I want to highlight this analysis that mirrors my own comments on the show. From Business Insider, OpenAI has signed deals worth a staggering $1 trillion with major technology firms in 2025, marking a significant acceleration in the race for computing power to support its artificial intelligence infrastructure. High-profile partnerships include agreements with Advanced Micro Devices, which amount to $270 billion, Nvidia at $500 billion, Oracle at $300 billion, and CoreWeave at $22 billion. Despite the impressive revenue projections, analysts have raised concerns about the gap between these commitments and OpenAI’s actual cash flow. For instance, financial analysts from DA Davidson noted that OpenAI may struggle to meet these obligations, raising questions about the sustainability of its aggressive expansion strategy. The company has projected a cash burn of $155 billion, an increase of $80 billion from previous estimates, indicating the immense financial pressure it faces in scaling its operations.
Why do we care?
So, OpenAI’s got a trillion dollars in deals lined up — Nvidia, AMD, Oracle, CoreWeave — sounds massive, right? But let’s be real. That’s not money coming in, that’s money they owe. These are compute commitments — basically, reservations for GPU and cloud power — and it shows how wild this AI arms race has gotten. It’s also entirely circular, which is a bubble in motion.
Here’s the problem: they’re burning $155 billion a year. You can’t scale that forever. This isn’t a business model; it’s a capital sink.
For the rest of us in IT services? This matters because those costs are coming downstream. If you’re running AI tools or helping clients deploy them, your compute bill’s going up too. The smart play is to focus on efficiency — smaller models, smarter workloads, and helping customers understand what their AI costs to run.
AI’s new constraint isn’t innovation. It’s power and price. Whoever helps clients manage that — wins.