[News] NVIDIA H100 Rentals in China Reportedly Up 20%–30% as Token Demand Surges, H-Series Also Rises

As AI-driven compute demand surges, supply tightness is no longer limited to GPUs or foundry capacity. According to Chinese media outlet Calian Press, token economics is emerging as a key industry theme, with sources noting that NVIDIA H100 rental prices have climbed roughly 20%–30% since October last year.

Amid market chatter that one-year leasing contracts for H100 may have surged around 40% from October 2025 to March 2026, sources cited by Calian Press suggest the actual price trajectory is somewhat more moderate.

The report, citing a source from Berlin Cloud Servers, notes that typical H100 rental rates were previously in the RMB 50,000- 60,000 range, with occasional lows of RMB 40,000- 50,000. Including rack costs, current pricing has now climbed to roughly RMB 80,000- 90,000, the report adds.

Notably, the report points out that the rental price increases are not limited to H100 alone, noting that pricing across the broader H-series GPU lineup has also seen a general upward trend. H200 system purchased for RMB 2.45 million in February 2025 is now valued at RMB 3 million after more than a year in operation, the report adds.

The report attributes the rise in server pricing to broad-based cost inflation across key components, including memory, storage, GPUs, CPUs, and optical modules, with memory seeing the most pronounced increases but far from the only pressure point.

China’s surging token usage is also a key driver behind the spike. Calian Press, citing China’s National Data Administration, reports daily token calls jumping from 100 billion in early 2024 to over 140 trillion by March 2026—an increase of more than 1,000x—fueled by large-scale image and video generation via tools like Seedance and Nano Banana, as well as high-concurrency, multi-step workflow agents such as OpenClaw, which require continuous iteration.

Cloud Price Hikes Spread Across China

Against this backdrop, cloud vendors in China have started to raise prices. Security Times reports that following its March 11 move to end free trials for select models and adjust pricing, Tencent Cloud announced on April 9 that it will revise list prices for AI compute, container services, and Elastic MapReduce (EMR), effective May 9, 2026. Prices for all three categories will increase by 5%.

Prior to Tencent Cloud’s price adjustment, Alibaba Could announced on March 18, 2026, price increases across select services, with the new rates taking effect on April 18, the report notes, adding that pricing for compute card-related offerings, including T-Head Zhenwu 810E, will rise by 5% to 34%, while CPFS will see a 30% increase.

Interestingly, the report points out that not all cloud providers are raising prices. On March 18, China’s JD Cloud issued a statement noting that while global AI industry growth is driving explosive demand for compute and storage infrastructure—and pushing up hardware procurement and supply chain costs, with some peers already raising prices—it is committing to keep its core product lineup unchanged in price.

Read more

• [Insights] AI-Driven Memory Boom Ripples Through Hardware Markets, From Price Hikes to Product Delays
• [News] Chinese AI Models Reportedly Hit ~15% Global Share in Nov. 2025, Fueled by DeepSeek Open-Source Push

(Photo credit: NVIDIA)