Skip to content
Technology

Byte’s 2029 and Tencent’s “Half People”

Measuring an Internet company used to mainly look at user duration. Around this indicator, there was a recent debate in the industry that lasted for ten days, focusing on the data caliber and conceptual boundaries.

Apart from the controversy, ByteDance and Tencent each gave another set of numbers during the same period, answering another question:In the AI ​​era, what do Internet companies rely on to buy themselves time?。

According to QuestMobile’s monthly tracking estimates quoted by a third-party agency, among the top 50 mobile apps in July, Byte’s time accounted for 40.9% and Tencent’s 29.1%. This is the conventional basis for outsiders to judge the two companies. This caliber only covers mobile head applications and does not include long-tail applications and PC applications. This is the starting point of this debate.

Soon after, the two companies disclosed their respective investments in AI. According to reports, Zhang Yiming gave a deadline for catching up with the forefront at an internal meeting of the Byte AI team: 2029 or 2030. During the same period, rumors that Tencent was “laying off 50% of its workforce” were on the hot search. Public relations director Zhang Jun called it a misunderstanding and rumor. This rumor came from Tencent’s 28th anniversary speech. Ma Huateng proposed that half of the manpower be used for core business efficiency improvement, and the other half was invested in AI Native exploration.

Bytes: Convert the distance to the frontier into a year

According to reports, Zhang Yiming said at an internal meeting of the Byte AI team that Byte has the ability to compete with OpenAI and Anthropic, and it will not be until 2029 or 2030 at the earliest to truly catch up with the frontier. The report also mentioned that the pre-training data set of the Seed series model is several orders of magnitude smaller than its top competitors in the United States. Byte does not rely on distillation in exchange for short-term ranking results. On November 12, other media obtained the same statement from multiple sources familiar with the matter. These are all paraphrases and there is no official confirmation from Byte.

This period corresponds to both gap and rhythm, with model catch-up measured in years and application iterations measured in weeks. Zhang Yiming said at the meetingAI assistance for Meta Muse will be launched in a few weeks手,Byte intends to use its application and engineering capabilities to maintain product rhythm while making long-term investment in the bottom layer.。

To determine whether a statement is serious, you can look at the amount of money the company has invested in it.Byte has raised its 2026 AI infrastructure budget to more than 200 billion yuan, and signed a syndicated loan of approximately US$29.6 billion in September. According to media disclosures, Byte’s net profit in the first half of 2026 is approximately US$20 billion, a single-digit decrease year-on-year. The cash flow of its main business remains stable, and the execution of the budget is supported by profits and cash flow.。

Whether it will be accurate in 2029 cannot be verified at the moment. What is confirmed is that thisThis is the first time Byte has publicly given the time to catch up with the cutting edge. Previously, no company was willing to announce the gap with the cutting edge. The outside world could only rely on the list to guess.。

After the year is announced, there will be a reference for the speed of progress, and the investor, the person selecting the company, and the talents considering joining all have an additional basis for checking.

Whether this coordinate can work depends on whether Byte continues to disclose progress. If it is just a one-time statement, no one may remember it by 2029. The outcome is still determined by products and models. What changes in the year of disclosure is the evaluation method. From now on, the outside world can compare Byte’s progress year by year.

Tencent: It’s resource scheduling

Tencent’s statement of “half of the people” came from the 28th anniversary Linktime online live broadcast on November 11. According to the transcript of the speech disclosed by the media, Ma Huateng said in his speech that Tencent used to do an asset-light business. This is the first time that Tencent has done such an asset-heavy business. The financial and organizational structure are different. He proposed that half of the people are like the army, guarding the core business and improving efficiency, and the other half are like the navy, exploring AI Native.

People close to Tencent told the media that this is a reallocation of the company’s resources to adapt to the AI ​​era. Another Tencent insider told the media that this was an internal speech, an overview, conveying development direction, and not an internal document.

This set of ratios has been simplified into layoffs in the communication. The most direct evidence to test this reading is the total number of employees: As of June, Tencent had 115,927 employees, compared with 111,221 in the same period last year, a net increase of about 4,700 people in a year.

The actual actions mentioned in the speech are to compress the management level and give decision-making power to the front line, and more importantly, to adjust the reporting relationship. In the same live broadcast, according to reports from participating employees, President Liu Chiping added,The AI ​​window period is very short, and the decision-making chain must be shortened。

Lu Shanze, who is in charge of the technical engineering business group, proposed,Don’t use existing business KPIs to constrain the AI ​​innovation team, but leave room for fault tolerance for innovation.. The previous sentence explains the reason for compressing the hierarchy. The decision-making chain is too long to catch up with the window period. The latter sentence shows that there is no corresponding assessment method for the AI ​​innovation team. Whether it is doing well or not, it is difficult to have a unified judgment standard in the short term.

Organizational actions have already taken place before. In May, multiple departments of the WeChat business group piloted the project responsibility system, and in July, the IEG Ecological Development Department canceled the L1 and L2 management ranks. Capital expenditures expanded simultaneously. According to Tencent’s interim results, capital expenditures in the first half of the year were 84.72 billion yuan, and in the second quarter, capital expenditures were 52.78 billion yuan, a year-on-year increase of 176%. Ma Huateng said in his speech that those who choose to adapt to the new style of play must be decisive.

The risk for “half the people” ultimately lies in cashing out. If the half of the people who invested in AI Native cannot produce demonstrable assets within a reasonable period of time, the layoff narrative may become a true prophecy without anyone pushing it. Referring to peers, Zuckerberg’s “Year of Efficiency” in 2022 also used an internal speech to rearrange resources, and subsequent layoffs and AI transitions began from this.

From user duration to resource investment

Byte and Tencent regard time and manpower as resources that can be proactively allocated, rather than expenses that naturally grow with existing business. User time is still generating cash income. Whether these incomes will remain in the existing business or transferred to AI may determine the next stage of the two companies.

This configuration logic is not limited to Byte and Tencent. On November 12, Robin Li proposed at Baidu World 2026, the next scene of AI is intelligent organization, he converged future jobs into two categories:Builder, who is responsible for creating new things, and Seller, who is responsible for connection and trust.. The data disclosed by Baidu is that the average daily active agents at the headquarters increased from 5,000 to 37,000. The judgments of the three companies are consistent: AI will change not only the products, but also the company itself.

The directions of organizational changes are becoming consistent. Tencent has compressed its middle management. In April, Baidu canceled the T, P, E, and M rank labels that have been used for many years. Byte CEO Liang Rubo’s all-employee letter in June included “going deep into the front line” in managers’ evaluation. The logic behind it is that AI compresses the value of information uploaded and delivered. The thicker the middle layer, the slower the company’s response to technological changes.

The scale of the capital market is also adjusting.Some institutions call the third quarter of this year the profit turning point verification season for Hong Kong technology and Internet stocks. They also suggest that the recovery in profits is mainly due to the decline of takeaway subsidies, and there has been no substantial improvement on the demand side.。

The expectations compiled by institutions are that Tencent’s adjusted net profit in the third quarter will fall by about 5% year-on-year. Other institutions judge that the actual results may be better than this expectation, and give a single-quarter capital expenditure estimate of 66 billion yuan. Tencent’s price-to-earnings ratio is about 14 times, which is at the bottom of the past decade. Taken together, the three sets of numbers conclude that the market’s cycle for evaluating Internet companies is lengthening: AI spending is suppressing short-term profits, the price-to-earnings ratio has been pushed to historical lows, and the market is beginning to price companies based on the payback cycle of AI investment. Single-quarter profits are no longer the only coordinate.

For the industry, this kind of adjustment is necessary. Asset-heavy, long-term, and organizational rearrangement are all actions that Internet companies are unfamiliar with, and measurement methods also need to be updated accordingly. This set of criteria can be summarized into three criteria:Whether AI revenue is listed separately, whether the depreciation life of capital expenditures is disclosed, and whether the ratio of new business personnel and computing power is disclosed. Various calibers at the enterprise level, industry level, and practitioner level ultimately fall on them.

“Buying time”At the operational level, the transformation progress depends on three sets of verifiable figures: whether AI revenue is listed separately, capital expenditures are depreciated over several years, and what actual outputs are brought about by organizational adjustments.

There is still a lack of a unified effect verification framework at the industry level. The number of token calls and the number of active agents are only process indicators, and the ratio of conversion to revenue reflects the results. Caliber discipline is therefore important. The share of Byte’s Volcano Engine in China’s public cloud MaaS market is 49.5% based on the number of calls, and the revenue caliber is more than 40%.

The flow of resources is also traceable. The number of Tencent employees is still increasing year-on-year. Changes have occurred in reporting lines and job definitions. How many people and computing power are assigned to the new battlefield will ultimately be reflected inRecruitment, financial reporting and organizational adjustments are underway. These clues have corresponding observation windows. Tencent will release its third-quarter financial report on November 18. Whether the actual figures of capital expenditures, changes in free cash flow, and AI revenue are disclosed separately will provide “half of the people” with a verifiable reading.

About Us · 關於我們