GPU MODELS / ASCEND 910B
Ascend 910B cloud GPU pricing
3 listings across 3 providers. Cheapest published rate: $0.594/GPU/hr at Gitee AI. Sorted by per-GPU hourly price. Every row links to the provider's own published price.
WHAT IS THE ASCEND 910B
The Ascend 910B is Huawei's domestic answer to NVIDIA's China-market cards: 64 GB of HBM2e and roughly A100-class FP16 throughput on SMIC 7nm. It exists in every China sovereign-AI conversation, but the software stack is CANN/MindSpore, not CUDA - the real comparison is ecosystem cost, not FLOPS.
ARCHITECTURE
Da Vinci v2 (Ascend)
TENSOR COMPUTE
FP16 ~320 TFLOPS class (third-party compiles; Huawei does not publish a full datasheet)
INTERCONNECT
OAM, HCCS 392 GB/s
WHAT IT IS GOOD AT
Sovereign China AI training and inference
The design target - domestic-model training runs where NVIDIA supply is restricted.
Inference at A100-class scale
64 GB and ~320 TFLOPS FP16 serve 13B-70B models if the stack ports.
CUDA-native workloads
The blocker. CANN has matured but porting cost is the deciding line item.
WHO SHOULD NOT RENT THE ASCEND 910B
Anything that must run CUDA - this is not an NVIDIA GPU and no amount of spec-sheet comparison changes that.
Buyers who need published datasheet certainty: key figures are third-party compiled, and variants (B1/B2/B3) differ.
THE ASCEND 910B MARKET, BY THE NUMBERS
3 providers publish 3 Ascend 910B configurations on the panel today. Published per-GPU hourly rates run from $0.594 (Gitee AI) to $5.71 (CTYun), a 10x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $0.95/GPU/hr. 1 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.
ASCEND 910B PRICE TREND, DAILY PANEL SNAPSHOTS
Across every Ascend 910B row on the panel, the daily floor moved from $0.950 to $0.594/GPU/hr over 10 days of tracking (-37.4%). Snapshots run daily; the full history is public in the repo.
ASCEND 910B PROVIDER NOTES
Gitee AI · 1 config · 1 regionSets the panel floor for this model.
from $0.594/hrCloudGPU · 1 config · 1 in stock1 of 1 configs reported in stock at last check.
from $0.950/hrCTYun · 1 config · 1 regionPublished catalog rates.
from $5.714/hrBUYING ASCEND 910B: WHAT THE PANEL SAYS
At the current floor, one Ascend 910B running around the clock costs about $434 a month (730 hours of on-demand arithmetic). Compare the alternatives below before committing - the same budget often buys more than one chip class.
ASCEND 910B VS THE ALTERNATIVES
The two China-market answers head to head: the H20 has more memory (96 GB) and bandwidth (4.0 TB/s) with the CUDA ecosystem; the 910B has roughly double the FP16 compute (320 vs 148 TFLOPS class) with a domestic stack. Compute-bound buyers lean Ascend; ecosystem-bound buyers lean H20.
The 910B's FP16 class (~320 TFLOPS) sits near the A100's 312 TFLOPS, with less memory bandwidth (1.6 vs 2.0 TB/s). On paper it is A100-tier; in practice the comparison is CANN versus CUDA, not silicon.
Compare live rates: Ascend 910B vs H20 · Ascend 910B vs A100
ASCEND 910B PRICING FAQ
What is the cheapest Ascend 910B cloud GPU right now?
The cheapest published Ascend 910B rate on the panel is $0.594 per GPU-hour at Gitee AI (Ascend 910B x 1, China). Every price links to the provider's own published rate.
How much does a Ascend 910B cost per hour?
Across 3 providers tracked today, published per-GPU hourly rates for Ascend 910B run from $0.59 to $5.71, with a median of $0.95. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent Ascend 910B GPUs?
3 providers publish 3 distinct Ascend 910B configurations on Compute Cafe. 1 of those rows were reported in stock by the provider's live endpoint at last observation.
Are Ascend 910B prices going up or down?
Over the 10 days Compute Cafe has tracked the panel, the Ascend 910B floor is falling: $0.950 to $0.594/GPU/hr (-37.4%). Daily snapshots are public in the repo, so the trend is auditable.
What does a Ascend 910B cost per month?
At the current floor of $0.594/GPU/hr, a Ascend 910B running 24/7 costs about $434 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
How fast is the Ascend 910B really?
Third-party compiles put FP16 around 320 TFLOPS with 64 GB HBM2e at 1.6 TB/s - A100-class on paper. Huawei does not publish a clean datasheet and variants differ, so treat exact figures as approximate and verify against the specific deployment.
Should non-China buyers care about the 910B?
Only as market context: it anchors the domestic floor for China AI compute pricing. It does not appear on international rental panels, and running it requires the CANN/MindSpore stack.
See also alternatives to the Ascend 910B · cheapest GPU-hour per model · inference chips · training chips · by location
Gitee AI
CloudGPU