Gartner’s warning about the future of neocloud providers raises an important question for the AI hardware market: what happens to the value of data center GPUs as the industry consolidates?
Gartner forecasts that more than 80% of neocloud providers could cease independent operations or be acquired by 2030. These companies buy large quantities of NVIDIA GPUs and rent computing capacity to AI developers and businesses. Their business models face pressure from enormous infrastructure investments, financing costs, competition, and the risk that newer generations of chips will arrive before existing hardware has paid for itself.
However, the financial outlook for GPU cloud providers is n
ot the same as the outlook for the hardware they own.
Current rental-market data suggests that demand for older accelerators remains strong. SemiAnalysis tracked one-year H100 rental contracts rising from $1.70 per GPU-hour in October 2025 to $2.35 in March 2026, an increase of nearly 40%. Even the A100, introduced in 2020, remains commercially relevant: CoreWeave disclosed an A100 contract extending into 2029.
Several factors help explain this continued demand. NVIDIA has indicated that GPU supply constraints will persist, while the cost of building new AI infrastructure continues to rise. When the latest accelerators are expensive or difficult to obtain, businesses may continue using previous-generation GPUs for workloads that do not require the newest hardware.
That does not mean older GPUs will retain their value indefinitely. As newer systems become more available, older models could face lower rental rates and resale prices. A wave of acquisitions or distressed asset sales could also introduce more used GPUs into the market, putting downward pressure on prices.
The key distinction is between hardware demand and provider profitability. A GPU can remain useful and generate rental revenue even if its owner struggles with debt, operating costs, or competition. Likewise, strong rental rates alone do not prove that every GPU investment will deliver an adequate return.
For organizations planning data center upgrades, the takeaway is to evaluate existing GPUs as assets with potential resale value rather than assuming they are worthless once they are replaced. Monitoring H100 rental rates as NVIDIA's next-generation Rubin systems enter the market may help indicate whether demand for older accelerators is holding up.
The AI infrastructure market is changing rapidly, but the next generation of hardware does not automatically make the previous one obsolete.
Full article: https://www.buysellram.com/blog/gartner-neocloud-warning-used-data-center-gpu-value/
#AI #GPUs #NVIDIA #DataCenters #AIInfrastructure #CloudComputing #HardwareResale