Nvidia is making one thing very clear: the hyperscalers aren’t done opening their wallets. The company says Amazon, Microsoft, Alphabet, Meta, and their peers will continue pouring money into AI hardware for the foreseeable future, with one important caveat. The spending continues “as long as the bond market allows.”

The trillion-dollar runway

Amazon, Microsoft, Alphabet, and Meta are expected to spend a combined $710B on AI infrastructure capex in 2024 alone. Analysts project that Nvidia’s data-center build-out market could reach $1 trillion annually by 2028, representing roughly a threefold increase in just three years. Global AI infrastructure spending nearly doubled year-on-year to $47.4B in the first half of 2024, and forecasts suggest it will exceed $200B annually by 2028.

The bond market condition

By tying hyperscaler spending to bond market conditions, Nvidia is acknowledging that this spending isn’t happening in a vacuum. Hyperscalers fund their massive capital expenditure programs through a combination of operating cash flow, retained earnings, and debt issuance. When bond markets tighten significantly, borrowing costs rise, and the bar for new projects gets higher.

The depreciation problem nobody wants to talk about

The economic life of AI GPUs is approximately three years. Each new generation of chips, from Nvidia’s Hopper to Blackwell to whatever comes next, renders the previous generation less competitive for cutting-edge AI workloads. This creates what some analysts have described as a potential “depreciation crisis” for hyperscalers, forcing them to recycle older hardware prematurely and reinvest in the latest generation.

What this means for investors

The depreciation cycle is a key variable worth watching. If hyperscalers start feeling the financial strain of replacing hardware every three years at increasingly large scale, they may push back on pricing, seek alternative suppliers more aggressively, or develop more of their own custom silicon. Amazon’s Trainium chips and Google’s TPUs are already examples of this diversification strategy in action.