Web3 平台应该收多少服务费?

深潮Опубліковано о 2025-09-08Востаннє оновлено о 2025-09-09

设计合理的费用结构并非与去中心化相悖,而是构建可运转的去中心化市场的核心要素。

撰文:Gérard Cachon、Tolga Dizdarer、Gerry Tsoukalas

编译:Luffy,Foresight News

Web3 旨在减少对中间机构的依赖,从而降低服务费用,并让用户对自身数据及资产拥有更强的控制权。例如,Gensyn(去中心化 AI 算力平台)提供的人工智能算力服务价格仅为亚马逊云服务(AWS)的一小部分;Drife(去中心化出行平台)则承诺帮助司机摆脱优步高达 30% 的佣金剥削。

然而,尽管为用户降低成本的理念颇具吸引力,制定合理的费用与定价标准却需要平台在多方利益间找到平衡。最成功的去中心化市场绝非完全放弃费用,而是会将 「去中心化定价」 与经过深思熟虑、能创造附加价值的费用结构相结合,以此实现供需平衡。

基于我们的研究,本文将阐述以下内容:定价控制权与费用结构在平台经济及治理中扮演的角色;为何 「零费用」 模式无论设计者初衷多么良好,最终都注定失败;以及区块链平台应如何制定定价策略。我们提出了一种基于交易量的 「仿射定价」新模式,该机制能够解决私有信息与市场协同之间的矛盾。

定价与费用为何重要

数字平台的兴衰取决于其对两大核心杠杆的管理能力:定价控制与费用结构(即平台向使用其服务的买卖双方收取费用的多少)。这两者不仅是创收工具,更是塑造用户行为、决定市场结果的市场设计工具。

定价控制权决定了 「谁来制定交易价格」。例如,优步通过中心化算法设定车费,以实现供需平衡与定价稳定性的优化;与之相反,爱彼迎赋予房东自主定价权,仅通过算法建议对其进行适度引导。两种模式各有解决重点:中心化定价确保大规模市场中的协同效率;去中心化定价让服务供应商能将私有信息(如成本、服务质量、差异化优势等)纳入定价策略。两种模式并无绝对优劣之分,其有效性取决于具体应用场景。

费用结构的影响不仅限于平台收入,更决定了哪些参与者会进入市场以及市场如何运转。苹果应用商店收取最高 30% 的佣金,这笔费用既用于筛选优质应用供给、为平台基础设施提供资金支持,也可能令应用开发者不满,但通常不会直接影响用户;与之相反,票务平台 Ticketmaster 的高额费用若存在替代选择,会推动艺人与粉丝转向其他渠道。从低费用端来看,Facebook Marketplace 的免费商品上架服务滋生了诈骗问题;多个近乎零费用的 NFT 平台则因涌入大量低质量 NFT,导致用户体验混乱。

规律显而易见:费用过高会导致供给方流失;费用过低则会损害服务 / 商品质量。

许多区块链项目采用零佣金模式,其逻辑是:平台放弃价值提取能力,就能为供给方与用户带来更优结果。但这种观点忽视了 「设计合理的费用」 对市场有效运转的关键作用:费用绝非单纯的抽成工具,更可以是一种协同机制。

信息与协同的权衡

平台设计的核心矛盾在于:如何在 「利用服务供应商的私人信息」 与 「为提升效率而协同市场」 之间找到平衡。我们的研究表明,定价控制与费用结构的互动方式,决定了这一矛盾是被化解还是加剧。

当平台直接制定价格时,虽能更轻松地实现供给侧协同及服务供应商间的竞争协调,但由于无法掌握每个供应商的 私人成本(如运营成本、边际成本等),定价往往对供需双方造成 错配:对部分用户而言价格过高,对部分供应商而言则过低。而平台通常按交易金额收取佣金,这种低效定价最终会导致 利润流失。

如果由服务供应商自主定价,理论上其价格能反映真实成本与服务能力:低成本供应商可通过降价获得竞争优势,从而实现更优的供需匹配与市场效率。但缺乏协同的定价模式可能会在两个方面适得其反。

当产品或服务同质化严重时,易引发低价竞赛。高成本供应商被迫退出市场,导致供给量减少;而此时需求往往处于上升阶段,最终削弱平台满足市场需求的能力。同时,平均价格下降虽可能让消费者受益,却会直接冲击平台基于佣金的收入模式。

当产品或服务需相互搭配才能发挥最大价值时,供应商往往会定价过高。尽管大量供应商会涌入平台,但各自设定的高价会推高市场平均价格,最终将用户驱离。

这并非纯理论推断:2020 年,优步在加利福尼亚州测试了 「路易吉计划」,允许司机自主定价。结果显示,司机设定的车费普遍过高,导致用户转向其他出行平台,该计划仅实施约一年便被终止。

关键结论:上述结果并非偶然,而是 标准佣金合约下的 均衡结果,即便对佣金合约进行优化,仍可能导致此类持续性市场失灵。因此,核心问题并非 「平台应收取多少佣金」,而是 「如何设计费用结构,确保市场对所有参与者都有效」。

如何解决问题

我们的研究发现,一种针对性的费用结构能巧妙解决市场协同问题,同时保留 「定价个性化」 的优势。这种仿射费用模式采用 「两部分收费」 机制,服务供应商需向平台支付:

  • 每笔交易的固定基础费用;

  • 浮动费用:随交易量增加而提高(附加费),或随交易量增加而降低(折扣费)。

该模式会根据供应商的成本与市场定位,对其产生差异化影响。

在这类市场中,供应商的成本存在显著差异:部分供应商因拥有更先进的技术、可接入可再生能源或具备高效散热系统,成本天然较低;另一些供应商虽成本较高,但能提供高可靠性等溢价服务。

在传统佣金模式下,若市场竞争过度,低成本的 GPU 供应商会设定极具攻击性的低价,占据过大市场份额,进而引发前文所述的市场扭曲:部分供给方退出导致成交量受限,同时市场平均价格被拉低。

针对此场景,最优策略是 「交易量附加费」:供应商服务的客户越多,每笔交易需支付的费用就越高。

这种机制能对激进的低成本供应商形成 「自然约束」,防止其以不可持续的低价占据过多市场份额,从而维护市场平衡。

当市场竞争程度适中或不足时,最优策略转为 「交易量折扣费」:供应商服务的客户越多,每笔交易需支付的费用就越低。这一机制会激励供应商通过降价扩大交易量,在避免价格低于可持续水平的前提下,有效提升市场竞争性。

例如,在去中心化社交平台中,可对 「用户互动量更高的创作者」 收取更低费用,鼓励其为付费内容设定更具竞争力的价格,同时吸引更多用户参与。

仿射费用机制的精妙之处在于,它无需平台掌握每个供应商的具体成本,费用结构会形成正向激励,引导供应商根据自身私有成本信息实现自我调节。低成本供应商仍可通过低于高成本竞争者的价格获得优势,但费用结构会防止其以损害整个生态健康的方式 垄断市场。

我们通过数学模拟验证:经过合理校准的 「基于交易量的费用结构」,能使平台实现超过 99% 的理论最优市场效率。在理论框架中,其表现远超 「中心化定价」 与 「零佣金」 模式。最终形成的市场将具备以下特征:

  • 低成本供应商保留竞争优势,但不会占据过度市场份额;

  • 高成本供应商可通过聚焦 「差异化服务的细分市场」 持续参与;

  • 整体市场达到更平衡的均衡状态,价格差异合理;

  • 平台在提升市场功能的同时,实现可持续的收入。

此外,分析表明:最优费用结构取决于 「可观测的市场特征」,而非每个供应商的 「私人成本信息」。在设计合约时,平台可将 「价格」 与 「交易量」 等可观测信号作为 「隐性成本」 的代理指标,既让供应商保留基于私有信息的定价权,又能解决完全去中心化系统中固有的协同失灵问题。

区块链项目的未来发展路径

许多区块链项目因采用传统佣金模式或零费用模式,既损害了自身的财务可持续性,也降低了市场效率。

我们的研究证实,设计合理的费用结构并非与去中心化相悖,而是构建可运转的去中心化市场的核心要素。

Пов'язані матеріали

Kelp DAO's $400 Million Bad Debt Was Covered, But at a $12 Billion Cost to Aave

On May 26th, Kelp DAO successfully transferred its final batch of rsETH, completing the 37-day process of fully backing rsETH 1:1 after a security incident. However, the resolution came at a significant cost to Aave. The protocol's TVL plummeted by over $12 billion in the following month. Furthermore, a separate legal battle over 30,766 frozen ETH continues in court, posing ongoing reputational risk. The recovery was enabled by an unprecedented, one-time coalition dubbed "DeFi United," involving major contributions from Aave's founder, treasury, Consensys, Mantle, and others. Despite this, the event triggered a major outflow of funds, with whales like Justin Sun moving capital to competitors like Spark. Aave's path to regaining its position relies heavily on the successful execution of its multi-pronged strategy. Its new V4 protocol, designed for open, heterogeneous asset markets, faces delays due to internal governance disputes. Meanwhile, the V3 version remains the core revenue generator, and the permissioned Horizon fork is targeting institutional RWA (Real-World Assets) growth—a segment less impacted by the rsETH incident but dependent on traditional finance adoption timelines. The key takeaway is that while the immediate bad debt was covered, Aave paid a steep price in lost trust and capital. Recovering market share depends on accelerating V4's rollout and advancing its institutional RWA offerings, both of which face external and internal hurdles. The "DeFi United" safety net is unlikely to be replicable for future crises.

marsbit33 хв тому

Kelp DAO's $400 Million Bad Debt Was Covered, But at a $12 Billion Cost to Aave

marsbit33 хв тому

TechFlow Intelligence Report: Xiaomi Announces 200 Billion HKD Stock Buyback Plan, Spot Gold Falls Nearly 1%

TechFlow Report: Xiaomi announced a HK$200 billion stock buyback plan, while spot gold fell nearly 1%. A wider range of tech headlines includes Google unveiling its powerful video editing model Gemini Omni and the original "Attention is All You Need" authors advocating for a move beyond Transformer architecture. In other AI news, IBM reported its first successful use of a quantum computer to train an AI model, and Qwen3.5 released uncensored local model versions. The crypto/Web3 sector saw discussions on opaque stablecoin products and DEX fee changes. Major tech companies are under scrutiny: Uber's COO publicly questioned the ROI of AI investments, Motorola was accused of hijacking Amazon app links for affiliate codes, and Google faced criticism for using web data to fuel its AI. U.S. markets are focused on high S&P 500 valuations (31.8x P/E) and an intense concentration of capital in semiconductor stocks, with warnings about the sustainability of the AI data center boom. Geopolitical tensions, featuring simultaneous U.S. airstrikes on Iran and peace talks, caused significant oil price volatility. Other notable developments include Ferrari's first pure EV priced at 4.35 million yuan and Boston Dynamics' Atlas robot learning soccer from videos. The underlying theme suggests the AI narrative is shifting from boundless potential to requiring tangible results, while traditional geopolitical risks remain a powerful force in markets.

marsbit36 хв тому

TechFlow Intelligence Report: Xiaomi Announces 200 Billion HKD Stock Buyback Plan, Spot Gold Falls Nearly 1%

marsbit36 хв тому

Coin & Stock Barometer: Bitcoin Miner MARA Holdings Spends Over $860,000 on Bulletproof Vehicle Services for Executives; Bitmine Included in Preliminary List for FTSE Russell 1000 Index (May 19)

Crypto Market Wrap & Key Corporate Updates (May 19) The crypto market saw a decline followed by a minor rebound, while U.S. crypto-related stocks fell broadly. In corporate news: **MARA Holdings**, a Bitcoin miner, disclosed spending over $869,000 on vehicle ballistic armor services for its CEO and CFO under its security program. The board cited higher risks associated with the company's public disclosure of holding substantial Bitcoin assets. According to BitcoinTreasuries.NET, Elon Musk's **SpaceX and Tesla** collectively hold 30,221 BTC ($2.3B), which would rank them as the fifth-largest public company holder if combined. **DDC Enterprise Limited** increased its Bitcoin holdings by 200 BTC, bringing its total to 2,583 BTC. The firm stated it plans to continue accumulating BTC based on liquidity, not short-term price movements. Bitcoin treasury company **Nakamoto** announced a 1-for-40 reverse stock split to regain compliance with Nasdaq's minimum bid price requirement. The company reported a Q1 2026 net loss of $238.8M, partly due to a $102.5M unrealized loss on its Bitcoin holdings. **Tether** acquired SoftBank's stake in **Twenty One Capital (XXI)**, increasing its control. Tether's CEO expressed strengthened confidence in XXI's long-term Bitcoin strategy. Fundstrat's **Tom Lee** stated that **Bitmine (BMNR)** has been included in the preliminary list for the FTSE Russell 1000 Index. Concurrently, two new wallets suspected to be linked to Bitmine withdrew 60,000 ETH ($126M) from Bitgo and Kraken. Solana treasury company **Solmate Infrastructure** announced a registered direct offering of shares to raise approximately $11.4 million. **AI Financial**, a WLFI treasury company, reported a Q1 2026 net loss of $271.5M and raised substantial doubt about its ability to continue as a going concern, partly due to unrealized losses on its WLFI token holdings. **SUI Group** disclosed it holds over 108.7 million SUI tokens (~$115M), with its market cap to net asset value ratio at 0.91x. *Disclaimer: This summary is for informational purposes only and does not constitute investment advice.*

marsbit52 хв тому

Coin & Stock Barometer: Bitcoin Miner MARA Holdings Spends Over $860,000 on Bulletproof Vehicle Services for Executives; Bitmine Included in Preliminary List for FTSE Russell 1000 Index (May 19)

marsbit52 хв тому

China's AI Fronts: From Yan'an to Midway

This article analyzes the competitive landscape of China's AI industry through a dual-front war analogy: the "Eastern Front" of business model competition and the "Western Front" of global strategic positioning. **The Eastern Front: The Scramble for Supply Lines and Monetization** The "Eastern Front" examines the contrasting strategies of three Chinese tech giants—Tencent, Alibaba, and ByteDance—in the face of AI's high marginal costs. Tencent integrates AI as a catalyst within its existing ecosystems (advertising, gaming, cloud) for monetization, prioritizing high-value scenarios over user growth. Alibaba bets on a full-stack, self-developed approach from chips to applications, aiming to control costs and ecosystem, though this requires immense patience and resources. ByteDance, with Doubao as its flagship, pursues a traditional traffic-driven, "super app" strategy but faces severe monetization challenges as its massive user base incurs unsustainable operational costs. The central challenge for all is building a reliable "supply line" (sustainable funding/profit) and achieving efficient monetization, moving beyond being mere "token factories." **The Western Front: "Preserving Land" vs. "Preserving People"** The "Western Front" frames a global strategic divergence. The U.S. model ("preserving land") focuses on closed-source, high-premium models (e.g., Anthropic) targeting lucrative enterprise markets. China's strategy ("preserving people") leverages open-source models (e.g., Alibaba's Qwen, DeepSeek) and extremely low pricing to attract global developers and capture long-tail markets, akin to a "surround the cities from the countryside" approach. The goal is to make Chinese models the default infrastructure, locking in future ecosystem value. However, the critical test is whether this open-source ecosystem can achieve a commercial闭环, converting developer adoption into tangible revenue (e.g., via cloud services), and bridging the monetization gap with Western models that charge for value, not just tokens. **Conclusion: The Long March from Factory to Brand** The article concludes that China's AI industry possesses technology, users, and scenarios but must integrate them to create and capture value. Its ultimate success depends on navigating both fronts: companies must establish sustainable monetization on the Eastern Front, while the industry's Western strategy must evolve from simply "preserving people" (developer adoption) to truly "preserving both people and land" — transforming open-source ecosystem dominance into commercial success and premium brand value. This journey from being a "token factory" to a "value highland" will require strategic patience and the ability to outlast competitors in a prolonged contest.

marsbit1 год тому

China's AI Fronts: From Yan'an to Midway

marsbit1 год тому

Торгівля

Спот
Ф'ючерси
活动图片