Game companies dominate China's AI investment landscape in 2026

Text | Beyond the Layout, Author | HuaHua, BanJun

In China’s large model赛道 in 2026, one group no one expected—game companies—was among the most profitable investors.

MiHoYo invested in MiniMax during its angel round in December 2021, with a pre-investment valuation of just $170 million; four institutions collectively invested $31 million.

In January 2026, MiniMax listed on the Hong Kong Stock Exchange at an offering price of HK$165, surging 109% on its first trading day, with a market capitalization exceeding HK$100 billion.

Subsequently, the stock price surged to a historical high of HK$1,330; prior to miHoYo’s IPO, the stake was approximately 6.4%, yielding a conservative paper gain of HK$10 billion.

Return multiple: over 100x.

In the same month, Zhipu AI went public amid the acclaim of being the world's first listed large model company. It listed on the Hong Kong Stock Exchange on January 8, 2026, at an issue price of HK$116.20; over the following three months, its stock price continued to surge, peaking at HK$1,010 in April.

Among its early shareholders is another gaming company—37Games.

Sanqi Interactive's recently released Q1 2026 financial report shows that investment income surged 981% year-over-year, generating an additional RMB 326 million. The primary source of this gain is the fair value change following Zhipu's listing.

One invested in MiniMax, another in Zhipu; two gaming companies have nearly secured the two largest tickets to China’s large model赛道.

What’s even more surprising is that they invested in more than just these two companies.

In China's large model industry, there is a widely recognized term: "The Four AI Giants"—Zhipu, Moonshot AI, Baichuan Intelligence, and MiniMax.

Sanqi Interactive invested in all three at once.

First is Zhipu. In 2023, 37 Entertainment increased its capital contribution by RMB 25 million to the Jiaxing Beishenghengwo Fund, which directly or indirectly holds equity in Zhipu AI.

In January 2026, Zhipu listed on the Hong Kong Stock Exchange at an issue price of HK$116.20; its stock price subsequently surged to HK$1,010, reaching a peak market capitalization of nearly HK$380 billion. Sanqi Interactive recognized an investment gain of HK$326 million in just the first quarter.

Then there is Moonshot AI. Sanqi Interactive made an early move by investing in the Shixiang Global fund series, positioning itself in this company. In March 2026, market rumors suggested the company was advancing its next funding round, with industry expectations placing its pre-money valuation at approximately $8–10 billion.

This isn’t the end—Baichuan Intelligence, a company specializing in the healthcare sector with a foundation in general-purpose large models, reached a valuation of RMB 20 billion after its Series A+ round in April 2026, with Sanqi Interactive also listed as a shareholder.

The Four Little Dragons invested in three companies, but missed MiniMax—because that seat had been held by miHoYo since the angel round.

But Sanqi Interactive's investment portfolio extends far beyond large models.

According to publicly available information, 37 Entertainment has invested in more than 30 hard tech companies, covering areas such as computing chips (StarSky Technology, Fangqing Technology), brain-computer interfaces (StrongBrain Technology), embodied intelligence (LightWheel Intelligence), semiconductor materials (Jingzheng Electronics), and XR hardware (Ray-Ban Glasses).

Without exaggeration, Sanqi Interactive has virtually covered the entire AI industry chain, with investments spanning half of the core AI ecosystem.

A Chinese A-share company, which generated revenue of RMB 15.966 billion in 2025 and began with game publishing, has been more active and broader in its AI investments than most market-driven venture capital firms.

Unlike Sanqi Interactive’s “broad net” approach, miHoYo’s strategy is completely the opposite—it is highly focused.

In the large model space, miHoYo made its core bet on MiniMax—and this investment became one of the most iconic deals in Chinese AI investment history.

Back to the end of 2021, MiniMax had just been founded. Its founder, Yan Junjie, had left SenseTime to start his own venture, focusing on general-purpose large models. At that time, ChatGPT had not yet gained widespread popularity, and the market’s understanding of large models was still relatively unclear.

MiHoYo joined in the angel round, investing $31 million alongside Hillhouse Capital, IDG Capital, and Yunqi Partners, with a pre-money valuation of just $170 million. Yan Junjie has a close personal relationship with Liu Wei, Chairman of MiHoYo, who still serves as a non-executive director of MiniMax.

More than three years later, MiniMax became one of the fastest-growing AI companies in the world. Its AI companion app, Talkie (Xingye), saw explosive growth in the North American market, reaching over 212 million registered users globally.

Listed on the Hong Kong Stock Exchange in January 2026 at an offering price of HK$165, the stock surged 109% on its first day, pushing its market capitalization past HK$100 billion. It continued to climb sharply, reaching a peak price of HK$1,097 and a maximum market capitalization of over HK$313 billion.

Prior to miHoYo's IPO, it held approximately 6.4%, and based on MiniMax's historical peak market capitalization, its stake's book value once exceeded HK$18 billion, and even by conservative estimates, it far surpassed HK$10 billion.

Since entering at the angel round in late 2021, this investment has achieved a return multiple of over 100x.

More interestingly, miHoYo is not only a shareholder of MiniMax but also its client.

Since 2023, MiniMax’s large model technology has been applied to Honkai: Star Rail, with core use cases including NPC intelligent dialogue and dynamic story generation. In MiniMax’s prospectus, miHoYo is listed as a “leading senior independent investor,” having both invested in and actively utilized the technology, establishing a deep partnership.

MiHoYo's investments in AI extend beyond this.

Internally, the company established the "Anti-Entropy Research Department" as early as 2018 and developed its own AI large model, Glossa; the digital human "Lu Ming" has approximately 1.7 million followers on Bilibili, with over 100 million total views across 32 videos.

The self-developed subsidiary Wudinggu Technology was established in July 2025 with a registered capital of RMB 500 million; its founder, Cai Haoyu, stepped down from his key position at the company in 2023 and subsequently founded Anuttacon overseas, specializing in the exploration of AI-native games.

In addition, miHoYo has also participated in establishing several AI investment funds, indirectly investing in AI applications, semiconductors, XR, and other fields through funds such as Shanghai Yuanchuang Future, Xiamen Wawu Phase I, and Shanghai Tongge Phase II.

A sniper, a bulk buyer. Two styles, same arena, both reaping massive profits.

Sanqi Interactive and miHoYo are not isolated cases. Broadening the perspective, the entire gaming industry is pouring significant investment into the large model赛道, with each company adopting its own unique approach.

Century Huatong directly targets computing power. As early as 2022, it invested in the domestic GPU company Moore Threads through an industrial fund.

Moore Threads listed on the STAR Market in December 2025, surging 468% on its first trading day. Century Huatong generated approximately RMB 640 million in net profit from this investment in the fourth quarter of 2025, accounting for 53% of its full-year 2024 attributable net profit.

Shijie Huatong has also partnered with Tencent to build the Yangtze River Delta AI Computing Center, with a planned capacity of 40,000 racks, of which 10,000 have already been delivered. A gaming company has stepped into the business of computing infrastructure.

Kunlun Wanwei is even more aggressive, directly entering the space themselves.

Since September 2023, we have invested a total of RMB 680 million to acquire controlling interest in Aijikexin, an AI computing chip company, fully integrating the entire chain of computing power, large models, and applications. Our Skywork series of large models spans multiple domains including multimodal reasoning, search, and music, making us one of the few Chinese gaming companies truly committed to AGI.

Youzu Network is betting on domestic computing power. In 2025, it invested in two domestic GPU companies, Birun Technology and Xiwang. Birun Technology has already gone public on the Hong Kong Stock Exchange, with a pre-investment valuation of HK$20.9 billion. Youzu also invested RMB 1 billion to co-establish Wuxi Yunxing Intelligent Computing, entering the AI computing power sector.

Juren Network is betting on the product side.

Invested in Aishi Technology, an AI video generation company, and LiblibAI, an AI image generation platform. Its product, "Supernatural Operations Unit," is China's first offering to deeply integrate AI large models into a high-DAU game, with NPCs powered in real time by large models—over 25 million AI-assisted gameplay sessions occurred within its first week of launch.

Tencent, the undisputed leader in the gaming industry, is also a major player in the AI investment space. It participated in MiniMax’s Series B round (holding approximately 5.7% post-IPO) and invested in Zhipu AI. Tencent has also developed its own Hunyuan large model and new AI products such as Yuanbao, with planned full-year investments of RMB 18 billion in 2025 and at least double that amount targeted for 2026.

But Tencent is more like both an athlete and a referee, which is a story on another level entirely.

An interesting comparison: According to IT Juzi, from January to October 2025, there were 139 financing rounds exceeding 100 million yuan in China’s AI sector, with a total amount surpassing 60 billion yuan.

Traditional VC funding is contracting, with AI model layer investment amounts declining by approximately 52.9% year-over-year and the number of investment deals dropping by about 35%, as capital accelerates toward leading players.

While VCs were hesitating, the game company had already invested the money.

Why a gaming company, and not another industry?

The most direct reason: they are too wealthy.

Gaming is one of the few cash-generating industries in China's internet sector. According to public media reports, miHoYo's gaming revenue exceeded RMB 78 billion in 2025, with net profits projected to surpass RMB 40 billion—its cash reserves now exceed those of most venture capital funds.

Since 2020, Sanqi Interactive has distributed over RMB 10.26 billion in cash dividends and made over RMB 20 billion in external investments. These funds were generated internally, requiring no exit timeline to LPs and no pressure to achieve DPI.

VCs investing in large models must go through investment committees, calculate fund lifespans, and consider IRR (internal rate of return); gaming companies, on the other hand, use their own profits and can invest whenever they want.

But money alone isn’t enough. What truly enables game companies to outpace VCs in the large model arena is another factor—they are the largest buyers of large models themselves.

The financial results of 37 Interactive Entertainment are already visible.

More than 70% of advertising creatives are generated with deep AI involvement, and AI-assisted ad targeting reaches 50%. The Moonshot AI Kimi large model is deeply integrated into 37Games’ customer service system and AI assistant “Ask Xiaoqi.” Zhipu’s GLM large model is directly used for expanding game narratives.

MiHoYo is more direct: MiniMax’s models run directly within their own gaming products, powering everything from NPC dialogue to story generation—not as a proof of concept, but as live, shipped products.

Game companies investing in AI don’t just gain equity returns—they also achieve tangible cost reductions. By investing once, they first capture efficiency gains, then realize additional returns upon exit.

VCs can't do this. Sequoia invested in MiniMax, but Sequoia itself doesn't use tokens or have use cases for large models. Its only return mechanism is to wait for an IPO and sell shares.

Another, deeper force driving game companies forward is anxiety.

The good days in the gaming industry aren't as bright as they seem. With stricter game licensing, skyrocketing user acquisition costs, and stagnant user growth, the traditional model of spending heavily on traffic no longer works.

In the first quarter of 2026, Sanqi Interactive's revenue declined by 12.3% year-over-year, yet its net profit attributable to shareholders increased by 59%. The decline in revenue alongside higher profits reflects AI-driven cost reductions and contributions from strategic investments.

As the gaming industry's growth logic shifts from who spends the most to who operates most efficiently, AI is no longer an elective—it's a necessity.

With money, use cases, and anxiety, game companies have entered the large model赛道 faster than anyone anticipated.

Look at the bigger picture: game companies investing in large models reflects a larger trend. On China’s AI investment board, industrial capital is pushing venture capital to the sidelines.

After MiniMax's IPO, Alibaba ranked first among investors with returns exceeding $1 billion (holding approximately 13.66% prior to IPO), miHoYo ranked second (holding approximately 6.4% prior to IPO), and Sequoia China ranked third (holding approximately 3.81% prior to IPO).

Post-listing, investors with returns exceeding $1 billion include Junlian Capital, Meituan, and Ant Group, with leading industrial and venture capital firms among the top backers.

Between 2025 and 2026, the largest exits in China’s large model sector—Zhipu, MiniMax, and Moore Threads—were primarily led by industrial capital, which reaped the greatest returns. It’s not that venture capital failed; rather, industrial capital holds a身份 that VC can never have: it is also a buyer of AI.

Venture capital investing in AI is betting on the future; industrial capital investing in AI is solving immediate problems. This distinction isn’t apparent at the exit stage, but at the moment of investment decision-making, the gap is enormous.

When miHoYo invested in MiniMax at the end of 2021, few in the market believed large models could become a viable business.

But miHoYo believed in it—not because of any advanced AI insights, but because it knew exactly what it needed: a foundational technology capable of driving NPCs, generating narratives, and making virtual worlds more authentic. MiniMax is precisely doing this.

Sanqi Interactive invested in Zhipu and Moonshot because it doesn’t need sophisticated market analysis—it needs AI to cut its annual billions spent on user acquisition. Zhipu’s GLM models and Kimi’s long-context capabilities are exactly what it needs.

This logic—"I invest because I myself need it"—cannot be replicated by any VC.

The stories of 37 Interactive Entertainment, miHoYo, and even gamers themselves actually address a larger question: In the AI era, what types of companies are best suited for early-stage investment?

The answer may not be institutions specifically focused on investments, but rather companies that are themselves the largest users of AI.

They can recognize the true value of technology earlier, make decisions faster, and engage more deeply in the growth of portfolio companies, because they are not just investing—they are investing in their own future.

This time, the gaming company moved ahead of VCs—and ahead of the times.

[Outside the layout]:

The gaming industry has always been an awkward presence in China's business world.

Making money, but not quite right. Has technical aspects, but doesn’t feel hardcore. Significant scale, yet categorized under entertainment, not high tech.

But in this round of large models, game companies have unexpectedly taken center stage. With abundant cash, real-world scenarios, strong technological intuition, and a global perspective that typical industrial capital lacks, game companies have been expanding overseas for a decade.

This is likely the greatest advantage of industrial capital—it doesn’t need others to explain whether AI is useful, because it uses AI every day.