I recently found out about Ed Zitron and read his article "Am I meant to be impressed?" that was deep dive into the problems with AI surrounding the 5 hyperscalers.
That got me thinking if there would ever be a point of 'No Return' where the 5 AI hyperscalers spend so much capex on AI that eventually becoming profitable from AI-driven revenue becomes mathematically impossible.
Google Gemini says yes there is and this point will occur sometime between late 2026 and 2028.
The numbers that Gemini used was the $1.5T AI capex total and that generally a data center needs to generate around $2 for every $1 of capex to turn a profit. So $3T needed in revenue to justify that. In total AI-driven revenue so far is $130B. So there is a shortfall of $2.87T.
Experts call this the 'AI revenue gap' and it has increased from $600B in 2024 when the term was first coined.
If the gap is real (even if my numbers are off) will that be the thing that pops the AI bubble?