The RTX 5090 average hourly rate across 28 reporting marketplaces rose 145.9% week-over-week, from $0.39 to $0.97, the largest single-SKU percentage move AIMC tracked during the May 18–24 window. The move was not isolated: nine of the ten largest weekly movers were consumer-grade GeForce and workstation SKUs, with the RTX 5060 Ti (+141.2%), RTX 5070 Ti (+109.3%), and RTX 3070 (+96.3%) all posting near-triple-digit gains from low absolute bases. According to AIMC observations across 26,789 total data points and 161 distinct SKUs, the week's structural story was a broad compression of the spread between entry-tier and mid-tier consumer GPU rental rates, while datacenter segments — H100 averaging $1.00–$10.00 and A100 spanning $0.38–$4.21 across 18 marketplaces — showed comparatively muted movement. What follows examines each segment in turn.
This Week in Numbers
- 28 authorized marketplace partners contributed pricing data this week. Full list at aiminingco.com/data-partners.
- Largest week-over-week decliner: L40S at -10.5% ($1.75 → $1.57 average across covered marketplaces).
- Largest week-over-week gainer: RTX 5090 at +145.9% ($0.39 → $0.97 average across covered marketplaces).
Headline Movements
| GPU | Last Week Avg | This Week Avg | Change |
|---|---|---|---|
| RTX 5090 | $0.39 | $0.97 | +145.9% |
| RTX 5060 Ti | $0.05 | $0.13 | +141.2% |
| RTX 5070 Ti | $0.12 | $0.25 | +109.3% |
| RTX 3070 | $0.03 | $0.07 | +96.3% |
| RTX 4080 | $0.09 | $0.17 | +87.5% |
H100 This Week
| SKU | Marketplace | Min | Avg | Max | Observations |
|---|---|---|---|---|---|
| H100 NVL | ionet | $1.00 | $1.00 | $1.00 | 7 |
| H100 NVL | hyperstack | $1.95 | $1.95 | $1.95 | 7 |
| H100 NVL | massedcompute | $2.58 | $2.86 | $3.11 | 32 |
| H100 NVL | runpod | $2.59 | $2.94 | $3.19 | 52 |
| H100 NVL | aime | $3.79 | $4.13 | $4.43 | 21 |
| H100 NVL | vast | $2.67 | $4.82 | $11.76 | 13 |
| H100 PCIe | ionet | $1.00 | $1.00 | $1.00 | 14 |
| H100 PCIe | dataoorts | $0.72 | $1.23 | $1.74 | 14 |
| H100 PCIe | hyperstack | $1.90 | $1.90 | $1.90 | 21 |
| H100 PCIe | massedcompute | $1.98 | $2.34 | $2.40 | 14 |
| H100 PCIe | runpod | $1.99 | $2.46 | $2.89 | 19 |
| H100 PCIe | latitude | $3.37 | $3.37 | $3.37 | 7 |
| H100 PCIe | ovhcloud | $2.79 | $4.27 | $15.00 | 231 |
| H100 SXM | dataoorts | $0.88 | $1.40 | $1.92 | 14 |
| H100 SXM | akash | $1.48 | $1.78 | $2.56 | 11 |
| H100 SXM | vectorlay | $1.80 | $1.80 | $1.80 | 11 |
| H100 SXM | ionet | $2.09 | $2.09 | $2.09 | 7 |
| H100 SXM | verda | $2.29 | $2.29 | $2.29 | 35 |
| H100 SXM | hyperstack | $2.40 | $2.40 | $2.40 | 14 |
| H100 SXM | massedcompute | $2.75 | $2.75 | $2.75 | 7 |
| H100 SXM | synpixcloud | $2.78 | $2.78 | $2.79 | 35 |
| H100 SXM | runpod | $2.69 | $3.01 | $3.29 | 169 |
| H100 SXM | cyfuture | $3.30 | $3.37 | $3.44 | 28 |
| H100 SXM | qubrid | $3.83 | $3.83 | $3.83 | 7 |
| H100 SXM | cirrascale | $4.28 | $4.28 | $4.28 | 7 |
| H100 SXM | vast | $1.47 | $7.37 | $27.32 | 166 |
| H100 SXM | huggingface | $10.00 | $10.00 | $10.00 | 21 |
B200 This Week
| SKU | Marketplace | Min | Avg | Max | Observations |
|---|---|---|---|---|---|
| B200 SXM | hyperstack | $3.50 | $3.50 | $3.50 | 7 |
| B200 SXM | verda | $4.89 | $4.89 | $4.89 | 28 |
| B200 SXM | ionet | $4.99 | $4.99 | $4.99 | 7 |
| B200 SXM | qubrid | $5.63 | $5.63 | $5.63 | 7 |
| B200 SXM | runpod | $5.49 | $5.86 | $5.98 | 52 |
| B200 SXM | cirrascale | $5.99 | $5.99 | $5.99 | 7 |
L40S This Week
| SKU | Marketplace | Min | Avg | Max | Observations |
|---|---|---|---|---|---|
| L40S | clore.ai | $0.13 | $0.21 | $0.50 | 32 |
| L40S | tensordock | $0.48 | $0.48 | $0.48 | 4 |
| L40S | cloudrift | $0.63 | $0.63 | $0.63 | 7 |
| L40S | akash | $0.80 | $0.80 | $0.80 | 6 |
| L40S | runpod | $0.79 | $0.83 | $0.86 | 77 |
| L40S | massedcompute | $0.88 | $0.88 | $0.88 | 30 |
| L40S | ionet | $0.89 | $0.89 | $0.89 | 7 |
| L40S | verda | $0.91 | $0.91 | $0.91 | 28 |
| L40S | packet | $0.92 | $0.92 | $0.92 | 7 |
| L40S | cyfuture | $1.24 | $1.27 | $1.30 | 28 |
| L40S | vast | $0.43 | $1.81 | $8.03 | 57 |
| L40S | ovhcloud | $1.46 | $2.11 | $7.45 | 252 |
| L40S | huggingface | $1.80 | $2.27 | $2.94 | 21 |
| L40S | qubrid | $2.48 | $2.48 | $2.48 | 7 |
A100 This Week
| SKU | Marketplace | Min | Avg | Max | Observations |
|---|---|---|---|---|---|
| A100 PCIe 40GB | dataoorts | $0.41 | $0.41 | $0.41 | 7 |
| A100 PCIe 40GB | ionet | $1.29 | $1.29 | $1.29 | 7 |
| A100 PCIe 40GB | aime | $1.21 | $1.32 | $1.42 | 21 |
| A100 PCIe 40GB | qubrid | $1.50 | $1.50 | $1.50 | 7 |
| A100 PCIe 80GB | dataoorts | $0.59 | $0.92 | $1.24 | 14 |
| A100 PCIe 80GB | tensordock | $1.25 | $1.25 | $1.25 | 3 |
| A100 PCIe 80GB | massedcompute | $1.25 | $1.25 | $1.25 | 12 |
| A100 PCIe 80GB | verda | $1.29 | $1.29 | $1.29 | 28 |
| A100 PCIe 80GB | runpod | $1.19 | $1.32 | $1.39 | 36 |
| A100 PCIe 80GB | hyperstack | $1.35 | $1.38 | $1.40 | 28 |
| A100 PCIe 80GB | ionet | $1.39 | $1.39 | $1.39 | 7 |
| A100 PCIe 80GB | packet | $1.43 | $1.43 | $1.43 | 7 |
| A100 PCIe 80GB | aime | $2.24 | $2.44 | $2.61 | 21 |
| A100 PCIe 80GB | ovhcloud | $2.84 | $4.21 | $14.73 | 231 |
| A100 SXM | tensordock | $0.85 | $0.86 | $0.88 | 6 |
| A100 SXM | vectorlay | $0.80 | $0.90 | $1.00 | 4 |
| A100 SXM | synpixcloud | $0.70 | $1.17 | $1.70 | 20 |
| A100 SXM | massedcompute | $1.28 | $1.28 | $1.28 | 9 |
| A100 SXM | runpod | $1.39 | $1.44 | $1.49 | 28 |
| A100 SXM 40GB | synpixcloud | $0.70 | $0.71 | $0.72 | 25 |
| A100 SXM 40GB | verda | $0.72 | $0.72 | $0.72 | 14 |
| A100 SXM 40GB | ionet | $0.79 | $0.79 | $0.79 | 7 |
| A100 SXM 40GB | cirrascale | $2.87 | $2.87 | $2.87 | 4 |
| A100 SXM 80GB | clore.ai | $0.38 | $0.38 | $0.38 | 1 |
| A100 SXM 80GB | tensordock | $0.85 | $0.86 | $0.88 | 14 |
| A100 SXM 80GB | vectorlay | $0.80 | $0.94 | $1.00 | 7 |
| A100 SXM 80GB | dataoorts | $0.64 | $1.05 | $1.46 | 14 |
| A100 SXM 80GB | akash | $1.20 | $1.21 | $1.22 | 7 |
| A100 SXM 80GB | massedcompute | $1.02 | $1.23 | $1.28 | 16 |
| A100 SXM 80GB | ionet | $1.39 | $1.39 | $1.39 | 7 |
| A100 SXM 80GB | runpod | $1.39 | $1.46 | $1.49 | 63 |
| A100 SXM 80GB | hyperstack | $1.60 | $1.60 | $1.60 | 7 |
| A100 SXM 80GB | synpixcloud | $1.38 | $1.63 | $1.70 | 25 |
| A100 SXM 80GB | qubrid | $1.69 | $1.69 | $1.69 | 7 |
| A100 SXM 80GB | cyfuture | $1.98 | $2.03 | $2.07 | 28 |
| A100 SXM 80GB | huggingface | $2.50 | $2.80 | $3.60 | 77 |
| A100 SXM 80GB | cirrascale | $3.25 | $3.34 | $3.42 | 14 |
Marketplace Activity
This week's data spans 28 marketplaces with 26,789 total listing observations across 161 distinct GPU SKUs.
H100 pricing ranged from $1.00/hour at ionet to $10.00/hour at huggingface across 17 marketplaces.
B200 pricing ranged from $3.50/hour at hyperstack to $5.99/hour at cirrascale across 6 marketplaces.
L40S pricing ranged from $0.21/hour at clore.ai to $2.48/hour at qubrid across 14 marketplaces.
A100 pricing ranged from $0.38/hour at clore.ai to $4.21/hour at ovhcloud across 18 marketplaces.
What This Means
The most striking feature of this week's data is the breadth and magnitude of consumer-tier GPU price movement. AIMC's data shows that across 26,789 observations collected between May 18–24, 2026, ten GPU models posted week-over-week average price increases of 50% or more, led by the RTX 5090 jumping from $0.39 to $0.97 per GPU-hour — a 145.9% single-week move — and the RTX 3080 rising a comparatively modest 50.3%, from $0.07 to $0.10. Ai Mining Co. tracked this pattern across 28 reporting marketplaces, and the simultaneity across both current-generation cards like the RTX 5070 Ti and older models like the RTX 3070 and RTX 3090 is notable: it suggests the pricing movement was not isolated to a single SKU or a single platform, but reflected something broader in spot-market conditions during this window.
What makes the market structure picture more complex is how differently the enterprise tier behaved. The H100 remained priced between $1.00 and $10.00 per GPU-hour across 17 marketplaces, and the A100 spanned $0.38 to $4.21 across 18 — wide ranges that reflect persistent tier fragmentation rather than unified pricing pressure. The L40S showed a similarly wide band of $0.21 to $2.48 across 14 marketplaces. This spread suggests that hyperscaler and neocloud tiers are not moving in lockstep with the consumer-GPU spot market, and that the two segments may be responding to different demand signals entirely.
With only one week of data, Ai Mining Co. cannot distinguish whether the consumer-tier surge represents a durable shift in spot demand, a temporary liquidity event on specific platforms, or a data artifact from changing marketplace coverage. The pattern warrants close monitoring against next week's observations before drawing structural conclusions.
Methodology & Coverage
This weekly report is generated from pricing data collected daily across all authorized data partners listed at aiminingco.com/data-partners. Coverage includes on-demand rates published by each marketplace. Spot pricing, reserved instance rates, and enterprise contracts are excluded.
For complete methodology, data collection schedule, and quality controls, see aiminingco.com/methodology.
This report contains observable market data only. No estimates, no forecasts, no buy/sell recommendations.