the broader AI IPO market, MiniMax's listing will serve as a data point in an ongoing debate: at what revenue multiple should AI companies with rapid growth but significant losses be valued? The answer Hong Kong's institutional investors provide will influence how other Chinese AI companies — and potentially some Western ones — think about their own IPO timing and venue selection.
The $614M raise also gives MiniMax a war chest for the next phase of growth: international marketing, compute infrastructure, and potentially acquisitions of smaller AI tools companies that can expand its product surface area without requiring internal development time.
MiniMax's growth story is real, but the risks are equally real and deserve clear articulation.
Regulatory risk cuts both ways. Chinese AI companies face potential restrictions from Western governments — particularly in the US, where the regulatory environment toward Chinese technology companies has tightened consistently over the past several years. App store removals, data residency mandates, or outright bans in certain markets remain plausible scenarios. MiniMax's strategic decision to avoid the US market is prudent given this, but it limits the company's total addressable market.
Domestically, Chinese AI regulation is evolving rapidly. The Cyberspace Administration of China has issued multiple AI governance frameworks, and compliance requirements for generative AI products continue to expand. MiniMax must navigate these requirements while simultaneously competing against well-capitalized domestic rivals.
Competitive intensity from both sides is severe. Internationally, OpenAI, Meta, and Google are not standing still. Character.AI has deep user engagement in the character AI space. Runway and Adobe continue to invest heavily in video generation quality. Domestically, ByteDance's resources and distribution make it a formidable competitor in every consumer AI category MiniMax operates in.
Model training costs remain a headwind. MiniMax must continue investing in next-generation models to maintain competitive quality — and those training runs are expensive. The company's revenue is not yet at the scale that makes self-funded frontier model training comfortable. It remains dependent on external capital for major model development cycles.
International expansion costs are also non-trivial. Building trust, establishing payment infrastructure, and localizing products across Southeast Asia and the Middle East requires sustained operational investment that is easy to underestimate.
Western AI companies — particularly those with consumer-facing products — should treat MiniMax's results as a competitive signal rather than a curiosity. Three implications are worth considering seriously.
First, international markets are not guaranteed territory for US AI companies. MiniMax's success in Southeast Asia and the Middle East demonstrates that early movers in those markets are establishing user habits and brand loyalty that will be expensive to dislodge. US companies that assumed international expansion could follow domestic success are watching that assumption tested in real time.
Second, multi-modal product suites outperform single-modality offerings. MiniMax's ability to offer text, voice, and video AI through both consumer products and a unified API is increasingly attractive to enterprise buyers who want to consolidate vendor relationships. OpenAI has recognized this with its own voice and image capabilities, but MiniMax offers this stack at competitive pricing with less geopolitical friction in many markets.
Third, the consumer AI subscription model is more durable than the industry gave it credit for. MiniMax's 143.4% growth in consumer subscriptions is not an anomaly — it reflects a genuine willingness among users globally to pay monthly for AI experiences that they value. Western companies that ceded the consumer subscription space to focus exclusively on B2B enterprise sales may need to reconsider that allocation.
Yes — and the question no longer deserves debate. $79M in revenue growing at 159% year-over-year, with more than 70% coming from international markets, is not the profile of a domestically subsidized Chinese tech company papering over weakness with state contracts. It is the profile of a company that has built products that people outside China willingly pay for, at scale, with accelerating momentum.
MiniMax's $614M Hong Kong IPO validates the business model and provides capital for the next growth phase. The $1.87B net loss is an accounting artifact, not an operational crisis. The 143.4% consumer subscription growth signals genuine product-market fit across multiple international markets.
The company faces real risks — regulatory, competitive, and operational. Sustaining triple-digit growth as the base grows larger is mathematically harder than maintaining it from a small base. International expansion is expensive. Model training is capital-intensive. And the geopolitical environment creates persistent uncertainty about market access in some geographies.
But MiniMax has cleared the most important hurdle for any AI startup: it has demonstrated that people will pay for its products, in volume, across multiple countries, in a competitive market. That is the foundation on which durable technology companies are built.
For anyone tracking the global AI race, MiniMax is no longer a footnote to the Chinese AI story. It is a primary chapter.
MiniMax is a Chinese AI startup founded in 2021 by Yan Junjie. The company builds and operates consumer AI products — including Talkie (character AI platform) and Hailuo AI (video generation) — alongside an enterprise API platform that provides developers access to its foundation models for text, speech, and video. MiniMax went public on the Hong Kong Stock Exchange in January 2026 with a $614M IPO.
Most Chinese AI companies derive the large majority of their revenue from domestic Chinese markets or enterprise clients. MiniMax's claim that more than 70% of its revenue comes from outside China is unusual in the sector and demonstrates that the company has built genuinely competitive consumer products that international users will pay for — a meaningful proof point for the global competitiveness of Chinese AI.
No. The $1.87B net loss is predominantly composed of non-cash fair value adjustments on preferred share liabilities — a standard accounting treatment for pre-IPO preferred equity that records increasing valuations as income statement charges. These charges disappear after IPO when preferred shares convert to ordinary shares. MiniMax's operational loss profile is materially different from the headline number.
DeepSeek gained Western attention through its cost-efficient open-source reasoning models but has not built substantial consumer or enterprise revenue. MiniMax is a revenue-generating business with consumer products and an API platform. DeepSeek is primarily a research organization; MiniMax is a product company. Both demonstrate different dimensions of Chinese AI capability, but they are not direct competitors.
Hailuo AI is MiniMax's text-to-video generation platform, which competes with OpenAI's Sora, Runway, and Kling. The platform has been benchmarked favorably on motion consistency and physics realism. It operates on a freemium subscription model with paid tiers that unlock longer durations, higher resolutions, and commercial usage rights — a structure similar to Western competitors but at pricing calibrated for international markets including Southeast Asia and the Middle East.
The US regulatory environment toward Chinese technology companies has tightened significantly over the past several years, creating substantial uncertainty for a Chinese AI company seeking a US listing. Hong Kong offers access to institutional capital from Asia and Europe, a regulatory environment MiniMax can navigate with confidence, and a growing track record of successfully hosting large technology listings following recent HKEX reforms. The choice is pragmatic rather than ideological.
Three signals matter most. First, international consumer AI markets — particularly Southeast Asia and the Middle East — are not guaranteed territory for US AI companies; MiniMax has established early positions that will be expensive to dislodge. Second, multi-modal product suites (text + voice + video) are increasingly preferred by enterprise buyers. Third, consumer AI subscriptions are more durable than the industry initially believed, with MiniMax's 143.4% consumer subscription growth demonstrating sustained willingness to pay across multiple global markets.
MiniMax released M2.5 — a 230 billion parameter mixture-of-experts model with 80.2% on SWE-Bench Verified, matching Claude Opus 4.6 speed at $0.29/M input tokens, trained across 200,000 real-world coding environments.
ByteDance is deploying its own AI chips at scale — targeting 100,000 units by EOY 2026 — while launching Seedream 5, an image generation model that outperforms Google's best. China's biggest AI sovereignty move yet.
AI chipmaker Cerebras raises $1 billion at $23 billion valuation and prepares for a Q2 2026 IPO, positioning as the most credible NVIDIA inference competitor.