400亿美元。年化。
三年前，OpenAI还只是一个靠捐款维持的非营利研究机构。三年后的今天，它的年化营收已经突破400亿美元——较去年底翻倍，7月单月环比增长超过20%。
这个数字是什么概念？谷歌用了12年才做到年营收400亿，Facebook用了11年，亚马逊用了16年。OpenAI从ChatGPT发布到400亿年化营收，用了不到三年。
更惊人的是增长速度。一个已经400亿体量的公司，单月环比还能增长20%以上——这不是线性增长，这是指数级增长。按这个速度，年底年化营收可能摸到500亿甚至600亿美元。
资本市场已经给出了定价：万亿估值的IPO正在加速推进。如果按这个估值上市，OpenAI将成为人类历史上最大规模的IPO之一，也将是AI行业第一个真正意义上的万亿级公司。
AI的商业化，正在以远超所有人预期的速度兑现。
400亿美元是怎么来的
400亿不是一个单一业务撑起来的。它是一个立体的收入结构——API、订阅、企业、编码、广告——多条腿同时在跑。
第一条腿：API业务。这是OpenAI最基础也是最稳定的收入来源。全球数百万开发者调用GPT系列模型，按token付费。从GPT-4到GPT-5，模型能力不断升级，调用量持续增长。企业级客户越来越多地把核心工作流迁移到OpenAI的API上，这部分收入的粘性非常高。
第二条腿：ChatGPT订阅。ChatGPT Plus、ChatGPT Pro、ChatGPT Enterprise——从个人用户到大型企业，订阅制已经被验证是AI产品最有效的商业模式。9亿周活用户的庞大基数，哪怕只有一小部分转化为付费用户，也是一个惊人的数字。
第三条腿：AI编码软件。这是近期增长最快的业务线，也是7月营收暴增的核心推动力之一。Cursor被SpaceX收购后市场格局生变，但OpenAI自己的编码产品——Codex、GPT Code Interpreter、以及面向开发者的各种工具——正在快速抢占市场。开发者是付费意愿最强的用户群体，也是人均ARPU最高的群体。
第四条腿：广告业务。虽然还在起步阶段，但ChatGPT搜索功能中嵌入的广告，已经开始贡献收入。考虑到ChatGPT的用户规模和使用时长，广告业务的天花板可能比很多人想象的要高——它本质上是一个新的搜索入口，而搜索广告是一个每年几千亿美元的市场。
"400亿不是终点，甚至不是起点的终点。这只是AI商业化第一幕的中场休息。"—— 一位AI行业分析师
这个收入结构最健康的地方在于：没有单一业务占比过高。API、订阅、企业、编码、广告——五条腿走路，任何一条腿暂时放慢，其他几条都能补上。这跟很多AI公司只有单一收入来源的脆弱结构，完全不在一个层级。
更重要的是，这些业务之间还能形成正向飞轮：API的开发者生态为ChatGPT输送功能，ChatGPT的用户数据反过来改进模型，更强的模型又吸引更多企业客户，企业客户的付费能力支撑研发投入——整个飞轮越转越快。
万亿估值的底气
一万亿美元。这个数字放在三年前，任何人说出来都会被当成疯子。
但今天，资本市场正在认真地把它当成一个合理的目标。为什么？
第一个理由是增长速度。400亿年化营收，同比增速超过100%——也就是一年翻一倍。如果明年再翻一倍，就是800亿。后年再翻一倍呢？1600亿。按照SaaS公司常见的10-15倍市销率，万亿估值对应的就是大约700-1000亿的营收规模——按当前的增长速度，可能就在明年后年。
第二个理由是市场天花板。OpenAI做的不是一个具体的产品，而是一个底层平台。它的潜在市场有多大？搜索广告（~3000亿美元/年）、软件行业（~7000亿美元/年）、云服务（~6000亿美元/年）、办公软件（~2000亿美元/年）——这些市场，OpenAI都在进入。它不是在抢某一块蛋糕，它是在重新定义所有蛋糕的切法。
第三个理由是生态壁垒。现在的开发者生态，有多少是建立在OpenAI的API之上的？无数的AI产品、插件、工具、应用——它们的底层都是GPT模型。这些产品的用户、数据、商业模式，反过来又在巩固OpenAI的平台地位。这跟当年Windows的生态护城河是一个逻辑——当所有人都基于你做开发的时候，你就赢了。
"OpenAI的万亿估值不是泡沫。它不是靠概念撑起来的，是靠400亿营收、100%增速、和一个正在形成的开发者生态撑起来的。"—— 一位科技行业投资人
当然，也有不同的声音。质疑者认为，AI的竞争格局还远远没有定型——Anthropic在抢企业市场，Google有搜索和云的双重护城河，国产大模型在快速追赶。OpenAI现在的领先优势，能维持多久？谁也说不准。
但有一点是确定的：AI商业化的速度，比任何预测都快。两年前还有人说"大模型不知道怎么赚钱"，一年前还有人说"AI公司都是亏损的"，现在——400亿营收摆在那里。
三年走完别人三十年的路
让我们把时间往回拨三年。
2022年11月30日，ChatGPT发布。那时候的OpenAI，还是一个靠微软输血的研究机构，年营收大概只有几千万美元，主要靠API和DALL·E。没有人能想到，三年后它会变成一个400亿体量的商业帝国。
2023年，是AI的"技术启蒙年"。所有人都在惊叹ChatGPT的能力，所有人都在讨论AGI，但几乎没有人在认真讨论商业化。那时候的主流观点是：大模型很厉害，但怎么赚钱还不知道。
2024年，是AI的"产品落地年"。ChatGPT Plus付费用户快速增长，企业版推出，API调用量飙升。市场开始意识到：AI不仅能聊天，还真的能赚钱。但400亿？想都不敢想。
2025年，是AI的"商业化爆发年"。营收从几十亿冲到几百亿，Agent、编码、企业服务多点开花。也是在这一年，OpenAI的估值从几百亿跳到几千亿。
2026年，也就是现在——400亿年化营收，万亿估值IPO。
三年时间，走完了传统科技公司二三十年的路。
这背后的原因是什么？简单说：AI不是一个新行业，它是对所有行业的重构。互联网用了二三十年重构了零售、媒体、广告、社交——AI可能只用五到十年，就能重构软件、办公、教育、医疗、金融……每一个被它触碰的行业，效率都会被重新定义，价值都会被重新分配。
而OpenAI，站在了这场重构的最中心。
隐忧与变量
但400亿的故事，也不是没有隐忧。
第一个隐忧是竞争。Anthropic正在企业市场快速抢占份额，预计10月上市估值2万亿。Google的Gemini背靠搜索和云服务，有天然的分发优势。国产大模型的进步速度也越来越快，DeepSeek、智谱这些公司在某些领域已经能跟海外旗舰掰手腕。OpenAI现在的领先地位，不是牢不可破的。
第二个隐忧是成本。大模型的训练和推理成本极其高昂。营收在增长，成本也在增长。如果成本增长的速度超过营收增长的速度，那利润率就会承压。AI公司最终能不能赚钱、能赚多少钱，现在还没有最终答案。
第三个隐忧是监管。欧盟AI法已经进入实质执法阶段，美国也在酝酿各种AI监管政策。数据隐私、内容安全、算法透明度——每一项监管政策的变化，都可能对OpenAI的商业模式产生重大影响。
第四个隐忧是增长的可持续性。20%的环比增长，不可能永远持续下去。基数越大，增长越难。当营收到了500亿、600亿、800亿的时候，还能保持多快的增速？这是资本市场最关心的问题。
"AI行业最确定的事是增长，最不确定的事是谁最终赢。今天的领跑者，未必是明天的赢家。"—— 一位AI行业观察者
但不管怎样，有一个事实已经无法否认：AI商业化已经全面兑现了。它不再是一个概念、一个愿景、一个未来时。它是一个正在发生的、400亿美元规模的、高速增长的真实市场。
三年前，ChatGPT发布的时候，很多人问：AI到底能干什么？
三年后，400亿美元的营收给出了第一个答案。但这只是第一个答案——AI对商业世界的重构，才刚刚开始。
未来三年，我们会看到更多答案。有些会超出预期，有些会低于预期。但有一点可以确定：这个行业的变化速度，会比你以为的更快。
明天见。
$40 billion. Annualized.
Three years ago, OpenAI was still a nonprofit research institution surviving on donations. Today, its annualized revenue has surpassed $40 billion - doubling from the end of last year, with July alone growing over 20% month-over-month.
What does that number mean? Google took 12 years to hit $40B in annual revenue. Facebook took 11. Amazon took 16. OpenAI went from ChatGPT launch to $40B annualized in less than three years.
Even more staggering is the growth rate. A company already at $40B scale still growing 20%+ month-over-month - that's not linear growth, that's exponential. At this pace, annualized revenue could hit $50B or even $60B by year's end.
The capital markets have already priced it in: a trillion-dollar IPO is accelerating. If it goes public at that valuation, OpenAI will become one of the largest IPOs in human history, and the AI industry's first truly trillion-dollar company.
AI commercialization is materializing faster than anyone expected.
Where the $40B Comes From
$40B isn't propped up by a single business. It's a three-dimensional revenue structure - API, subscriptions, enterprise, coding, advertising - multiple legs running simultaneously.
First leg: API business. This is OpenAI's most foundational and most stable revenue source. Millions of developers worldwide call GPT-series models, paying by the token. From GPT-4 to GPT-5, model capabilities keep upgrading, and call volumes keep growing. Enterprise customers are increasingly migrating core workflows to OpenAI's APIs, and this revenue is extremely sticky.
Second leg: ChatGPT subscriptions. ChatGPT Plus, ChatGPT Pro, ChatGPT Enterprise - from individual users to large enterprises, subscription has proven to be AI products' most effective business model. With 900 million weekly active users, even if only a fraction convert to paid, it's a staggering number.
Third leg: AI coding software. This is the fastest-growing business line recently, and the core driver of July's revenue surge. After SpaceX's acquisition of Cursor shifted market dynamics, OpenAI's own coding products - Codex, GPT Code Interpreter, and various developer-facing tools - are rapidly capturing market share. Developers are the highest-willingness-to-pay user group, and they have the highest per-capita ARPU.
Fourth leg: advertising business. Still in early stages, but ads embedded in ChatGPT search functionality are already contributing revenue. Given ChatGPT's user scale and usage time, the ceiling for ad revenue might be higher than many think - it's essentially a new search entry point, and search advertising is a multi-hundred-billion-dollar annual market.
"$40B isn't the finish line. It's not even the end of the beginning. This is just halftime of the first act of AI commercialization." - An AI Industry Analyst
The healthiest part of this revenue structure is that no single business dominates. API, subscriptions, enterprise, coding, advertising - five legs walking together. If any one leg slows temporarily, the others make up for it. This is on a completely different level from the fragile structures of many AI companies with only a single revenue source.
More importantly, these businesses form a positive flywheel: the API developer ecosystem feeds features into ChatGPT, ChatGPT's user data in turn improves the model, stronger models attract more enterprise customers, and enterprise paying power funds R&D - the whole flywheel spins faster and faster.
The Case for a Trillion-Dollar Valuation
One trillion dollars. Three years ago, anyone saying that would've been called crazy.
But today, the capital markets are seriously treating it as a reasonable target. Why?
First reason: growth rate. $40B annualized revenue, growing over 100% year-over-year - meaning doubling every year. If it doubles again next year, that's $80B. The year after? $160B. At the 10-15x price-to-sales ratio common for SaaS companies, a trillion-dollar valuation corresponds to roughly $70-100B in revenue - and at current growth rates, that could be next year or the year after.
Second reason: market ceiling. OpenAI isn't building a specific product - it's building a foundational platform. How big is its potential market? Search advertising (~$300B/year), software industry (~$700B/year), cloud services (~$600B/year), office software (~$200B/year) - OpenAI is entering all of these markets. It's not fighting for one slice of the pie. It's redefining how every pie gets sliced.
Third reason: ecosystem moat. How much of today's developer ecosystem is built on OpenAI's APIs? Countless AI products, plugins, tools, applications - they all run on GPT models underneath. Those products' users, data, and business models in turn reinforce OpenAI's platform position. It's the same logic as Windows' ecosystem moat back in the day: when everyone builds on top of you, you've already won.
"OpenAI's trillion-dollar valuation isn't a bubble. It's not propped up by hype - it's propped up by $40B in revenue, 100% growth, and a developer ecosystem that's forming." - A Tech Industry Investor
Of course, there are dissenting voices. Skeptics argue that AI's competitive landscape is far from settled - Anthropic is grabbing enterprise market share fast, Google has the dual moat of search and cloud, and Chinese foundation models are catching up quickly. How long OpenAI's current lead can last is anyone's guess.
But one thing is certain: AI commercialization is moving faster than any prediction. Two years ago people were saying "nobody knows how LLMs will make money." A year ago people were saying "AI companies all lose money." Now - $40B in revenue is there for everyone to see.
Three Years to Cover What Took Others Thirty
Let's rewind three years.
November 30, 2022. ChatGPT launches. Back then, OpenAI was still a research institution funded by Microsoft, with annual revenue of maybe a few tens of millions, mainly from APIs and DALL·E. Nobody could've imagined that three years later it'd be a $40B commercial empire.
2023 was AI's "year of technical enlightenment." Everyone was marveling at ChatGPT's capabilities, everyone was talking about AGI, but almost nobody was seriously discussing commercialization. The mainstream view back then: LLMs are impressive, but we don't know how they'll make money.
2024 was AI's "year of product落地." ChatGPT Plus paid users grew rapidly, enterprise edition launched, API call volumes soared. The market started realizing: AI doesn't just chat - it can actually make money. But $40B? Unthinkable.
2025 was AI's "year of commercial explosion." Revenue shot from single-digit billions to tens of billions, with agents, coding, and enterprise services blooming across the board. It was also the year OpenAI's valuation jumped from tens of billions to hundreds of billions.
2026 - right now - $40B annualized revenue, trillion-dollar IPO.
Three years to cover what traditional tech companies take 20-30 years to achieve.
What's the reason behind this? Put simply: AI isn't a new industry - it's a restructuring of all industries. The internet took 20-30 years to restructure retail, media, advertising, social - AI might only need 5-10 years to restructure software, office work, education, healthcare, finance... every industry it touches gets its efficiency redefined and its value redistributed.
And OpenAI stands at the very center of this restructuring.
Risks and Variables
But the $40B story isn't without its risks.
First risk: competition. Anthropic is rapidly gaining share in the enterprise market, reportedly targeting a $2T valuation for its October IPO. Google's Gemini has the natural distribution advantage of search and cloud services. Chinese foundation models are also improving faster and faster - companies like DeepSeek and Zhipu can already go toe-to-toe with overseas flagships in certain areas. OpenAI's current lead isn't unassailable.
Second risk: costs. Training and inference costs for large models are extremely high. Revenue is growing, but so are costs. If cost growth outpaces revenue growth, margins come under pressure. Whether AI companies can ultimately be profitable, and how profitable, still has no final answer.
Third risk: regulation. The EU AI Act has entered substantive enforcement, and the US is also crafting various AI regulatory policies. Data privacy, content safety, algorithmic transparency - every regulatory policy shift could significantly impact OpenAI's business model.
Fourth risk: sustainability of growth. 20% month-over-month growth can't continue forever. The bigger the base, the harder growth becomes. When revenue hits $50B, $60B, $80B - how fast can it still grow? That's the question capital markets care about most.
"The most certain thing in AI is growth. The most uncertain thing is who ultimately wins. Today's leader might not be tomorrow's winner." - An AI Industry Observer
But regardless, one fact is undeniable: AI commercialization has fully arrived. It's no longer a concept, a vision, or a future tense. It's a real, happening, $40-billion-scale, fast-growing market.
Three years ago, when ChatGPT launched, many people asked: what can AI actually do?
Three years later, $40 billion in revenue gives the first answer. But it's only the first answer - AI's restructuring of the business world has barely begun.
Over the next three years, we'll see many more answers. Some will exceed expectations, some will fall short. But one thing is certain: this industry will change faster than you think.
See you tomorrow.
