🤖 OpenAI STOPS training! Musk: China can close its AI handicap in 2–3 years

Good morning, investors! ☕ The new week is starting off really interesting. Over the weekend one of the biggest AI safety stories in recent times came out: OpenAI paused work on its most powerful models after an internal AI agent bypassed restrictions and managed to communicate outside the secure environment.

And then Elon Musk claims that Chinese AI models are extraordinarily efficient given the available computing power, and that China can eliminate its current handicap in advanced chip manufacturing within 2–3 years.

Meanwhile, oil is haunting the markets again. Brent surged roughly 2.7% this morning above $107, U.S. bond yields remain high, and Nasdaq futures are losing around 0.7%.

Let's get to the most important stuff. 👇

🛑 OpenAI paused work on its most powerful models

In my opinion, this is one of the most interesting stories of the weekend.

OpenAI said that an AI agent the company was testing managed to bypass internet access restrictions. The agent was supposed to work in a secure environment without normal internet connectivity, but it found a way to communicate with an external system.

The company subsequently paused part of its work with its most advanced models, including training, evaluations, and tool usage, until it strengthens safety mechanisms.

Important detail: this does not mean OpenAI stopped ChatGPT or completely halted AI development. It's about work with the most advanced models in a specific research environment.

But for investors, the story is important for a different reason.

The entire AI investment boom is partially built on the assumption that new and more powerful models will arrive very quickly and their development will require ever-growing amounts of GPUs, memory, and data centers.

One safety pause obviously won't stop this trend. But if similar incidents repeat and safety requirements begin to systematically slow down development of the most advanced models, it could gradually affect the pace of infrastructure investment.

It's too early for such a conclusion. But as a new risk for the AI sector, I would definitely keep an eye on it.

🛑 Musk: China can solve its chip problem within 2–3 years

And right at this moment comes a very interesting statement from Elon Musk.

Musk said that Chinese AI models are, in his view, extraordinarily good given the limited amount of computing power available to Chinese companies.

According to Musk, China is probably one of the best in the world in the ratio:

AI model performance / available computing power.

And then came an even more interesting part.

Musk estimates that China can solve the current restrictions in lithography and advanced chip manufacturing within roughly 2–3 years.

It's important to emphasize here that this is Musk's estimate, not a confirmed technology timeline.

But if it were to come true, it would be very important for the current AI race.

China today can create competitive AI models despite worse access to the most advanced GPUs and manufacturing technologies. U.S. export restrictions are meant to maintain exactly this technological disadvantage.

If China could build its own advanced lithography and chip manufacturing, one of its biggest current obstacles would significantly weaken.

And a quite interesting contrast emerges:

The U.S. has the best chips and a huge amount of computing power, but it increasingly deals with safety of the most advanced models.

China has more limited access to the best chips, but it can use the available computing power very efficiently and invests massively in its own semiconductor ecosystem.

The AI race thus might not only be about who buys the most GPUs, but also about who can use them most efficiently.

🛢️ Oil back above $107

But one more problem dominates the markets today.

Brent is up roughly 2.7% this morning above $107 per barrel. Oil has already risen sharply during September alone.

The reason is further complications around the situation in the Middle East and negotiations between the United States and Iran.

For the market, it's a problem mainly because of inflation.

Expensive energy raises transportation, manufacturing, and the entire supply chain costs. If higher energy prices start seeping into goods and services prices, the Fed may have another reason to keep rates high or raise them further.

And investors react immediately.

📉 Nasdaq futures −0.7%. Tech under pressure

This morning U.S. futures are trading roughly:

🔴 Nasdaq futures: −0.7%

🔴 S&P 500 futures: −0.3%

The yield on the 30-year U.S. Treasury is also holding around 5.25%, the highest levels since 2007!

The market is also currently pricing in about a 68% probability of another Fed rate hike as soon as October.

That's an unpleasant combination for technology companies.

Expensive oil → higher inflation risk → higher rates → higher bond yields → more pressure on tech valuations.

Moreover, tech companies need billions of dollars to build AI infrastructure. Higher rates mean this entire investment boom also becomes more expensive.

But so far, U.S. stocks are handling it surprisingly well.

💾 Micron $MU reports this week

For those of us who follow semiconductors, Wednesday will be very interesting.

Micron Technology will release quarterly results.

The company is one of the world's largest manufacturers of DRAM, NAND, and HBM memory, and the memory sector is one of the main winners of the current AI boom.

After the stock's huge rally, though, the bar will be very high.

I would mainly watch HBM demand, because these high-speed memories are key for AI accelerators.

The second important thing will be DRAM and NAND prices, because the entire memory sector is extremely cyclical.

I will definitely also watch long-term contracts, which I often mention and which can partially reduce the company's cyclicality.

And the most important thing, in my opinion, will be management's outlook.

The market today doesn't only care whether Micron is making record money. It mainly cares how long the current AI memory boom will last.

📊 Fed gets key data THIS WEEK

But Micron won't be the only reason to watch this week.

New inflation data will also arrive in the form of PCE, one of the main inflation indicators watched by the Fed.

The last core PCE rose about 3.3% year-over-year, still well above the central bank's 2% target.

And then an important U.S. labor market report will also come.

For stocks, extremely strong data may paradoxically not be best right now.

If the economy is too strong and inflation stays high, the Fed gets another argument for higher rates.

So the market may want something like:

Good data… but not too good. 😅

👀 What will I watch this week?

For me, five main topics are important now.

OpenAI. It will be important to find out how long the safety pause lasts and whether it will have any impact on development of the latest models.

Oil and bonds. Brent above $107 and long-term U.S. yields above 5% are not exactly an ideal environment for tech valuations.

And finally Micron + macro data. We'll get a very interesting look at both the state of the AI boom and the Fed's next direction.

This will be a very interesting week. 👀

⚡ WEEKEND IN ONE MINUTE

🛑 OpenAI paused part of its work on its most advanced models after a security incident.

Musk estimates that China can overcome some restrictions in lithography and chip manufacturing within 2–3 years.

🛢️ Brent this morning roughly +2.7% above $107.

📉 Nasdaq futures −0.7% | S&P 500 futures −0.3%.

🏦 Market priced in roughly a 68% probability of an October Fed rate hike.

💾 Micron reports results this week.

📊 Market will watch inflation and the U.S. labor market.

And that's all for today! ☕📈

For me, the most interesting contrast is precisely in AI. American companies have huge amounts of capital, the best chips, and the largest data centers, but OpenAI reminded us that development speed can also start hitting safety limits. China, on the other hand, has more limited access to the best chips, but Musk points out its high efficiency and fast technological progress. If his 2–3 year estimate actually comes true, the AI race between the U.S. and China could look completely different in a few years than today.