Google (GOOGL.US) DeepMind Executive Warns: AI Revenue Can't Support $700 Billion in Capital Expenses, but RSI Will Disrupt Everything

The Zhitong Finance App learned that Google (GOOGL.US) DeepMind executive Jasjit Saihon said that the technology industry's huge investment in capital expenditure related to artificial intelligence is a prelude to the next stage of AI.

The next stage of AI is called “recursive self-improvement” (recursive self-improvement, RSI for short), which essentially means that AI can automatically create a better version of itself. The concept has been viewed as part of General Artificial Intelligence (AGI) for years, but the tech industry, and the business community in particular, has begun to adopt the term RSI in recent weeks.

Sejon said that revenue from AI “cannot currently sustain our ongoing capital expenditure.” However, Seyhon added that it “seems unwise” to be bearish on the tech industry's current and future technology, and notes that the prototype of RSI has already emerged.

He added that this shouldn't come as a surprise to anyone because “steam engines were used back then to make the next generation of steam engines.”

Despite this, Saihon, who joined Google from the Bridgewater Fund in April of this year, did point out that there may be an “AI air pocket (AI air pocket, or fault period), where expenses have already occurred, but revenue has not been shown.” He estimated that after spending more than $400 billion in 2025, the construction scale is expected to exceed 700 billion US dollars in 2026, compared to projects larger than the Apollo lunar landing program, the Manhattan Project, and the Internet construction in the 1960s.

Major hyperscale companies, including Google, Amazon (AMZN.US), Microsoft (MSFT.US), and Meta Platforms (META.US), have stated many times that they expect 2027 capital expenditure to be higher than 2026.

Later last month, Google raised its 2026 capital expenditure forecast again. Currently, it is expected to spend between $19.5 billion and $205 billion, higher than the previous forecast of $180 billion to $190 billion.