📊 Full opportunity report: The Neocloud Cartel: How the AI Industry Started Renting Compute From Itself on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
AI firms are increasingly renting compute from each other, creating a cartel led by Nvidia. This shift decouples ownership from use and concentrates power, but also introduces fragility.
In 2026, a small group of AI firms is leasing compute from each other, forming a cartel that is centered around Nvidia, which controls the majority of GPU supply and financing. This development marks a fundamental shift in how AI infrastructure is accessed and controlled, with ownership of hardware becoming decoupled from its use, and the power concentrated among a few dominant firms.
Recent reports reveal that major AI companies, including OpenAI, Anthropic, and xAI, are leasing hundreds of millions of dollars worth of GPU compute from each other and from a handful of key suppliers, notably Nvidia. This interconnected leasing network has evolved into a de facto cartel, where a small number of firms finance, lease, and control access to critical hardware, effectively creating a chokehold on AI compute resources.
Nvidia plays a central role, with estimates indicating that about 70% of AI data center costs flow to the chipmaker. Nvidia’s investments and equity stakes in firms like CoreWeave, Anthropic, and others further reinforce its control. The leasing agreements often include clauses that grant Nvidia or other landlords the ability to revoke or reprice access, making compute supply highly reconfigurable and subject to control by a few firms.
This system emerged partly due to a GPU shortage in 2024–2025, which made owning hardware less feasible for many labs and companies. As a result, renting became the dominant model, with firms increasingly leasing from each other rather than owning their own machines. This shift has led to a tightly interconnected network where the flow of money, chips, and compute capacity is circular and heavily concentrated.
The Neocloud Cartel
Almost no one racing to build AI owns the machine it runs on. They rent — increasingly from each other — and the money loops back to one chip maker that’s also an investor in nearly everyone at the table.
The cartel isn’t a conspiracy — it’s the endpoint of extreme capital intensity, real scarcity, and one dominant supplier. But the same circularity that makes it powerful makes it a fuse: each cancelled order is someone else’s missing revenue. Don’t be a price-taker at the bottom of a loop you don’t control — own your inference, keep an open-weight fallback, diversify silicon.
Why the AI Compute Cartel Concentrates Power
This development signifies a major concentration of power within the AI industry, as a small group of firms now controls the majority of compute resources through leasing agreements. Nvidia’s dominant position as a supplier and financier means it effectively controls who can access the hardware needed to train and run large AI models. This control could influence industry competition, innovation, and even governance of AI development, as access to compute becomes a gatekeeper.
However, this cartel-like structure also introduces fragility. The circular financing and leasing agreements create dependencies that could unravel if key firms face financial or operational disruptions. The power dynamics hinge on Nvidia’s continued control over supply and allocation, making the entire system susceptible to shocks if that control is challenged or if supply chains break.
NVIDIA Tesla L4 24GB PCIe Graphics ACELLERATOR HH/HL 75W GPU 900-2G193-0000-000
24GB Video Memory
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Industry Shift Toward Leasing and Centralized Control
Historically, AI companies built or owned their own hardware infrastructure. The GPU shortage of 2024–2025 shifted the industry toward a leasing model, with companies like CoreWeave, Meta, and OpenAI relying heavily on Nvidia hardware through third-party providers. The emergence of the ‘neocloud’—a hyperscaler model dedicated solely to AI compute—accelerated this trend, with firms renting GPU time rather than owning physical assets.
In 2026, the leasing ecosystem became more interconnected when firms like xAI began leasing their own supercomputers to other companies, including competitors, blurring the lines between supplier and customer. This interconnected leasing network has created a small, powerful cartel that controls a choke point in the AI industry’s infrastructure, with Nvidia at its core.
Prior to this, industry dynamics were more fragmented; now, the flow of capital and hardware is circular, with a handful of firms financing and leasing among themselves, inflating valuations and reinforcing their dominance.
“A gigawatt of AI data center capacity costs roughly $50 billion, with Nvidia capturing the majority of that in supply and financing.”
— Jensen Huang, Nvidia CEO
AI compute leasing hardware
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unclear Risks and Potential Disruptions to the Cartel
It remains uncertain how sustainable this tightly interconnected leasing network is, given its fragility. If Nvidia or a major firm faces financial difficulties, or if supply chains are disrupted, the entire system could unravel. Additionally, regulatory scrutiny or geopolitical tensions could challenge the current concentration of power. The long-term impact on industry competition and innovation is still unclear, as are potential responses from smaller or new entrants seeking to break the cartel.
Local AI on Linux in Practice: Build Private LLM Servers, GPU Workstations, Ollama Apps, Dockerized AI Services, and Self-Hosted AI Infrastructure with CUDA, ROCm, vLLM, and Open WebUI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps for Industry Regulation and Supply Chain Stability
Industry analysts expect increased scrutiny of leasing agreements and Nvidia’s dominant role, possibly prompting regulatory review. Firms may seek to diversify supply sources or develop alternative infrastructure to reduce dependency. Further, the industry could see the emergence of new competitors or technological shifts that challenge the current leasing-dominated model. Monitoring how these dynamics evolve will be critical in understanding whether the cartel persists or dissolves.
AI Hardware Engineering: Designing GPUs, TPUs, and Neural Processing Units for High-Throughput Machine Learning Workloads (AI Infrastructure, Hardware & Compiler Engineering Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How does Nvidia control the AI compute market?
Nvidia controls the majority of GPU supply and financing, holding equity stakes in key firms, and setting allocation policies that determine who gets access to hardware, effectively acting as a gatekeeper.
Why are companies leasing compute instead of owning hardware?
The GPU shortage in 2024–2025 made hardware ownership costly and impractical for many firms, leading to a shift toward leasing as the primary model for access to compute resources.
What risks does this leasing cartel pose to the industry?
The system’s reliance on a few firms and contractual control introduces fragility; disruptions or regulatory actions could threaten the stability of the entire ecosystem.
Could this concentration of control impact AI innovation?
Yes, if access to compute remains tightly controlled, it could limit competition and slow innovation from smaller or new entrants.
What might change the current industry structure?
Emergence of alternative hardware sources, regulatory interventions, or technological breakthroughs could break the current leasing and control model, dispersing power more broadly.
Source: ThorstenMeyerAI.com
