On July 2, Kuaishou made a public announcement via the Hong Kong Stock Exchange, revealing that its video generation model, Kling AI, had secured independent financing, achieving a post-investment valuation of $18 billion. Tencent emerged as a notable investor in this round.

Just four days later, on July 6, Kuaishou issued another statement on the Hong Kong Stock Exchange, confirming that Tencent had divested 273 million Class B shares through off-market block trades, reducing its stake from 15.68% to 9.37% and relinquishing its major shareholder status. Based on the day's closing price of HK$46, this divestiture amounted to approximately HK$12.6 billion.

In a swift maneuver within four days, Tencent invested in Kuaishou's AI subsidiary with one hand while offloading shares in the parent company with the other. Its investment was in AI; its divestiture, in the broader internet sector.

This strategic shift is not unique to Tencent.

I. Tech Giants Follow Suit

Prior to Tencent's move, this 'parent-child separation' strategy had been quietly gaining traction in China's internet industry for over six months.

On January 8 and 9, 2026, Zhipu and MiniMax made their debut on the Hong Kong Stock Exchange, marking the world's first listings of foundational large model companies. Zhipu's offering price was set at HK$116.2, surging 13% on its first trading day with a market capitalization exceeding HK$57 billion. MiniMax fared even better, with its share price skyrocketing nearly 110% on debut, attracting 420,000 subscribers and witnessing intense competition for shares, with 1,209 bidders vying for every HK$1 worth of stock.

Both companies, once part of tech giants' investment portfolios, opted for independent public listings. Zhipu is supported by Tsinghua University and the National Social Security Fund, while MiniMax counts Tencent, Alibaba, and miHoYo among its shareholders. Rather than being bundled into a tech giant's financial statements, they chose to carve out their own narratives in the capital market.

Facts have borne out the wisdom of this decision. The capital market employs vastly different valuation metrics for an AI company experiencing 300% growth but still incurring losses versus an internet platform growing at 10% with stable profits. The former focuses on ARR (Annual Recurring Revenue), token consumption, and model iteration speed; the latter, on DAU (Daily Active Users), ARPU (Average Revenue Per User), and operating profit margins. Merging these metrics into a single financial report results in the faster-growing entity being held back by the slower one, preventing the market from offering a premium.

Baidu recognized this dynamic. On January 1, 2026, Baidu announced that its AI chip subsidiary, Kunlunxin, had confidentially submitted a listing application to the Hong Kong Stock Exchange. On May 7, Kunlunxin initiated listing guidance for the STAR Market, embarking on a dual-track 'A+H' listing process. Baidu clarified in its announcement that the spin-off aims to 'more comprehensively reflect Kunlunxin's value based on its unique strengths,' enabling investors to 'clearly distinguish Kunlunxin from the parent company and independently assess its performance and potential.' In essence, if Kunlunxin remained within Baidu's financial statements, the market would struggle to recognize—or be unwilling to pay a premium for—its true worth.

Goldman Sachs provided a telling figure: If valued at Cambricon's multiple, Baidu's stake in Kunlunxin would be worth $22 billion. Yet, Baidu's entire market capitalization at the time was just over HK$200 billion. An asset potentially worth a quarter of the parent company was obscured within a conglomerate priced as a search engine company—a classic 'valuation trap.'

Alibaba took a different tact, opting for a more aggressive internal restructuring instead of spinning off its AI business for listing. In December 2024, Tongyi Laboratory was separated from Alibaba Cloud and merged into the Smart Information Business Group. In March 2026, Alibaba established a new business group, Alibaba Token Hub (ATH), consolidating all core AI assets, including the Tongyi large model, Qianwen App, and Wukong Platform. Meanwhile, rumors swirled about the independent listing of its chip business, Pingtouge. Essentially, Alibaba is pursuing the same strategy as Baidu: granting AI assets an independent identity within the parent company, whether as a separately listed entity or an independently accounted business group.

By July, Kuaishou had spun off Kling for independent financing, with a valuation of $18 billion, nearly on par with Kuaishou's parent company's market capitalization of approximately HK$199 billion (based on the July 6 closing total market value). The reason a subsidiary commands a valuation nearly equal to its parent lies in the market's inability to assign an AI company's deserved valuation when it remains within the parent's fold; only through independent financing can it be priced according to AI company logic.

II. Not the Entire Kuaishou, Just Its Crown Jewel

Returning to Tencent.

Against this backdrop, Tencent's four-day maneuver ceases to be puzzling. Essentially, Tencent didn't want the entire Kuaishou; it sought only Kuaishou's most valuable asset.

When Tencent invested in Kuaishou in 2017, its primary motive was defensive. That year, Douyin was gaining momentum, and ByteDance's aggressive foray into the short-video market raised concerns across the industry. Tencent needed a 'second entrance' to counterbalance Douyin, and Kuaishou emerged as the optimal choice. From Series D to E to F rounds, Tencent steadily increased its stake, holding over 21% before the IPO—making it Kuaishou's largest institutional shareholder.

Nine years have elapsed. The short-video landscape has stabilized, with Douyin and Kuaishou carving out their respective territories. The strategic defensive purpose has long been fulfilled. However, Kuaishou's growth trajectory began to flatten significantly in 2025. In the first quarter of 2026, Kuaishou's revenue reached RMB 33.7 billion, with year-over-year growth slowing to 3.4%. Adjusted net profit fell 26.3% year-over-year, and live-streaming revenue shrank 13.5% year-over-year. Its stock price plummeted from a yearly high of HK$85 to below HK$40.

Continuing to hold a 15% stake in a company with single-digit growth and a halved stock price is clearly less capital-efficient than reallocating the funds. Tencent's capital expenditures in the first quarter of this year surged to RMB 31.9 billion, up 63% quarter-over-quarter. President Martin Lau stated at the March earnings call that Tencent invested RMB 18 billion in new AI products in 2025 and plans to at least double that in 2026.

Where would the funds come from? From mature financial investments. The HK$12.6 billion realized from reducing its Kuaishou stake, roughly equivalent to RMB 10.9 billion, is sufficient to cover a significant portion of the quarterly AI investment increase.

Yet, Tencent isn't entirely pessimistic about Kuaishou's future. What it favors is Kuaishou's most dynamic segment—Kling AI. In the first quarter of 2026, Kling's revenue exceeded RMB 650 million, growing over 300% year-over-year. By March, its annualized revenue run rate neared $500 million. With over 60 million global users and more than 600 million videos generated cumulatively, this data set would be the crown jewel in any AI company's financing pitch.

Thus, Tencent reduced its stake in the parent company while increasing its investment in the subsidiary. It exchanged HK$12.6 billion in cash-out for a precise RMB 1.36 billion bet. The capital amounts differ by nearly 10 times, but Tencent sold a company growing at 3% and bought one growing at 300%. The calculation isn't about the amount; it's about which growth curve each dollar is invested in.

Tencent has made this calculation abundantly clear.

III. A Four-Year Trend

This isn't Tencent's first foray into such strategic realignment.

In late 2021, Tencent distributed approximately 460 million JD.com shares it held as a dividend in kind to its own shareholders, reducing its stake from 17% to 2.3%, nearly clearing its position. Martin Lau also resigned as a JD.com director at the same time.

In November 2022, the same script unfolded with Meituan. Tencent distributed about 90% of its Meituan shares to shareholders, reducing its stake from 17% to less than 2%, with Lau resigning as a Meituan non-executive director. Coupled with the reductions in Weimob and UBTECH in early 2025 and this latest move with Kuaishou, Tencent is systematically divesting its internet investment portfolio.

Tencent itself explained when reducing its Weimob stake: 'We proactively review our investment portfolio and assess potential adjustments to allocate funds to shareholder returns or new investment projects.' This may sound like standard PR language, but it's actually quite candid. Tencent is indeed transitioning from an 'investment conglomerate holding substantial internet equities' back to a 'technology company concentrating its bets on AI.'

JD.com, Meituan, Weimob, UBTECH, and Kuaishou share a commonality: They are all well-established companies with barriers in their respective sectors, but their growth stories have matured. The e-commerce landscape is set, the local services landscape is set, and the short-video landscape is set. For Tencent, these assets have shifted from 'strategic control' to 'financial holdings,' and the logic for financial holdings is straightforward: Allocate funds where returns are highest.

This doesn't imply Tencent is bearish on the internet. As China's largest internet company, Tencent has no reason to be pessimistic about the industry it relies on. What it's doing is restructuring its assets, swapping mature internet equities for burgeoning AI equities.

Epilogue

This is the pricing logic of the 2026 capital market: Growth is paramount; AI commands a premium. Profitability is secondary; what matters is whether you're on the 300% growth curve.

Is this logic sound? No one knows for certain. But Tencent, Baidu, Alibaba, and Kuaishou—China's largest internet companies—are voting with their wallets. They're spinning off AI assets from their parent companies to let them be priced independently as AI companies while reducing or reallocating internet assets with slowing growth.

In the eyes of those who understand China's internet best, the ticket to the next decade isn't in short videos, e-commerce, or local services. It's in large models, AI chips, and the tokens consumed every second.

Whether this judgment ultimately proves visionary or fervent, time will tell. But at this moment, industrial capital has already cast its vote.