Captured September 19, 2026 at 1440px wide. The page continues below the crop (10,196px in total).
CoreWeave plans and prices
| Plan | Price | Billing |
|---|---|---|
| NVIDIA GB200 NVL72 | $42.00 | Per Hour |
| NVIDIA HGX B200 | $68.80 | Per Hour |
| NVIDIA RTX PRO 6000 Blackwell Server Edition (High Memory) | $20.00 | Per Hour |
| NVIDIA GH200 | $6.50 | Per Hour |
| NVIDIA HGX H100 | $49.24 | Per Hour |
| NVIDIA L40 | $10.00 | Per Hour |
| NVIDIA HGX H200 | $50.44 | Per Hour |
| NVIDIA L40S | $18.00 | Per Hour |
As printed on the page captured September 19, 2026.
How does CoreWeave pricing work?
The value metric is the instance-hour, and each row defines exactly what the hour contains. An NVIDIA HGX B200 row states 8 GPUs, 180 GB VRAM, 128 vCPUs, 2,048 GB system RAM and 61.44 TB local storage before it states $68.80 on-demand and $34.11 spot, so the buyer prices a configured node. Packaging is by purchase mode: On-Demand for burst work, Spot at roughly half the on-demand rate, and Reserved Compute Capacity at 'up to 60% discounts over our On-Demand prices for committed usage'. Storage runs a four-tier ladder at $0.06, $0.03, $0.015 and $0.0125 per GB per month, with Distributed File Storage at $0.070, and a footnote defines a GB as a gibibyte.
The motion is sales-led, and the page reads as a reference document for an infrastructure buyer. A hero promising 'Flexible, portfolio-wide pricing for AI compute, storage, networking, and more' leads into Compute, with GPU and CPU tables for North America and then Europe, a Reserved Compute Capacity paragraph, then Storage, Networking and Managed Kubernetes. Contact sales sits in the tables wherever capacity is allocated by conversation, and the close pairs 'Start building today' with Talk to an expert and a link to CoreWeave classic pricing. For a buyer sizing clusters in eight-GPU nodes on committed terms, a conversation is the natural next step, and the published rates give that conversation a starting point.
Under the hood, the page shows several billing units working together. Training capacity is billed per node-hour, while an 'Inference Single GPU Price' column, which footnote 2 applies to CoreWeave inference platform customers, bills serving work per GPU, at $1.25 for L40 up to $10.50 for GB200 NVL72. On-demand rates match across regions and spot rates vary by region, so spot tracks local spare capacity. Networking is mostly Free, Public IP Address is $4.00 per IP / month, and Direct Connect lists $1,250 to $50,000 per month, with cross connects and Equinix or Megaport port charges billed separately as pass-through costs. SUNK and the CoreWeave Kubernetes Service control plane are Free, with SUNK's CPU usage billed as compute, so the platform layer is priced through the resources it runs on.
What does CoreWeave's pricing page do well?
- Every instance row states GPU count, VRAM, vCPUs, system RAM and local storage next to the hourly rate, so the billed unit is fully specified
- On-demand and spot prices sit in adjacent columns, so the discount reads in one glance, for example $68.80 against $34.11 for HGX B200
- Storage is a four-tier ladder from Hot at $0.06/GB/mo to Archive at $0.0125/GB/mo, with a footnote defining GB as a gibibyte
- Networking prints Free for NAT Gateway, VPC, internet transfer and internal transfer, turning a cost line into a selling point
- Direct Connect list rates are published, up to $50,000 per month for 400G, with footnotes naming the third-party charges billed separately
What does CoreWeave's pricing page reveal about its pricing strategy?
- Three purchase modes, On-Demand, Spot and Reserved, are the packaging: the same hardware sold at different levels of commitment and guarantee
- A per-GPU inference rate beside per-node training rates points to two billing units for two workloads, serving and training
- Matching on-demand rates across regions with region-specific spot rates shows spot pricing tied to local spare capacity
- Reserved capacity at up to 60% off on-demand is where term, volume and timing are set per customer, the commercial room in a capacity business
- Free egress, free control planes and free core networking put the whole bill on compute and storage, the resources CoreWeave sells
Learn with Orb
- GuideMigrating from subscriptions to usage-based billing The rate card bills fully specified instance-hours, a four-tier storage ladder and networking, and Orb's guide covers billing a cloud product on compute, storage and bandwidth as customer consumption varies.
- DocsConfiguring plans: minimums and plan phases Reserved Compute Capacity offers up to 60% off on-demand for committed usage; in Orb, a dollar commitment is a minimum, and plan phases step committed spend up on a schedule.
- GuideAn expert's guide to enterprise pricing for SaaS Reserved capacity and the newest silicon run through Contact sales, where term, volume and timing are set per customer, the deal-by-deal pricing Orb's enterprise guide covers.
Questions about CoreWeave pricing
How much does CoreWeave cost?
CoreWeave's pricing starts at $6.50/instance/hr. Paying annually saves up to 60% for committed use.
What pricing model does CoreWeave use?
CoreWeave prices by usage. The page is built around talking to the sales team. Its plans are aimed mainly at enterprises.
Does CoreWeave have a free plan?
CoreWeave's pricing page shows no free plan or free trial. The lowest paid plan starts at $6.50/instance/hr.
Does CoreWeave offer an annual discount?
Yes, paying annually saves up to 60% for committed use. The listed prices start at $6.50/instance/hr. Terms for the plans priced through sales are set with the sales team.
How is CoreWeave's pricing page laid out?
The page shows its pricing and includes a comparison table. The breakdown above marks 14 points on it: 7 things it does well and 7 insights into the pricing.
Does CoreWeave offer discounts for reserved GPU capacity?
Yes. CoreWeave offers "up to 60% discounts over our On-Demand prices for committed usage" through its Reserved Compute Capacity option, arranged directly with the team. For flexible short-term work, Spot pricing is listed beside every on-demand rate. For example, the NVIDIA HGX B200 is $68.80 per hour on-demand and $34.11 per hour spot in North America.
Does CoreWeave charge egress or data transfer fees?
CoreWeave states that its storage comes with "no ingress, egress, or transfer fees" so customers can move data as their workloads demand. The networking table lists data transfer between CoreWeave and the internet, data transfer within CoreWeave, NAT Gateway and VPC as Free. A Public IP Address is $4.00 per IP / month, and Bring Your Own IP carries a $4.00 one-time setup fee per IP.
What is CoreWeave's inference single GPU price?
The North America GPU table includes an Inference Single GPU Price column, which ranges from $1.25 per hour for the NVIDIA L40 to $10.50 per hour for the NVIDIA GB200 NVL72. Other rates include $6.16 for HGX H100 and $8.60 for HGX B200. A footnote explains that this per-GPU pricing is exclusively applicable to CoreWeave inference platform customers, who can contact their account executive for more information.
From the page as captured September 19, 2026. Current prices are on coreweave.com
Chroma
$250/mo + usageChroma prints all four metered rates in the hero, Write at $2.50 per GiB, Storage at $0.33 per GiB per month, Query at $0.0075 per TiB and Network at $0.09 per GiB returned, then hands the reader a calculator that turns documents and queries into a $79 monthly total. Plans sit on top of the meter: Starter at $0 plus usage, Team at $250 a month plus usage with $100 in credits, and Enterprise on custom terms.
ClickHouse
$0.2181/compute unit/hrClickHouse opens with its model in one line, 'Pay only for the compute and storage you use, metered separately and scaled to zero when idle.', sets the cloud and region, then gives every tier the same two meters: storage flat at '$25.30 Per 1TB / mo' and compute at $0.2181, $0.2985 and $0.3903 per unit per hour. The estimator opens pre-filled with a 1 TB real-time analytics workload at '$658' a month, and an eight-question FAQ explains each line that can appear on the bill.
Confluent
~$385/moConfluent prices serverless Kafka on a composite meter, eCKU-hours plus data in and out per GB plus storage per GB-month, and closes its rate table with an 'Estimated Monthly Starting Cost' of $0, $385, $895 and $2,300 that turns an hourly unit into a budget number. Behind it sit two discount structures and an annual commitment that can be spent across the whole Confluent Cloud stack.
Databricks
From $0.069/CUDatabricks prices a whole data platform on one consumption unit, the DBU, with seven workload tiles that each print a 'Starting at' rate and link to a full SKU price list. Billing runs 'at per second granularity' on pay-as-you-go, and Committed Use Contracts trade larger commitments for discounts that can flex across clouds. The page works as the front door to a rate table, organized by the job a customer is doing.
Datadog
$15/host/moDatadog publishes a rate card per product, with every billable SKU listed in a left-hand navigator from AI Credits to Workflow Automation. Infrastructure Monitoring prices per host across three host levels: Basic, set with sales, beside Pro starting at $ 15 and Enterprise at $ 23 billed annually, with on-demand rates printed beside them and allotments of containers, custom metrics and custom events attached to each paid host. The bill is the sum of the products a customer adopts, and Basic hosts let one account price noncritical workloads at their own level.
Dremio
$0.20/DCUDremio prices its lakehouse on one compute meter, '$0.20 per DCU (Dremio Compute Unit)', and organizes the page around a deployment choice: Dremio Cloud, fully managed on AWS, or Dremio Enterprise, self-managed anywhere. Both options are 'Consumption based', so the decision is about who runs the infrastructure. Every path leads to a conversation, including a Cloud trial that comes with '$400 credit for 30 days'.