🎙️ Anthropic 營收超車 OpenAI？Google 用百億股票買晶片的背後盤算｜AI 日報 Podcast

Opening

Welcome to "Mark's Tech Insights," hosted by Muyan! Today's most shocking news in the tech world is definitely Anthropic's annualized revenue officially overtaking OpenAI, as they prepare for an October IPO at a whopping $2 trillion valuation. Additionally, Google has surprisingly used a bizarre "buy chips, get stock" contract to procure tens of billions of dollars in chips from Marvell. What brilliant business calculation lies behind this? I'll tell you all about it in just a moment!

Today's Top Stories

1. Anthropic's Annualized Revenue Surpasses $65 Billion, Overtaking OpenAI to Prepare for October IPO

Source: Bloomberg (https://www.bloomberg.com/news/articles/2026-08-17/anthropic-revenue-run-rate-surpasses-65-billion-ahead-of-ipo）
Summary: AI unicorn Anthropic has reported stellar news, with its annualized revenue recently surging to an incredible $65 b i l l i o n, g r o w i n g m o r e t h a n s e v e n f o l d i n j u s t a s i n g l e y e a r . E v e n m o r e s u r p r i s i n g l y, t h e i r c u r r e n t q u a r t e r l y a d j u s t e d o p e r a t i n g p r o f i t h a s s u c c e s s f u l l y t u r n e d p o s i t i v e, a n d t h i s r e v e n u e r u n r a t e h a s a l r e a d y l e f t c o m p e t i t o r O p e n A I^{'} s$ 40 billion in the dust. Currently, giants like Morgan Stanley are co-underwriting their upcoming October IPO, which is targeting a $2 trillion valuation.
Most Surprising Point: Not only has Anthropic's revenue grown sevenfold, but its revenue run rate has actually surpassed the former leader, OpenAI, and its operating profit has already turned positive.
Taiwan Perspective: This indicates that the enterprise AI market is highly lucrative. Taiwan's server ODMs and cooling supply chains must collaborate even more closely with the cloud giants backing Anthropic, such as AWS, to capture this wave of explosive growth.
Key Discussion Points:
1. How did Anthropic quietly overtake OpenAI in the foreign investment and enterprise sectors?
2. If the $2 trillion valuation becomes a reality, what kind of capital crowding-out effect will it have on the upcoming AI investment market?
Script Suggestions: You might think OpenAI is still the undisputed leader, but today's data is absolutely jaw-dropping! Anthropic has quietly relied on stable enterprise services to drive its annualized revenue to $65 billion, and they've even turned an operating profit—which is practically a miracle in the cash-burning AI sector. This tells us that rather than flashy marketing, focusing on secure, stable enterprise-grade solutions is the real way to monetize traffic. Our supply chain partners in Taiwan need to read the wind correctly; the arms race between these two camps is only going to get more white-hot from here!

2. Google and Marvell Sign $12.2 Billion Chip Deal, Pioneering a "Buy Chips, Get Equity" Model

Source: CNBC (https://www.cnbc.com/2026/08/19/marvell-google-ai-chips.html）
Summary: To develop custom TPU chips, Google has partnered with Marvell for co-design, utilizing an extremely rare payment structure. Google has received Marvell stock warrants valued at up to $12.2 b i l l i o n . U n d e r t h e a g r e e m e n t, e v e r y t i m e G o o g l e p u r c h a s e s$ 500 million worth of chips, it unlocks the next tranche of shares. The contract runs through fiscal year 2033 and is projected to bring up to $120 billion in revenue for Marvell.
Most Surprising Point: A customer buying chips while simultaneously acquiring the supplier's equity based on procurement milestones is an incredibly rare and deeply integrated business model.
Taiwan Perspective: In the past, Google's custom chips relied heavily on Broadcom. Now that Marvell has made a strong entry, TSMC—which handles the actual manufacturing in Taiwan—remains the ultimate winner regardless of who comes out on top, as all orders will ultimately flow to Taiwanese wafer foundries.
Key Discussion Points:
1. How will a "buy more, unlock more equity" contract change the competitive rules of the future chip design services industry?
2. Broadcom's stock price tumbled as a result; does this indicate that Google is consciously decentralizing its supply chain?
Script Suggestions: Google's move this time is incredibly clever—it's basically "I'll buy your products, and I'll become your shareholder too!" This strategy of tying stock warrants to procurement milestones essentially locks the interests of both companies together. Marvell gets the revenue, and Google gets the stock appreciation. For our semiconductor industry in Taiwan, this is undoubtedly great news. Whether Broadcom or Marvell snatches up this lucrative business, the most critical advanced processes and advanced packaging behind it still rely on TSMC. Taiwan's position in the AI silicon ecosystem remains rock solid.

3. Beijing Robot Olympics Debuts: Chinese Humanoid Robot Runs 100m in 9.39 Seconds, Breaking Human World Record

Source: CGTN (https://news.cgtn.com/news/2026-08-23/Second-edition-of-World-Humanoid-Robot-Games-gets-underway-in-Beijing-1PPXjuMlQqc/p.html）
Summary: The 2nd World Humanoid Robot Games kicked off with great fanfare in Beijing, with the number of participating teams and robots surging threefold compared to last year. During the competition, a Chinese-made humanoid robot completed the 100-meter dash in 9.39 seconds, shattering the human world record, while the high jump record was pushed to an astonishing 2.88 meters. However, pre-competition training footage also emerged showing a robot crashing into a wall at full speed and snapping in half.
Most Surprising Point: Not only did the robot surpass elite human sprinters in speed, but the absurd pre-competition training video of a robot crashing into a wall at full speed and breaking in half also went viral.
Taiwan Perspective: The breakthrough in humanoid robot athletic performance requires high-density precision electrical machinery and motor control. This is precisely the massive business opportunity that Taiwanese precision machinery and industrial PC manufacturers can tap into.
Key Discussion Points:
1. Does the rapid evolution of robot athletic performance mean that their practical application and entry into factories and homes will be significantly accelerated?
2. With 96% of the participating teams coming from China, what does this reflect about the current state of the industry ecosystem and government subsidies?
Script Suggestions: Just imagine a metal-skeleton robot sprinting down a track, actually running faster than the Jamaican lightning bolt, Usain Bolt—it's just so surreal! But the funny thing is, in the pre-competition training videos circulating online, there's a robot that couldn't brake in time, crashed straight into a wall at full speed, and snapped right in half. It was both shocking and hilarious to watch. This shows that while hardware limits have already surpassed humans, robots still have a long way to go in terms of algorithms and environmental awareness. Taiwan has strong capabilities in motor control and sensor modules, so we absolutely cannot afford to miss out on this wave of robotics.

4. OpenAI Launches ChatGPT for Teens, Algorithmic Misclassifications Trigger Collective Meltdown Among Adults

Source: Axios (https://www.axios.com/2026/08/18/openai-chatgpt-for-teens）
Summary: OpenAI has launched a dedicated version of ChatGPT for teens aged 13 to 17, which automatically filters out sensitive conversations related to self-harm, romance, and sex. However, the system uses an "age prediction" algorithm that judges age based on users' chat topics, word choices, and usage hours. This has led to a large number of adult users being misclassified as teens and locked out by the system.
Most Surprising Point: Many adults were flagged as minors simply because their chatting tone sounded too young or they had irregular sleep schedules, forcing them to upload ID for age verification to unlock their accounts.
Taiwan Perspective: This demonstrates that behavior-based AI classification algorithms still suffer from significant bias. When developing local AI applications in Taiwan, we should pay closer attention to privacy and the customer complaints and regulatory risks brought about by algorithmic misclassifications.
Key Discussion Points:
1. Does predicting user age based on linguistic style and daily routines violate personal privacy?
2. In the face of collective algorithmic misclassifications, what kind of more convenient appeal and verification channels should platforms provide?
Script Suggestions: This news is honestly both hilarious and frustrating. Many working professionals might just be chatting with ChatGPT late at night in a more relaxed tone, or asking some naive questions, only to wake up the next day and find themselves classified as "teens" by the system—even getting blocked from discussing romantic topics! This shows us that using AI as a "hall monitor" isn't that easy. While protecting teens is a good intention, labeling users based on behavioral traits is not only inaccurate but also causes major headaches for users. It also reminds us that AI system design still needs to retain a touch of human warmth, rather than relying on a one-size-fits-all approach.

5. Latest Pew Survey: Over Half of Americans Are Concerned About AI, with Adoption and Favorability Inversely Correlated

Source: TechCrunch (https://techcrunch.com/2026/08/19/ai-was-supposed-to-win-people-over-by-now-it-hasnt/）
Summary: According to the latest poll released by the Pew Research Center, 52% of Americans feel more concerned than excited about AI being integrated into daily life, a significant increase from 37% in 2021. Historically, whether it was personal computers or the internet, favorability rose during the early stages of adoption. AI is unique in showing a bizarre phenomenon where higher usage correlates with deeper resentment.
Most Surprising Point: Unlike all past disruptive technologies, AI is the only one where social anxiety and resentment have steadily deepened the more the public uses it.
Taiwan Perspective: Taiwanese society generally has a high acceptance of new technology. However, as AI integrates deeper into workplaces and replaces certain jobs, worker anxiety is also rising. When promoting AI transformation, governments and enterprises must simultaneously prepare supporting measures and retraining programs.
Key Discussion Points:
1. Why does the public fear AI more the more they use it? Is it due to deepfakes, misinformation, or the fear of unemployment?
2. Are tech giants blindly pursuing technological breakthroughs while ignoring the importance of building trust with the general public?
Script Suggestions: This survey result is a massive wake-up call for Silicon Valley! In the past, whether it was smartphones or the internet, people loved them more the more they used them. Why is it that with artificial intelligence, people are getting more scared the more they use it? I think it's because this time, AI is challenging humanity's core "intelligence" and "creativity." Combined with deepfake scams and the potential wave of unemployment, people are feeling a very real threat. Tech companies can't just hide in their ivory towers declaring how great their technology is. If they can't address the public's concerns about safety and trust, this AI revolution might eventually face massive social resistance.

6. OpenAI Hosts First Luxury Influencer Trip: Beekeeping Itinerary Contrasts Ironically with Hundred-Billion-Dollar Data Center Plans

Source: TechCrunch (https://techcrunch.com/2026/08/03/influencers-draw-backlash-for-attending-openais-first-luxury-trip/）
Summary: OpenAI hosted its first influencer brand event at a luxury resort in New York costing $2, 000 a n i g h t, i n v i t i n g i n f l u e n c e r s t o e x p e r i e n c e n a t u r e a n d e v e n g o b e e k e e p i n g t o g e t h e r . H o w e v e r, t h i s P R e v e n t d r e w o v e r w h e l m i n g c r i t i c i s m o n l i n e, w i t h n e t i z e n s a c c u s i n g i t o f o u t r i g h t " g r e e n w a s h i n g " b e c a u s e t h e c o m p a n y i s s i m u l t a n e o u s l y p l a n n i n g a$ 500 billion investment to build extremely energy-intensive data centers in Ohio.
Most Surprising Point: Promoting green sustainability and beekeeping at a luxury cabin on one hand, while undertaking unprecedentedly energy-consuming infrastructure on the other—this stark contrast was slammed by netizens as hypocritical.
Taiwan Perspective: Data center energy consumption has become a global pain point. Taiwan similarly faces a tug-of-war between tech industry development and power supply. How to utilize green energy and improve energy efficiency is an urgent priority for Taiwanese supply chains.
Key Discussion Points:
1. How can tech giants avoid accusations of "greenwashing" during PR campaigns and achieve true consistency between their words and actions?
2. In the face of controversies over AI's high energy consumption, what substantive green solutions do tech companies have on a technical level besides marketing tactics?
Script Suggestions: This PR event was a total trainwreck! OpenAI wanted to shape a friendly, eco-friendly image by taking influencers to a forest cabin for beekeeping, but netizens are sharp-eyed. Everyone immediately connected it to their plans to drop $500 billion on super data centers in Ohio—the power consumption of which is astronomical! This stark contrast of guzzling electricity on one hand while teaching people to love nature on the other is exactly why netizens mocked it as "dystopian" greenwashing. It seems that while pursuing massive computing power, tech giants really need to face the big question of energy head-on, rather than just brushing it off with PR packaging.

Closing

That's all for today's episode of "Mark's Tech Insights"! Today we talked about Anthropic's revenue miracle, Google's clever chip-buying strategy, and those fast-running but wall-crashing robots. AI technology is developing faster than imagined, but how to win public trust is clearly the biggest homework for tech giants next. I'm Muyan, see you next time, bye-bye!