Ahead of its initial public offering (IPO), Anthropic released a 42-minute video to persuade investors. Executives, including CEO Dario Amodei, appeared directly to explain why the company deserves a valuation exceeding $1 trillion, showing investors the competitive strengths and growth potential Anthropic, a company that has consistently emphasized AI safety, is presenting.
On the 10th local time, Anthropic began its investor roadshow and posted a presentation video on its U.S.-only IPO website.
A roadshow is an event at which executives introduce their company to investors and seek investment. Recently, roadshows have also used pre-produced videos alongside in-person presentations. While a securities registration statement is a document that broadly discloses such information as the company’s business operations and risk factors, a roadshow is a forum in which executives directly convey the messages they wish to emphasize to investors.
The first thing Anthropic emphasized was its argument that AI will create the greatest economic opportunity in human history. CEO Amodei introduced AI as “the most promising technology in human history” and forecast that it will transform not only software development and professional work but also scientific and medical research. The argument is that because the market AI will create is so enormous, the company’s current high valuation can also be justified.
It also emphasized “computing efficiency” as a technological competitive advantage. Anthropic argued that it can develop better AI with the same computing resources by using a variety of semiconductors, not just Nvidia’s, and by improving its algorithms. It cited the example of using “Mythos” to switch “Claude” to run on AMD GPUs in just three days. Given that enormous computing costs are the greatest burden on AI companies, Anthropic highlighted the fact that the same investment can produce greater results.
The company also presented the commercial results it has achieved in the enterprise market. It said that Japan’s Rakuten used Claude to shorten the time required to launch software features from 24 days to five, while the U.S. insurance company AIG reduced its quote-processing time by 55%. This emphasized the fact that AI is not merely a possibility for the future but a productivity tool that companies are already paying for and using.
Revenue growth and future profitability were also key messages. Anthropic said third-quarter revenue reached approximately $17 billion, more than 12 times the figure for the same period last year. It reported a cumulative gross profit margin of 85% for the first three quarters of the year and a margin of 43% after accounting for computing and infrastructure costs. It also forecast that it could achieve an adjusted operating profit margin of more than 35% in the long term. The company’s argument is that while AI infrastructure requires enormous capital, it will be able to generate sufficient profits going forward.
Particularly notable was Anthropic’s framing of AI safety as a business competitive advantage. Anthropic drew attention by warning in IPO-related materials released last September that AI could threaten human survival. CEO Amodei has also argued that the pace of AI development needs to be regulated, maintaining that Anthropic must preserve its competitiveness to take the lead in safe AI development.
In this video as well, Amodei mentioned the danger of developing biological weapons using AI within roughly the first 30 seconds. Throughout much of the presentation, however, he linked safety to the trust of enterprise customers and to business competitiveness.Anthropic explained that because safety does not impede technological progress and Claude’s work can be trusted, the results of AI can be applied to real-world operations. It presented safety not as a constraint on technological development, but as a reason for companies to choose Claude.
An analysis of the video script by Business Insider also shows these characteristics. “Claude” appeared 65 times, “model” 58 times, and “computing” was the next most frequently mentioned term at 49 times. “Customer” appeared 42 times, while “work” and “AI” each appeared 39 times, and safety-related expressions appeared 27 times. When the related words were grouped by topic, models and training appeared 83 times, computing and infrastructure 81 times, financial performance and growth 77 times, and safety, trust, security, and risk 55 times.
Anthropic also disclosed in a regulatory filing nearly $895 billion in total commitments related to computing and infrastructure. Its roadshow appears to have emphasized the logic that it can secure growth and profitability despite bearing such a massive investment burden. On Wall Street, the astronomical computing costs ahead and the resulting profitability concerns are being identified as major investment risks.
Ultimately, Anthropic’s message is that the AI market has enormous growth potential and that the company can lead that market on the basis of its technological capabilities, computing efficiency, and the trust of corporate customers. In other words, a company that had long led the way in warning about the risks of AI is now positioning safety as a competitive advantage in its business and seeking to persuade investors.
Next, here are the top news stories for the 10th.
■ Anthropic’s “Fable 5.2” Regains No. 1 in AI Performance Rankings…API Prices Also Cut by 20%
As expected, “Claude Fable 5.2” dethroned Google’s “Gemini 4 Carbon” and reclaimed the top spot in AI performance rankings. More noteworthy is the fact that Anthropic is actively pursuing price competitiveness as well as performance by continually lowering the API prices of its state-of-the-art models.
■ “Hiku” Performance at One-Third the Cost…Meta Launches “Muse Ember” for Agents
Following the high-performance model “Muse Wonder Full,” Meta released the inexpensive, lightweight model “Muse Ember,” which is specialized for repetitive agent tasks. The move seeks to improve agents’ cost efficiency through a strategy of using top-tier models for complex tasks and fast, inexpensive models for simple, repetitive ones.
■ Tencent Begins Full-Scale Research into “AI Self-Improvement”…Joins the U.S. in the RSI Race
The Chinese AI industry, which has long been pursuing the cutting-edge AI models of the United States, has recently begun venturing into research into “recursive self-improvement (RSI),” where AI improves its own performance. Following Tencent’s unveiling of “Terminal-RSI,” Baidu also introduced “AnyEvolve,” which automates AI research.
AI Times news@aitimes.com
