美国AI大牛市会崩吗?

marsbitPubblicato 2026-05-29Pubblicato ultima volta 2026-05-29

Introduzione

软银已向OpenAI投入巨额资金,并计划继续增持,驱动孙正义的是OpenAI估值飙升带来的巨大账面浮盈。目前,资本正不计成本地涌入AI领域,OpenAI和Anthropic的估值已分别达到8520亿美元和约9000亿美元。市场假设AI将像火箭般直线上升,但这引发了对定价合理性的质疑。 文章回顾了1990年代末的互联网泡沫,将当时雅虎作为门户网站的估值逻辑与当前大模型公司类比。两者都被视为不可逾越的“入口”或“收费站”,但历史表明,技术普及后,护城河可能被更精准的解决方案侵蚀。雅虎股价最终暴跌并被低价收购。 从财务角度看,支撑万亿美元市值需要极高的盈利水平,但OpenAI和Anthropic的年度经常性收入(ARR)与期望存在差距。此外,激烈的价格战正在将“智能”变为廉价商品。大模型的高估值本质上是提前透支了下游应用层的未来利润,而目前AI尚未催生出能创造巨额营收的“超级应用”。如果下游企业因回报不足而暂停投入,上游估值可能瞬间坍塌。 文章推演了两种可能的终局:一是像雅虎一样估值崩盘,大模型公司回归普通基础设施的利润率;二是成功重构商业闭环,例如通过价值抽成或实现AGI来支撑估值。但资本市场缺乏耐心,可能在黎明前踩踏出局。 文章最后强调,必须将宏观趋势的确定性与个体公司的命运区分开。行业前景光明不代表当前领跑者一定能成功。早期互联网泡沫破灭曾造成严重的经济创伤,当前AI估值若坍塌,其引发的系统性风险可能更大。技术革命往往由泡沫堆叠而成,AI最终会像水电煤一样成为基础设施,但过程中的狂热需要警惕。

346亿美元。这是软银已砸进OpenAI的真金白银。为了凑钱,孙正义清空了英伟达、德意志电信、阿里巴巴和T-Mobile的股权。

但他还嫌不够,计划今年再追加300亿美元,把持股从11%推到13%,为此甚至不惜举债。

驱动这位69岁投资家大干一场的力量很纯粹——随着OpenAI估值在2月冲上8520亿美元,软银的账面浮盈已超450亿美元。

巨大、快速的纸面财富效应,让资本不计成本地抢筹AI。在OpenAI最新融资三个月后,据CNBC、彭博社等媒体报道,Anthropic正在与投资人谈判新一轮融资,目标估值约9000亿美元。一年前,它的估值还只有615亿美元。

从硅谷到华尔街,都在假设AI的未来会像火箭升空般笔直兑现,但这究竟是对一场伟大科技革命的合理定价,还是对AI概念的一场万亿级豪赌?

雅虎时刻

26年前的硅谷,有一幕似曾相识。

那一年,互联网如日中天,有一家公司上市不过四年,市值却冲到了1280亿美元,高过了伯克希尔。

它不铺光缆,也不造路由器,而是头号门户网站雅虎。

雅虎的商业模式很清晰:所有人上网,都要先上雅虎。互联网流量永续增长,雅虎作为入口,将永远向用户和广告商收税。

这也是华尔街为雅虎给出高估值的底层逻辑,它不负众望,股价从上市时13美元的发行价,一路上涨到近500美元,与众多带着“ .com”的公司一起,把纳斯达克指数合力推到了5132点。

华尔街看对了大趋势,互联网确实改变了世界,流量确实爆炸,但也犯了一个致命的错误:

他们认为门户网站的护城河不可逾越。

市场忽略了一件事:随着互联网基础设施的普及,用户总会找到更精准的入口,雅虎那种大而全的目录模式,一旦遇到精准的广告推荐,原本的广告分发模式就会从根本上动摇。

2000年3月11日,雅虎的股价在顶部拐了个弯,随后一路下跌;一年后,纳斯达克跌破2000点;2017年,雅虎被Verizon收购,价格只有48亿美元。

今天,卖出Token的OpenAI与Anthropic,正像是AI时代的雅虎。

这两家万亿级巨头,向企业提供API接口,向开发者出售Token。资本给出万亿估值的底层逻辑与当年如出一辙:所有AI应用的爆发,都必须向大模型公司购买算力与智能;大模型就是未来AI时代的收费站,且这个收费站永远存在。

▲图源:新智元

这和当年的忽略如出一辙。大模型正在以惊人的速度,将人工智能变为一种廉价的、无处不在的水电煤。

就像当年的雅虎无法阻止互联网走向纵深一样,当人工智能无处不在,当众多大模型同时竞速,单纯的“模型入口”是没有人工智能定价权的。

万亿算术

历史案例,只是参照的镜子;财务算术,才是真正刺破泡沫的针。

▲图源:新智元

如果我们把OpenAI和Anthropic看作未来的科技巨头,为它们匹配30-40倍这样的高市盈率,那么要支撑万亿美元市值,这两家公司每年必须净赚250到300亿美元,对应500到800亿美元的营收。

但现实比期望差得远。

衡量这些订阅制AI公司的经营状况有个关键指标,叫作ARR(年度经常性收入。2026年4月,OpenAI的ARR是250亿美元,Anthropic是300亿美元。

如果有一家公司的ARR达到300亿美元,并不意味着它在过去一年收入300亿美元,只是它过去一个月里,靠用户订阅费用达到了25亿美元收入,然后乘以12。

以及,ARR很大程度上是按照当前API调用的消耗速度推算出来的,它包含了大量短期的、试探性的开发需求,甚至包含了买算力送Token的促销水分。这种收入的粘性,远远不及SaaS时代按年订阅的软件服务。

让盈利希望变得更渺茫的,还有大模型公司的促销。为了竞争、为了抢占开发者生态,OpenAI和Anthropic在过去一年里大幅压低Token价格。

当年,雅虎为了保住流量入口的地位,不得不提供海量免费内容,却发现流量根本无法转化为对等利润。今天,大模型公司为了保住开发者生态,主动把智能变成廉价商品,却又享受垄断级的高估值溢价。

在这相似的情形背后,藏着一条残酷的产业链利润定律:

整个AI产业链的利润总额,最终是由终端应用层的商业化决定的。大模型的高估值,本质上是对下游应用层未来利润的提前透支。

放眼如今的应用层,除了编程与部分文案辅助,AI尚未催生出产生巨量营收的“超级应用”,大多数企业端对AI的付费,依然停留在降本增效的试探期——以AI替代一部分初级文案或客服,而不是创造增量营收的爆发期——通过AI创造出全新的、价值十亿百亿美元量级的市场。

这是一个埋藏着很大风险的环节,如果下游应用企业发现,购买AI算力无法带来超额回报,他们会随时按下暂停键,这可以来自一次模型迭代的性能瓶颈,可以来自一阵舆论思潮;也可以来自一批老板出于好恶的决策,可以来自新一轮经济周期的下行或震荡,让下游应用公司在一个阶段内对AI购置预算的集体暂停......

无论是哪一种,一旦下游断流,上游万亿市值的地基就会瞬间塌方。

终局推演

在资本狂热的顶点,任何风吹草动都可能引发雪崩。

如当年的雅虎一样,当增长的故事讲不下去时,万亿估值就会面临均值回归的残酷绞杀。这意味着,大模型公司必然要走上一条终局之路。

第一条路,是雅虎式崩盘。

因为应用层迟迟无法形成商业闭环,Token价格战彻底失控,智能彻底沦为廉价商品。失去耐心的资本开始寻求变现,万亿估值开启腰斩甚至脚踝斩。

失去资本光环的大模型公司,因此被迫收缩战线,裁员止损,回归普通的软件基础设施利润率。

就像今天的门户网站依然存在,雅虎也还在运营,但它再也不是资本市场的宠儿,只能赚取合理的经营利润。大模型企业将沦为和云存储、云计算一样的“正常生意”。

第二条路,是重构商业闭环。

可能有一天,大模型公司真的找到了让企业大量付费的终局模式。也许是取代整个行业的SaaS软件,按创造的价值抽成;也许是AGI真的降临,接管了全球的数字劳动力......

这条路如果走通,万亿市值就有了坚实的支撑。这需要时间,也充满了不确定性。但以VC资金性质为核心的资本市场,最缺乏的就是耐心,在黎明到来之前,他们可能已经开始在黑夜中踩踏出局。

然而,无论走哪条路,我们都必须警惕一个极易被狂热掩盖的常识:宏观趋势的确定性,从来不等于个体命运的确定性。

绝大多数人都相信,AI最终一定会改变世界,就像当年大多数人都相信互联网会改变世界一样。但问题在于,行业的前景是一回事,能在漫长的马拉松中跑到终点并摘得桂冠,是另一回事。

2000年,许多人看对了互联网的方向,却买中了网景、雅虎、Pets.com,最终血本无归。今天的OpenAI与Anthropic,虽然此刻领跑,但它们是否能跨越商业闭环的鸿沟,是否会像当年的门户网站一样被更垂直、更高效的后来者颠覆,仍然是一个巨大的未知数。把行业的终局红利,提前折现给当下的领跑者,正是泡沫最典型的特征。

更值得警惕的是,早期互联网泡沫的破灭,虽然长远来看没有阻挡互联网的步伐,但它却在当时造成了纳斯达克市值蒸发超四分之三,大批企业倒闭连带员工失业,对美国乃至全球经济造成了一段相当长的真实创伤。

当OpenAI、Anthropic这些大模型公司纷纷突破万亿美元估值,当软银不计成本地加杠杆入场,当整个硅谷的资本链条都绑死在“AI必须立刻兑现”的预期上时,这种潜在的系统性风险已到了必须高度警觉的时刻。一旦万亿估值坍塌,其引发的负向财富效应与信用收缩,对宏观经济的反噬力度将远超二十年前。

技术革命,从来都不是一条平滑的上升线,而是由无数泡沫的生灭堆叠而成的。

2000年,雅虎暴跌超80%,无数投资者的财富灰飞烟灭,但互联网并没有死,它以更深入、更高效的方式重塑了世界。今天,如果AI大模型的万亿估值崩塌,也不意味着AI的终结。它只意味着,AI将褪去金融炒作的色彩,真正作为一种通用技术,沉默而深刻地融入千行百业。

前景是光明的,道路是曲折的。人工智能注定会和水电煤一样,成为这个时代不起眼又不可或缺的基础设施,但过程中必然有波澜起伏,当起得太快太猛,就有必要警惕落的危机了。

本文来自微信公众号 “华商韬略”(ID:hstl8888),作者:华商韬略

Domande pertinenti

Q当前美国AI大牛市的繁荣与上世纪90年代末的互联网泡沫有哪些相似之处?

A文章认为两者存在显著相似性,主要体现在市场预期的盲目乐观上。就像当年雅虎被普遍视为不可逾越的互联网入口并享有极高估值一样,当前OpenAI和Anthropic等AI大模型公司也被市场视为未来AI时代不可或缺的“收费站”,从而获得了万亿级别的估值。这种相似性在于,市场都过度相信了先行者的垄断地位和商业模式(门户网站/模型API)的永续性,而忽略了技术扩散、竞争加剧和商业模式本身可能存在的脆弱性。

Q文章认为支撑OpenAI或Anthropic万亿估值需要达到怎样的财务指标?现实情况如何?

A文章指出,若以30-40倍的高市盈率来匹配其万亿市值,公司年净利润需达到250到300亿美元,对应500到800亿美元的营收。然而现实情况是,其关键收入指标ARR(年度经常性收入)虽分别达到250亿和300亿美元,但这部分收入主要来自API订阅和Token消耗,且包含了大量短期、试探性的需求以及促销水分,收入粘性低。同时,为了竞争而进行的Token价格战进一步压缩了盈利空间,使得当前实际盈利能力与支撑高估值所需的水平相距甚远。

Q文章提到的AI产业链的“利润定律”是什么?这对大模型公司的高估值意味着什么?

A文章提出的“利润定律”是:整个AI产业链的利润总额,最终是由终端应用层的商业化成功与否决定的。这意味着,大模型公司的高估值本质上是对下游应用层未来可能创造出的利润的提前透支。如果下游企业无法利用AI创造出足够多的增量价值或超额回报,它们可能会削减AI算力支出,这将直接动摇上游大模型公司高估值的基础,使其面临巨大风险。

Q关于AI大模型公司的未来终局,文章推演了哪两条可能的路径?

A文章推演了两种可能的终局路径:第一条是“雅虎式崩盘”,即由于应用层商业闭环难成、价格战失控,智能沦为廉价商品,资本失去耐心,导致估值泡沫破裂,大模型公司最终回归为提供普通基础设施服务的正常生意。第二条是“重构商业闭环”,即大模型公司找到了让企业大规模付费的终极模式(如取代SaaS按价值抽成,或AGI实现),从而为高估值提供坚实支撑。但文章同时指出,资本市场缺乏耐心,可能在第二条路的曙光到来前就发生踩踏出局。

Q作者对当前AI投资热潮的核心警示是什么?

A作者的核心警示在于两点:首先,宏观趋势的确定性不等于个体命运的确定性。坚信AI将改变世界是正确的,但将整个行业的未来红利提前折现并押注于当前的一两家领跑公司,是典型的泡沫特征。其次,当前以万亿估值和杠杆投资为特征的AI狂热,蕴含着巨大的系统性风险。一旦估值崩塌,其引发的财富负效应和信用收缩对宏观经济的冲击,可能远超二十年前的互联网泡沫破裂,因此必须高度警觉。

Letture associate

Single-Day Plunge of 30%, Arthur Hayes Suddenly Liquidates: Why Did ZEC Get Exploded by Security Issues?

On June 5th, Zcash founder Zooko Wilcox disclosed a critical soundness vulnerability in the project's latest Orchard privacy pool. This flaw, found in the elliptic curve multiplication constraints, could allow an attacker to create unlimited counterfeit ZEC within the shielded pool, with transactions appearing valid. The vulnerability was discovered in late May by security researcher Taylor Hornby, who utilized Anthropic's new Opus 4.8 AI model for a targeted audit. The Zcash ecosystem had already performed an emergency network upgrade to patch the issue. However, the detailed disclosure triggered severe market panic, causing ZEC's price to plummet over 30% in a single day. Notably, prominent investor Arthur Hayes announced he had sold his entire ZEC position following the news. The incident starkly challenges the "technological trust" narrative central to privacy coins. Despite years of top-tier cryptographic audits, the bug persisted until uncovered with advanced AI-assisted research. This highlights the growing gap between theoretical perfection and practical implementation in privacy technology. The event serves as a industry-wide warning: in an AI-driven security landscape, the assumption that "undiscovered equals safe" is obsolete. It underscores the urgent need for continuous, proactive security practices combining AI audits, formal verification, and rapid response mechanisms.

foresightnews_api1 h fa

Single-Day Plunge of 30%, Arthur Hayes Suddenly Liquidates: Why Did ZEC Get Exploded by Security Issues?

foresightnews_api1 h fa

Breaking the Curse of DeFi Cascading Liquidations, Vitalik Proposes a New Solution

**Vitalik Buterin Proposes New DeFi Design to Eliminate Forced Liquidations** Ethereum co-founder Vitalik Buterin has published a proposal for a new decentralized finance (DeFi) architecture aimed at removing the automatic liquidation mechanisms prevalent in current lending protocols. The core idea involves creating synthetic assets using options as building blocks, fundamentally avoiding the抵押借贷结构 that triggers forced sell-offs. The proposal responds to a recurring flaw in DeFi: during sharp market downturns, mass自动清算 of under-collateralized positions can exacerbate price declines, creating systemic selling pressure and market instability, as evidenced by recent crypto market volatility. Buterin's model would split an asset like 1 ETH into two option-like derivatives, P and N, pegged to a price index with a set strike price and expiration. At expiry, an oracle determines the settlement price to allocate the underlying ETH between P and N holders. This design eliminates the "cliff" of instant liquidation. Instead, a position's value would gradually drift from its target peg if not actively rebalanced by the user, transferring the rebalancing decision from the protocol to the user or automated tools. A key advantage is the reduced reliance on high-frequency, real-time oracle price feeds, which are vulnerable to manipulation and errors in current systems. The delayed settlement in the options model allows for more robust, fault-tolerant oracle designs. However, significant challenges remain for practical adoption. High transaction costs (slippage) from frequent rebalancing on automated market makers (AMMs) could erode user funds. The model may not be suitable for stablecoins requiring a strict 1:1 dollar peg, as it inherently allows for value drift. Success would depend on developing new liquidity provisioning models and deep markets for these synthetic assets. The proposal represents a fundamental rethinking of DeFi risk management, challenging the industry to explore alternatives to被动集中平仓 rather than merely optimizing existing liquidation processes. It remains a theoretical framework awaiting implementation and testing by development teams.

foresightnews_api1 h fa

Breaking the Curse of DeFi Cascading Liquidations, Vitalik Proposes a New Solution

foresightnews_api1 h fa

Bitcoin's Decline Marks the Transformation of Crypto

Title: The Decline of Bitcoin Marks the Transformation of Crypto While Bitcoin's price recently fell below $70,000, down approximately 45% from its peak, the broader crypto industry is not following it into decline. Instead, crypto is maturing and evolving beyond its dependence on Bitcoin's price movements. Two of Bitcoin's core functions are being usurped. First, AI has captured its role as the primary speculative asset. AI, with its tangible revenue, explosive demand, and massive capital inflows ($700-830 billion in 2024), is siphoning off the speculative "hot money" that once drove Bitcoin. It also contributes to a sustained high-interest-rate environment, further tightening liquidity for assets like Bitcoin. Second, dollar-pegged stablecoins like USDC and USDT have replaced Bitcoin as the crypto market's foundational currency and primary on/off-ramp. Most trading pairs and on-chain transactions are now settled in stablecoins, severing the historical link where all capital inflows had to pass through Bitcoin first. This decoupling allows projects to thrive based on their own fundamentals rather than Bitcoin's price. Examples include Hyperliquid, an on-chain derivatives exchange with annual revenues of $8-13 billion, and prediction market platform Polymarket, valued at $200 billion with $3.65 billion in annual fees. These projects are evaluated on traditional metrics like revenue and user growth. New opportunities are emerging, particularly around privacy. Privacy coins like Zcash (ZEC) are seeing surging demand, while infrastructure like NEAR enables private, cross-chain asset transfers without requiring users to hold a specific token—privacy becomes a universal service layer. In this new paradigm, stablecoins are the universal cash, various project tokens represent equity, and privacy-enabled cross-chain coordination layers (like NEAR) act as the critical infrastructure connecting a fragmented, multi-chain ecosystem. Bitcoin is now just one asset among many. The era where the entire crypto market moved in lockstep with Bitcoin is over. The industry's health should now be judged by project fundamentals—real revenue, active users, and tokenomics that capture value—and the development of the underlying infrastructure enabling a mature, dollar-denominated crypto economy.

foresightnews_api1 h fa

Bitcoin's Decline Marks the Transformation of Crypto

foresightnews_api1 h fa

Lightspark CEO: In Ten Years, Bitcoin Will Be as Invisible as TCP/IP, Yet Power Trillions in Daily Transactions

A decade from now, Bitcoin will function like TCP/IP — invisible yet foundational, supporting trillions in daily transactions globally, according to Lightspark CEO David Marcus. In this future, a coffee shop in Lagos receives instant payment, a manufacturer in São Paulo settles an invoice with a supplier in Ho Chi Minh City, and a freelancer in Bangalore gets paid weekly from an Austin startup — all via Bitcoin's settlement layer, with none of the parties consciously interacting with it. This vision parallels the adoption of open protocols: first driven by necessity where existing systems fail, then scaling rapidly as tools mature and economic benefits become clear. The structural shift begins with wallets. Modern non-custodial wallets, like Spark, allow users to hold dollars, local currency, and Bitcoin in a single address, seamlessly switching between them. This eliminates friction and revolutionizes global custody, moving significant deposits to user-controlled keys not by ideology, but by superior utility. As a result, Bitcoin becomes the default savings layer for billions, as its fixed supply and appreciating value make it a rational choice for savers holding it alongside stablecoins in their everyday wallets. Businesses follow a similar path, from small companies in emerging markets to multinational corporations, holding Bitcoin alongside operational stablecoins. The latest trend is direct Bitcoin transactions for commerce. When both parties hold Bitcoin, transacting in it becomes the simplest option — no conversions, no intermediary currency. This starts in niche areas like high-value B2B settlements but grows as infrastructure makes sending Bitcoin as easy as stablecoins. An accelerating force is AI agents. By 2036, AI agents conducting commerce on behalf of individuals and firms will increasingly choose Bitcoin for settlement. Optimizing for speed, finality, and minimal counterparty risk across jurisdictions, they find Bitcoin's global, neutral, and programmable network ideal for netting and settling obligations. Thus, Bitcoin is becoming the native currency for machine commerce, just as it has become a native savings asset for humans. The global monetary system is being rebuilt from the protocol layer: open infrastructure, default self-custody, Bitcoin settling everything underneath, with stablecoins as the interface. Most users won't think about Bitcoin when they transact — and they won't need to.

foresightnews_api1 h fa

Lightspark CEO: In Ten Years, Bitcoin Will Be as Invisible as TCP/IP, Yet Power Trillions in Daily Transactions

foresightnews_api1 h fa

Trading

Spot
Futures
活动图片