🎙️ AI 日報 Podcast — IPO 大戰、Agentic AI、中美競爭

Opening Remarks

Hello everyone, and welcome to "Mark's Tech Insights"! I'm Mark. Wow, the AI world has been hit with a series of bombshell news today. It feels like the rules of the entire industry are about to be rewritten. We'll be diving into the IPO battle between the two giants, OpenAI and Anthropic, whose valuations are skyrocketing. At the same time, Google has shockingly announced the end of the search experience we've known for over two decades! Without further ado, let's jump right into today's topics.

Today's Top Stories

1. OpenAI Confidentially Files S-1, Targeting a September 2026 IPO with a Nearly Trillion-Dollar Valuation

Source: OpenTools AI (Original article)
Summary: The long-rumored OpenAI IPO is finally in motion! They have confidentially filed an S-1 with the U.S. Securities and Exchange Commission, aiming for a public listing this September with a valuation approaching one trillion dollars. That number is just staggering.
Taiwanese Perspective: For investors and the tech industry in Taiwan, this is a landmark event. Besides watching to see if hardware supply chains like TSMC will benefit, it's more important to consider how Taiwanese software startups can find their niche in this "behemoth-level" capital game.
Discussion Points:
- What kind of revenue projections support a one-trillion-dollar valuation? Can the market truly sustain it?
- How will the transition from a non-profit to a profit-driven public company affect OpenAI's mission to "benefit all of humanity"?
- After the IPO, how will the relationship between Microsoft and OpenAI change?
Script Suggestion: "Hey everyone, what does a trillion dollars even mean? That's about 2.5 times the market cap of TSMC. An AI company founded just a few years ago has a chance to reach this scale. It just goes to show that the wave AI is creating is even crazier than we imagined. This also means OpenAI must prove to Wall Street that it can not only build models but also make money. We've seen their recent moves, like launching an ad platform, all paving the way for this trillion-dollar valuation. But it also makes you worry: when the pressure to turn a profit overrides everything else, will the idealistic OpenAI we once knew still exist? This is definitely the biggest thing to watch in the tech world in the second half of this year."

2. Anthropic Follows Close Behind, Targeting an October IPO with a $900 Billion Valuation

Source: Tech Journal (Original article)
Summary: OpenAI's biggest rival, Anthropic, is not to be outdone. They are planning an IPO in October with a target valuation of a whopping $900 b i l l i o n . E v e n m o r e a s t o n i s h i n g, t h e y p r o j e c t t h e i r Q 2 r e v e n u e t o e x p l o d e t o$ 10.9 billion, up from $4.8 billion in Q1—a terrifying growth rate.
Taiwanese Perspective: Taiwan's server, cooling, and chip distribution companies must be ecstatic. Two AI giants rushing to raise funds via IPO means they'll need more firepower (i.e., computing power) for their arms race, which is a huge boon for Taiwan's hardware supply chain.
Discussion Points:
- Anthropic's 130% quarter-over-quarter revenue growth—how strong does this indicate enterprise demand for large language models is?
- They mentioned a "60% chance of AI beginning recursive self-improvement by 2028." What are the technical implications and risks behind this statement?
- Anthropic has always championed "AI safety." Will the financial pressure of being a public company force them to compromise on this front?
Script Suggestion: "No sooner did OpenAI make its move than Anthropic followed suit. This war for the AI throne has escalated from technology to the capital markets. What strikes me most is their revenue projection—jumping from $4.8 b i l l i o n t o$ 10.9 billion in a single quarter. That's not a startup's growth curve; that's a money-printing machine's growth curve! This explains why Wall Street is willing to offer such a high valuation. But what I find deeply unsettling is their executives publicly predicting that AI is on the verge of 'self-evolution.' It sounds like science fiction, but coming from the people who understand AI best, perhaps we should take the potential risks and regulatory issues of AI more seriously."

Source: CNBC (Original article)
Summary: Just as everyone is cheering for the West Coast AI giants, data shows that AI models from China are quietly capturing the market through a "price war." On the API routing platform OpenRouter, traffic from models by Chinese companies like MiniMax and DeepSeek now accounts for 60% of total usage, and their prices are only 1/10th to 1/20th of their American counterparts.
Taiwanese Perspective: For the vast number of developers and small to medium-sized enterprises (SMEs) in Taiwan, this is definitely good news. More high-cost-performance model options can significantly reduce the cost of developing AI applications. However, this could also impact Taiwanese companies that act as distributors for high-priced European and American models.
Discussion Points:
- Will "price disruption" become the main theme in the AI field in the coming years?
- How can America's high-cost, high-performance models compete with China's low-price, "good enough" model strategy?
- What is the impact on the global AI landscape? Will it lead to a decoupling of the high-end and low-to-mid-range markets?
Script Suggestion: "Just as OpenAI and Anthropic are preparing to go public with astronomical valuations, this CNBC report is like a bucket of cold water. It reveals a harsh reality: the market might not care how cutting-edge your model is, but rather how cheap it is. If you're a developer and the cost of an API call is 10 times different, which would you choose? This is what Chinese models are doing—using incredibly disruptive pricing to rapidly gain market share. This has led analysts to worry whether that nearly trillion-dollar valuation is built on an overly optimistic 'winner-take-all' assumption. This US-China AI war seems to be not just about technology, but also about whose business model is more grounded and practical."

4. Google I/O Bombshell: Farewell to the "Ten Blue Links," a Full Embrace of AI Agents

Source: Google Blog (Original article)
Summary: Google dropped a bombshell at this year's I/O conference, announcing the biggest change to search in the company's history: the traditional "ten blue links" page will become a thing of the past, replaced by AI Agents that can autonomously browse the web and synthesize information. They also unveiled a series of new models, including Gemini 3.5 Flash and Gemini Omni, as well as a personal AI assistant, Gemini Spark.
Taiwanese Perspective: For everyone in Taiwan involved in SEO, content marketing, and digital advertising, this is nothing short of a devastating blow. Two decades of accumulated knowledge and strategies could be wiped out overnight. Everyone will have to relearn how to be "seen" in the world of AI Agents.
Discussion Points:
- Is the saying "SEO is dead" finally coming true this time? Where does the future lie for content creators?
- When traffic is intercepted by AI Agents, how will websites' advertising business models survive?
- How will personalized AI Agents (like Gemini Spark) change the way we interact with the digital world?
Script Suggestion: "I have a lot of friends in marketing who are losing sleep over this news. The 'Googling' we're used to might no longer mean getting a list of links to click through. Instead, Google might just send an AI assistant to read everything on the internet for you and directly generate a comprehensive report, complete with text and images. It sounds convenient, but for everyone who relies on website traffic for their livelihood, it's practically the apocalypse. This means Google is transforming from a 'traffic cop' into a 'destination.' In the future, the target of our optimization efforts might not be the search engine, but Google's AI Agent itself. This is a whole new ball game."

5. Anthropic's Claude Mythos Preview Autonomously Discovers Thousands of Zero-Day Vulnerabilities

Source: Dark Reading (Original article)
Summary: Anthropic has once again demonstrated its astonishing technical prowess. Its latest model, Claude Mythos, autonomously identified thousands of previously unknown "zero-day" vulnerabilities across all major operating systems and browsers, all without human guidance. However, they have not released this model publicly. Instead, they initiated "Project Glasswing" to provide it to cybersecurity professionals for priority patching.
Taiwanese Perspective: Taiwan has a very strong cybersecurity community and industry. This technology is a double-edged sword. On one hand, it can be used to train stronger AI defense systems. On the other, it reveals the potential for AI to be used in devastating cyberattacks, posing an even greater challenge to the cybersecurity industry.
Discussion Points:
- When AI can be both the "strongest lock" and the "strongest key," how will the balance of cyber offense and defense be disrupted?
- Anthropic chose "responsible disclosure" rather than direct release. Will this become the new normal for releasing cutting-edge AI capabilities in the future?
- How should governments regulate AI technologies with potential weapon-grade capabilities?
Script Suggestion: "This piece of news honestly gives me the creeps. A 'zero-day vulnerability' is a flaw unknown even to the software company itself—a hacker's favorite weapon. And now, an AI can find thousands of them as easily as breathing. I admire Anthropic's responsible approach this time, letting the defenders patch the holes first instead of just publishing the nuclear launch codes for all to see. But this raises a serious question: what if the next model with this capability is developed by someone with malicious intent? This shows that AI's capabilities have far surpassed chatting or drawing pictures; it's becoming a powerful tool that can impact national security."

6. OpenAI Launches Self-Serve Ad Platform in ChatGPT, Aiming for $100 Billion in Ad Revenue by 2030

Source: Build Fast With AI (Original article)
Summary: To justify its trillion-dollar valuation, OpenAI is accelerating its commercialization efforts. They have officially launched a self-serve ad management platform within ChatGPT, removing the previous $50,000 minimum spend and allowing small and medium-sized businesses to easily place "native ads" in AI conversations.
Taiwanese Perspective: For e-commerce businesses, app developers, or any merchant in Taiwan looking to reach new customers, this is a new and potentially massive traffic channel. Just like getting in early on Facebook or Google ads a decade ago, early adopters have the chance to enjoy a low-cost traffic dividend.
Discussion Points:
- What will AI-native advertising look like? Will it be more effective than traditional keyword ads?
- Will inserting ads into conversations severely impact the ChatGPT user experience?
- How will OpenAI balance using user conversation data for targeted advertising with protecting user privacy?
Script Suggestion: "Earlier, we were just asking how OpenAI would justify its trillion-dollar valuation, and here's the answer: advertising. And we're not talking about traditional banner ads, but 'AI-native ads.' Imagine asking ChatGPT to help plan a family weekend trip, and while suggesting an itinerary, it also recommends a family-friendly restaurant or a car rental service. That's an AI-native ad. It's scarier than keyword ads because the AI understands your full conversational context and intent. By removing the minimum spend, they're opening the floodgates to millions of SMEs worldwide. This could create a brand new advertising empire that rivals Google and Meta."

Closing Paragraph

Today's news was an absolute information overload. From the capital frenzy of AI giants and the paradigm shift in search engines to the US-China AI price war, every single story has the potential to change our digital lives in the coming years. It seems that AI development in 2026 will only get faster and more exciting. Thanks for tuning in today. This is "Mark's Tech Insights," and we'll see you next time!