The Numbers ByteDance Wins, and the Ledgers It Doesn't

"ByteDance is double-short Hang Seng Tech." The line has circulated in Hong Kong stock narratives in 2026. Li Liang, vice president of Douyin Group, publicly denied it, calling it organized hype by marketing accounts. The denial is accurate: ByteDance is unlisted and not in the index. The metaphor has life because it captures a structural asymmetry. Six of the top ten weighted stocks in Hang Seng Tech compete directly with ByteDance, with a combined weight around 40 percent. Alibaba, Meituan, Tencent, Kuaishou, JD.com, and Baidu. To buy Hang Seng Tech is to indirectly hold a portfolio that ByteDance keeps eroding, without being able to hold the eroder.

The metaphor is also too smooth. It folds competitive pressure and financial health into the same thing. Unfolded, ByteDance's offensive has sharp edges. The defenders are not collapsing.

Local Services: GTV Has Caught Up, the Money Has Not

Douyin local services is where ByteDance's offensive is easiest to see. In 2025, in-store payment GTV before redemption was RMB 800 billion. In the first half of 2026 it exceeded RMB 600 billion, with the full year expected to land at RMB 1.2 trillion. Nomura estimates Meituan's in-store GTV at RMB 1.18 trillion in 2025 and about RMB 1.3 trillion in 2026. On a payment basis, Douyin has spent five years putting itself in the same order of magnitude as Meituan. Meituan, counting from its group-buying business in 2010, took sixteen years.

The redemption rate is the step in this arithmetic that gets quietly skipped. Douyin local services' redemption rate is in the 40 to 50 percent range. Meituan's is steadily above 80 percent. Converted into actual transaction value after redemption, Douyin is about RMB 600 billion for the year. Meituan is above RMB 900 billion. A gap of around 30 percent.

Douyin knows this. In 2026 its strategy shifted from content reach to intent capture. Search GMV grew 55 percent year on year. The shift is making up for the redemption lesson: from "scroll past and casually stockpile a voucher" to "think it through, then come search for it."

The redemption gap is not an operational detail. It points to two different structures of consumer intent. The content field stimulates impulse orders driven by emotion. The shelf field receives consumption plans with clear purposes. Payment GTV can be piled up quickly through traffic inertia. The redemption rate requires a migration of user intent, and that migration is much slower.

Douyin has smashed through Meituan's excess-profit pool. Meituan's in-store profit margin fell from above 40 percent before 2024 to around 25 percent in the second quarter of 2026. But between smashing the profit pool and replacing the owner of the profit pool lies a hard constraint called the redemption rate.

Douyin Pay: Third in Usage, Another World in Value Share

Douyin Pay is the evidence most often cited for the squeezing-WeChat-and-Alipay narrative. iiMedia's 2026 data shows Douyin Pay at a 36.82 percent usage rate. Behind WeChat at 67.85 percent and Alipay at 66.40 percent. In third place. In a market long ruled by a duopoly, a third player at 36.82 percent usage is a development the incumbents have noticed.

But usage rate and market share answer two different questions. By transaction value, Alipay accounts for 54.6 percent and WeChat Pay 33.4 percent. Together nearly 88 percent. Douyin Pay's high usage rate largely results from being embedded in scenarios. When users buy group-buy vouchers or place e-commerce orders inside Douyin, they casually complete payment. This is incidental penetration within a closed-loop scenario. It is different in nature from WeChat Pay's cross-scenario invocation through the social graph, or Alipay's cross-institution clearing through e-commerce and financial infrastructure.

A self-operated checkout counter in a shopping mall may have a high usage rate. That does not make the counter the city's payment infrastructure. Douyin Pay has taken a share of incremental scenarios. It has not begun to shake the duopoly's clearing foundation.

Feishu: Strong ARR Growth, a Persistent Profit Gap

Feishu's ARR exceeded RMB 3 billion in 2025. In the second quarter of 2026 it grew more than 100 percent year on year. In the first half, ARR growth reached 2.5 times the year-earlier level. Nine out of ten new customers also purchased Feishu AI products. Within ByteDance's map, Feishu is the hard-but-right story. Enterprise services is a slow, heavy, patience-demanding track. ByteDance used its consumer-internet organizational capacity to push it to high revenue growth.

But ARR is not profit. After layering on compute, marketing, and other expenses, Feishu's profit gap remains. Doubling ARR in SaaS first reflects customer-acquisition momentum and sales efficiency. It does not reflect an inflection point where scale effects are realized. The profit-margin curve in enterprise services naturally lags the revenue curve. Feishu still has distance to travel before becoming a profit contributor. The ARR number at the celebration and the negative line on the financial statement are two sides of the same thing. Looking at only one side yields the conclusion that Feishu has won. Looking at both reveals that it is still on the way.

One Hundred and Twenty-Five Minutes, and the Thousand Impressions It Cannot Sell

Hongguo Short Drama's DAU broke 100 million in January 2026. By July it reached 168 million, with per-user time of about 125 minutes. Its total usage time exceeds the four major platforms combined. Fanqie Novel has 245 million MAU. Hongguo takes its IP and builds a free short-drama closed loop. The logic works. ByteDance's overall user-time share reached 40.9 percent in July 2026. Douyin's single-app time reached 19.4 percent, surpassing WeChat's 18.8 percent for the first time.

Time advantage is the battlefield ByteDance least needs to defend. But the path from time to profit is advertising. Advertising's cyclicality and unit monetization efficiency are not in the same order of magnitude as WeChat's social, payment, and fintech combination. Hongguo's free short-drama model relies on advertising. Its ad load per unit of time is a world away from WeChat Moments and Channels. A user watches 125 minutes of Hongguo a day, and advertisers bid on a CPM basis. The same user scrolling Moments is worth another number to advertisers. Time is a traffic asset statement, not a profit statement. To directly convert Hongguo's DAU and time into making money hand over fist skips the intermediate variable of monetization efficiency.

The Pinned Family Group, and the Unanswered "You There?"

Douyin's single-app time surpassing WeChat happened in 2026. But time surpassing reflects content consumption crowding out fragmented time. It does not reflect migration of the communication social graph. WeChat's DAU is about 931 million. Douyin's is about 714 million. WeChat's DAU/MAU ratio is about 90.2 percent. Douyin's is about 80.6 percent. WeChat's moat has never been how much time users spend inside it. The moat is the lock-in effect of the relationship chain and the low substitutability of communication as an essential need. You can stop scrolling Douyin. It is hard to stop using WeChat to reply to messages, take voice calls, scan to pay, and log into all kinds of accounts.

Douyin's chat and group functions have so far produced no observable breakthrough in relationship-chain density. A person's WeChat pinned items are the family group and the work group. In Douyin DMs lie a few "you there?" messages with no reply. Using two apps and moving the relationship chain from A to B are separated by a distance far greater than time data can measure. Time surpassing is a fact. Social replacement is another story that has not yet happened.

RMB 52.8 Billion, and RMB 52.7 Billion

Tencent's free cash flow in the second quarter of 2026 came in at negative RMB 13.8 billion. Rare in recent years. In the same earnings report, net operating cash flow was RMB 52.7 billion. First-half revenue was RMB 401.2 billion and profit for the period was RMB 117.3 billion, still growing at a double-digit rate year on year. The driver of negative FCF was capex shifting from cautious to catch-up. Second-quarter capex was RMB 52.8 billion, up 176 percent year on year, mainly into AI compute.

Look at these two numbers. Quarterly capex of RMB 52.8 billion. Quarterly operating cash flow of RMB 52.7 billion. They are almost equal. A cash cow does not contract because cash flow has dried up. It chooses to pledge its current cash-conversion capacity to AI infrastructure. The market's repricing is understandable. Tencent's PE center moved down from 20x to 15 to 17x. Foreign investors reclassified it from a cash cow with high cash flow and high buybacks to a growth stock that keeps burning cash on AI. The valuation logic has changed. The fundamentals have not collapsed. Whether a company turns FCF-negative because it cannot make money or because it chooses to spend money is two completely different financial situations. Tencent is the latter.

Before Alibaba's HK$80 billion placement, it had RMB 474.5 billion in cash and short-term investments. After deducting interest-bearing debt it was still a net-cash company, with Fitch A rating and a stable outlook. All placement proceeds went into full-stack AI. Quarterly capex was RMB 67.678 billion and FCF outflow was RMB 44.670 billion. The company did not raise money because it ran out. It used equity financing instead of debt financing to protect the net-cash position required for its A rating, making a front-loaded arrangement for a steep capex curve. The 8.4 percent placement discount is unfriendly to old shareholders. The choice can be debated. Reading it as a sign of collapsing fundamentals confuses financial discipline with financial distress.

Yuanbao is another matter. Its MAU is about 57.35 million, an order of magnitude below Doubao's 345 million. Criticism of the user experience is not groundless. But Tencent AI's financial returns are not written only on Yuanbao. Channels' total time growth is driven by AI recommendation. Marketing services revenue growth outperforming the industry relies on AI ad models. Game revenue growth also has AI contributions. Using Yuanbao's setback to infer that Tencent AI has failed across the board ignores the AI contributions already booked in the profit statement of the existing businesses.

Six Companies on the Weight Table

The accurate part of the double-short line lies in the index structure. Among the top ten weighted stocks in Hang Seng Tech, six compete directly with ByteDance. Combined weight around 40 percent. Alibaba, Meituan, Tencent, Kuaishou, JD.com, and Baidu. To buy Hang Seng Tech is to indirectly hold a portfolio that ByteDance keeps eroding, without being able to hold the eroder. Li Liang publicly denied the claim, calling it organized hype by marketing accounts. What he denied was the act of shorting. What he could not deny was the competitive relationship on the weight table.

What the line gets wrong is causality. Hang Seng Tech's moves are driven by global liquidity, the profit erosion of the food-delivery war, and the repricing caused by AI capex. Of the nearly 20 percent drawdown in Chinese concept stocks in early 2026, there was the factor of ByteDance competition. There was also liquidity tightening from a hawkish Fed nomination, and the earnings shock of Meituan swinging from RMB 35.8 billion net profit to RMB 24 billion net loss. Attributing all of this to one unlisted company compresses a multivariate world into a story that is easy to tell.

Back to That Voucher

ByteDance's offensive has data-verifiable sharpness. Local-services GTV has caught up to Meituan's order of magnitude. Hongguo's time advantage is structural. Feishu's ARR growth is rare in enterprise services. Douyin Pay has found its place in in-scenario penetration. These are not narratives. They are numbers on statements.

But the nature of winning needs to be separated. ByteDance has taken surface indicators: payment-basis GTV, usage rate, time, revenue growth. The deeper ledger tells a different story. WeChat's clearing foundation remains intact. Meituan's post-redemption transaction scale is still larger. The profit structure of enterprise services has not shifted. Winning GTV is not the same as winning post-redemption transaction value. A high usage rate does not mean a high transaction-value share. Time does not move a relationship chain. ARR growth does not become profit.

That expired hotpot voucher is still in the user's coupon wallet. Douyin's GTV booked it. Meituan's redemption volume did not. Both ledgers are honest. They are simply not recording the same thing.