2026-11-14 18:06

Lead: The Common Foundation Behind Zhang Yiming and Pony Ma’s Bets: Cash Flow from Legacy Businesses. Traditionally, evaluating an internet company centered on user engagement time.

Evaluating an internet company, the traditional metric has been user engagement duration. A recent ten-day debate within the industry focused on data methodology and conceptual boundaries around this metric.

Beyond the debate, ByteDance and Tencent both released another set of figures during the same period—answering a different question: In the AI era, what resources do internet companies need to secure strategic time?

According to third-party estimates based on QuestMobile’s monthly tracking, in the top 50 mobile apps by usage duration in July, ByteDance’s share reached 40.9%, while Tencent’s stood at 29.1%—the conventional benchmark used externally to compare the two firms. However, this metric only covers leading mobile applications and excludes long-tail apps and PC clients, marking the starting point of this controversy.

Shortly thereafter, both companies publicly disclosed their respective AI investments. According to reports, Zhang Yiming set a target for ByteDance’s AI team to catch up with frontier players by either 2029 or 2030. At the same time, rumors of Tencent “laying off 50% of staff” trended on social media. PR Director Zhang Jun dismissed it as a misunderstanding and misinformation, clarifying that the statement originated from Ma Huateng’s address during Tencent’s 28th anniversary event, where he announced allocating half of human capital to efficiency optimization in core businesses and the other half to AI-Native exploration.

ByteDance: Converting the Gap to a Year

Reports indicate that at an internal AI meeting, Zhang Yiming stated that ByteDance has the capability to compete with OpenAI and Anthropic, but full parity with the cutting edge will not be achieved until at least 2029 or 2030. The report also noted that the pre-training dataset for ByteDance’s Seed series models is several orders of magnitude smaller than those of top U.S. competitors, and ByteDance does not rely on model distillation for short-term leaderboard performance. On November 12, additional media sources confirmed similar claims through multiple informed insiders—though these remain unverified by official statements from ByteDance.

This timeline reflects both the current gap and the pace of development: model advancement is measured in years, while application iteration occurs on a weekly cadence. At the meeting, Zhang mentioned that a product comparable to Meta’s Muse would be launched within weeks. ByteDance aims to maintain product velocity through application and engineering strength, while simultaneously investing in foundational research over longer cycles.

To assess whether such declarations are credible, one must examine the financial commitment. ByteDance has raised its 2026 AI infrastructure budget to over RMB 20 billion, and in September signed a syndicated loan worth approximately USD 29.6 billion. Media disclosures reveal that ByteDance’s net profit for the first half of 2026 was around USD 20 billion, down slightly year-on-year, but with stable cash flow from core operations—providing solid backing for budget execution.

Whether 2029 is accurate remains unverifiable today. What is certain is that this marks the first time ByteDance has publicly disclosed a timeline for catching up with the frontier. Previously, no company had revealed its gap relative to the cutting edge; external observers could only speculate based on rankings.

With a specific year now established, progress can be tracked against a clear benchmark. Investors, decision-makers, and potential talent now have a verifiable reference point.

The effectiveness of this benchmark depends on ByteDance’s continued transparency. If this is merely a one-off statement, few may remember it by 2029. Ultimately, success will still hinge on product and model performance. Publicly stating a target changes the evaluation framework—now the external world can annually assess ByteDance’s actual progress.

Tencent: Resource Re-allocation

The phrase “half the workforce” originated from Ma Huateng’s livestream address at Tencent’s 28th anniversary Linktime event on November 11. According to media-transcribed remarks, Ma emphasized that Tencent previously operated light-asset businesses and is now entering a heavy-asset domain—requiring fundamentally different financial and organizational strategies. He proposed deploying half the workforce like an army to defend and optimize core businesses, while the other half acts like a navy exploring AI-Native frontiers.

Insiders close to Tencent told media outlets this represents a strategic reallocation of corporate resources in response to the AI era. Another internal source clarified that this was an internal speech—a high-level overview conveying direction, not an official directive issued via memo.

This ratio was oversimplified in public discourse as “layoffs.” The most direct evidence lies in headcount: As of June, Tencent’s employee count stood at 115,927, up from 111,221 a year earlier—a net increase of nearly 4,700 employees.

The actual measures discussed were reducing management layers and empowering frontline teams with decision-making authority—essentially restructuring reporting lines. During the same livestream, attendees reported that CEO Liu Chicheng added that the AI window is narrow, and decision-making chains must be shortened urgently.

Lu Shan, who leads the Technology and Engineering Business Group, stressed that innovation teams should not be constrained by KPIs from legacy operations, emphasizing the need for tolerance of failure. The former point explains why layer reduction is necessary—long decision chains cannot keep pace with fast-moving technological windows. The latter highlights that no standardized evaluation framework exists yet for AI innovation teams, making short-term performance assessment difficult and subjective.

Organizational shifts have already begun. In May, multiple departments within WeChat’s business group piloted project-based accountability. In July, the IEG Ecosystem Development Department eliminated L1 and L2 management levels. Capital expenditure is expanding in parallel: According to Tencent’s interim results, capex reached RMB 84.72 billion in the first half of the year, with Q2 alone at RMB 52.78 billion—an increase of 176% year-on-year. Ma Huateng emphasized in his speech that choosing individuals aligned with new operational models should be decisive.

The risk of the “half-workforce” strategy ultimately rests on delivery. If the half allocated to AI-Native initiatives fails to produce demonstrable assets within a reasonable timeframe, the narrative of layoffs may become self-fulfilling without any external push. Drawing parallels, Zuckerberg’s 2022 “Year of Efficiency” similarly began with an internal speech reorganizing resources, followed by subsequent layoffs and strategic pivot toward AI.

From User Engagement to Capital Allocation

Both ByteDance and Tencent treat time and human capital as actively allocatable resources, not passive costs growing organically from existing operations. User engagement continues generating cash revenue—whether this cash is reinvested into current businesses or redirected toward AI will likely define each company’s next phase.

This logic extends beyond just these two firms. On November 12, Li Yanhong announced at Baidu World 2026 that the next stage of AI is intelligent organizations, consolidating future roles into two categories: Builders (creators of new value) and Sellers (connectors and trust architects). Baidu disclosed that daily active intelligent agents at headquarters increased from 5,000 to 37,000. The consensus among the three companies is clear: AI isn’t just transforming products—it’s reshaping the enterprise itself.

Organizational trends are converging: Tencent is streamlining middle management, Baidu abolished its long-standing T/P/E/M job level system in April, and ByteDance CEO Liang Rubo incorporated “deep engagement with frontline operations” into managerial KPIs in a June company-wide message. Underlying this shift is the recognition that AI compresses the value of information flow—thicker middle layers slow down a company’s response to technological change.

Market valuation metrics are also evolving. Some institutions refer to Q3 as the “inflection point validation quarter” for Hong Kong tech stocks, cautioning that improved profitability mainly stems from reduced food delivery subsidies, with no substantial demand-side recovery yet observed.

Consensus expectations suggest Tencent’s adjusted net profit for Q3 declined by about 5% year-on-year. Other analysts believe actual results may exceed this estimate, forecasting a Q3 capex of RMB 66 billion. Tencent’s P/E ratio is around 14x, near the bottom of its decade-long range. Together, these indicators signal a prolonged evaluation cycle for internet companies: AI spending suppresses short-term profits, driving P/E ratios to historical lows. Markets are now pricing companies based on the expected recovery cycle of AI investments—not quarterly earnings alone.

For the industry, such adjustments are necessary. Heavy asset investment, long payback cycles, and organizational restructuring are unfamiliar moves for internet firms—evaluation frameworks must evolve accordingly. This can be distilled into three criteria: whether AI revenue is separately disclosed, whether capital expenditure depreciation periods are disclosed, and whether staffing-to-compute ratios for new ventures are made public. These standards apply across enterprise, industry, and individual practitioner levels.

On the operational level, “securing time” hinges on three verifiable metrics: whether AI revenue is separately itemized, how many years capital expenditures are depreciated over, and what tangible outputs result from organizational restructuring.

At the industry level, a unified outcome verification framework is still missing. Token usage volume and active intelligent agent counts are process indicators only—the conversion rate to revenue reveals true outcomes. Consistent methodology is therefore critical. For example, Volcano Engine’s market share in China’s public cloud MaaS segment is 49.5% by token call volume, but exceeds 40% by revenue.

Resource allocation patterns are also traceable. Tencent’s headcount continues growing year-on-year, with changes occurring in reporting lines and role definitions. How many people and how much compute power are assigned to new fronts will eventually manifest in hiring, financial reports, and organizational adjustments. These signals offer observable windows: Tencent’s Q3 earnings release on November 18 will provide verifiable data on actual capex, free cash flow movements, and whether AI revenue is disclosed separately—offering a concrete check on the “half-workforce” claim.

For ByteDance, whether its AI assistant matching Meta’s Muse is delivered as promised, when the next version of the Seed main model will launch (current version remains at 2.1 from June), and whether the developer for Phase II of the Brazil data center has been confirmed—all remain on the timeline.

Whether these commitments materialize depends on one fundamental condition: sustained cash flow from legacy businesses must continue to cover new ventures’ burn rates. Every investment ByteDance plans for 2029 relies on advertising-driven cash flow. Tencent’s allocation of half its workforce to new directions is funded by stable game and fintech operations.

This implies that the sustainability of these investments hinges not on rankings like 40.9% or 29.1%, but on the surplus between cash inflows and outflows.

(This article was first published on Titanium Media APP, Author | AGI-Signal,

Source: AGI-Signal