Over the past two decades, the most valuable assets of the internet have been two things: user time and advertising space.
Whoever can get users to scroll longer and click more will capture the biggest slice of the digital economy. Traffic is the hardest currency of this era.
But today, a new signal is emerging.
From January to May 2026, Anthropic's annualized revenue surged from $9 billion to $45 billion.
Meanwhile, ChatGPT's personal subscription growth has stagnated, and the global 2C paid conversion rate for AI applications is generally below 5%. The idea that users will switch to Doubao for just one dollar isn't a joke, but a repeatedly verified reality.
On one side is ice for consumers (C-end), and on the other side is fire for businesses (B-end).
This is not a contradiction, but a clear structural shift: the focus of AI commercialization is shifting from serving consumers to saving companies labor costs.
In the internet age, money is made from traffic.
In the AI era, you'll earn a salary.
Ice and Fire: The Commercialization of AI Experiences Two Extremes
Let's look at the other side first. Over the past year, many consumer-facing AI products have faced growth anxiety. ChatGPT's monthly active user growth has slowed significantly, and conversion rates for both free and paid versions have remained low. Large-scale domestic AI apps have entered a price war, with API prices approaching free. The user mentality is: use whichever is free; paying? Forget about it.
The predicament of AI for consumers is not accidental. The differences in AI's chat, writing, and drawing capabilities are becoming increasingly smaller, with switching costs virtually zero. No single company can truly replace yours. According to SearchLab data, ChatGPT Plus subscription conversion rates have consistently been below 5%, while the quality of free alternatives is approaching GPT-4 levels. Users have calculated it clearly: paying $20 per month for a 10% capability improvement is not worthwhile.
Now let's look at the other side of the story. Anthropic's ARR jumped from $9 billion to $45 billion in just five months. Over 90% of this came from enterprise API and agent deployments, not individual subscriptions. Claude Code's programming agent became the core growth engine, with enterprise customers growing from 500 with annual spending exceeding $1 million in February to over 1,000 in May. OpenAI's enterprise revenue continued to climb, Microsoft Copilot's penetration rate among Fortune 500 companies jumped to 55%, and Salesforce and ServiceNow are both using AI agents as a core selling point for price increases.
Why are companies paying so much? The core logic is ROI. One Claude Code agent can replace the workload of hundreds of junior programmers. Companies spend $3 on AI and save $10 in salaries. This formula is so clear it doesn't need sales pitches. Industry estimates suggest that the average ROI for enterprise clients is 3.7 times, with some exceeding 10 times. Against the backdrop of macro-level cost reduction and efficiency improvement, such a certain return is irresistible.
This isn't just a phenomenon among a few leading companies, but a collective shift across the entire industry. According to PitchBook data, venture capital investment flowing to enterprise AI startups increased by 210% year-on-year in Q1 2026, while funding for consumer-facing AI decreased by 35%. Talent is also migrating: industry observers note that over 40% of founders of consumer-facing AI products have announced a shift to the enterprise track. While this appears to be a split, it's essentially the first time that AI commercialization has truly established a closed loop regarding who pays the price and why.
Moreover, B2B is not a low-margin business. Anthropic boasts a gross margin exceeding 70%, a customer retention rate of 140%, and is projected to be profitable in Q2 2026. Companies pay high prices because the savings far outweigh the costs. This isn't a price war, but a positive ROI cycle driven by productivity premiums. The global pool of annual human resource costs for back-office, customer service, and junior R&D positions exceeds $5 trillion. Even if AI replacement rates are only 10%, it's still a $500 billion market. Anthropic's $45 billion ARR represents less than 10%, meaning its ceiling is far from being reached.
Destruction and Construction: The Clash Between Traffic Logic and Cost Logic
Many people are used to understanding AI through the logic of the internet: acquiring customers for free and then monetizing through advertising and value-added services. But AI is not the internet. Confusing these two sets of logic is the biggest misconception in understanding the commercialization of AI.
Why can't consumers make money in the consumer market? Because there are insurmountable structural obstacles:
First, productivity tools struggle to capture entertainment time. Short videos and games satisfy emotional needs, and users are willing to pay for enjoyment. AI, on the other hand, solves specific tasks and is used and then discarded. ChatGPT's average conversation lasts about 7 minutes, while TikTok's exceeds 30 minutes. AI is inherently at a disadvantage in the battle for user time.
The second issue is homogenized competition and extremely low migration costs. AI capabilities are rapidly homogenizing; in 2024, GPT-4 stood out, but by 2026, open-source models had caught up to the same level. When performance is roughly the same, price becomes the only differentiator, ultimately leading to free services and price wars. This has already been demonstrated in the fields of text-to-image processing and translation.
Of course, the lack of network effects leading to a weakened competitive advantage is also a significant issue. Whether you use ChatGPT or Claude doesn't affect anyone. User migration only requires changing a bookmark once. User scale is not a competitive advantage; even OpenAI's hundreds of millions of monthly active users cannot lock in users.
Most importantly, there's a ceiling effect to paid subscriptions for individual consumers. Users are only willing to pay a limited amount for productivity tools, not exceeding their replacement cost. Low-frequency users only accept free services, while high-frequency users turn to bulk purchases by enterprises. Squeezed from both ends, consumer subscriptions have become a double-edged sword.
Conversely, the explosive growth of the B2B market is precisely because its business DNA perfectly aligns with AI.
It's important to understand that businesses only consider ROI when purchasing AI. Consumers might pay for an attractive interface, but enterprise purchasing decision-makers only calculate the cost: spending 3 yuan saves 10 yuan, so they buy it. A Goldman Sachs report shows that the customer lifetime value of enterprise-grade AI software is 8 times the customer acquisition cost, far exceeding the SaaS average, and exhibiting extremely high customer stickiness.
In the B2B sector, AI doesn't just replace a few people, but rather entire job functions. As companies gradually delegate customer service, initial financial review, and code generation to AI, they save on the human resource costs of these entire functional modules. One large e-commerce company, after introducing AI customer service, reduced its 500-person team to 80, and its response time from 5 minutes to 30 seconds. AI replaces workflows, not just human heads.
The extremely high switching costs associated with deep integration mean that after a company has deeply integrated AI into its CRM, CI/CD, and data warehouse, migrating to another model requires re-optimization and transformation, which in itself constitutes a competitive advantage. Business-related fine-tuning data and prompt templates are also assets.
Of course, there's also the reason why B2B companies have stronger pricing power. A company with annual revenue of 1 billion can save 10 million in labor costs by spending only 3 million on AI, representing just 0.3% of its revenue. Companies won't sacrifice quality and stability to save a few cents on token price. This is precisely why Anthropic boasts a gross profit margin of over 70%—based on value-driven pricing, not cost-plus pricing.
The consumer (C-end) market operates on a traffic-driven logic, while the business (B-end) market operates on a cost-substitution logic. The failure of the consumer (C-end) market is not due to the inadequacy of AI itself, but rather a mismatch in business models. AI commercialization is shifting from the former to the latter; this is not a short-term shift, but a fundamental change in the underlying logic.
Virtual and Real: The Evolution from Digital Tools to Digital Workforce
What does Anthropic's 45 billion ARR truly validate? It's not just about making money in the B2B sector, but a more fundamental shift: AI is evolving from a digital tool into a digital workforce.
First, AI is no longer auxiliary software, but the main driver of productivity. For the past forty years, the logic of enterprise software has been to improve human efficiency. Excel made calculations faster, but accountants still exist. Photoshop made designers more efficient, but designers still exist. All software is a tool; humans are the decision-makers. But AI agents are different: Claude Code writes code directly, and customer service agents directly respond to users. AI has transformed from a tool into an executor, and humans have transformed from operators into supervisors. This is a qualitative change.
Secondly, B2B revenue and the AGI narrative are not contradictory, but rather form a symbiotic closed loop. Some question whether AGI is a bubble since revenue primarily comes from enterprise tools rather than AGI. The opposite is true. B2B revenue feeds back into model training; 45 billion in ARR is invested in the next generation of models, and the stronger the model, the more willing enterprises are to pay. Model progress sustains the AGI belief; the market doesn't need AGI to be achieved today, but only to see it continuously approaching. The AGI belief supports high valuations, which in turn attract funding, which is then reinvested in R&D. This is a complete positive cycle. Today's agents, in a commercial sense, are the prototypes of AGI. The market wants the path, not the destination, and B2B revenue is the cornerstone that paves this path.
Third, AI is replicating the essential logic of the Industrial Revolution. More than two hundred years ago, the steam engine replaced human and animal power, becoming the new core of productivity. Those companies that first adopted the steam engine gained an overwhelming efficiency advantage. The Industrial Revolution was essentially a revolution in labor substitution, replacing manual labor with machines and liberating productivity from the limitations of biological organisms.
Today, AI is doing the same thing, only replacing mental labor. Programmers, customer service representatives, data analysts, accountants—white-collar jobs are being gradually penetrated by AI. This isn't a gradual increase in efficiency, but a structural replacement of labor. Those companies that were among the first to integrate AI agents into their business processes are reaping the dual benefits of cost advantages and faster response times.
In the internet age, the most valuable assets were traffic and user attention. That was the logic of the consumer internet. In the AI age, the most valuable assets are digital labor, algorithms, and computing power capable of performing mental labor at extremely low cost. This is the logic of the productivity internet. Global annual wages exceed $50 trillion; even if AI replaces only 10% of that, it would still be a $5 trillion annual market. Meanwhile, the global market for internet advertising and subscriptions combined is just over $1 trillion.
Therefore, AI is not the next Facebook, nor the next Google. It's not a traffic business. It's the next steam engine, a new factor of production that redefines labor and costs. When it massively replaces human labor, the market value it creates will far exceed that of the internet. Wages, much larger than traffic.
Looking back , we may have been using the wrong analogy to understand AI. In the internet age, the most valuable asset was traffic. Whoever captured users' time and attention could build an empire. But AI is not a traffic business. Its true value lies not in getting users to scroll for a few more minutes, but in replacing human labor and improving organizational efficiency.
This is more like the Industrial Revolution. Over two hundred years ago, the steam engine emerged, replacing human and animal power to become the new core of productivity. Today, AI is doing the same thing. It's not the next Facebook, nor is it the next Google. It's the next steam engine, a new factor of production that redefines labor and costs.
When an agent replaces not just 10 people, but an entire job function; when companies save 10 yuan for every 3 yuan spent; when AI's ARR (Average Revenue Per Transaction) skyrockets from tens of billions to hundreds of billions... then we will truly understand: in the internet era, we earned traffic; in the AI era, we earn wages. And wages are far more valuable than traffic.
AI is not replicating the internet. It is replicating the Industrial Revolution.
References:
36Kr, "For the First Time in History, Anthropic is Going to Make Money," May 2026 https://www.36kr.com/p/3819897940562307
PitchBook, Q12026AIVCTrendsReport
https://pitchbook.com/news/reports/q12026aivctrends
NetEase News, "The More AI is Used, the More Money It Makes: Read Goldman Sachs' Intelligent Agent Economics Report," May 2026
http://www.163.com/dy/article/KSAL8CLK05568W0A.html
Caizhong News Agency, "Haitong International: Anthropic Achieves Profitability Two Years Ahead of Schedule, Establishing a Milestone in AI Commercialization," May 2026.
https://www.caizhongshe.cn/article7465239590204012512.html
This article is from the WeChat Official Account "Tech Insights" (ID: kejixinzhi), author: Orange
