阿里腾讯字节吃掉93%流量后,360借力算力交易中心杀入:1亿Token是企业智能体抢占市场的硬通货

The competitive landscape of the 2026 AI office agent track presents an extremely brutal oligarch pattern.


After Three Giants Seize 93% of Traffic, 360 Breaks In With 100 Million Tokens

English Version

The competitive landscape of the 2026 AI office agent track presents an extremely brutal oligarch pattern.

Among the total 60 million desktop visits across the industry, Tencent’s four core products contribute 32.62 million visits, ByteDance’s ecosystem accounts for 14.41 million, and Alibaba’s system takes 9.19 million. The three tech giants have occupied 93% of the entire market traffic, leaving merely 4.4 million visits for all independent innovative AI products. This is no longer a fiercely competitive market, but a mature oligarch monopoly track with extremely narrow survival space for new players.

Against this rigid pattern of traffic monopoly, 360 officially stepped into the enterprise AI track with its new-generation intelligent agent workspace — NanoWork, opening a differentiated breakthrough path beyond traffic competition.

100 Million Tokens: Not a Simple Subsidy, But a Premium Industry Admission Ticket

On July 28, 2026, Zhou Hongyi, founder of 360 Group, officially launched NanoWork. At the launch event, he announced a landmark support policy for early users: providing 100 million exclusive token trial quota per user, paired with professional systematic training and one-on-one full-process operation coaching services, to help the first batch of 1,000 small and micro enterprises complete AI business implementation and operational efficiency upgrading at zero cost.

Converted by the pricing standards of current mainstream large models, 100 million tokens are equivalent to tens of thousands of US dollars in high-quality computing resources. In total, 360 released billions of tokens in one go, with unprecedented support intensity.

It is superficial to regard this move as simple subsidy-based customer acquisition. Behind NanoWork’s massive token benefit lies a mature wholesale-to-retail computing power industrial chain: the upstream relies on national-level computing trading centers for bulk standardized computing supply, while the downstream leverages token benefits to accurately empower end enterprises, reshaping the underlying competition logic of the AI industry.

NanoWork builds five core capability systems: multi-agent engine, multi-model base, cloud office workspace, diverse working modes and multi-expert intelligent library. It supports seamless switching of 16 top-tier global AI models and integrates an ecological matrix of more than 50,000 professional AI skills.

Enterprises do not need to spend resources researching model selection, agent frameworks or tool adaptation. Users only need to clarify business goals, and the platform will automatically disassemble complex tasks, organize collaborative work of multiple AI expert agents, and match the optimal model and execution solution intelligently.

Its core advantage is the three-tier automatic intelligent computing scheduling mechanism, which precisely allocates lightweight, cost-effective and full-performance computing resources according to task complexity, avoiding blind resource consumption.

This design solves a universal industry pain point that most AI products avoid: every token consumption has real costs, yet most enterprise users lack the ability to calculate and control expenses accurately, resulting in unpredictable token waste and out-of-control bills in complex scenarios.

NanoWork fundamentally reverses this situation by taking over the decision-making rights of "model selection and token consumption" from users. All computing allocation is completed through platform intelligent scheduling, realizing full-process refined cost control.

Zhou Hongyi shared a thought-provoking viewpoint: "For enterprises to adopt AI, business leaders must use it first."

This is not a marketing slogan, but an accurate insight into industrial pain points. The biggest obstacle to enterprise AI implementation is never insufficient technical capability, but decision-makers’ ignorance of the AI cost structure. Most enterprise leaders cannot clarify token unit prices, model-task matching logic and the actual business value of massive tokens. The 100 million-token trial quota is designed to let enterprise decision-makers experience the whole process personally, verify AI’s true value, and thoroughly figure out the AI cost accounting logic.

Computing Power Trading Centers: Tokens Evolve From Consumables to Circulating Currency

Combining NanoWork’s layout with a series of national industrial policy upgrades and infrastructure constructions in 2026, the strategic pattern of token industrialization becomes extremely clear.

Tokens have officially obtained national industrial recognition. Ahead of the 2026 Zhongguancun Forum, the National Data Bureau officially standardized the Chinese translation of Token as "ciyuan", defining it as a measurable, priced and tradable core asset in the intelligent era. The Ministry of Industry and Information Technology also explicitly proposed innovative service models such as "computing power bank", "computing power supermarket" and "token billing" in special action notices, making tokens the standardized hard currency for domestic computing power transactions.

A number of national-level computing trading hubs have been successively launched and implemented in 2026:

On July 19, the Tonglu DeepLink Token Sales and Settlement Headquarters project officially settled in Tonglu, building a national unified hub for computing service measurement, settlement and trading with five core functions: supply-demand matching, real-time settlement, green power traceability, cross-border clearing and compliance supervision.

On July 1, the Nanjing Token Collection Store launched in Qilin Science and Innovation Park, taking tokens as the unified measurement standard for computing power and connecting the whole industrial chain of computing supply, model development, industrial application and financial services.

On June 1, the Inner Mongolia Ciyuan Trading Platform, the country’s first full-stack green computing AI platform, officially operated, integrating one-stop functions such as computing scheduling trading, convenient model calling and token settlement.

The core logic of these platforms is to realize standardized market circulation of computing power, just like public utilities such as water and electricity. The new trading mechanism has three disruptive core designs:

Real-time settlement: It reduces the cash flow pressure on small and medium-sized computing suppliers to zero. Suppliers can withdraw funds immediately after providing token resources, completely solving the long-standing industry pain point of advance capital occupation and slow repayment in traditional computing transactions.

Inclusive access mechanism: Platform entry and usage thresholds are greatly reduced through incentive policies and supporting financial services. The Nanjing Token Collection Store cooperates with Bank of Nanjing to launch exclusive computing financial products. The "Computing Super Ecosystem Co-construction Plan" initiated by the China Academy of Information and Communications Technology further promotes inclusive service modes of on-demand access and credit support.

Market-oriented dynamic pricing: API calling prices float in real time according to market supply and demand. Individual developers enjoy lower costs during off-peak periods, while suppliers gain higher profit margins during peak demand. This is the key landing practice of national top-level token trading design, completely breaking the monopoly pricing of head platforms.

All layouts confirm the core industrial trend:tokens are completing a fundamental transformation from one-time consumables to universal circulating industrial currency.

In the traditional model, tokens flow in a single direction of "purchase-consume-exhaust" with no circulation value. The new computing trading ecosystem forms a closed-loop two-way flow: supplier supply, user on-demand purchase, dynamic market pricing and real-time fund settlement, evolving from a simple API store to a complete financial trading market for AI resources.

Two Paths, One Ultimate Industrial Endgame

360 NanoWork and national computing power trading centers adopt different industrial layouts but converge at the same core goal, forming a two-way empowered pattern of demand-side activation and supply-side restructuring.

360 NanoWork: Activating the demand side

By providing 100 million-token subsidies for small and micro enterprises, it greatly reduces the trial-and-error cost of enterprise AI adoption. The three-tier intelligent scheduling mechanism realizes full transparency of AI costs. Users do not need professional model knowledge, and can accurately match cost-saving or full-performance computing modes according to scenarios, making AI cost accounting simple and intuitive.

Computing Power Trading Centers: Restructuring the supply side

It returns token pricing power from platform monopoly to the market. The dynamic supply-demand pricing mechanism benefits both developers and suppliers: low-cost usage in off-peak periods and high-profit returns in peak periods, fully activating industrial vitality.

The integration of the two paths spawns a new industrial logic: tokens are no longer rigidly priced commodities, but tradable assets with market-oriented floating values.

The New Accounting Rule for Explosive Computing Demand

The 2026 AI market shows a distinctive structural change: while the unit price of large models continues to decline, enterprises’ overall AI computing bills keep rising. The core reason is that the growth rate of token usage far exceeds the decline rate of unit prices.

According to data disclosed by the National Data Bureau, China’s daily average token calling volume has soared from 100 billion in early 2024 to 100 trillion by the end of 2025, and exceeded 140 trillion by March 2026, a surge of more than 1,000 times in two years.

The explosive growth of usage exposes two fatal bottlenecks of the traditional pay-as-you-go model: price rigidity under platform monopoly and severe cash flow pressure on suppliers. The market-oriented pricing and real-time settlement mechanisms of computing trading centers precisely solve these two pain points.

NanoWork complements the supply-side reform from the demand side: realizing on-demand payment through refined scheduling and lowering enterprise AI decision thresholds through large token subsidies. The industry forms a healthy iterative pattern of supply-side marketization and demand-side transparency.

The Core Ace of Breaking the Monopoly

Zhou Hongyi pointed out: "Models determine how intelligent AI is, and agents determine whether AI can truly work." Supplemented with the token industrial trend: Computing power trading mechanisms determine whether AI can be universally affordable and popularized.

Before the launch of NanoWork, 360 completed rigorous internal verification: deploying self-developed agents such as OpenClaw in more than 1,000 real internal business scenarios, collecting over 56,000 user feedback entries, and completing 166 version iterations within 5 months, accumulating mature enterprise-level AI service capabilities.

This fully proves that 360’s core output is not simple token resource sales, but a complete set of systematic capabilities to help enterprises "spend tokens reasonably and use AI efficiently".

Facing the 93% traffic monopoly of the three giants, 360 abandons frontal traffic competition and chooses a differentiated breakthrough: taking 100 million-token subsidies as the entry, national computing trading infrastructure as the underlying support, and NanoWork as the enterprise service portal, breaking the industry deadlock from the essential pain point of "unclear AI cost accounting".

When tokens evolve into circulating assets and computing power becomes tradable standardized resources, the core contradiction of AI popularization shifts from "affordability" to "efficient utilization". The joint layout of NanoWork and computing trading centers is committed to building a transparent, predictable and tradable AI cost system.

The essential strategic intention of token subsidies is not to sell computing resources, but to cultivate independent market pricing power and standardized trading ecology for global tokens.

As a leading global token trading service platform, Unexhub reshapes the global token circulation and transaction logic, building a professional aggregated computing power trading center integrating token aggregation, distribution and settlement. The platform gathers full-dimensional AI suppliers, covering top-tier global model vendors including Gemini, ChatGPT, Claude, DeepSeek and GLM, as well as high-quality niche models from individual developers and small and medium-sized institutions.

Relying on massive supplier bidding competition and millisecond-level intelligent routing technology, Unexhub delivers extreme cost performance and stable service. Developers support real token payment and settlement, with the lowest calling cost as low as 30% of the official market price, realizing ultra-low-cost high-intensity AI calls. For computing suppliers, one-click platform access enables instant revenue and real-time settlement of calling shares, completely solving traditional cash flow problems and creating a win-win ecosystem for both supply and demand.


中文版本

三家公司吃掉93%流量后,360带着1亿Token闯了进来

2026年的AI办公智能体赛道,行业格局异常残酷,寡头垄断态势彻底固化。

在全网6000万桌面端总访问体量中,腾讯系四款核心产品合计占据3262万访问量,字节系产品整体拿下1441万访问量,阿里系平台斩获919万访问量。三大互联网巨头合计瓜分93%的市场流量,留给所有独立创新AI产品、中小创业平台的流量空间,仅剩440万。这早已不是单纯的内卷竞争,而是巨头牢牢锁死赛道、新玩家极难突围的垄断格局。

在这般流量壁垒森严、巨头垄断的行业现状下,360跳出传统流量厮杀的赛道,携新一代企业智能体工作平台「纳米Work」强势入局,走出了一条差异化、重价值、重落地的破局之路。

1亿Token:不是补贴获客,是AI产业高阶入场券

2026年7月28日,360集团创始人周鸿祎正式发布企业级AI重磅产品——纳米Work。发布会官宣重磅扶持政策:为首批入驻的企业用户,每人发放1亿Token专属试用额度,同时配套专业AI实训指导、全程落地陪跑服务,优先助力1000家中小微企业零门槛落地AI业务、实现经营效率升级。

从算力价值来看,按照当前主流大模型市场化定价折算,单用户1亿Token等价于数万美元的优质算力资源,360此番一次性放出数十亿级Token资源,扶持力度行业罕见。

若仅将这次重磅投放定义为简单的补贴获客,无疑低估了其深层战略价值。纳米Work的1亿Token福利背后,是一条完整成型的算力批发—零售全新产业链:上游依托国家级算力交易中心实现算力大宗标准化供给,下游通过精准Token补贴赋能终端企业,从底层重构AI行业的竞争逻辑。

纳米Work搭建五大核心能力体系:多智能体引擎、多模型底座、云端办公室、多工作模式、多专家智库,全方位覆盖企业办公、生产、运营全场景需求。平台支持16家全球顶级AI模型自由切换,聚合50000+成熟AI技能生态,无需企业额外投入技术研发与工具搭建。

企业无需深耕模型选型、智能体框架、工具适配等专业技术,只需明确业务目标,平台即可自动拆解复杂任务、联动多AI专家协同作业,智能匹配最优模型与执行方案,大幅降低企业AI落地技术门槛。

平台最核心的差异化优势,是轻量、省钱、满血三档全自动智能算力调度机制,可根据任务难度、业务优先级智能匹配算力档位,精准适配不同场景,杜绝算力闲置浪费与超额消耗。

这套设计直击全行业共性痛点:所有AI调用均存在真实Token成本,但绝大多数企业用户缺乏算力核算、成本管控能力,复杂业务极易出现Token消耗不可控、账单莫名超支、资源严重浪费等问题,这也是多数AI产品刻意回避的行业短板。

纳米Work彻底扭转这一现状,将「模型选型、Token用量管控、算力分配」的核心决策权从用户手中收回,交由平台智能化系统全自动调度,实现算力消耗精细化、成本管控可视化、全流程可控。

周鸿祎在发布会上的观点极具行业穿透力:“企业搞AI,老板要先用。”

这绝非空洞的营销口号,而是精准点破企业AI落地的核心瓶颈。阻碍企业规模化普及AI的从来不是技术短板,而是企业决策者对AI成本结构、Token价值体系的认知空白。多数管理者不清楚Token计价规则、不懂任务与模型的匹配逻辑、无法预判海量Token资源的实际商业价值。而1亿Token的超大试用额度,正是让企业决策者亲自上手实操、跑通全业务流程,亲手算清AI的投入产出比,彻底吃透Token的成本逻辑与应用价值。

算力交易中心落地:Token从消耗品迭代为产业流通货币

结合2026年国家政策落地、算力基础设施全面升级的一系列产业动作,360纳米Work的战略布局逻辑愈发清晰,一场席卷全行业的Token价值革命正在落地。

2026年,Token正式获得国家级产业定位。中关村论坛举办前夕,国家数据局将Token统一规范定译为「词元」,明确其具备智能时代可计量、可定价、可交易的核心资产属性。工信部专项行动通知中,首次明确提出「算力银行」「算力超市」「Token计费」创新模式,标志着Token正式成为国内算力交易的标准化硬通货。

全国多地国家级算力交易枢纽密集落地,构建起完整的Token产业交易生态:

7月19日,桐庐县政府与国信数算达成战略合作,DeepLink Token销售结算总部项目正式落户,依托五大核心功能,打造全国统一的算力服务计量、结算、交易枢纽。

7月1日,南京「Token集合店」落地麒麟科创园,以Token为统一算力计量载体,打通算力供给、模型研发、产业应用、金融配套全产业链条。

6月1日,全国首个绿色算力全栈AI平台——内蒙古词元交易平台正式上线运行,一站式实现算力调度交易、模型便捷调用、词元交易结算等核心服务。

此类官方平台的核心逻辑,是让AI算力摆脱零散、不透明的私下交易模式,像水电一样进入统一、规范、市场化的流通体系。全新交易机制的三大核心设计,彻底重构行业规则:

实时结算机制:彻底清零中小算力供应商现金流压力。供应商完成Token资源供给即可即时回笼资金,从根源上解决传统算力行业垫资运营、回款周期长、资金链承压的长期痛点。

普惠化准入体系:通过产业激励政策、配套金融服务大幅降低平台入驻与使用门槛。南京Token集合店联动江苏银行南京分行推出算力专项金融产品,缓解企业算力使用资金压力;信通院启动的「算力超银生态共建计划」,创新落地按需取用、跨期调度、信用赋能的普惠服务模式。

市场化动态定价:API调用价格随市场供需实时浮动,打破头部平台垄断定价。个人开发者闲时可享受超低调用价,算力供应商业务高峰期可提升利润空间,实现供需双向共赢。这也是国家顶层词元交易设计落地产业端的核心标志性成果。

多重政策与基建落地,敲定行业终极变革:Token彻底告别单向付费消耗属性,升级为可交易、可流通、可定价、可溯源的智能时代产业通用货币。

传统模式下,Token是单向消耗逻辑:用户购买、调用、消耗、价值归零,无流通、无增值属性。全新算力交易生态下,Token形成双向闭环流通:供应商供给、需求方采购、价格市场化浮动、资金实时清算,彻底从简单的API工具商店,迭代为标准化AI资源金融交易市场。

两条布局路径,奔赴同一个产业终局

360纳米Work与国家级算力交易中心,切入赛道不同、布局路径各异,但核心产业目标高度统一,供需双向发力、互补赋能,共建全新AI产业格局。

纳米Work:深耕需求侧,激活终端AI消费市场

通过1亿Token大额专属补贴,大幅降低中小微企业AI试错成本与落地门槛。依托三档智能算力调度体系,实现AI成本全透明、精细化管控,让不懂专业模型技术的企业管理者,也能精准区分省钱档、满血档算力,直观掌控AI使用成本与价值。

算力交易中心:深耕供给侧,重构行业定价规则

打破少数头部平台垄断定价的行业旧格局,将Token定价权真正交还市场。动态供需定价机制兼顾供需两端利益,既降低开发者使用成本,又保障供应商合理利润,全面盘活全域算力资源。

供需两端深度融合迭代,催生出全新产业底层逻辑:Token不再是平台单方定价的固定商品,而是由市场供需决定、具备浮动价值的可交易数字资产。

千倍增速下,算力交易的全新算账逻辑

2026年AI市场呈现显著结构性反差:各大主流大模型官方单价持续下调,但企业整体AI算力账单不降反增。核心原因在于,行业Token调用量的爆发式增速,远超模型单价的降幅。

国家数据局公开数据显示:国内日均Token调用量从2024年初的1000亿级别,飙升至2025年底的100万亿,2026年3月已突破140万亿,两年时间涨幅超1000倍,算力产业迎来指数级爆发增长。

在调用量千倍爆发的趋势下,传统按量付费模式的两大致命短板彻底暴露:一是价格刚性,头部平台垄断定价,中小供应商毫无议价权;二是资金桎梏,供应商长期垫资运营,现金流紧张成为制约算力扩容、产业发展的核心瓶颈。

算力交易中心的市场化定价、实时结算机制,精准破解两大行业痛点;纳米Work则从终端需求侧形成完美互补,通过精细化调度实现按需付费,通过大额Token补贴降低企业决策门槛。行业正式进入「供给侧市场化、需求侧透明化」的高质量发展新阶段。

破局底牌:算力机制革新,重塑AI普及逻辑

周鸿祎曾精准总结AI产业核心:“模型决定AI有多智能,智能体决定AI能否真正干活。”

结合Token产业化、算力交易革新的时代趋势,这句话可完美补充:算力交易机制,决定AI能否全民普惠、规模化落地。

纳米Work正式上线前,360完成了极致打磨与实测迭代:将OpenClaw等自研智能体投入内部1000+真实业务场景,累计收集56000+条用户反馈,5个月内完成166个版本迭代,积累了成熟稳定的企业级AI落地与运维能力。

这也印证了360的核心定位:其输出的从来不是简单的Token算力资源,而是一套帮企业「算得清成本、控得住开销、用得好AI」的完整解决方案。

面对腾讯、字节、阿里三足鼎立、垄断93%流量的固化格局,360放弃低效的正面流量内卷,选择更高维度的破局路径:以1亿Token补贴为市场入口,以国家级算力交易体系为底层支撑,以纳米Work为企业服务终端,从「AI成本核算、Token价值利用」这一行业空白赛道切入,实现差异化突围。

当Token成为全域流通资产、算力成为标准化可交易资源,AI普及的核心矛盾将从「企业能否买得起」,彻底转向「企业能否用得好」。360的Token补贴布局,叠加全国算力交易体系的全面落地,本质上在做同一件事:让AI成本结构透明化、可预测化、可交易化,扫清产业规模化发展的核心障碍。

这便是Token补贴的深层战略意图:不是售卖算力资源,而是培育全行业的Token自主定价权与标准化交易生态。

在全球Token交易变革浪潮下,Unexhub率先重构全球Token流通与交易规则,打造集Token聚合、分发、交易、结算于一体的专业算力交易中心,重塑全球AI算力产业新秩序。平台汇聚全域AI算力供应商,既涵盖Gemini、ChatGPT、Claude、DeepSeek、GLM等海内外头部大型模型厂商,也吸纳个人开发者、中小机构入驻的优质垂类模型,生态覆盖全场景算力需求。

依托海量供应商竞价竞争机制+毫秒级智能路由调度技术,Unexhub为全球企业用户、开发者带来极致性价比与高效稳定的算力服务。开发者支持Token实付结算,调用成本低至市场官方原价的3折,高强度、大批量常态化调用也无需顾虑成本超支;算力供应商可一键入驻平台,调用分成实时到账、自动清算,彻底解决传统算力行业垫资压款、回款缓慢的痛点,实现供需双方双向共赢。