The AI Gold Rush could be floating the US economy, but its current trajectory is not sustainable, according to Deutsche Bank.
New research notes from German lenders warn that AI capital expenditures have reached extremely high levels, and extraordinary heights that have prevented the US from falling into a recession.
Deutsche Bank is not the only bank to realize the huge impact AI is having on the economy. Kobeissi’s letter was posted a chart by Arch Global Economies. This shows that software and technology investments are contributing to the company, with real GDP growth exceeding 1% points for the first time in history. It also surpassed the previous peak that reached during the 1998 dot-com bubble.
“This is unprecedented… the AI boom is driving economic growth.”
But by racing ahead of actual productivity gains, Deutsche Bank sees Storm Clouds on the horizon.
Deutsche Bank cites the growth of CAPEX fuels rather than software output
The scale seems daunting. Goldman Sachs estimates that global AI-related CAPEX reached $368 billion between early 2023 and August 2025. Most of this money went into physical infrastructure, including building data centers, upgrading power supplies, installing high-quality equipment.
However, the actual output from AI software remains limited to the promised leap of productivity and efficiency. In fact, Deutsche Bank points out that removing tech-driven spending means that actual GDP growth in the US hovered about 0% in 2024 and 2025. Is there a translation? Without data centers, the economy would already be in a recession.
According to Deutsche Bank, to continue to contribute fresh points to GDP, you will need to accelerate the “parabolic” quarter after the quarter. Such an infinite upward gradient is mathematically unlikely, if not impossible.
Instead, the current AI boom looks like a sprint. At an unsustainable speed, the structure is frontloaded and destined to slow the infrastructure build-out plateau. The risk is not limited to GDP, as tech stocks are responsible for about half of the S&P 500’s profit this year. They expand directly into financial markets.
A shortage of $800 billion
Consultant company Bain & Co. adds more fuel to skeptic fires. Their estimates suggest that by 2030 the AI sector will need $2 trillion a year to rely on demand for computing power. But even considering increased efficiency and reduced costs, the world is still staring at a revenue shortage of $800 billion.
That gap raises an offensive question: Who is putting the bill on its feet? If demand for AI calculations is not lined with revenue, the industry could face calculations of overpower and squeezed margins that are creepy reminiscent of the dot-com era.
However, there is a more measured outlook. Goldman Sachs believes that AI productivity will ultimately increase, with US GDP increasing by about 0.4 percentage points per year in the short term and about 1.5% in the long term. It’s not a “parabove” but can provide a softer landing than a dramatic AI bust.
Deutsche Bank said that “balanced” readings are actually coming in productivity improvements, not yet at a pace that justifies today’s runaway spending. In other words, AI may change the economy quite well, but the timeline is not in line with the ongoing fanatic buildings.
For now, AI Capex has kept construction workers busy, with utility investments and stock markets still buoyant. However, long-term questions remain. Is this foundation durable or is there a risk that the world will build a house with trillion dollar cards?
It is mentioned in this article
