Something Strange Is Happening in AI.

If you told me on Friday that the biggest AI leaders are gonna come together and tell the world to “slow down on AI”, I would be like “no way”.

But just a few hours ago, Anthropic CEO Dario Amodei published an article warning about how quickly AI is advancing and why the industry may need to slow down.

Anthropic, the developers of Claude, is one of the companies leading the AI race. When its own CEO starts publicly warning about the technology he is building, people are naturally going to pay attention.

Now coming from him alone is already significant.

But things took an interesting turn when competitors Sam Altman (OpenAI) and Elon Musk (xAI) came out and support the same concern.

I mean these are some of the people who have spent years competing to build more powerful models one after another. Competition is still hot to see who comes out on top.

And now almost at the same time, they are telling us the race may be moving too quickly… very unusual in my opinion.

This news is still fresh out of the oven, and what happens when the market opens on Monday is anyone's guess.

Some investors believe AI stocks could sell off on Monday. Others are wondering whether this is the first sign that the AI boom is beginning to slow.

I actually had another article planned for this week.

But after seeing all of this unfold, I decided to push that article back and write this one first. So let’s dive right in.

My first thought after digesting it all was not, "AI is finished".

It was, "Why are all of them saying this now?"

Now, the concern is not that ChatGPT will suddenly wake up tomorrow and take over the world.

The fear is about what comes next - Recursive Self-Improvement, or RSI.

In simple language, it is where an AI helps create a stronger version of itself, which then helps create an even stronger one. An infinite feedback loop.

If that process moves faster than humans can understand or control, we may not have enough time to respond (*cues Terminator music*)

This also comes after several AI researchers resigned, which made people question the timing even more.

Anthropic researcher Jacob Coxon resigned after accusing Anthropic and OpenAI of racing towards self-improving superintelligence and "gambling with our lives".

Another researcher, Joe Benton, left for independent AI evaluator METR because he believes competition makes it almost impossible for one company to slow down alone.

Anthropic alignment researcher Evan Hubinger went further. He personally estimates a greater than 10% chance that AI could cause human extinction within the next decade.

If you ask me, I would not make an investment decision based on that number alone. What matters to me is what the industry does next.

Will companies stop training models? Do governments impose hard limits? Or do the frontier labs simply take longer to test and release them?

Building advanced AI models is incredibly expensive. It requires huge amounts of computing power, electricity and engineering talent. At the same time, each new model is expected to be much better than the last.

OpenAI and Anthropic are spending enormous amounts of money to build and run these models.

And as they move closer towards an IPO, that matters.

Once a company goes public, it has to open up its books. Investors will finally see more clearly how much money is coming in, how much is going out and whether the business model actually makes sense.

Could this slowdown simply be a way to buy time?

More time to improve their financials. More time to show investors a clearer path towards profitability before they have to reveal their numbers publicly.

In fact, Sam Altman just confirmed that OpenAI IPO will not happen in 2026. Anthropic, meanwhile, has reportedly been preparing for a possible IPO as early as October 2026. Whether that goes ahead is anyone’s guess.

But from an investor's perspective, I think this financial angle is far more important than all the noise around it.

There is one major reason I do not believe the AI race will slow down for long.

China.

The United States cannot afford to give up AI sovereignty while Chinese companies are progressing at breakneck speed.

If American companies halt for too long while China continues building, the race does not become safer. It simply changes who may win it.

Palantir CEO Alex Karp put it clearly:

Even Dario's own explanation makes the distinction clear:

So I expect the models to keep improving and the infrastructure buildout to continue.

America simply cannot afford to abandon the AI race.

As usual, I expect the headlines to be painted in the most frightening way first.

So yes, I would not be surprised by a knee-jerk reaction on Monday. Fear normally gets priced faster than nuance.

But one red day would not prove that the AI thesis is broken.

Investors need to separate three things:

Model releases may slow. Companies could spend more time testing their most advanced systems.

AI adoption can keep growing. Businesses can continue using existing models to write software, analyze data and automate work.

Infrastructure spending may continue. Training, inference, cybersecurity, safety evaluations and competition with China still require enormous computing capacity.

This is where my view differs from the most bearish reactions.

Slowing the release of the next model is not the same as cancelling the data centers already being built.

Safety testing itself requires computing power. Cheaper and more efficient models may also attract far more users, just as cheaper internet bandwidth eventually led to much greater total usage.

That is why my long-term AI infrastructure thesis has not changed.

What would change my mind?

Big Tech cutting its AI spending, major chip-order cancellations, data-center projects being cancelled or frontier labs reducing their long-term computing commitments.

One frightening weekend and one red market session are mostly noise.

“AI is a bubble, so sell every AI stock”

“AI will change the world, so buy anything connected to it”

Markets are rarely that straightforward.

My view is that this proposed slowdown is a ruse to buy time, not a sign that America is giving up the AI race.

OpenAI and Anthropic gain time to bring costs down, improve their financials and prepare for the public markets. The United States keeps moving forward instead of surrendering its AI lead to China. Model releases may be paced more carefully, but training and technical development continue.

Could AI stocks react negatively in the short term? Yes.

But the bigger question is whether the companies funding this race begin cancelling the infrastructure behind it.

Until that happens, this looks like a change in how AI is released, not the end of the AI buildout.

And as always -

Patience builds wealth,
Bjorn

For more real-time market updates and insights, join over 900 investors in my Telegram community

Follow my personal investing journey on Instagram

For full breakdowns and long-form investing videos, subscribe to my YouTube