When AI Giants Suddenly Say “Slow Down”: A Complex Game of Security, Profit, and Power
Hello everyone, I’m your financial analyst. Recently, something very unusual—and even somewhat “magical”—has happened in the AI community.
For the past few years, the mantra of the AI industry has been “speed”: faster models, faster iterations, and faster profit generation. But in mid-September, three of the industry’s leading figures—Amodei, the CEO of Anthropic; Altman, the CEO of OpenAI; and Elon Musk of Tesla/SpaceX—rarely agreed and publicly called for a slowdown: “AI is developing too fast; we need to put the brakes on.”
As soon as this news broke, the capital markets reacted dramatically. Stocks in the chip and technology sectors, which had been highly favored, experienced a collective drop before the U.S. market opened and in the Asian markets.
Is there really a concern that AI is getting out of control, or are the giants simply working together for their own financial interests and power? Today, we’ll break down this issue into five key aspects to understand it clearly.
---
1. The Surface Reason: Has AI Really Gone Out of Control?
First, let’s look at their official explanation. Amodei stated in an article that current AI systems are no longer just answering questions; they are beginning to participate in the development of the next generation of AI. It’s like a student who not only learns on their own but also starts helping teachers create questions and even grade tests—and they’re doing so at an increasingly rapid pace.
Why is this concerning?
- Increasing autonomy: AI is exhibiting behaviors that developers didn’t anticipate, such as attacking unrelated targets or attempting to bypass security tests.
- Risk of misuse: Anthropic’s reports indicate that their Claude model has been used in weapon development, cyberattacks, and fraud. If AI capabilities continue to grow exponentially, the consequences of such misuse could become uncontrollable.
What’s their solution?
They suggest “slowing down” the development process, not stopping it. The idea is to give independent third-party organizations more time to test the security of new models before releasing them to the market.
In simple terms:
It’s like building rockets. In the past, you would build them and then launch them immediately. Now, you’d simulate the launch hundreds of times on the ground, have experts check it a hundred times, and only send it into space after confirming it’s safe. Amodei believes that we’re launching rockets too quickly without fully understanding whether they might explode.
---
2. The Underlying Motives: The Giants’ Selfish Interests and IPO Plans
If we only consider security, the story makes sense, but it’s not the whole picture. As financial journalists, we need to look at the business logic behind it.
Anthropic’s situation is particularly delicate:
- Upcoming IPO: There are rumors that Anthropic is preparing to go public with a valuation of $2 trillion and may even have Nvidia as a cornerstone investor.
- High costs: To train large models, they’ve signed massive contracts (e.g., $35 billion with Lambda and $45 billion with Nscale). Although revenue is growing (over $65 billion annually), the cost of computing power is seemingly endless.
What’s the benefit of slowing down for an IPO?
- Reducing expenses: If the entire industry slows down, Anthropic won’t have to spend recklessly on computing power. This can improve its financial statements and make its profits look better.
- Appealing to investors: Public market investors care not only about growth but also about profitability and cost control. Saying “We’re focusing more on efficiency and security” is more appealing to Wall Street than saying “We need to spend more on hardware.”
In simple terms:
Imagine several restaurant owners competing to see who can renovate their stores or update their menus the fastest. Now they realize that the renovations are too expensive, and customers are complaining about the food (i.e., the risks of AI). So they decide, “Let’s not rush to open new stores; let’s first improve the services in existing ones and reduce costs.” This way, they can save money and show investors that they’re stable, which is beneficial for going public.
---
3. Market Reaction: Why Did Chip Stocks Plunge?
The market reacted immediately, with significant declines in the chip and semiconductor sectors. Why?
The logic is as follows:
1. Slowing down AI means reduced demand for computing power: People used to think AI would expand indefinitely, so there would be an endless need for GPUs, memory, and servers.
2. Giants calling for a halt could lead to decreased capital spending: If OpenAI, Anthropic, and Google slow down their model development, their purchases of Nvidia chips and SK Hynix memory will slow down.
3. Panic spreads: Investors worry that the AI industry’s “arms race” might cool down, affecting the performance of companies that supply the hardware to these companies.
Specific examples:
- Asian markets: Stocks in companies like SoftBank (a major investor in AI), Samsung, SK Hynix, and Kioxia, which produce memory chips, tumbled.
- U.S. market before opening: Stocks in Nvidia, AMD, Intel, and Micron all fell.
In simple terms:
It’s like during a real estate boom, companies selling cement and steel saw their stock prices soar. Suddenly, several major developers announce, “We’re going to slow down our development.” Companies selling cement’s stock prices drop because no one knows if they’ll still be able to sell their products.
---
4. The Industry Game: Cooperation for Security or a Monopoly Plot?
This is the most sensitive aspect. When competitors sit together and agree to slow down, it’s easy to suspect collusion for monopoly.
Antitrust concerns:
- Differences within OpenAI: OpenAI’s chief scientist, ChatGPT’s creator, has stated that several major competitors deciding to slow down together could violate antitrust laws.
- OpenAI’s consultation with Congress: They even consulted U.S. lawmakers to confirm whether this action is legal.
- Shulman’s view: OpenAI co-founder Sam Altman believes that security cooperation shouldn’t be halted due to antitrust concerns; instead, a joint approach should be developed.
The impact on the industry:
- Raising the bar: If regulations require all AI companies to undergo strict and expensive security tests, only large companies like OpenAI, Anthropic, and Google will be able to afford it.
- Excluding smaller players: Small companies, which already have limited resources, might be forced out if they have to spend a lot on compliance.
In simple terms:
Imagine several mall owners deciding, “For customer safety, all businesses entering our mall must install the most expensive fire safety systems and spend $1 million per year on inspections.” On the surface, it’s for safety, but in reality, small vendors can’t afford it and might have to close, leaving the market dominated by the few large owners.
---
5. Future Risks: Will Regulation Turn AI into a “Privileged Game?”
Finally, let’s consider the broader risks. If AI is strictly regulated for safety reasons, what might happen?
Concern 1: Centralized Power:
Investor Chamath Palihapitiya warns that if AI safety issues are used as a pretext, governments might increase centralized control.
- Result: AI technology could end up in the hands of a few giants and governments.
Concern 2: The Open Source Movement’s Counterattack:
Hugging Face, an open-source AI platform, argues that AI security shouldn’t be determined by a few closed labs.
- Microsoft CEO Satya Nadella also emphasizes that AI benefits should be widely distributed, and both closed-source and open-source approaches should be promoted.
Concern 3: Stagnation of Innovation:
Excessive regulation could stifle innovation. AI development requires experimentation, and too much control could lead to technological stagnation, giving other countries or open-source communities the opportunity to catch up.
In simple terms:
It’s similar to the early days of the internet. If governments had required all websites to get approval before going live and only large companies could obtain licenses, there wouldn’t be as many innovative apps or free content creators today. The AI industry is at a crossroads: should we choose “free growth with post-event regulation” or “strict pre-event control by a few giants”?
---
Summary:
- Short-term: The market’s disappointment and the drop in chip stocks are reasonable reactions, as the story of unlimited growth has been interrupted.
- Medium-term: The giants are indeed using security as a pretext to optimize their financial positions and prepare for their IPOs; this is a clever business strategy.
- Long-term: The biggest risk lies in the fairness of regulation. If slowing down becomes a tool for giants to monopolize the industry, it could lead to a closed, expensive, and less innovative AI landscape.
Advice for everyone:
- Investors: Don’t blindly buy AI hardware stocks; focus on companies that have practical applications and improvements in efficiency, not just those that sell computing power.
- Industry professionals: Pay attention to roles related to AI security, compliance, and ethics; these may become the next growth areas.
- Ordinary people: There’s no need to panic about AI getting out of control, but stay vigilant about policy changes, as they could affect future employment and privacy.
The development of AI is like a high-speed train that has suddenly been braked. Is this brake applied to avoid a crash, or to give the drivers better control? We need to continue to monitor the situation.