Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
CoreWeave has the lowest listed eight-H100 node rate in a third-party comparison checked in September 2026, but the figures do not establish that it is the cheapest or fastest choice for a real workload. The compared systems differ in configuration and terms, and there is no comparable evidence here for live capacity, provisioning time, or cross-provider performance. CoreWeave’s own North America price list gives more specific on-demand and spot prices for several configurations; those are useful starting points, not a guarantee that a matching node is available to your account.
What does the available price comparison show?
The September 2026 comparison reports an eight-H100 node rate of $49.24 per hour for CoreWeave, compared with $55.04 for AWS, $88.49 for Google Cloud, and $98.32 for Azure. On those reported figures alone, CoreWeave is the least expensive listing. But the comparison warns that the node bundles are not identical, so the table is a dated price snapshot—not a like-for-like performance or total-cost result.
As an Amazon Associate I earn from qualifying purchases.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match| Provider | Reported eight-H100 node rate | Source and qualification |
|---|---|---|
| CoreWeave | $49.24/hour | iTechGuides comparison checked September 2026; bundle differs from other compared nodes |
| AWS | $55.04/hour | iTechGuides comparison checked September 2026; bundle differs from other compared nodes |
| Google Cloud | $88.49/hour | iTechGuides comparison checked September 2026; bundle differs from other compared nodes |
| Azure | $98.32/hour | iTechGuides comparison checked September 2026; bundle differs from other compared nodes |
Calculated directly from those reported rates, CoreWeave’s listed node is $5.80/hour below AWS, $39.25/hour below Google Cloud, and $49.08/hour below Azure. These are arithmetic differences between the cited listings, not savings guarantees: the comparison does not establish that the nodes have equal CPU, host memory, storage, networking, region, or scheduling terms.
The practical takeaway is that CoreWeave is the lower-priced entry in this particular published comparison. It is not evidence that a training run, inference service, or completed job will cost less there. A provider’s relevant price is the cost of completing your own workload to its required quality and deadline, including the time the workload actually takes and any costs or interruptions tied to its pricing terms.
#1 Best Overall
- [ Maximum AI Compute Power ] Dominate complex workloads with the ASUS ESC8000A-E13. This 4U rack server is a powerhouse engineered for mass-scale AI, machine learning, and deep training. Featuring support for dual AMD EPYC 9005/9004 processors and up to eight dual-slot GPUs, it delivers the raw computational muscle required to train LLMs and run complex simulations effortlessly. Accelerate your data science pipeline and transform raw data into actionable intelligence faster than ever.
- [ Advanced Thermal Efficiency ] High performance demands elite cooling. The ESC8000A-E13 features a cutting-edge aerodynamic design with independent CPU and GPU airflow tunnels. Equipped with redundant hot-swap fans and optimized for liquid cooling integrations, this 4U server ensures maximum uptime under heavy, sustained workloads. Keep your data center running cool, quiet, and highly efficient while preventing thermal throttling during mission-critical enterprise operations.
- [ Scale with Flexible Storage ] Future-proof your infrastructure with unmatched storage and expansion flexibility. This offers comprehensive front-panel drive bays supporting Gen5 NVMe, SAS, or SATA drives alongside multiple PCIe 5.0 slots. Designed as a high-density 4U server capable of housing eight dual-slot GPUs: NVD H200, RTX PRO 6000 Blackwell, RTX PRO 4500 Blackwell or AMD Instinct MI350P PCIe Card, each supporting up to 600 watts.
- [ Enterprise-Grade Reliability ] Minimize downtime and secure your ecosystem with server-grade redundancy. The ESC8000A-E13 is built for 24/7 continuous operation, boasting 2+2 redundant (3200W total) 80 PLUS Titanium power supplies and integrated ASUS ASMB11-iKVM for comprehensive out-of-band management. Ideal for cloud service providers, rendering farms, and large enterprise infrastructure, it combines robust physical hardware with smart remote monitoring to safeguard your digital assets.
- [Reliability Guaranteed] Shop with total peace of mind knowing that every new computer component we sell is backed by our EPC 3-year warranty. Whether you are investing in high-speed DDR5 RAM or a powerhouse GPU, we protect your build against defects and performance failures. We stand firmly behind the quality of our hardware, ensuring that your setup remains fast, stable, and secure for years to come.
What GPU prices does CoreWeave publish?
CoreWeave’s official pricing page, accessed October 25, 2026, lists North American rates for several GPU systems. It also lists configurations in North America and Europe, although the specific figures below are the North America prices in the page’s public listing.
| Listed North America configuration | On-demand rate | Spot rate |
|---|---|---|
| HGX H100, eight GPUs | $49.24/hour | $19.71/hour |
| HGX H200 | $50.44/hour | $20.93/hour |
| A100 | $21.60/hour | $9.65/hour |
The H100 listing is explicitly an eight-GPU configuration. The listed H200 and A100 rates are configuration prices on the pricing page; do not assume their GPU count or full system bundle matches the H100 entry or a competitor’s node without checking the associated configuration details. Some systems have public prices, while others require a sales inquiry. A published rate is not a live capacity reservation or a quote tailored to your account.
How should you read spot prices?
CoreWeave’s listed spot rates are materially lower than the corresponding on-demand rates. For the listed H100, the arithmetic difference is about 60%; for H200, about 59%; and for A100, about 55%. Those percentages describe the gap between the published rates only. They do not establish the chance of interruption, the notice period, whether a particular region has spot capacity, or the cost of recovering a job after interruption.
Before putting a workload on spot capacity, establish the provider’s current interruption and recovery terms for the exact configuration and region. Then assess whether your application can checkpoint state, restart safely, and tolerate uncertain completion time. A lower hourly rate does not compensate for lost progress if the workload cannot recover efficiently.
How do CoreWeave’s prices compare across GPU models?
A separate StackPath compilation gives per-GPU figures as of October 18, 2026. These are compiled prices, not direct workload measurements, and they should not be mixed with node-level rates as if they were the same billing unit or configuration.
| GPU model | Compiled CoreWeave rate per GPU | Compilation date |
|---|---|---|
| A100 | $2.70/hour | October 18, 2026 |
| H100 | $6.16/hour | October 18, 2026 |
| H200 | $6.30/hour | October 18, 2026 |
| B200 | $8.60/hour | October 18, 2026 |
| GB200 | $10.50/hour | October 18, 2026 |
These per-GPU rates are best treated as a discovery aid. The compilation describes a broad set of vendor prices and notes that Google Cloud was absent from its comparison because the script could not read a public price list without a key. A compiled per-GPU number does not by itself tell you the number of GPUs required, system overhead, the full node price, regional availability, or whether the price applies to your use case.
Rank #2
- NVIDIA Volta GV100 Architecture — 4,608 CUDA Cores, 640 1st-Gen Tensor Cores delivering 14 TFLOPS FP32 and 112 TFLOPS deep learning performance for AI training, inference, HPC, and scientific computing workloads
- 32GB HBM2 ECC Memory — 900 GB/s Bandwidth — High-bandwidth memory on a 4096-bit bus with ECC error correction provides the memory capacity and throughput required for the largest AI models, simulations, and datasets
- PCIe 3.0 x16 Interface — 250W TDP — Standard PCIe Gen3 connectivity with passive cooling designed for enterprise rack server deployment in HPE ProLiant, Dell PowerEdge, and Supermicro platforms with adequate chassis airflow
- NVLink — Scale to 96GB Unified Memory — Connect two V100 GPUs via NVLink at 300 GB/s bi-directional bandwidth to scale GPU memory from 32GB to 96GB for larger AI training and HPC workloads
- Multi-Precision Computing — Supports FP64 (7 TFLOPS), FP32 (14 TFLOPS), FP16 (112 TFLOPS) and INT8 precision modes for flexible deployment across training, inference, and scientific simulation workloads
Does the lowest hourly rate mean the lowest training cost?
No. Hourly rental price is only one input to workload cost. A less expensive node can still cost more to complete a job if its configuration or performance makes the job take longer, requires more GPUs, or adds other costs. The available price comparison does not supply a controlled benchmark that would resolve those differences.
For a meaningful comparison, calculate the cost of the same job under matched conditions. Keep the model, dataset, software stack, precision, batch settings, target quality, and completion criteria consistent. Record the total runtime and billed resources, not just the GPU rate.
Build a comparable workload estimate
- Define the work. Specify the training run or inference workload, dataset or request volume, software and framework versions, quality target, and deadline. Without a fixed target, two providers may appear to perform differently because they completed different work.
- Match the hardware bundle. Record GPU model and count, accelerator memory, CPU count, host memory, local storage, and networking. If the provider cannot offer an exact match, document the difference rather than treating hourly rates as directly comparable.
- Match location and terms. Compare the region needed for your data and users, plus on-demand, spot or preemptible, and committed-price terms where offered. A regional price or discount is useful only if it applies to your workload and can be provisioned under your conditions.
- Measure completion time. Run the same workload with the same settings and collect elapsed time, interruptions, restarts, and any failed runs. The evidence available here does not identify a performance winner, so a workload-specific measurement is needed.
- Calculate total cost to completion. Add the billed compute time for all required resources and include any other workload charges that your quote or billing records show. Compare the cost for a successful run or agreed volume of output—not only the nominal hourly GPU price.
This method separates three questions that are often collapsed into one: what a provider lists per hour, how quickly the chosen configuration completes the work, and how much the completed workload costs. Only the first question has a cross-provider number in the September 2026 H100 snapshot.
What should you compare besides GPU model and price?
GPU names alone do not describe a cloud system. The iTechGuides comparison specifically cautions that the compared bundles differ in CPU, memory, storage, networking, region, and scheduling terms. Those distinctions can change whether a configuration fits your application and whether its hourly rate is useful as a comparison.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →- Accelerator memory and count: Check the exact GPU model, how many GPUs are in the billed unit, and the accelerator memory available to your job. A price per GPU and a price per eight-GPU node are different units.
- CPU and host memory: Confirm the system resources paired with the GPUs. A workload can be limited by data loading or other host-side work, so a GPU count alone is not a complete system specification.
- Storage: Identify what storage is included in the configuration and what your workload needs for datasets, checkpoints, and outputs. The available price snapshot does not normalize storage across providers.
- Networking: Compare the networking details relevant to your workload, particularly where data movement or multi-GPU work makes them material. The cited node rates do not establish equivalent networking.
- Region and data location: Compare the actual region you can use, not only a provider’s general regional footprint. CoreWeave’s price page lists North American and European configurations, but the cited public figures do not establish a like-for-like regional price comparison.
- Scheduling and billing terms: Clarify whether a rate is on-demand, spot or otherwise interruptible, or tied to a commitment. Do not treat a spot rate as equivalent to an on-demand rate.
- Capacity and startup: Ask whether the exact configuration is available to your account, how provisioning works, and what timing to expect. Published prices do not show live capacity or queue time.
Can you compare availability and provisioning today?
Not from the cited price lists. The available material does not establish live GPU capacity, queue times, provisioning performance, or the likelihood that a particular account can obtain a specific configuration in a specific region. CoreWeave’s public price listing confirms that prices are published for some configurations; it does not guarantee that those systems are currently available.
Availability is a time- and account-specific question. Before committing a deadline-sensitive project, confirm capacity for the exact GPU model, count, region, and pricing type with each provider. Ask how long provisioning is expected to take and whether the quote represents reserved capacity or merely a listed rate. For spot or preemptible capacity, also ask what happens if capacity is interrupted.
Rank #3
- AI-Optimized: Designed to support up to 4 GPUs, it is perfect for handling intensive AI and machine learning tasks, ensuring high performance and scalability for advanced computational needs.
- Intelligent Storage: Equipped with 8 hot-swappable 3.5" SATA/SAS drives (12Gbps), featuring SGPIO and temperature control, it ensures efficient data management and reliable storage performance.
- Robust Cooling: The system includes 3x 12038 hot-swap PWM fans and 2x 8038 rear fans, providing advanced thermal management to maintain optimal temperatures and ensure stable operation under heavy workloads.
- Rack-Ready: Comes with a pre-installed rail kit, allowing for quick and easy installation in standard 19-inch server racks, making it ideal for data center environments and enterprise setups.
- Versatile Connectivity: Offers USB 3.0 and the latest USB 3.2 Type-C ports, ensuring high-speed data transfer and compatibility with a wide range of peripherals and devices for enhanced connectivity options.
These checks matter because a low rate has little value if the needed system cannot be provisioned in time. Conversely, this evidence does not establish that any one provider is generally more available; that conclusion would require current, comparable capacity information for the regions and configurations in question.
Is there a performance winner among these providers?
No performance winner is established by the available comparison. The reported rates are not accompanied by a controlled test of the same workload on equivalent systems. Differences in CPU, memory, storage, networking, region, and scheduling prevent an hourly-rate table from acting as a benchmark.
Recommended Free Tools
For a procurement decision, benchmark the workload that matters to you. Use the same inputs, software versions, and completion criteria; record both throughput and time to a successful result. For training, define a quality target rather than comparing runs that stop at different points. For inference, use a representative request mix and measure the latency or throughput targets your service actually needs. These are ways to create evidence for your decision; they are not performance results supplied by the cited sources.
If a vendor provides benchmark data, check whether it discloses the full configuration, software stack, workload, measurement method, and date. A result that omits those details may not transfer to your workload, and a benchmark from one model or region cannot automatically stand in for another.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How should you choose between CoreWeave and a major cloud provider?
Start with requirements that could disqualify a system, then compare price and measured workload results among the viable choices. The September 2026 price snapshot makes CoreWeave worth including in an H100 quote comparison, but it does not settle the provider decision.
CoreWeave is a candidate when
- Its listed configuration matches your GPU and system requirements.
- The exact region and capacity can be confirmed for your account and delivery schedule.
- The applicable on-demand or spot terms fit your tolerance for commitment or interruption.
- A matched workload test supports the cost and completion-time outcome you need.
Compare AWS, Google Cloud, and Azure when
- Your existing architecture, data location, or operational requirements make one of those environments relevant.
- You can obtain a quote for a configuration that is sufficiently comparable to the CoreWeave option.
- You can verify region, scheduling terms, and capacity rather than relying on the dated H100 price snapshot.
- You can test the same workload and calculate cost to completion for each viable configuration.
The best choice is the system that meets your workload’s performance, availability, and operational requirements at an acceptable total cost—not automatically the provider with the lowest published hourly figure.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesRank #4
- Professional AI & Creator Workstation: AMD Radeon AI PRO R9700 GPU with 32GB GDDR6 is engineered for AI development, professional content creation, and compute-intensive workloads.
- Massive 32GB Memory Capacity: 32GB of GDDR6 memory on a 256-bit bus provides ample bandwidth for large AI models, 8K video editing, and complex 3D rendering.
- Advanced RDNA 4 with AI Accelerators: 64 Compute Units with 3rd Gen Ray Tracing and dedicated 2nd Gen AI Accelerators for groundbreaking AI performance and visual computing.
- Professional Blower Cooling: Efficient single blower design exhausts heat directly out of the chassis, ideal for multi-GPU workstation and server configurations.
- Enterprise-Grade Thermal Solution: Vapor chamber heatsink with industrial Honeywell PTM7950 thermal interface material ensures reliable cooling under sustained professional loads.
What should you verify before placing an order?
Prices and capacity can change, and the dated figures above should be rechecked directly before purchase. CoreWeave’s pricing page says that GPU-based pricing for inference applies exclusively to customers of its inference platform and directs readers to contact an account executive. If you are evaluating inference pricing, confirm whether that restriction applies to your intended product and account rather than assuming the listed GPU rate is a general inference offer.
- Ask for the exact configuration, including GPU count, CPU, system memory, storage, and networking.
- Confirm the region, rate type, billing unit, and any minimum or commitment terms in the current quote.
- Verify whether the requested capacity is available to your account and when it can be provisioned.
- For spot or preemptible capacity, obtain the current interruption and recovery terms.
- Check whether platform-specific terms apply to your workload, especially for inference.
- Recalculate total cost using measured runtime and the provider’s actual billing details.
For vendor-to-vendor comparison, use the third-party H100 figures as a prompt to request and test comparable offers, not as a substitute for current quotes. The StackPath compilation is dated October 18, 2026, while the iTechGuides H100 comparison was checked in September 2026; neither establishes today’s live capacity or a guaranteed purchase price.
How strong is the evidence behind the comparison?
The figures answer a narrow but useful question: what public or compiled rates were reported at particular dates? CoreWeave’s official page supplies its listed North America on-demand and spot rates, and the third-party September comparison supplies a cross-provider eight-H100 snapshot. StackPath adds compiled per-GPU prices dated October 18, 2026.
Those sources do not answer which provider currently has capacity for your account, how long provisioning will take, or which provider completes a matched workload fastest. The cross-provider node bundles are explicitly non-identical, and the compiled per-GPU rates are not independent workload measurements. Treat those distinctions as part of the comparison, not as fine print that can be ignored when selecting a cloud.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Frequently Asked Questions
Does CoreWeave have GPU prices for Europe?
Its official pricing page lists GPU configurations in both North America and Europe, but the figures cited here are North America rates. Ask for the current European price and configuration for the exact region you intend to use.
Can I use CoreWeave’s GPU price for any inference workload?
The pricing-page footnote says GPU-based inference pricing applies exclusively to customers of CoreWeave’s inference platform and directs readers to contact an account executive. Confirm eligibility and terms for your account before using a listed GPU rate in an inference budget.
Why does a cloud price compilation omit Google Cloud?
StackPath’s compilation dated October 18, 2026 says its script could not read a public Google Cloud price list without a key. That omission reflects the compilation method and should not be read as evidence that Google Cloud has no GPU pricing.
Are these prices industry-wide average GPU rates?
No. The figures come from specific provider listings and dated third-party compilations, not an industry-wide survey. They should be used as reference points for current provider quotes, not as market averages.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
