CEO Column | Claude’s Parent Company Going Public Is Not Just an AI Company IPO — It Is the Start of Human Civilization Securitizing “Judgment”
Claude’s Parent Company Prepares to Go Public: What the Capital Market Is Really Pricing Is Not AI but Humanity’s Second Kind of Judgment
Many people, on seeing that Anthropic, Claude’s parent company, is preparing to go public, react first with: AI is about to be hyped yet again. But I see this not as hype but as a signal that the structure of civilization is turning.
Anthropic has confidentially submitted an IPO filing to the US SEC; foreign media report a valuation as high as US$965 billion and say annualized revenue has reached US$47 billion. At the same time, Anthropic itself has always stressed that it is an AI safety and research company whose core goal is to build reliable, interpretable and steerable AI systems. Many people read these words as technical, but I read them as commercial and very real: reliable, explainable, controllable — this is not just the language of AI technology; it is the language of future corporate governance and of future national competitiveness.
What did the capital market price in the past? It priced land, factories, equipment, brands, channels and cash flow.
Later, it priced platforms, traffic, data and algorithms.
But if Anthropic lists successfully, what it truly represents is this: the capital market beginning to price “judgment.” This is what I think is most critical. Claude is not just a chatbot; it is a kind of plug-in brain for enterprises. It can write code, read documents, analyze contracts, organize knowledge, assist decisions and enter workflows. So what the capital market is really buying is not how beautifully Claude answers today, but whether it can in the future become the “second-layer judgment system” within the systems of enterprises, governments, finance, healthcare, education, defense and scientific research. This is a very big matter — so big that many people have not yet understood it.
1. An AI Listing Means “Compute” Has Officially Become the Capital Market’s Oil
In the past, when we talked about oil, it was energy. Later, when we talked about semiconductors, it was industrial grain. Now when we talk about compute, it is no longer a technical term; it is the currency of a new era. Without compute there are no models. Without models there are no intelligent agents. Without intelligent agents there are no next-generation enterprise processes. So Anthropic’s planned listing looks, on the surface, like a model company heading to the capital market, but underneath, compute, chips, cloud, data centers, power, cooling, cybersecurity and data governance are all being re-linked into one value chain. This is also why PSF has kept building out AI compute centers. We are not buying a few GPUs just to rent them out — that would be too shallow; if it were only renting GPUs, that would be an equipment business. But if you connect the compute center to enterprise AI adoption, government cooperation, international nodes, academia-industry talent, medical applications, video production and the capital market, then it is not a single project; it is a territory.
In the Czech Republic, we see that Europe needs AI compute sovereignty.
In Japan, we see that regional cities need AI transformation and a gateway for tourism.
In Singapore, we see capital, family offices and technological sovereignty being recombined.
In Taiwan, we see that SMEs do not lack the need for AI; what they lack is an adoption system that truly runs on the shop floor.
So when I look at Anthropic, I am looking not only at its listing but at the global AI infrastructure war behind it.
2. Claude’s Real Value Is Not Chatting Better but Fitting Better into the Enterprise Floor
When many people compare AIs, they like to ask: which model is smartest? Which gives the best answers? Which codes best? These all matter, but they are not everything. What enterprises truly need is not an AI that speaks well; they need an AI that does not talk nonsense, can be traced, managed and controlled, and can enter workflows. This is where Anthropic’s narrative is so clever: it pulls AI back from “showing off” to “governance.” From “model capability” back to “enterprise trust.” From “benchmark races” back to “controllable systems.” This is very important for Taiwanese companies. The biggest mistake many Taiwanese companies make now when adopting AI is buying AI as a tool: buying accounts, software and platforms, and telling employees to use them on their own — then discovering three months later that the data is still scattered, the processes are still messy, responsibility is still unclear, the AI still answers carelessly and the boss still dares not use AI for decisions. This is not because AI is bad; it is because the company has no knowledge foundation of its own. Without a company knowledge foundation, AI is like a very smart outsider: it speaks fast but does not necessarily understand your company; it answers well but does not necessarily know your real structure of interests. So I always say: the first step in adopting AI is not buying tools but organizing your own brain. A company’s contracts, quotations, customer data, product knowledge, SOPs, meeting minutes, financial models and risk judgments are its real moat. For AI to enter a company, you do not let it improvise freely; you turn the company’s knowledge, processes, permissions, reviews and responsibilities into a system that AI can read, cite and trace. This is the real battlefield of enterprise AI.
3. Anthropic’s Listing Means AI Companies Are Moving from “Technology Companies” to “Governance Companies”
This is my own judgment: the AI companies that will be truly valuable in the future will not just be those with the strongest models, but those that can enter governance systems. What is governance? Corporate governance, data governance, risk governance, government governance, military governance, medical governance, financial governance, educational governance. Once AI enters these fields, it is no longer just a production tool; it begins to influence decisions, and wherever decisions are influenced, power is involved. Wherever power is involved, governance is necessarily involved. Anthropic placing AI safety, reliability and interpretability at the core of its narrative is in fact a contest for this position. It does not want merely to be a useful AI; it wants to be the AI that enterprises and governments dare to use — and that is very different. Useful is product value. Daring to use is institutional value. Being something you can rely on long-term is infrastructure value. What the capital market will truly give high valuations to is the third kind.
4. CEO Yang’s View: What AI Ultimately Sells Is Not Answers but the “Ability to Reduce Wrong Decisions”
Many people think AI’s value is improving efficiency; I think that says only half of it. AI’s greatest value is reducing wrong decisions. A business owner makes many judgments in a day: invest or not? Sign or not? Hire this person or not? Enter this market or not? Go with this partner or not? Where are the risks? Is cash flow enough? Are there traps in the contract? Getting these wrong once does not just mean earning a little less; sometimes it means a company working in vain for three years, sometimes a project capsizing outright, sometimes the wrong partner spoiling the whole game. So I say AI’s biggest cost is not the API fee. And AI’s greatest value is not saving you labor; AI’s greatest value is letting the company make fewer wrong decisions. What models like Claude, ChatGPT and Gemini will really compete on in the future is not merely who answers better, but who can become the second judgment system at the business owner’s side — one that never sleeps, can always organize the context and can always raise risk reminders. This is why Anthropic’s listing matters so much: it means the capital market is buying a new capability — scalable judgment.
5. Taiwan Cannot Only Be an AI Supply Chain; It Must Build AI Application Sovereignty
Taiwan is strong. Chips, servers, cooling, power supplies, racks, PCBs, contract manufacturing — all are strong. But the problem is that we cannot forever stand underneath someone else’s infrastructure. If the next stage of AI is application systems, knowledge foundations, enterprise processes and intelligent agents, then Taiwan must move up the stack. It cannot only make hardware for others; it must build its own AI application scenarios. It cannot only make equipment; it must run operations. It cannot only do supply chains; it must do system chains. This is also PSF’s direction going forward. What we want to build is not only AI compute centers, but compute centers turned into an entry point: connecting downward to chips, equipment, server rooms and energy, and upward to enterprise adoption, AI courses, an AI Lab, a video base, urban applications, medical applications, educational applications and the capital market, with international nodes in between — Taiwan, Japan, the Czech Republic and Singapore. Only then is there strategic depth. Otherwise, in this round of AI, Taiwan could easily once again become the most important yet hardest-working supplier: earning hardware money and bearing manufacturing pressure while others take the models, platforms, data and valuation. This is not because we are not strong enough; it is because our position needs to be redesigned.
6. What Anthropic Teaches PSF: Don’t Just Do Projects — Build a System the Capital Market Can Understand
The capital market does not like fragmentation; it likes structure. If you do a compute project, the market will ask how much equipment, how much gross margin, how long the payback is. But if you build an AI battle system, the market will start asking: where are your nodes? Where are your customer scenarios? Where is your data entry point? Where is your enterprise adoption methodology? Where is your education system? Where is your international cooperation? How does your cash flow replicate? That is the difference. Anthropic does not only sell Claude. It sells a vision of AI infrastructure that future enterprises and governments can trust. PSF cannot only sell a single project either; what we want to sell is: the PSF International AI Battle System. Compute is the foundation. The knowledge base is the brain. The AI Lab is R&D. The video base is content and traffic. Enterprise adoption is cash flow. International nodes are strategic position. The capital market is the amplifier. Only when these pieces are connected does a company have real value.
7. The Real Question Is Not How Much Anthropic Is Worth, but Whether We Have the Ability to Be Re-Priced
Many people, seeing a US$965 billion valuation, feel it is outrageous. I will not look at it only in terms of expensive or cheap. The capital market sometimes overheats, which is normal; but every truly major trend is called a bubble at the start. The internet was like that, electric vehicles were like that, the cloud was like that, and AI will be like that too. Bubbles burst, but infrastructure remains; speculation disappears, but a new order forms. So what I really care about is not what valuation Anthropic finally lists at, but this: can Taiwanese companies find their own position within this wave of AI re-pricing? Can PSF integrate the projects it has in hand, from single-point collaborations into one international AI system? Can CEO Yang turn his experience from business judgment into a replicable methodology? That is the point.
Conclusion: The Next Round of AI Is Not a Battle of Tools but a Battle of Position.