Deutsche Bank says AI spending may be the factor propping up the whole economy, but there are big questions about what comes next

Business Insider ^ | 10/10/2025 | Samuel O'Brient

[SeekAndFind]

• Deutsche Bank analyst George Saravelos said AI spending is offsetting other economic headwinds.
• He recently said that without the huge capex from Big Tech, a recession would be more likely.
• However, he also said there are key questions about what happens when the spending spree ends.

AI spending has grown to be so enormous that it might be a driving force for the entire economy, not just an engine for stock-market gains.

In a note on Tuesday, George Saravelos, Deutsche's global head of FX research, argued that if US tech companies hadn't dramatically ramped AI capex this year, the economy would be in a recession or on the edge of one.

His view comes a day after Nvidia announced a $100 billion investment in OpenAI as part of a partnership that will help it build and scale data centers using Nvidia's hardware.

"The good news to all of this is that the AI super-cycle may be helping mute the negative demand (tariffs) and supply (immigration) shocks hitting the US economy right now," Saravelos wrote. "It may not be an exaggeration to write that NVIDIA — the key supplier of capital goods for the AI investment cycle — is currently carrying the weight of US economic growth."

However, he also raised questions about the future of such spending, and what it could mean once the colossal capex in recent years begins to wane.

"The bad news is that in order for the tech cycle to continue contributing to GDP growth, capital investment needs to remain parabolic. This is highly unlikely."

That raises some difficult questions for investors to consider as they prepare for the economic shifts of 2026, and pencil in AI as a driver of both economic growth and stock market returns.

---

TOPICS: Business/Economy; Computers/Internet; Society
KEYWORDS: ai; consumers; economy; spending

[SeekAndFind]

THE ALL IMPORTANT QUESTION:

Growth in 2025 is being driven by the buildout of the underlying AI infrastructure, but there will come a day when those projects are completed.

"Once the factories have been built, will the productivity gains from AI take over?" Saravelos asked. "And how globally disseminated will those benefits be as opposed to the location of the factories themselves?"

[SeekAndFind]

This year, AI spending has added more to the US GDP than consumer spending itself. But the growth isn’t coming from revolutionary AI products changing how we work — it’s from a frantic buildout of data centers and chips in the hope that such products will arrive someday.

This distinction matters because once the data centers are built and the chips are installed, what happens if the promised productivity gains never arrive?

[Raycpa]

If spending wasn’t going to AI, the assumption in this chart is business would not spend on development of any kind. That’s an absurd assumption.

[RoosterRedux]

Bfl

[Jonty30]

Humans win.

[RoosterRedux]

Growth in 2025 is being driven by the buildout of the underlying AI infrastructure, but there will come a day when those projects are completed.

"Once the factories have been built, will the productivity gains from AI take over?" Saravelos asked. "And how globally disseminated will those benefits be as opposed to the location of the factories themselves?"

Saravelos misunderstands AI.

Data centers aren’t static “factories” with a finish line. They’re in perpetual upgrade mode.

The “factory buildout then plateau” logic misreads the nature of AI infrastructure. These aren’t fixed capital goods but self-renewing cognitive engines that evolve with each new model and application.

[Openurmind]

“This distinction matters because once the data centers are built and the chips are installed, what happens if the promised productivity gains never arrive?”

That is an excellent discussion. Because it will not. Already the costs are far outweighing the revenue. And this is still not even taking into account hidden variables. One of the things not being discussed is the maintenance costs to even keep them working. It will be double what they are expecting. Those electronics are not “install and forget” or reliable.

The integrity of the chips and memory degrade just like Solid State Memory does. And AI puts such a load on them the life expectancy is reduced by 50%. They are touting a life expectancy of ten years for these electronics when the true reality will only be three to five years. So in the long term their overhead is actually going to be double what they are selling to everyone.

It is already a sinking ship with too many holes to plug... It is a huge mistake to invest so much of our economy into this sinking ship.