Will the AI bubble burst before AI kills us?

Two themes have dominated the talks about AI in the last 2 weeks. The first has to do with the financing of AI companies and datacenters. The other with the warning of some techbro’s that AI might wipe out mankind before the end of the decade. Although I wil definitely not pose myself as an expert, after having read a lot about this and listened to some in-depth podcasts I would like to share some of my thoughts.

Will the AI bubble burst?

In the past software companies like Alphabet, Meta and Amazon were generating enormous amounts of cash. sitting on these piles of cash every investment they wanted to do could be easily paid using their own money. Basically they had so much cash that they didn’t know what to do with it. Hence the share buybacks and other tricks to enrich the owners. But now with huge AI datacenter investments even these companies have to lend money. And with higher interest rates and their lending need of 800 billion usd a year they are competing with the US government lending need of 2.000 billion usd. Some of the lending is circular money (Nvidia lending money to datacenter companies for buying Nvidia chips) which makes the risks even bigger.

Some would say lending a lot of money and paying interest is no issues as long as the corporations keep growing and make profits. But have companies actually found out how to make enough money to pay off these enormous debts. Are people willing to pay enough for AI? And are companies seeing enough productivity gain from AI to allow for big spending. All of this is still a big question mark. If enough revenue will come and when it will come. It all contributes to the risk of the bubble bursting.

The US is investing 23x the amount of money China is investing in AI. But the Chinese models are only slightly behind the US models. And powerful enough for most every day cases. Chances are price erosion starts, especially on the consumer side even before people start thinking about paying for AI.

Almost all economic growth can currently be attributed to AI companies. So if this bubble burst the impact will be huge. Not only for the tech companies that will see their value shrink. But also for everyone that has invested in these companies. Banks, pension funds and individuals. And not just the US, but globally.

Will AI kill us?

Not sure about the motivation of the techbro’s to call out the dangers of AI. Are they looking to protect their interest and exclude others? Or are they afraid of companies and investors turning their back to AI because of the risks? Maybe even avoiding future legal claims may play a role.

Although the risk of Recurring Self Improvement or RSI (AI systems building their own AI systems) is no reality yet, the potential destructive powers are definitely there.

The biggest risk is in my opinion the infrastructure cybersecurity aspect. Bad actors using AI to hack banks, or physical infrastructures like energy, railway or watermanagement. Or the power of AI used for autonomous weapons. The hack of Hugging Face, although being done by AI agents specifically trained for cybersecurity hacks shows how AI agents can take more power than they have been given.

Another risk is about the environment. Imagine the water and power consumption of a trillion dollar of datacenters. The actual consumption, but also electricity and water shortages for individuals living in the areas where datacenters are being built.

The concentration of power is a risk that a limited number of people will benefit from AI while others will suffer. The techbro’s will get rich while workers lose their jobs.

What’s next?

So what’s next? Regulation would be best. Governments standing up and regulate AI. Create a governance model, audit & control. But chances that will happen are small. With US and China in an arms race these two are not going to put their feet on the brake. And while the EU has regulation, the influence is small.

In that sense the AI bubble bursting and an economic crises limiting the amount of money poured into AI might be the necessary foot on the brake to save the world from AI disasters.