A new price comparison normalizes on-demand H100 GPU rental rates to dollars per GPU per hour across AWS, Azure and neocloud providers such as RunPod, grouping only identical chip variants to avoid mixing SXM, PCIe and NVL models. It finds RunPod Community pricing at $2.69/hour versus Azure's ND96isr_H100_v5 at $12.29/hour, a 4.57x spread that translates to roughly $1,964 versus $8,972 for a month of continuous use on one GPU.
github.com
· 2026-11-09
According to the Financial Times, Oracle has agreed to lease about 100,000 advanced AI chips to Tencent across Southeast Asian data centers over five years, in a deal valued near $7 billion with roughly 30% paid upfront. Neither company has confirmed the arrangement, and the specific chip model involved has not been disclosed, though the FT notes such leases are permitted under current US export rules since the hardware is unavailable in China.
tomshardware.com
· 2026-10-05
A new Ornn Data paper argues that NVIDIA's older Ampere-generation GPUs, particularly the A100, remain economically useful far longer than assumed because open-weight models can be run cheaply on them. The analysis found that self-hosted open-weight models like gpt-oss-120b produce output more cheaply on A100 chips than on newer H100 hardware, and that A100 rental prices have stayed unusually stable over a five-year contract period compared to Hopper and Blackwell chips.
data.ornn.com
· 2026-09-22
Nvidia-backed startup Starcloud says it will fly dedicated Bitcoin mining hardware on its second satellite, Starcloud-2, targeting an early 2027 launch. CEO Philip Johnston argues that ASIC chips are far cheaper per kilowatt than GPUs, and that abundant orbital solar power could make space mining economically appealing, following the company's earlier Starcloud-1 test that put an Nvidia H100 GPU into orbit.
techspot.com
· 2026-09-22
GoKawiil