Just now, news broke that Anthropic, the maker of Claude chatbots, is preparing for an IPO with a potential valuation of over $300 billion.
According to the latest reports from foreign media, Anthropic has engaged the Silicon Valley law firm Wilson Sonsini to assist with its initial public offering (IPO) as early as the beginning of next year. The firm is also working on an internal checklist to pave the way for what could be one of the largest IPOs in history. The firm has extensive experience leading IPOs of large technology companies, having served companies such as Google, LinkedIn, Lyft, and Square.
The report indicates that Anthropic is pursuing a private funding round ahead of its potential IPO, targeting a valuation of $350 billion. The company has already begun discussions with several major investment banks, but negotiations are still in the early stages. Anthropic's move may be aimed at gaining an advantage over its competitor OpenAI, which is considering an IPO in the second half of 2026 with a potential valuation of $1 trillion, a figure previously unimaginable.
In a recent podcast interview, Anthropic CEO Dario Amodei not only revealed Anthropic's actual revenue figures, but also discussed the current "dilemma" and competitive moats facing the AI industry, as well as multifaceted strategies for addressing job loss risks in the post-AGI era.
It's worth mentioning that Amodei ridiculed OpenAI's decision to activate a "red alert" status and repeatedly criticized its CEO Sam Altman's management philosophy and the huge financial investment in its bid to compete for the top position in the AI industry.
What justifies a valuation of 350 billion?
CEO: Revenue increased tenfold annually.
The news that Anthropic is preparing for a large IPO is a prime example of the explosive growth of AI companies. At the earliest IPO date of early next year, Anthropic will be only 5 years old. In comparison, Google went public 6 years after its founding, with a valuation of approximately $23 billion; Meta (formerly Facebook) took 8 years to go public, with a valuation of approximately $100 billion at the time of its IPO; and even the "veteran" Microsoft waited 11 years, with a valuation of only about $800 million when it went public in 1986.
Behind the hype, Anthropic has solid business backing. Its core product is the chatbot Claude, which has already generated revenue. Amodei projects its annualized revenue will more than double next year, reaching approximately $26 billion, and serving over 300,000 enterprise clients. Last month, Microsoft and Nvidia announced plans to invest up to $15 billion in Anthropic, while Anthropic pledged $30 billion to utilize Microsoft's cloud infrastructure, forging a close partnership. Also in the same month, Anthropic released the Claude Opus 4.5 model, claiming it to be the most advanced AI product currently available, featuring upgrades in computer code generation and workplace document processing capabilities.
In a recent interview, Amodei revealed that Anthropic's revenue has grown tenfold annually for the past three years, from $0 to $100 million in 2023 and from $100 million to $1 billion in 2024. By the end of this year, the company expects to grow from $1 billion to between $8 billion and $10 billion. "Admittedly, this growth trend will certainly slow down, but the growth rate will still be very rapid."
According to foreign media reports, Anthropic expects its sales to reach $70 billion by 2028, which means its proposed valuation is equivalent to five times that sales. In contrast, Meta's valuation at its IPO in 2012 was six times its sales three years later, and Palantir's was ten times.
Perhaps the most difficult asset to value is what Anthropic sees as its core asset: its core principles. The company was founded with the aim of becoming a safer alternative to OpenAI, creating AI robots that are "useful, honest, and harmless." In fact, the Center for AI Safety believes that among mainstream AI models, Anthropic's products are the least likely to "blatantly lie" and the least likely to provide answers to "dangerous expert-level virological questions."
“Our core focus is on the problem-solving needs of consumers in their work and daily lives. Since we don't deal with passive consumer applications or image/video generation, these areas have been explicitly excluded from our mission. Our core goal is to create beneficial products while minimizing potential risks, because any technology is a double-edged sword. We believe that empowering human growth, expanding the boundaries of capabilities, creating value, and solving problems represent the best balance between risk and reward,” said Joel Levinstein, Design Director at Anthropic, in an October interview.
Whether investors will pay a premium or demand a discount remains to be seen. Meanwhile, the future is fraught with uncertainty. If Anthropic ultimately succeeds in its IPO, AI technology may have already achieved another leap forward, or it may run into trouble due to excessive hype. Therefore, it is necessary to consider how Anthropic can justify its $350 billion valuation, and also to be prepared to overturn all assumptions and reassess the situation.
It's worth noting that Anthropic's potential IPO comes at a time of growing market concerns about overheated asset valuations. The Bank of England warned on Tuesday that valuations of technology companies, particularly those focused on AI, "remain significantly overvalued." The bank noted that US stock valuations are nearing levels seen during the dot-com bubble, and "the risk of a significant correction is rising."
Former Reserve Bank of India Governor and current University of Chicago professor Raghuram Rajan warned that the current ample credit environment and the Federal Reserve's interest rate cuts are creating market conditions that are "accelerating the accumulation of risks." Pablo Hernández de Cos, Managing Director of the Bank for International Settlements, also issued a similar warning about excess liquidity in financial markets. Adding to the complexity, reports indicate that National Economic Council Director Kevin Hassett has become President Trump's top choice to replace Federal Reserve Chairman Jerome Powell next year, potentially signaling a continuation of loose monetary policy.
The Battle of Red Alert
"Claude's traffic is only about 1% of ChatGPT's, but its profitability far surpasses ChatGPT's. I agree with Amodei's point of view; he runs his company very responsibly, while Altman is extravagant and tries to play the role of a salesman," one netizen commented.
On December 2nd, an internal memo obtained by foreign media revealed that OpenAI CEO Sam Altman officially announced a "Code Red" state to all employees on December 1st. This comes as Google, a major competitor in the AI field, has released its Gemini 3 model with great fanfare, and three years ago, when ChatGPT was launched, Google also issued its own "Code Red" announcement in response to this situation.
In an interview, Amode stated that Anthropic never issued a "red alert." Both Google and OpenAI are essentially consumer-centric. While they have attempted to venture into enterprise business, their focus remains on competition in the consumer sector. This is the root cause of the frequent "red alerts" and fierce competition: Google wants to defend its search monopoly, while OpenAI's core business is also focused on consumers. The two are locked in direct combat, with enterprise services being secondary to both.
As for the so-called "model war," Amodei believes that maintaining a model advantage is crucial, but Anthropic is also developing in different directions and dimensions.
“We’re in a relatively advantageous position; all we need to do is continue to grow and develop our models,” he added. He noted that Anthropic feels relatively less competitive pressure, partly because the company’s products are more focused on enterprise clients than consumers. “Our models are increasingly being optimized for enterprise needs,” he stated, emphasizing the fundamental difference between enterprise-oriented model development and consumer-focused products. “The priorities are completely different; we focus less on user engagement and more on coding capabilities, support for high-end intellectual activities, and scientific research assistance.”
Amodei also revealed that although Anthropic has found a dominant market in enterprise coding, the company is expanding its focus to industries such as finance, biomedicine, retail, and energy. "I'm glad Anthropic chose a different path, with a core focus on the enterprise market. Over the years, our models have been continuously optimized, becoming increasingly tailored to enterprise needs. And I don't believe that all models will converge after AGI is implemented."
Furthermore, he pointed out that there are some common "moats" in the enterprise services sector: enterprises establish long-term partnerships with service providers and are accustomed to using specific models. Even though API businesses are essentially just selling native models, it is extremely difficult for enterprises to actually switch models. Their downstream customers have already adapted to the existing models, and the prompts, interaction methods, and "features" of different models are all different, making the switching cost very high.
“Looking at Anthropic itself, because we are deeply rooted in the enterprise market, I think we are a good trendsetter, and perhaps even more focused than other companies. After all, other companies' businesses are filtered through the consumer level to some extent, and consumers have their own usage habits and scenarios,” Amodei said.
In a dilemma,
What are the strategies of each company?
“Some players are adopting a ‘yol-or-nothing’ strategy, investing too aggressively,” Amodei said. Anthropic is trying to “manage in a responsible manner,” but he is skeptical of the huge sums of money that companies like Google, OpenAI, and Meta are investing in to compete for the top spot in the AI industry.
According to Amodei, the industry currently faces a genuine dilemma: on the one hand, the growth rate of economic value is uncertain; on the other hand, the construction of data centers to support technological development has a long time lag. This means that by early 2025, a decision must be made on how much computing resources need to be purchased to meet the operational needs of the model when corresponding revenue scales are reached by early 2027. This presents two interconnected risks: if insufficient computing resources are purchased, it will be impossible to serve all potential customers, forcing them to competitors; if too many are purchased, revenue may not be able to cover the cost of computing resources, and in extreme cases, even bankruptcy may be at risk.
"The buffer space within this uncertainty range essentially depends on our profit margin. If we have an 80% profit margin, then a $20 billion investment in computing resources can support $100 billion in revenue. However, due to the extremely large range, it is almost impossible to avoid decision-making errors in both of the above scenarios."
Amodei stated that the business models in the enterprise market are more robust and offer more substantial profit margins. However, if a different business model is adopted, such as a consumer-focused model where revenue sources are less stable and profit margins are uncertain, and if the company's leaders are inherently inclined to "go all in" or pursue exaggerated figures, they may become overly aggressive in their investments. He also emphasized that some participants are failing to manage these risks properly and are taking on unwise risks.
According to reports, Anthropic's core strategy is to procure sufficient computing resources to ensure it can cover costs even in the most pessimistic 10th percentile scenario (i.e., when the company might face difficulties). Of course, there is an extreme scenario where the company might be unable to afford the costs if things go terribly wrong—a tail risk that cannot be completely eliminated. However, Anthropic is working to manage this risk effectively while procuring sufficient computing resources to remain competitive with other players.
The IPO competition has reached a fever pitch.
Anthropic vs OpenAI: A Comparison in Action
Anthropic's anticipated IPO comes at a time of increasingly fierce competition in the AI industry. Once investors have the opportunity to directly purchase Anthropic stock (existing investors include Amazon, Google, Microsoft, and Nvidia), they will undoubtedly conduct a close comparison between Anthropic and OpenAI, the maker of ChatGPT.
OpenAI is reportedly preparing for what could be one of the largest IPOs in history, with a potential valuation of $1 trillion. While OpenAI CFO Sarah Friar downplayed short-term listing plans, the company could file for an IPO with regulators as early as the end of 2026. OpenAI's latest valuation is $500 billion, equivalent to five times its projected sales in 2028.
The outcome of this "battle" hinges on investor preferences. Founded in 2021 by a group of former OpenAI executives, Anthropic has secured billions of dollars in investment from Amazon and Google. Unlike its Microsoft-backed competitors, Anthropic focuses on commercial applications of its AI tools, rather than vying for a broad consumer market share through generative AI multimedia technologies. According to Menlo Ventures, as of the end of July, Anthropic held a 32% market share in the "enterprise" market, seemingly favored by enterprise clients. Menlo Ventures is also one of Anthropic's investors. This advantage is crucial because enterprise clients are far more willing to pay for AI than consumers.
Anthropic's business scope is also more focused: solely on model development. OpenAI, on the other hand, has diversified its operations, investing in data centers, portable devices, equity stakes in other companies, and even launching its own web browser. Some might call this "disorganized business expansion," but venture capitalists would likely prefer to call it "full-stack development." From this perspective, Anthropic is closer to Palantir or Salesforce, while OpenAI bears the imprint of Google's parent company, Alphabet, or Microsoft.
Furthermore, Deutsche Bank analysts pointed out that Anthropic's path to profitability may be smoother compared to its larger competitors.
OpenAI is reportedly facing significant headwinds in its consumer business growth. Deutsche Bank analysts Adrian Cox and Stefan Abrudan found that ChatGPT's subscription growth in key European markets has stagnated since June, suggesting the market may be nearing saturation. Meanwhile, Anthropic's subscription revenue has surged nearly sevenfold this year (despite a low base), while OpenAI's subscription revenue growth rate is only 18%.
How to deal with the challenges brought by AI
"Technical unemployment"?
"Perhaps half of the entry-level jobs will disappear."
It's worth noting that Amodei stated at the end of the interview that he has been deeply considering this issue. Reportedly, Claude has already begun writing a significant amount of code within Anthropic. Amodei believes the problem can be divided into multiple levels, encompassing both short-term responses and long-term planning, some of which may even require a greater investment of resources from society as a whole to solve.
The first layer is the private sector level, where some issues can be resolved, even in the collaboration between model suppliers and clients. Amodei noted that each company they work with faces a trade-off, and it's not an either-or choice.
On the one hand, businesses can improve efficiency by replacing human labor with AI. For example, processes such as insurance claims processing and Know Your Customer (KYC) can now be automated end-to-end using AI. This means that the number of people required for these positions will be significantly reduced, improving efficiency, saving costs, and achieving the same results with a lower human resource investment. On the other hand, businesses can also create a lot of new value with AI. Even if AI only completes 90% of the work (not 100%), human work efficiency can still be increased tenfold. Sometimes, it is even necessary to increase the number of people by tenfold to complete 100 times the amount of work previously done with the efficiency of AI.
“We encourage businesses to choose the latter over the former. We know that businesses will inevitably take the first approach, and we are not trying to stop them, but if the second approach can account for a larger share than the first, perhaps more jobs will be created than jobs will be replaced. Does this require government incentive policies? This is the first question we need to consider.”
The second level is government involvement. Amodei points out that retraining programs are not a panacea, but some form of retraining is necessary. Businesses will proactively undertake related work and must also cooperate with the government to promote it. From a fiscal perspective, the government ultimately needs to intervene, whether through tax policies or other means.
“Even current AI models are expected to increase annual productivity by 1.6%. That’s almost equivalent to doubling productivity, and the models are constantly being upgraded. I think future annual productivity could reach 5%, or even 10%. This is a huge ‘cake’ that we can distribute to those who don’t directly benefit. Even if wealth concentrates in the hands of a few, this ‘cake’ is still large enough for more people to share.”
The third layer is the level of social structure. In the long run, societies with powerful AI will inevitably undergo structural changes. Looking back at John Maynard Keynes's *The Economic Possibilities of Our Descendants*, he introduced the concept of "technological unemployment" and predicted that his grandchildren might only need to work 15 to 20 hours a week—a completely new model of social structure.
Amodei proposes: Is it possible to build a new world where work no longer holds the central place for many, and where people can find meaning in life elsewhere; or where the nature of work changes, becoming more about self-actualization than simply survival. Society is highly adaptable and capable of self-adjustment. He explains, "I'm not advocating top-down change, but rather that society needs to restructure itself. We all need to explore together how to coexist and thrive in the post-AGI era."
“These three levels—from quick actions that are easy for businesses to implement to lengthy social changes that require broad consensus—are all indispensable. In the coming years, we must implement work at each of these three levels one by one,” Amodei concluded.
Reference link:
https://www.ft.com/content/9c8c1bb7-b1aa-4c9c-9896-791998b506fd
https://www.businessinsider.com/anthropic-ceo-dario-amodei-drags-openai-and-google-code-red-2025-12
https://www.youtube.com/live/FEj7wAjwQIk
This article is from the WeChat public account "AI Frontline" , author: Hua Wei, and published with authorization from 36Kr.
