Annual recurring revenue
Reported January 20, 2026. ARR is an annualized revenue run rate, not profit or audited annual revenue. Company announcement ↗
BEP Research · GPU Price Watch · Public brief
RunPod GPU prices
moved higher.
Six tracked Secure Cloud configurations rose 12.6%–17.7%. A meaningful budget change at a provider reporting more than $120 million in annual recurring revenue.
Free to read and download. No sign-in required.
Our research call
Refresh the on-demand budget for these configurations. Our tracker matched the same GPU IDs and memory sizes before and after the change. This is evidence of higher advertised rates at RunPod; it does not establish market-wide scarcity, realized transaction prices or higher provider margins.
Why this provider matters
An established GPU cloud, with a broad developer audience.
Developers
Reported in the June 24 funding announcement. This is a developer count, not a count of paying customers. Company announcement ↗
Growth investment
Announced June 24, led by Summit Partners at a reported $1 billion private valuation. These figures are company-reported. Funding announcement ↗
RunPod supplies GPU infrastructure for development, training and inference. Its scale makes these increases relevant to a substantial provider’s users and worth tracking alongside other clouds. Scale alone does not tell us why prices rose, how much capacity is available or what customers actually pay.
The matched observations
Six configurations. The same price basis.
| GPU configuration | Before | After | Change |
|---|---|---|---|
| A100 PCIe 80GB | $1.59 | $1.79 | +12.6% |
| A100 SXM 80GB | $1.59 | $1.79 | +12.6% |
| H100 SXM 80GB | $3.49 | $3.99 | +14.3% |
| H200 SXM 141GB | $4.59 | $5.29 | +15.3% |
| B200 180GB | $6.79 | $7.99 | +17.7% |
| B300 288GB | $7.89 | $8.99 | +13.9% |
Before: . After: . Rechecked: . The exact effective time is unknown. Region, host CPU and RAM are unspecified in the catalog. Six is the tracked change set, not the entire catalog or a fleet census.
The budget implication
Same B200 hours.
$12,000 more at list.
At an illustrative 10,000 GPU-hours, the quoted GPU-rate component rises from $67,900 to $79,900.
Before · $6.79 × 10,000$67,900
After · $7.99 × 10,000$79,900
Increase+$12,000
Illustration only, holding GPU-hours and configuration constant. Excludes taxes, storage, networking, discounts and other charges. This is not a customer bill, forecast or measurement of cost per completed task.
For compute buyers
Check the deployment quote, then the work it buys.
Confirm the GPU variant, region, included resources, availability and contract term before adjusting a budget. Compare alternatives at the same quality and latency targets. A lower GPU-hour rate can still produce more expensive useful work.
For investors
A pricing signal that needs commercial follow-through.
Matched executable quotes or bills below list would weaken the budget inference. Broader transaction evidence and sustained productive utilization would strengthen a pricing-power inference. We make no earnings, target-price or margin revision from these six observations.
Keep the evidence
Take the brief with you.
The dated research package brings together the brief, chart and six-row price comparison for your own review.
Sources, scope and corrections
- RunPod public GraphQL catalog ↗ — historical comparisons use
data.gpuTypes[].securePrice, matched by provider GPU ID and memory. Before and after captures are dated above. - RunPod public pricing ↗ and Pods pricing documentation ↗ — checked October 12. Current pages corroborate the new quotes; they do not provide the historical baseline or exact effective time.
- RunPod ARR announcement, January 20, 2026 ↗ — source for the company’s reported revenue run rate.
- RunPod funding announcement, June 24, 2026 ↗ — source for developer count, growth investment and private valuation.
Secure Cloud and Community Cloud are different tiers. Community Cloud fields for these same six rows were unchanged across the two October 8 captures. This does not establish unchanged pricing elsewhere, bookable capacity or realized transaction prices.
This edition is frozen as of October 12, 2026. No corrections were recorded at publication. Send sourced corrections to [email protected]. Read our terms and research disclosures.
Download the source notes · Markdown ↓ · Dated source and calculation record · JSON ↓
Follow what changes next
Get the next compute signal
with the evidence behind it.
Subscribe for BEP’s AI infrastructure research. Paid membership includes the GPU Price Watch dashboard, with dated provider observations, source evidence and cost scenarios.
Need a decision-specific brief? Explore our institutional research pilot ↗