文 | Tech商业
本周，一场关于“中国互联网到底是谁的时间”的交锋，意外地撕开了中国科技产业最深层的结构变迁。
知名学者秦朔发文提出，中国互联网已经进入了“字节时间”；而抖音集团副总裁李亮则连夜发文回应，直指其数据口径存在常识性错误，并以字节系在Top 50 App中41.8%的时长份额未达《反垄断法》二分之一标准来否认垄断定性。李亮在回应中还特别提到，如果算上腾讯参股投资的App，总使用时长应该远超其他，按照同样的逻辑，中国互联网可以说是“腾讯时间”。
这场关于统计口径与业务边界的罗生门，看似是公关与学者的隔空对话，但其引发的巨大共鸣，恰恰揭示了一个无法回避的现实：无论“字节时间”这个定性是否严谨，字节跳动已经长成了一个在体量、财富创造以及技术纵深上，足以定义整个行业标准的超级巨头。
讨论这家公司的现在与未来，已经不能仅仅停留在单一产品的日活或某项数据的统计标准上。我们需要从更深层的商业逻辑、技术演进和产业生态来拆解：一个超级巨头是如何在移动互联网向AI时代过渡的夹缝中，完成体量的跃升，并重构整个行业的价值坐标的。
重写定价
讨论巨头的体量，最直接的标尺是资本市场的定价与创始人的财富值。
在2026年的私募市场中，字节跳动的估值在5000亿至6000亿美元区间；腾讯在二级市场的市值同期约5040亿美元，但需注意腾讯市值在2026年3月曾突破5万亿港元后大幅回调，字节的“逼近”部分受益于对手的低潮期。如果字节以6000亿美元左右的估值登陆公开市场，它将直接超越腾讯，成为中国大陆市值最高的互联网公司——但一级市场估值不等于上市市值，港股IPO常有折价，6000亿美元是“锚”而非“结果”。
伴随估值狂飙的，是创始人财富的登顶。2026胡润百富榜显示，43岁的张一鸣财富增长2000亿元，以6700亿元人民币第二次成为中国首富，并刷新了中国首富的财富纪录。胡润明确指出，其增长直接得益于AI业务进展推动的公司估值走高。而55岁的马化腾，财富缩水26%至3460亿元，排名降至第四。
这不仅仅是两个富豪排名的更替，更是中国互联网财富创造逻辑的底层切换。
在过去十年，腾讯的财富神话建立在一个无与伦比的“社交网络效应”之上。微信和QQ构建的熟人关系链，具有极高的迁移成本和网络锁定效应。这种效应为腾讯带来了丰厚的游戏代理收入和广告利润，但也意味着，基于社交关系的流量红利已经触及了物理上的天花板。
而字节的财富跃升，则代表了另一种更具爆发力的模式：算法驱动的注意力分发与全栈式技术红利。从今日头条到抖音，再到红果短剧和豆包，字节并不依赖用户的社交关系链，它依赖的是对用户多巴胺和好奇心的精准计算。
财务数据印证了这种爆发力。2026年上半年，字节营收约1200亿美元，同比增长约30%，营收规模已追平Meta同期的1171亿美元；净利润约200亿美元，同比下降个位数百分比，海外收入占比超过30%。即便由于AI基建的疯狂投入导致净利润率有所下降，其现金流和融资能力，依然足以支撑它在算力、人才和基础设施上维持极高强度的持续投入。
全栈闭环
字节之所以能在估值和财富上完成对传统巨头的追赶甚至超越，核心驱动力来自它在AI时代展现出的全栈式布局与极其强悍的工程化落地能力。这套布局并非停留在概念层面，它在每一个环节都留下了可被验证的数字。
在模型层，调用规模正在以指数级速度膨胀。截至2026年6月，豆包大模型的日均Tokens调用量已突破180万亿，相较2024年5月发布时的1200亿初始规模，两年间涨幅超过1500倍，过去一年增长超10倍。在公有云MaaS市场，火山引擎的Token份额达到49.5%——中国企业每消耗两个Token，就有一个由火山引擎提供。年累计Token消耗超过1万亿的企业客户，已从2025年12月的100家增至200家。
在应用层，国民级产品的密度构成了闭环的流量底座。截至2026年9月，豆包App月活跃用户达4.19亿，同比增长143.6%；日活跃用户规模的不同来源存在口径差异：QuestMobile监测的9月末日活约1.81亿，另有报道称日活已突破2亿。豆包稳居国内AI原生应用市场首位。红果短剧2026年7月日活跃用户达1.68亿，同比增长107%，已超过爱奇艺、腾讯视频、优酷、芒果TV四家长视频平台日活之和（约1.55亿）；其月活跃用户3.68亿对应的是2026年6月的数据。加上抖音、TikTok和番茄小说，字节在内容消费、社交娱乐和AI助手三个维度上，同时握有日活过亿的超级入口。
在商业化与工具层，收入引擎的增速同样可观。火山引擎2026年将MaaS营收目标上调至150亿元，相当于2025年实际营收（约15亿元）的10倍；其中仅Seedance 2.0视频生成模型，单月即可贡献超10亿元收入。支撑这一切的是字节2026年逾2000亿元的AI资本开支，以及一笔约296亿美元的银团贷款。
将这些数字放在一起，闭环的轮廓才真正清晰：180万亿日均Token调用量驱动着豆包和红果的模型能力；4.19亿月活和1.68亿日活产生海量真实交互数据，反哺模型迭代；Seedance的单月10亿收入和火山引擎150亿的MaaS目标，则将技术能力转化为现金流，支撑下一轮算力投入。模型、流量、数据与资本在这个体系内形成了一条高速循环的链路，而每一个环节的规模效应，都在为下一个环节降低边际成本。
这正是“强者愈强”在AI时代最直观的体现。当其他大模型公司还在为高昂的算力成本和难以闭环的商业化而焦虑时，字节已经通过极其丰富的落地场景，完成了大模型技术的自我造血和持续迭代。每一次内容的分发、每一次广告的点击、每一次API调用，都在为它的算法模型提供海量的数据反哺。
集中的张力
当一个组织同时拥有了最大的用户时长份额、最完整的AI产品矩阵、最雄厚的现金流以及最高效的变现机器时，资源集中就成为了一种自然演化的结果。
从积极的一面看，这种集中带来了极高的产业效率。字节用推荐算法重构了内容分发的成本结构，用AI大幅压低了短剧等品类的制作门槛，用一套统一的流量与商业化体系，让创意和生产的周转速度达到了传统工业难以企及的水平。
但资源集中同样会引发值得持续观察的张力。
当越来越多的注意力、创作产能和商业预算向同一个体系收敛时，生态的多样性、中小参与者的议价空间、以及平台规则对创作自由的边界，都会成为长期的公共议题。李亮在回应中以字节系Top 50 App合计41.8%的时长份额未达二分之一来否认垄断定性，这在法律和统计口径上是一个有效的辩护；但从产业生态的角度看，份额数字之外的影响力半径——对上下游的牵引力、对定价和分发规则的塑造力——同样值得被纳入讨论。
在互联网反垄断的实践中，“二分之一”是法定的量化门槛，但并非衡量市场影响力的唯一维度。当一家公司通过参股、孵化、技术输出和流量分发，构建起跨领域的网络时，它对市场竞争的实际影响，往往比单一App的份额数字更为复杂。字节并非第一个面临这种审视的巨头。腾讯、阿里、Meta、谷歌，在成长为基础设施级公司的过程中，都曾经历类似的追问。这是任何一家达到超级规模的企业，必然面对的结构性处境。
算法的折叠
如果我们更深一层去审视字节在AI时代的扩张，会发现它不仅在重构商业，也在重塑内容的生产关系。
以红果短剧为例，它之所以能在短时间内实现对传统影视行业的超越，核心在于它打通了“网文IP+AI制作+算法分发+广告变现”的完整链条。番茄小说提供经过数据验证的爽文IP，豆包大模型充当高效率的编剧，AI视频工具进一步压缩制作成本，最后通过红果和抖音的算法精准推给受众。
这套模式极其高效，但也带来了一种隐蔽的折叠。在算法的驱动下，内容创作越来越向“数据验证过的爽点”倾斜。什么题材留存率高，就批量生产什么；什么反转能刺激多巴胺，就疯狂堆叠什么。算法没有偏见，它只是忠实地执行效率最大化的原则。但在这个过程中，那些需要慢节奏铺垫、需要深度思考、或者不符合大众爽点的高质量小众内容，往往会因为ROI不够而被边缘化。
这并非字节一家的独有现象，而是整个算法推荐时代面临的共同挑战。当平台规则和流量分发成为一种无形的指挥棒时，创作者的独立性、内容的多样性，与平台追求的商业效率之间，存在一种天然的结构性张力。
巨头的边界
“字节时间”这个说法是否准确，或许并不重要。重要的是，字节跳动已经用实打实的估值、财富、营收和AI全栈能力，证明了自己是这个时代效率与规模的顶点之一。
张一鸣成为首富、字节估值逼近甚至可能超越腾讯，本身并不是需要被赞美或批判的事情——它们是市场、技术和组织能力共同作用的结果，是一个时代选择的一种呈现方式。
在商业史的漫漫长河中，我们往往会对庞然大物产生本能的警惕。但客观来看，每一个时代的更迭，必然伴随着超级巨头的诞生。PC时代的微软，移动互联网时代的腾讯，以及如今的字节跳动，都是那个时代最先进生产力的代表。
字节的崛起，代表了在极致人才密度和极致工程效率下，一家中国企业所能达到的商业高度。它靠着对用户需求的精准洞察、对AI技术的疯狂投入，以及极其残酷的内部竞争机制，硬生生打出了一片江山。它代表了中国科技企业在出海、在AI应用层极具竞争力的真实水平。
但与此同时，当一家公司成长到足以影响数亿人的信息获取习惯、决定海量创作者的生存状态时，它的商业决策就不再仅仅关乎财务报表，而具有了极其广泛的社会属性。
体量与边界的平衡，是所有超级巨头都避不开的终极命题。在追求极致的商业效率与维持生态的多样性之间，在技术扩张的冲动与商业伦理的约束之间，如何找到那个动态的平衡点，不仅考验着管理层的智慧，也决定着这家公司能否在更长的时间维度里，持续获得公众与市场的信任。
答案不在任何一篇文章里，而在市场规律的演进、行业规则的碰撞，以及每一次商业选择的长期回响之中。
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Man, Tech Business
This week, an exchange on “Who’s Time is China’s Internet” accidentally torn the deepest structural changes in China’s technology industry.
The leading scholar, Qin Soo, submitted that the Chinese Internet had entered “byte time”, while the Vice-President of the Douyin Group, Li Yang, responded by writing overnight, pointing to a common sense error in his data calibre and denying the monopoly characterization by byte ' s long-term share of 41.8 per cent in Top 50 App, which did not meet one half of the antimonopoly law criteria. In its response, Li Shun also mentioned in particular that, if the app of Tencent equity investment were included, the total use time should be much longer than other, and that, in the same logic, the Chinese Internet could be described as “Tencent time”.
This seemingly isolated dialogue between public relations and scholars on the statistical caliber and operational boundaries of Rome has generated a huge resonance that reveals an inescapable reality: Whether or not the term “byte time” is strict, the byte beat has grown to be a super-gravity in volume, wealth creation and technology depth sufficient to define the whole industry standard.
Discussion of the company ' s present and future can no longer be confined to a single product ' s daily life or a statistical standard for a particular data. We need to break down from deeper business logic, technological evolution and industrial ecology: how a super-grave can leapfrog the volume in the middle of moving the Internet to the age of AI and re-engineer the value coordinates of the entire industry.
Rewrite pricing
To discuss the size of the giant, the most direct rule is the price of the capital market and the wealth of its founders.
In the private market in 2026, byte jumps were valued at between $500 billion and $600 billion; the market value of Tencent in the secondary market was about $504 billion over the same period, but it should be noted that the market value of Tencent had been reversed in March 2026, after having exceeded HK$5 trillion, and the “closing” part of byte had benefited from a low level of rivals. If byte landed on the open market with a valuation of around $600 billion, it would go beyond Tencent and become the Internet company with the highest market value on mainland China — but the first-tier market valuation would not be equivalent to market value on the market, and the IPO would often be at a discount, with $600 billion being an “interlock” rather than a “outcome”.
With the valuations raging, the founders' wealth peaked. The list shows that 43-year-old Zhang Zhang Zheng Yi Zheng's wealth grew by 200 billion yuan, became China's first wealth for the second time with RMB 67 billion and refreshed China's wealth record of the richest. Hu Yun made it clear that his growth was directly due to higher corporate valuations driven by the progress of AI operations. At the age of 55, however, the wealth is reduced by 26 per cent to 34.6 billion yuan, down to the fourth.
This is not just a change in the ranking of the two wealthy, but rather a bottom-up shift in the logic of Internet wealth creation in China.
Over the past decade, the wealth myth of Tencent has been based on an unparalleled “social network effect”. WeChat and QQQ have a familiar relationship chain with extremely high migration costs and network lock-in effects. This effect has resulted in a high level of game agent income and advertising profits for Tencent, but it also means that the flow dividend based on social relationships has touched the physical ceiling.
byte's wealth surge represents another more explosive model: Algorithms-driven attention distribution and a whole-house technology dividend. From today ' s headlines to Douyin, to the red fruit dramas and Doubao, Z113Z does not depend on the user ' s social network, but on a precise calculation of dopamine and curiosity.
Financial data attest to such an outbreak. In the first half of 2026, the byte battalion received approximately $120 billion, an increase of about 30 per cent over the same period, and the scale of the collection was brought up to $117.1 billion over the same period in Meta; the net profit was about $20 billion, a percentage of the decline in the same period, with overseas revenues accounting for over 30 per cent. Even as a result of the decline in net profit margins caused by the mad investment in AI infrastructure, its cash flow and financing capacity are still sufficient to sustain its sustained investment in a very high level of computing, talent and infrastructure.
The whole house is closed.
byte's ability to catch up with traditional giants in terms of valuation and wealth has been driven even beyond its core by its full-knot layout and extremely strong engineering downland capability in the AI era. The layout is not at the conceptual level; it leaves verifiable figures at every point.
At the model level, the call scale is expanding at an exponential rate. By June 2026, the average daily tokens call of Doubao LLM had exceeded 1.8 trillion, an increase of more than 1,500 times over two years and 10 times over the past year, compared with the initial size of 120 billion at publish in May 2024. In the public cloud Maas market, the Token share of Volcengine has reached 49.5 per cent - for every two Token consumed by a Chinese enterprise, one is provided by Volcengine. The cumulative token consumption of over 1 trillion business clients per year increased from 100 in December 2025 to 200.
At the application level, the density of national-level products constitutes a closed-ring flow base. As of September 2026, DoubaoApp had 419 million active users per month, an increase of 143.6 per cent over the same period; there were calibration differences between different sources of the size of active users per day: QuestMobile monitored about 181 million people living on the end of September and there were reports of more than 200 million people living on the day. Doubao is at the top of the national AAI application market. In July 2026, the red fruit short show reached 168 million active users, an increase of 107 per cent over the same period, exceeding the number of Achi, Tencent videos, and the sum of parents ' video platforms (approximately 155 million) of the mango TV four; its monthly active users of 368 million corresponded to the data of June 2026. Together with Douyin, TikTok and tomato novels, byte has three dimensions of content consumption, social entertainment and AI ' s assistant, while holding a super-entry point where hundreds of millions live daily.
The growth of the revenue engine is as significant in commercialization as in the tool layer. Volcengine2026 increased the target for the Maas battalion to $15 billion, which is 10 times the actual collection in 2025 (about $1.5 billion); of these, only the Feedance 2.0 video generation model can contribute more than $1 billion a month. This was underpinned by byte ' s over $200 billion AI capital expenditure in 2026 and a co-financing loan of approximately $29.6 billion.
When these figures are put together, the contours of the closed circle are really clear: Token's call averaged 1.8 trillion days drives the modelling capacity of Doubao and the red fruit; 419 million monthly and 168 million daily live to produce real interactive data on the mass, and an iterative anti-nursing model; and Seedance's monthly income of 1 billion dollars and the Maas target of Volcengine15 billion, translates technological capabilities into cash flows to support the next round of calculus inputs. Models, flows, data and capital form a chain of high-speed cycles within the system, and the scale effect of each link is reducing marginal costs for the next.
This is the most intuitive manifestation of the "strength of power" in the AI era. While other LLM companies are still anxious about the high cost of computing and the unenclosed commercialization, byte has completed the self-made blood and continuity of LLM technology through an extremely rich landing scene. Every distribution of content, every click on an ad, every API call is providing a big data retrospect for its algorithm model.
Focused tension.
When an organization simultaneously has the largest user-time share, the most complete AI product matrix, the largest cash flow and the most efficient liquidity machine, resource concentration becomes the result of a natural evolution.
On the positive side, this concentration has led to very high industrial efficiency. byte reconstructs the cost structure of content distribution using recommended algorithms, significantly lowers the production threshold for short dramas, etc., with AI, and uses a unified flow and commercialization system to bring creativity and production up to a level that traditional industries cannot access.
However, the concentration of resources can also give rise to tensions that deserve sustained observation.
When more attention, creative capacity and commercial budgets are drawn to the same system, ecological diversity, the bargaining space of small and medium-sized participants and the boundaries of platform rules for freedom of creation become long-term public issues. In his response, Lee denied monopoly characterization by byte, which was less than one half of the total 41.8 per cent of Top 50 App, as a valid legal and statistical defence; however, from an industrial ecological point of view, the impact radius beyond the share figures — the pull of up and down, the shaping of pricing and distribution rules — is also worth discussing.
In Internet antimonopoly practice, “one in two” is the statutory quantization threshold, but it is not the only dimension of market impact. When a company builds cross-cutting networks through equity participation, incubation, technology export and distribution, its actual impact on market competition is often more complex than the single App share figure. byte is not the first giant to face such a review. Tencent, Ali, Meta, Google have all experienced similar inquiries as they developed into infrastructure companies. This is the structural situation inherent in any super-sized enterprise.
Collapse of algorithms
If we look deeper at the expansion of byte in the AI era, we find that it is reshaping not only business but also content production relationships.
In the case of the red fruit drama, for example, its ability to move beyond the traditional video industry in a short period of time lies at the heart of its complete chain of “web-based IP+AI production + algorithm distribution + advertising to cash”. Tomato novels provide data-valided cipher IP, Doubao LLM acts as an efficient script writer, and AI video tools further compress production costs, which are ultimately channelled to the audience by the algorithms of the nuts and Douyin.
The model is extremely efficient, but it also brings a hidden fold. Driven by algorithms, content creation is increasingly tilted towards " data-validated ". Any subject has a high retention rate, it's a mass production; whatever inverts dopamine, it's a mad pile. Algorithms are free of bias, but only faithfully implement the principle of maximizing efficiency. In this process, however, high-quality content that requires a slow pace, deep-thinking or that is not in line with popular good will tend to be marginalized by the lack of ROI.
This is not unique to the byte family, but is a common challenge to the entire recommended age of algorithms. There is a natural structural tension between the independence of creators, the diversity of content and the commercial efficiency pursued by the Platform when its rules and traffic are distributed as an invisible command bar.
The borders of giants.
It may not matter whether the term “byte Time” is accurate. Importantly, the byte jump has proved itself to be one of the culminations of the efficiency and scale of this era, using solid valuation, wealth, revenue and the full capacity of AI.
It's not in itself something to be praised or criticized - I don't know. They are the result of a combination of market, technological and organizational capacities, and a way of presenting an era of choice.
In the long rivers of business history, we are often instinctively wary of big things. But objectively, the transformation of every era must be accompanied by the birth of a super-grave. The Microsoft of the PC era, the Tencent of the mobile Internet age, and now the byte beat are the most productive representatives of that era.
The rise of byte represents the commercial height that can be achieved by a Chinese enterprise with a high level of talent density and engineering efficiency. It has struck a mountain with precision of insight into user needs, a mad investment in AI technology and an extremely cruel internal competitive mechanism. It represents the true level of competitiveness of Chinese science and technology enterprises at the level of the AI application.
At the same time, however, when a company grows sufficiently to influence hundreds of millions of people ' s access to information to determine the survival of its mass creators, its business decisions are no longer related to financial statements alone, but have very broad social attributes.
The balance between volume and boundary is the ultimate proposition that all super-Powers can avoid. Finding a balance between the pursuit of extreme commercial efficiency and the maintenance of ecological diversity, between the impulses of technological expansion and the constraints of business ethics, is not only a test of management wisdom, but also of whether the company can sustain public and market confidence in a longer time horizon.
The answer is not in any article, but in the evolution of market patterns, the collision of industry rules and the long-term resonance of every business choice.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.