Price gap between domestic and overseas for the same GPU

“Median price gap” = (domestic median − international median) ÷ international median. Negative values indicate domestic prices are lower; unit: RMB/hour.

The table below is not a price comparison—it isa factual snapshot of supply-side availability: These GPU models are currently listed for sale on platforms in only one region. China-exclusive models and domestically produced GPUs appear only domestically; latest-generation flagships and select competing models appear only internationally.

① Only same-day automatically collected quotes: Sourced from public pages or APIs of platforms (144 entries included in this round), with quotes older than 12 hours, expired, or flagged as questionable excluded.
② Platforms not yet integrated into auto-collection are excluded: This round includes Paratera and Verda (Finland), whose prices are manually verified only and not mixed with real-time pricing.
③ At least 2 platforms must offer the configuration to enter the comparison table, preventing a single platform’s price from representing an entire market.
④ Median—not average—is used: Price distributions exhibit strong long tails; medians remain unaffected by outliers.
⑤ All prices converted uniformly to RMB per hour: Only displayed prices are compared—bandwidth, storage, image disks, and coupon discounts excluded.
⑥ Quote-on-request platforms excluded: Azure, AWS, Google Cloud, etc., which do not publish unit prices, are listed for specs only—not priced.

Some card models are listed by only 2 platforms — making the median highly sensitive to individual pricing adjustments. When observing large gaps, please refer back to the card model page to verify specific platforms and billing methods. This page does not compare network quality, availability, GPU interconnect bandwidth, or shared-GPU configurations; nor does it include spot/preemptible instances or monthly discounts. Electricity, bandwidth, and compliance costs differ between domestic and overseas regions — here we answer only “what is the current displayed price for this exact GPU model”, not procurement advice.