Train ML models via state of the art clusters
Our A100 distributed training clusters leverage a rail-optimized design using NVIDIA Quantum Infiniband networking and in-network collections using NVIDIA SHARP to deliver the highest distributed training performance possible.
Serve ML models with the fastest spin up times and responsive auto-scaling
We help you serve models as efficiently as possible with our proprietary auto-scaling technology and spin up times in as little as 5 seconds. Data centers across the country minimize latency delivering superior performance for all of your end users.
Directly access "bare metal" Kubernetes environments without the hassle of managing infrastructure
Our GPUs are accessible by deploying containerized workloads for increased portability, less complexity and overall lower costs. Not a Kubernetes expert? Our engineers are here to help.
Artificial Intelligence & Machine Learning Cloud Solutions
Looking for a modern cloud, purpose-built for cutting edge AI?
CoreWeave empowers you to train, fine-tune, and serve models up to 35x faster while saving up to 80% on your current GPU costs.
Schedule a ChatUnparallelled performance for your most complex workloads
Not only do you get on-demand access to the industry’s broadest range of NVIDIA GPUs, but our Kubernetes-native infrastructure also delivers lightning quick spin-up times, responsive auto-scaling, and modern networking architecture.
Performance-Adjusted Cost Structure + Broadest Range of NVIDIA GPUs
- For on-demand workloads, we offer a la carte pricing (see below), where the total instance cost is a combination of a GPU component, the number of vCPU, and the amount of RAM allocated per hour.
- For workloads with relatively predictable usage patterns, we offer two types of volumetric discounts. Reserved Instances may be discounted up to 60% for 24/7 committed usage. Bulk Credits may be discounted up to 25% for larger-scale, burst compute use cases consumed on-demand.
- We DO NOT CHARGE for things like region-to-region transfers, workstation data, or egress in the vast majority of use cases.
GPU Model
VRAM (GB)
Max vCPUs per GPU
($0.01/hr)
($0.01/hr)
Max RAM (GB) per GPU
($0.005/hr)
($0.005/hr)
GPU Component
Cost Per Hour
Cost Per Hour
80 VRAM
48 Max CPUs
256 Max RAM
A100 80GB NVLINK
$2.21 / Hour
80 VRAM
48 Max CPUs
256 Max RAM
A100 80GB PCIe
$2.21 / Hour
80 VRAM
48 Max CPUs
256 Max RAM
A100 40GB NVLINK
$2.16 / Hour
40 VRAM
48 Max CPUs
256 Max RAM
A100 40GB PCIe
$2.06 / Hour
40 VRAM
48 Max CPUs
256 Max RAM
A40
$1.28 / Hour
48 VRAM
48 Max CPUs
256 Max RAM
RTX A6000
$1.28 / Hour
48 VRAM
48 Max CPUs
256 Max RAM
RTX A5000
$0.77 / Hour
24 VRAM
36 Max CPUs
128 Max RAM
RTX A4000
$0.61 / Hour
16 VRAM
36 Max CPUs
128 Max RAM
Quadro RTX 5000
$0.57 / Hour
16 VRAM
36 Max CPUs
128 Max RAM
Quadro RTX 4000
$0.24 / Hour
8 VRAM
36 Max CPUs
128 Max RAM
Tesla V100 NVLINK
$0.80 / Hour
16 VRAM
36 Max CPUs
128 Max RAM
What Others Say
From our clients to our partners, we strive to provide best-in-class solutions to
drive innovation and fast, flexible experiences.
drive innovation and fast, flexible experiences.
"Anyone can experience the power of a personal AI today based on our state-of-the-art large language model that was trained on CoreWeave’s powerful network of H100 GPUs."
“Always having access to GPUs on-demand has been a huge sanity saver. The availability and reliability of CoreWeave’s service allowed us to serve our current models and continuously build and test new ideas.”
“CoreWeave’s deployment architecture enables us to scale up extremely fast when there is more demand. We are able to serve requests 3x faster after migrating to CoreWeave, leading to a much better user experience while saving 75% in cloud costs. For the users, this means the generation speeds will never slow down, even when there is peak load.”
"We decided on CoreWeave because we had a good experience in the past, mostly for inference but also for training. And because CoreWeave manages everything for you, we don't need to deal with technicalities, we just write our Kubernetes configs and the rest is hassle-free"
“After a few months of struggling to keep up with demand at mega-cloud prices, we were able to seamlessly move our cloud infrastructure over to CoreWeave. Using CoreWeave’s Inference Optimized V100 instances, our inference latencies dropped by 50% and the cost savings from our partnership allowed us to continue delivering our free-tier experience and substantially reduce our cost-per-user.”
"We could not imagine a better cloud platform to realize this vision than what we’re creating with CoreWeave. We are humbled and honored to partner with the CoreWeave team to push the boundaries of modern AI computing and to build the infrastructure that will serve as the foundation of tomorrow’s AI-powered discoveries."
“CoreWeave is a valued, dedicated partner of NVIDIA. As such, they were named our first Elite Cloud Solutions Provider for Compute in the NVIDIA Partner Network. By offering their clients a tremendously broad range of compute options - from A100s to A40s - at unprecedented scale and their commitment to delivering world-class results in AI, machine learning, visual effects and more. NVIDIA is a proud supporter of CoreWeave.”
“We easily added support for H100s to our platform via integration with the NVIDIA Transformer Engine library, and are undergoing the system optimization process. Customers who can get access to H100s will be able to leverage this integration to get excellent performance from H100s in CoreWeave Cloud servers.”
Ready to get started?
A member of our team will reach out within 24 hours... but probably sooner