European Alternatives to CoreWeave
Looking for a European alternative to CoreWeave? CoreWeave is a US-headquartered GPU cloud (its own site lists Livingston, New Jersey), and its public price list is built around whole 8-GPU nodes, for example $49.24 per hour on demand for an HGX H100 node. Teams that want a European contracting party, or smaller GPU units billed by the hour, look elsewhere.
The European providers below range from per-minute GPU VMs to Slurm clusters, and each one signs customer contracts in the EU, the UK or Switzerland.
How we rank these tools — 4-step process
-
1
European ownership, verified
The company is headquartered and incorporated in the EU, EEA, the UK, Switzerland or Israel, and processes customer data in Europe or Israel. A US parent company disqualifies a tool from this page regardless of where its servers are.
-
2
Category fit and hands-on review
What the tool actually does, who it suits, and where it falls short — checked against the vendor’s own documentation, changelog and pricing page rather than its marketing copy.
-
3
Compliance and pricing check
GDPR posture, hosting location and the prices quoted on this page are verified against the vendor’s public pricing before publication, and re-checked when we revisit the category.
-
4
Position on this page
Placement on this page can be paid, and that can affect which tools appear here and the order they appear in. It never buys a good review: a tool that fails the checks above is not here at any price, and payment does not change the shortcomings we write about. A vendor can ask us to correct a factual error — not to remove a criticism.
Vendors can pay for visibility on this page. It never changes what an entry says about a product, including the criticism, and we earn nothing when you click through to a vendor. Vendors can pay for more visibility, higher up on the review pages; it never affects which tools are listed. Editorial policy
11 European Alternatives to CoreWeave
Verda
Finnish GPU cloud with instant clusters, spot pricing and confidential computing
Hyperstack
Per-minute billed GPU VMs from H100 to Blackwell, run by NexGen Cloud
Nscale
Vertically integrated AI cloud with Nordic and UK data centers
Nebius
AI cloud from Amsterdam with H100 to B300 and a public price list
evroc
Swedish sovereign cloud with dedicated L40S and B200 GPU VMs
Seeweb
Italian cloud with hourly NVIDIA and AMD GPU servers in four data centers
Civo
UK cloud with bare-metal and Kubernetes GPUs and free data transfer
Exoscale
Swiss cloud with per-second GPU servers and liquid-cooled racks
Leafcloud
Amsterdam GPU cloud on OpenStack that reuses server heat
Scaleway
French cloud with L4, L40S, H100 and B300 instances in Paris and Warsaw
OVHcloud
French provider with H200, H100, L40S and L4 GPU instances
Key takeaways
- CoreWeave lists its address as Livingston, New Jersey, so the contracting party is a US company even in the Europe table of its price list.
- The HGX H100 node costs $49.24 an hour on demand at CoreWeave, which is $6.16 per GPU, while Hyperstack lists one H100 SXM at $3.20 and Civo at $2.99.
- CoreWeave offers up to 60% off its on-demand rates for committed use, according to its pricing page.
- Nebius is the European provider on this page whose catalogue, from H100 to B300 with a public price list, comes closest to the node range at CoreWeave.
- Moving off CoreWeave means replacing its Kubernetes service, tracking tool and storage layer as well as the GPU hours.
Why people leave CoreWeave
CoreWeave is a GPU cloud that sells accelerators the way a wholesale supplier does: in whole nodes. The footer of its site gives the address as Livingston, New Jersey, and its price list is organised around 8-GPU HGX machines, with an HGX H100 node at $49.24 an hour on demand in both the North American and the European table. The company is built for training runs that fill racks, not for a single card on a Tuesday afternoon.
That design explains most of the reasons people go looking elsewhere. A small team that wants one H100 for a fine-tuning run finds a node-sized invoice. A company with a European customer base finds a US contracting party behind a European rack. And a buyer who has to explain the data path to a security reviewer finds that the legal question is about the company, not about the building the servers stand in.
None of this makes the platform weak. The pricing page lists GB200 and GB300 NVL72 systems and HGX B300, which most European providers do not show at all. The point of this page is narrower: for which workloads does a European contracting party with smaller billing units do the job, and what do you give up when you move?
- The price list is written in nodes Next to the eight-GPU rows, the pricing page carries a separate single-GPU inference column, and the H100 figure there, $6.16, is the node price divided by eight. For training that is a fair unit. For a lone fine-tuning run or a small inference service it means shopping inside a shape built for something larger. Hyperstack lists an H100 SXM at $3.20 an hour and Civo at $2.99, though network, memory and support differ.
- The contracting party is American The company states a New Jersey address, so a customer in Europe signs with a business under US jurisdiction even when the rack stands in the European region of the price list. For regulated buyers the question of who can be compelled to hand over data is usually settled by the company, not the postcode of the data centre. Every provider on this page contracts under EU, UK or Swiss law instead.
- Capacity is aimed at committed volume The page offers discounts of up to 60% over on-demand rates for committed use, and several newer systems, including GB300 NVL72 and some Blackwell rows, show Contact sales instead of a price. That suits a customer who plans spend a year ahead. It suits a start-up that cannot predict its next quarter less, and several European tools publish a price for every row instead.
- Regional choice is a smaller menu CoreWeave shows two regions on its pricing page, North America and Europe. A buyer who wants to say Stockholm, Amsterdam or Milan in the contract has more specific options elsewhere: evroc names Stockholm with three zones, Leafcloud names Amsterdam, and Nebius lets you pick Finland, France or the UK when you create a cluster.
What you have to replace, not just match
The GPU hours are the easy part to replace. Around them CoreWeave sells CoreWeave Kubernetes Service, a node health layer called Mission Control, bare-metal access, AI object storage and Weights & Biases Models for experiment tracking.
A migration plan should list which of these your jobs actually touch. A team that only rents nodes and runs its own scheduler can move quickly. A team whose training code depends on the managed Kubernetes layer and the tracking service has two more pieces to find a home for.
Storage is the second item. The pricing page says there are no ingress, egress or transfer fees on CoreWeave storage, and the company runs a Zero Egress Migration programme for data moving in. Leaving is a different direction, so ask for your own exit terms in writing and test a full export before you commit. On the European side, Civo states that data transfer is free in both directions, which removes the question for that provider. Others price transfer separately.
Finally replace the capacity you were promised. A reservation at CoreWeave is a statement about a specific GPU generation and node count. Moving to a European provider means checking whether the same generation exists there. Nebius and Verda list Blackwell-generation cards on their pages; evroc documents L40S and B200, and Seeweb lists H200 and AMD MI300X. Match memory per GPU and interconnect before the hourly price.
The alternatives compared
| Position | Tool | Headquarters | Pricing | Jurisdiction |
|---|---|---|---|---|
| #1 | Verda | Finland | Public hourly list in US dollars, for example $4.746/h for one H100 SXM5 80GB on demand and half that on spot (checked 29 October 2026) | EU (Finland) |
| #2 | Hyperstack | United Kingdom | Public hourly list in US dollars, for example H100 SXM $3.20/h and H200 SXM $3.99/h on demand (checked 29 October 2026) | United Kingdom (adequacy decision, outside the EEA) |
| #3 | Nscale | London, United Kingdom | See website; docs describe a prepaid self-serve model | United Kingdom (adequacy decision, outside the EEA) |
| #4 | Nebius AI Cloud | Amsterdam, Netherlands | GPU prices per hour on its price list, for example HGX H100 from $4.50 on demand; MLflow cost see website | EU (Netherlands) |
| #5 | evroc | Sweden | See website | EU (Sweden) |
| #6 | Seeweb | Frosinone, Italy | Public euro hourly list, for example A100 PCI 80GB from €0.99/h and L40S from €0.85/h (checked 29 October 2026) | EU (Italy) |
| #7 | Civo | London, United Kingdom | Public USD list, for example H100 SXM $2.99/h on demand and $2.49/h on a 36-month commitment (checked 29 October 2026) | United Kingdom (adequacy decision, outside the EEA) |
| #8 | Exoscale | Lausanne, Switzerland | From about CHF 5.50/month (Compute) | Switzerland (adequacy decision, outside the EEA) |
| #9 | Leafcloud | Netherlands | Public euro list excl. VAT, for example H100 80GB €3.15/h, A100 80GB €1.61/h (checked 29 October 2026) | EU (Netherlands) |
| #10 | Scaleway | Paris, France | From about €0.0006/hour (DEV1-S) | EU (France) |
| #11 | OVHcloud | Roubaix, France | PostgreSQL essential from about €0.16/hour (2 vCPU / 8GB RAM); MySQL, MongoDB, Kafka, Cassandra, OpenSearch, M3DB and Valkey priced separately by engine and tier | EU (France) |
How each alternative compares to CoreWeave
Best for: ML teams that want hourly H100 to B300 hardware under Finnish law, not a node-sized invoice
Verda sells the same class of hardware that CoreWeave does, from H100 to B300 and GB300 NVL72 trays, but prices single instances. Its list shows $4.746 an hour for one H100 SXM5 80GB on demand, where the CoreWeave node works out at $6.16 per card. Instant clusters with InfiniBand cover multi-node runs without a sales conversation, which narrows the gap on the training side.
The contract is with DataCrunch Oy under Finnish law, and disputes go to arbitration in Helsinki in English. Verda claims SOC 2 Type II and the ISO 27001 family and offers confidential computing. The pages checked do not name individual data centre sites, so ask which one your workload lands in before moving regulated data.
What Verda does better than CoreWeave
- Finnish company and Finnish law instead of a New Jersey contracting party
- Single H100 instances rather than eight-GPU nodes
- Spot rate displayed beside the on-demand rate on every row
- Confidential computing offer for hardware-attested inference
Where Verda is a step down from CoreWeave
- Smaller than CoreWeave, which sells at rack scale
- Prices in US dollars and data centre sites not named
- Object storage still marked beta
Standout against CoreWeave. Verda combines a self-service InfiniBand cluster with a confidential computing option under a Finnish contract.
Best for: Developers who run short GPU sessions and want to stop paying the minute a job ends
Hyperstack bills to the minute and lists an H100 SXM at $3.20 an hour and an H200 SXM at $3.99, with the H100 PCIe at $2.50. Against CoreWeave's node rows, a developer pays for the card and the time used. The menu reaches from L40 to GB300 NVL72, so the hardware range overlaps with CoreWeave's more than most European tools.
The seller is NexGen Cloud Limited, governed by the law of England and Wales, which is a UK party and not an EU one. Data centres sit in Europe and North America, so the region is a deployment choice, and the exact European sites are not named on the pages checked. A Terraform provider and Python and Go SDKs help with code-driven clusters.
What Hyperstack does better than CoreWeave
- Minute-level billing on single VMs
- Lower listed H100 SXM rate per card
- Terraform provider and SDKs on the public docs
- Reservation rates published next to on-demand
Where Hyperstack is a step down from CoreWeave
- UK entity, outside the EU, with North American sites also offered
- No named European sites on the pages we read
- Less capacity and fewer managed services than CoreWeave
Standout against CoreWeave. Hyperstack states that its billing cycles are accurate to the minute.
Best for: AI companies that want Norwegian or UK capacity plus managed Slurm or Kubernetes
Nscale runs its own data centers in Glomfjord and Narvik in Norway and Loughton in England, and offers bare-metal instances, managed Slurm, managed Kubernetes and inference endpoints. That stack is the nearest thing in this list to what CoreWeave sells around its GPUs. Docs describe a prepaid self-serve console with groups, roles and auto-recharge.
The contracting party is Nscale Limited, registered in England and Wales under number 16925886. Some listed sites are in the United States and some capacity is partner-run in Portugal, Iceland and elsewhere, so pin the region when you create resources. No price list was visible, which means a quote is the only way to compare cost.
What Nscale does better than CoreWeave
- Own data centers in Norway and England, not only a regional table
- Managed Slurm for teams that schedule training outside Kubernetes
- Hosted inference endpoints on the same account
- Company number and registered office published in the privacy policy
Where Nscale is a step down from CoreWeave
- Prepaid balance commits money before use
- No public price, unlike CoreWeave's table
- Some sites in the United States
Standout against CoreWeave. Nscale pairs its own Norwegian data centers with managed Slurm on a self-serve account.
Nebius AI Cloud
Dutch entity with the widest GPU range and a public price list, closest match on scale
Best for: Teams that outgrew single VMs and want H100 to B300 capacity with EU region choice
Nebius AI Cloud is the tool on this page whose catalogue looks most like CoreWeave's. It lists H100, H200, B200 and B300 nodes, an HGX H100 at $4.50 an hour on demand, and preemptible rates from $0.79 for some models. It also runs managed MLflow, which stands in for the experiment tracking that CoreWeave offers through Weights & Biases.
The contract is with Nebius B.V. in Amsterdam, and you choose Finland, France, the UK, Israel or Kansas City when you create a cluster. The privacy policy lists a US subsidiary, so select an EU region deliberately. Prices are in US dollars, and the MLflow service price was not found on the pages we read.
What Nebius AI Cloud does better than CoreWeave
- Dutch contracting entity and a Finnish or French region on request
- Listed H100 rate below CoreWeave's per-card figure
- Managed MLflow with standard logging code
- Preemptible GPUs on the public list
Where Nebius AI Cloud is a step down from CoreWeave
- A US subsidiary and a US region sit in the same console
- Smaller footprint than CoreWeave overall
- Pricing in dollars
Standout against CoreWeave. Nebius lets you pick Finland or France for a cluster and also hosts MLflow for the same runs.
Best for: Swedish and Nordic organisations with sensitive AI workloads that want a national provider
evroc dedicates each GPU to one VM instead of time-sharing it, using NVIDIA L40S and B200 profiles with local NVMe from 3.8 TB to 32 TB. Against CoreWeave, the offer is narrower: two documented GPU models, one region and no public price. What it adds is a company with its own legal pages under evroc AB and a Swedish footprint.
The region is se-sto in Stockholm with three zones that fail independently. The local disk is ephemeral and is lost if the VM is stopped through the API, deleted or its host fails, so checkpoints belong on object storage. The documentation covers CLI and API management, and an AI services platform sits on top of the VMs.
What evroc does better than CoreWeave
- Swedish company and Swedish residency in a single named region
- Dedicated GPU per VM
- Three independent zones in Stockholm
Where evroc is a step down from CoreWeave
- No public price on the pages checked
- Only L40S and B200 documented, no H100 or H200
- Single region, no multi-country spread
Standout against CoreWeave. evroc documents three zones inside one Stockholm region with a dedicated GPU behind each VM.
Best for: Italian and Southern European teams that want euro invoices and an Italian-law contract
Seeweb publishes its hourly prices in euros: A100 PCI 80GB from €0.99, L40S from €0.85, H200 SXM at €2.60 and AMD MI300X at €2.35, falling to €2.00 on a twelve-month reservation. CoreWeave quotes dollars and nodes, so a finance team in the euro area compares fewer unknowns. A note on the page says interruptible use can save up to 40%.
The supplier is SEEWEB srl, with Italian law and the court of Frosinone named in its terms. Data centers are in Milan, Frosinone, Sofia and Lugano, and the page does not say which GPU sits where. Lugano is in Switzerland, so confirm the site before you place EU-bound data. The page mixes node sizes, so read the configuration, not just the hourly figure.
What Seeweb does better than CoreWeave
- Prices in euros on the public page
- AMD MI300X option beside NVIDIA cards
- Italian-law contract with a named court
- Serverless GPU and NPU options in the same menu
Where Seeweb is a step down from CoreWeave
- GPU per data center not stated
- Much smaller scale than CoreWeave
- Dense, mixed price tables
Standout against CoreWeave. Seeweb lists AMD MI300X at a public euro price next to its NVIDIA cards.
Best for: UK and European teams that run GPU jobs on Kubernetes and want predictable transfer costs
Civo lists an H100 SXM at $2.99 an hour on demand and $2.49 on a 36-month commitment, an A100 80GB at $1.79 and an A100 40GB at $1.09. It sells GPUs as bare metal or through Kubernetes, and the Blackwell B200 and a B300 option appear on its GPU page. The page states free, unlimited transfer in both directions removes the egress line item.
The agreement is with Civo Limited, registered in England and Wales under number 09568551, and the privacy notice names affiliates in New York and India. The GPU page says data stays within UK borders, a UK statement and not an EU one. Some GPUs show commitment rates only, so the on-demand menu is shorter than the full list.
What Civo does better than CoreWeave
- Lowest listed H100 hourly rate on this page
- Free and unlimited data transfer, stated on the pricing page
- Bare-metal GPUs as well as Kubernetes
- Public table readable without a sales call
Where Civo is a step down from CoreWeave
- US and Indian affiliates named in the privacy notice
- Sovereignty wording refers to the UK, not the EU
- Fewer Blackwell on-demand prices than CoreWeave
Standout against CoreWeave. Civo pairs bare-metal GPUs with a stated policy of free data transfer.
Best for: Organisations that need Swiss jurisdiction and run GPUs next to ordinary cloud services
Exoscale bills GPU servers by the second and offers one to eight GPUs per server, plus inference endpoints and liquid-cooled racks according to its listing. Compared with CoreWeave this is a general cloud with GPUs inside it: managed Kubernetes (SKS), S3-compatible object storage and managed databases share the account. GPU prices are shown through a calculator, not a table.
Exoscale belongs to A1 Digital, part of the A1 Telekom Austria Group, and is based in Lausanne. Data centres are in Switzerland, Germany, Austria and Bulgaria. Switzerland holds an EU adequacy decision, so transfers are lawful, although the processing is not strictly inside the EEA. The service catalogue is smaller than the hyperscalers', as the tool page notes.
What Exoscale does better than CoreWeave
- Per-second billing on GPU servers
- Ordinary cloud services in the same account
- Swiss jurisdiction with sites in four countries
- Backing by a European telecom group
Where Exoscale is a step down from CoreWeave
- GPU prices only in a calculator
- Swiss franc pricing for non-Swiss buyers
- Smaller GPU fleet than CoreWeave
Standout against CoreWeave. Exoscale bills GPU servers by the second and keeps them in the same account as managed databases and Kubernetes.
Best for: Dutch and neighbouring teams with moderate GPU needs and a sustainability requirement
Leafcloud lists RTX 6000 Blackwell 96GB at €1.99, H100 80GB at €3.15, A100 80GB at €1.61 and A30 24GB at €0.60 an hour excluding VAT, with the full VM included in each price. The table sits beside rows for other providers, and the comparison is the vendor's own. For small and mid-sized jobs the euro price is simple to budget against CoreWeave's node rows.
The contract is with Leafcloud B.V., under Dutch law with the civil court of Amsterdam named, and the sovereignty pages address the US CLOUD Act directly. The platform is OpenStack with Gardener Kubernetes, and servers are placed where their heat warms buildings. The GPU table shows one region, ams-1, and no cluster fabric, so multi-node training is out of scope.
What Leafcloud does better than CoreWeave
- Euro prices on one public page
- OpenStack compatibility that eases moves from other providers
- Dutch law and an Amsterdam court
- Heat reuse model
Where Leafcloud is a step down from CoreWeave
- One listed GPU region
- No InfiniBand cluster offer
- Smaller fleet and unstated capacity for large reservations
Standout against CoreWeave. Leafcloud places servers where their heat is reused to warm buildings, and it prices GPUs in euros.
Best for: French and EU teams that want GPUs inside a broad cloud platform with EU-only sites
Scaleway lists L4, L40S, H100 and B300 instances in Paris and Warsaw, and it adds a Model-as-a-Service product on top. It is a general cloud first: Kubernetes Kapsule, S3-compatible object storage, serverless functions and managed databases sit in the same account, which CoreWeave does not try to replace. Hourly billing and the lack of egress surcharges are noted on its tool page.
Scaleway was founded in 1999 as Online.net and is a subsidiary of the Iliad Group. Its data centres are Paris, Amsterdam, Warsaw and Milan, and the tool page describes the platform as EU-only for data location. Against CoreWeave there are fewer regions and a thinner GPU range, and support at lower tiers is community-based.
What Scaleway does better than CoreWeave
- EU-only data centres under French law
- GPUs alongside Kubernetes, storage and databases
- Model-as-a-Service on the same platform
- Backing by the Iliad Group
Where Scaleway is a step down from CoreWeave
- Fewer GPU generations at the top end
- Documentation thinner for edge cases
- Support at lower tiers is self-service
Standout against CoreWeave. Scaleway offers GPU instances together with the usual cloud services, all in EU-only data centres.
Best for: Buyers who need a large French provider with certified ranges and a wide choice of regions
OVHcloud lists H200, H100, L40S and L4 GPU instances, with savings plans and a range that carries the SecNumCloud qualification from ANSSI. The company was founded in Roubaix in 1999, builds its own servers and runs 46 data centres on four continents, most of them in Europe. That is a bigger physical estate than any other European provider on this page lists.
The contracting entity is OVHcloud SAS under French law, and the tool page notes it is a founding member of Gaia-X. Pricing is granular per resource, which takes effort to estimate for a real bill, and the console is described as less polished than specialist platforms. The GPU range on the page stops at H200, with no Blackwell listed.
What OVHcloud does better than CoreWeave
- SecNumCloud-qualified range for French public and regulated buyers
- Own hardware and data centres
- Savings plans for committed use
- Wide choice of regions
Where OVHcloud is a step down from CoreWeave
- No Blackwell generation on the card summary
- Granular pricing is hard to forecast
- Console less polished than GPU specialists
Standout against CoreWeave. OVHcloud builds its own servers and runs a SecNumCloud-qualified range from French data centres.
What actually breaks when you switch
Expect the first week to go on data and scheduling. Datasets, checkpoints and model weights have to be copied out of CoreWeave storage and into the new provider's object store, and the transfer time is set by your own uplink, not by the vendor.
Check whether the destination bills ingress or egress: Civo states free transfer, Scaleway's tool page mentions no hidden egress fees, and for the others read the price list before you size the move. Keep the old environment running until one full training or inference cycle has completed on the new one.
The second source of friction is the tooling around the GPUs. If your team uses CoreWeave's Kubernetes service or its Weights & Biases integration, plan for the equivalent on the new side: managed Kubernetes at Hyperstack, Civo, Scaleway or Leafcloud, managed Slurm at Nscale, MLflow at Nebius.
Contracts matter too. Committed-use discounts at CoreWeave are tied to a term, so check the end date before you give notice, and ask the new vendor whether the GPU generation you need is available at your volume, since several of the tools here quote on request.
Can I move a CoreWeave training job to a European provider without rewriting it?
If the job is a container that asks Kubernetes or Slurm for GPUs, the code travels and the scheduler settings change. Nscale offers managed Slurm and Kubernetes, Hyperstack offers on-demand Kubernetes, Civo runs GPUs on Kubernetes and bare metal, and Leafcloud uses Gardener-managed Kubernetes. Jobs that depend on CoreWeave-specific services, such as its Kubernetes layer or Weights & Biases Models, need those pieces rebuilt or replaced.
Multi-node runs are the harder case. Verda describes self-service instant clusters with InfiniBand and Nebius lists Slurm integrations, but Leafcloud shows no cluster fabric on the pages checked. Test one short distributed run on the target before you move a month of work.
Is a European GPU cloud cheaper than CoreWeave?
Per card, the public lists are lower for several tools. CoreWeave divides to $6.16 for an H100 in its HGX node, against $4.746 at Verda, $4.50 at Nebius, $3.20 at Hyperstack, $2.99 at Civo and €3.15 at Leafcloud. The figures are not like for like: node size, interconnect, networking and tax treatment differ, and Nscale and evroc publish no price in the pages we checked.
Committed use changes the picture. CoreWeave states discounts of up to 60% off on-demand rates, and Civo shows $2.49 for an H100 on a 36-month term. Ask each shortlisted vendor for a quote at your real volume and term.
Which one to pick
For cluster-scale training where the node range matters, start with Nebius, Verda and Nscale. Nebius has the broadest catalogue and a public price list, Verda adds self-service InfiniBand clusters and spot pricing, and Nscale brings Nordic sites with managed Slurm. For hourly rental of single GPUs, Hyperstack, Civo, Seeweb and Leafcloud publish prices you can read before you talk to anyone, and the last two quote in euros.
For a buyer whose main requirement is jurisdiction, pick by contract: evroc for Sweden, Seeweb for Italy, Leafcloud for the Netherlands, Scaleway and OVHcloud for France, Exoscale for Switzerland. Stay with CoreWeave if you need GB300 NVL72 or HGX B300 capacity at rack scale and a US contract is acceptable; none of the European tools shows that range on the pages we checked.
Frequently Asked Questions
It depends on the workload. Nebius is the closest match for cluster-scale training, with a public GPU price list and a Dutch contracting entity. Verda adds instant InfiniBand clusters and spot rates. Nscale offers managed Slurm with Nordic and UK sites. Hyperstack and Seeweb suit smaller jobs billed by the minute or the hour, and evroc suits Swedish buyers.
Its pricing page has a Europe table next to the North America one, with the HGX H100 node at the same $49.24 an hour on demand. The company itself, however, gives a New Jersey address, so the contract and the corporate jurisdiction remain American even if the machines are in Europe.
Seeweb and Leafcloud. Seeweb lists A100 PCI 80GB from €0.99 an hour and L40S from €0.85, and Leafcloud lists H100 80GB at €3.15 excluding VAT. Verda, Hyperstack, Nebius and Civo quote in US dollars, and Nscale and evroc publish no price on the pages we read.
Yes at most of the providers listed here. Hyperstack bills to the minute, Exoscale bills by the second with 1 to 8 GPUs per server, Seeweb and Leafcloud bill hourly, and Verda lists a single H100 SXM5 on demand. CoreWeave itself shows a single-GPU inference price per card, but its on-demand table is built from nodes.
No. Hyperstack, Nscale and Civo contract under English law, Exoscale is Swiss, and Seeweb has one site in Lugano. Verda, Nebius, evroc, Leafcloud, Scaleway and OVHcloud are EU companies. Hyperstack, Nscale and Nebius also list sites outside the EU, so the region must be chosen when you provision.
Related comparisons
European tools head to head
About the author
Explore More European Alternatives
Discover privacy-focused European alternatives to other popular US tech services.