Citadel Securities, the market-making powerhouse run by Ken Griffin, projects that more than $500 billion in new public and private debt will be issued by 2028 to fund artificial intelligence chips and the sprawling data center campuses that house them.
AI capital expenditures are projected to hit roughly $600 billion in 2026, a 50% jump from the approximately $400 billion spent in 2025. The cash burn has gotten so severe that hyperscalers are hitting the debt markets at a pace that would make a leveraged buyout firm blush.
The debt machine is already running
During September and October of 2025 alone, hyperscalers issued about $75 billion in bonds and loans earmarked for AI data center construction. US data center debt hit $25.4 billion in 2025, a 112% year-over-year increase.
Morgan Stanley estimates the private credit market for AI data centers could balloon to around $800 billion by 2028. That sits within a broader financing gap of $1.5 trillion against an expected $2.9 trillion in global data center capital expenditures.
MediaTek, the Taiwanese semiconductor company, has approved a $5 billion financing package targeting AI chip production capacity.
Why crypto investors should pay attention
The core issue is resource competition. Centralized AI data centers and decentralized computing networks, including crypto mining operations, rely on many of the same hardware components. When half a trillion dollars in new capital floods into GPU procurement over the next few years, it creates enormous upward pressure on prices and downward pressure on availability for everyone else.
For projects in the decentralized compute space, companies like Render Network and Akash building marketplaces for distributed GPU power, higher hardware costs make it more expensive to onboard new supply. But they also make decentralized alternatives more attractive to users who can’t secure allocations from the major cloud providers.
What this means for markets
For crypto-native investors, the competitive landscape for mining economics deserves close scrutiny. Bitcoin miners have already been navigating tighter margins after the April 2024 halving. If GPU and general semiconductor costs climb further due to AI demand, proof-of-work operations face additional headwinds on the hardware side precisely when block rewards are already diminished.
As Morgan Stanley’s $800 billion estimate suggests, a massive new asset class is forming around AI infrastructure debt. Some of that capital might otherwise have found its way into digital asset strategies, venture funds, or crypto-adjacent investments.
Traders monitoring the intersection of traditional finance and crypto should keep an eye on GPU pricing indexes, hyperscaler earnings calls discussing capex guidance, and the corporate bond market for signs that the borrowing pace is accelerating or decelerating.