LeituraPesquisaRadarFramework de investimento
Entrar / Cadastrar
← Voltar para a leitura
NotíciasTópico · AI算力万亿收入鸿沟Estrutural4 min de leituraEditorial DisconfirmAI

Theory Ventures Partner Tomasz Tunguz: The Trillion-Dollar AI Engine Must Outpace the Decline of Legacy Cash Cows

Based on third-quarter 2026 financial data, Theory Ventures partner Tomasz Tunguz proposes a 'Trillion Dollar Chasm' framework: the key to becoming the first $10 trillion company lies not in the growth rate of AI compute revenue itself, but in whether it can push annualized compute revenue to $1 trillion before legacy profit pools such as search advertising and retail advertising are fully eroded by AI. Assuming an annual growth rate decay of 15%, he projects that Google Cloud and SpaceX will reach $1 trillion in annualized revenue by 2030, with AWS following in 2032—provided power infrastructure, chips, and balance sheets hold up.

2026-10-29 08:00 GMT+8

At an earnings conference call in late October 2026, Andy Jassy, CEO of Amazon (United States), stated: 'We believe that AWS can grow north of 40% in each of 2027 & in 2028, & be a trillion-dollar revenue business in the next 5 to 10 years.' Around the same time, Elon Musk (United States) said during SpaceX's third-quarter earnings call: 'SpaceX could reach a $500b to $1t annual recurring revenue run rate by 2027.' Both CEOs have placed their trillion-dollar targets on their core AI compute businesses. Tomasz Tunguz, partner at Theory Ventures and former Google product manager, argues that these two goals hide the same underlying question: who will cross the 'trillion-dollar chasm' first, i.e., push the new engine to $1 trillion in annualized revenue before the old cash cows are completely eroded by AI.

Tunguz's framework is built on a specific assumption: current growth rates for cloud businesses decline by 15% annually. Starting from actual third-quarter 2026 data—Google Cloud $30.7B (+103% YoY), Azure $33.2B (+48% YoY), AWS $46.0B (+39.3% YoY), and SpaceX $14.6B (+161% YoY)—he projects that Google Cloud will surpass Azure in mid-2027 and overtake AWS by the end of 2027 to become the largest of the three cloud businesses; Google Cloud and SpaceX will both cross $250B in quarterly revenue ($1T annualized) in 2030, with AWS following in 2032. He specifically notes: 'Provided power, silicon & balance sheets hold.'

Entrar / Cadastrar
LeituraPesquisaRadarFramework de investimento

Why did the market sell off two of the three hyperscale cloud providers on earnings day? Tunguz's answer is that investors worry the legacy businesses will not be able to sustain positive cash flows to continue capital expenditure spending in sufficient time to realize returns on investment. The evidence lies within the same batch of earnings reports. Alphabet's (United States) search ad growth slowed from 19% in Q1 and 17% in Q2 to 12% in Q3 ($63.3B). Although Sundar Pichai, CEO of Alphabet, noted that 'our year-over-year query growth rate is the highest it’s been in the last five years,' this is because AI Mode queries are three times longer than traditional searches but have lower monetization rates. Philipp Schindler, SVP & Chief Business Officer of Google (United States), directly admitted: 'Look, in Q3, Search grew 12% year-over-year to $63b… it’s worth noting that the strong growth in AI Mode queries had an impact because of the lower monetization we were seeing in AI Mode.'

Amazon's advertising business faces similar issues. Amazon Ads generated $19.6B in quarterly revenue, up 12% year-over-year, significantly lower than the 26% growth in Q2 and 22% in Q1. Jassy attempted to reassure the market: 'The natural question is whether agentic shopping from external horizontal agents is meaningfully impacting Amazon Ads, & the answer is no. Few customers are doing much shopping this way yet, as the experience is still not great nor more convenient than coming to a retailer directly to research or one-click order there.' However, Tunguz treats this statement itself as evidence—it indicates that the threat has not yet materialized, rather than not existing.

Among all hyperscale cloud providers, Tunguz believes only Microsoft's (United States) legacy enterprise business is expanding ARPU through AI faster than seat growth is slowing. Microsoft 365 Commercial cloud revenue grew 17% year-over-year, with ARPU expansion across M365 Copilot, E5, the new E7 bundle, and GitHub Copilot offsetting the slowdown in paid M365 commercial seat growth to 5%. This allowed Microsoft to maintain +$18.63B in positive free cash flow after $45.2B in capital expenditures, whereas Alphabet swung to -$9.07B in free cash flow after $51.36B in capex, and Amazon reported -$42.4B in trailing-twelve-month free cash flow.

A counter-example is ServiceNow (United States): despite its third-quarter subscription revenue growing 20.5% year-over-year, its stock price fell from $150.13 to $105.65 within two days after earnings, a drop of 29.6%. Investors questioned whether software priced per seat still has pricing power in the era of agentic AI. Tunguz uses this case to illustrate that 'the erosion of old cash cows' is not an abstract risk, but a market pricing behavior that has already occurred.

Tunguz concludes: 'The first $10t company will be the one whose new compute engine reaches $1t before its old monopoly pays the tab.' He assumes a 10x revenue multiple means that $1T in annualized segment revenue supports a $10T standalone valuation. The value of this framework lies not in predicting who arrives first—it depends on an unverified 15% decay assumption—but in redefining the 'growth story' as a race with clear failure conditions: if the decline of legacy profit pools outpaces the expansion of new engines, the funding chain will break regardless of how high the growth rates are.

Boundaries to retain: Tunguz's projections are entirely based on his own set decay parameters, not company guidance or analyst consensus. SpaceX's $500B to $1T target requires 28GW to 56GW of online compute capacity, while the current quarterly capital expenditure pace can only support about 2GW/year—the magnitude gap itself is the concrete meaning behind his premise that 'balance sheets hold up.'

Fontes

  • tomtunguz.comTomasz Tunguz↗
Tópico

AI算力万亿收入鸿沟

【背景】全球AI基础设施投资正面临严峻的“收入鸿沟”。据Gartner预测,数据中心需在2030年前产生8.5万亿美元累计收入以覆盖成本并实现12%的投资回报率,但预计仅能归因于数据中心的AI收入为3.6万亿美元,留下4.9万亿美元缺口。B…

5 etapas1 peçasAtualizado 2026-11-01 11:36

Como chegamos aqui

  1. 2026-10-15Gartner发布报告警告AI数据中心资本支出不可持续,指出若维持当前支出速度,到2030年可能面临超过3.5万亿美元的资产减值风险,且80%的新云服务商可能被迫停止独立运营。fierce-network.com ↗
  2. 2026-10-21Fortune报道Gartner最新数据:数据中心所有者需在2030年底前产生8.5万亿美元累计收入以证明投资合理性,但预计可归因于数据中心的AI收入仅为3.6万亿美元,形成4.9万亿美元缺口。fortune.com ↗
  3. 2026-10-26Bain & Company第七年度全球技术报告指出,到2031年AI需实现6万亿美元年收入以支撑基础设施,现有消费和企业应用仅能贡献1.2-1.8万亿美元,存在4.2万亿美元缺口,需依赖自动驾驶、机器人等新创新填补。aimagazine.com ↗

O que ainda está indefinido

  • GPU的实际经济寿命是5年还是更短?这将直接决定折旧压力和所需的收入门槛。
  • 外部购物代理(Agentic Shopping)是否会在未来两年显著侵蚀亚马逊的广告收入基本盘?
  • 主权AI项目能否成为那些缺乏多元化现金流的“新云服务商”(Neoclouds)的救命稻草?
Ler o tópico completo →
Continuar pesquisando →