Skip to content
State of GPU Compute 2026

Compare

Nexpod vs Oracle Cloud for GPU work.

Oracle Cloud is often chosen to sit next to existing Oracle workloads. For GPU compute on its own, compare it with Nexpod: whole 8-GPU servers, per-second billing and euro invoices.
Get startedTalk to an engineer

4× H200 · €128.00 /hrsame rate as one nodeper-second meter

A Nexpod Instant Cluster: four 8-GPU H200 servers in a two by two layout, each GPU board linked to an InfiniBand switch, 32 GPUs in total.
clusters / new
H200
Shape
1 node2 nodes4 nodes
node 0
node 1
node 2
node 3
fabricInfiniBand
32 GPUs · one cluster€4.00/GPU·hr, as on 1 node

Teams running GPU work on Nexpod

Placeholder
  • Customer
  • CUSTOMER
  • customer
  • Customer Co
  • customer.ai
  • CUSTOMER

Before you choose

Know what a GPU run involves.

Whichever cloud you pick, a production GPU workload pulls in capacity, data, networking and contracts. Settle this worksheet before you book.

Evaluation worksheet0 / 6
  • 01Cluster size and networking written down
  • 02GPU family available near your data
  • 03Storage and transfer costs estimated
  • 04Capacity confirmed for your dates
  • 05Contract term and minimums reviewed
  • 06Invoice currency agreed
1 node · 8 GPUs
4 nodes · 32 GPUs

Cluster shape

Decide whether you need one node or many, and how they have to connect.

Worksheet 01
datasetswhere they live€ ?GPUswhere they run

Data location

Check where your datasets live and what it costs to bring them to the GPUs.

Worksheet 02 · 03
Run window9 days
MTWTFSSMTWTFSS
GPUs free on these dates: ?

Capacity dates

Confirm the GPUs are free for the dates of your run, not only in general.

Worksheet 04
term length?
minimum commitment?
invoice currency?

Contract terms

Review term lengths, minimum commitments and the currency on the invoice.

Worksheet 05 · 06

Why Nexpod

Start on one server, finish on a cluster.

Start with a single 8-GPU server today. When the run grows, the same container moves to an Instant Cluster, and the rate per GPU stays the same.
Instant Cluster · H200
64GPUs
8 nodes × 8 GPUs
n0 = your first node
n0
n1
n2
n3
n4
n5
n6
n7
one fabric · torchrun across all nodes
Instant Clusters

Many nodes, one fabric

Join 8-GPU servers into one cluster and launch across all of them with one command. The rate per GPU does not change.

NODE8× H200 · one tenant
8/8 online
0
H200
1
H200
2
H200
3
H200
4
H200
5
H200
6
H200
7
H200
NVLink
all-to-all
Cloud Servers

A whole server, by the hour

Eight GPUs on NVLink, the host CPUs, the RAM and the NVMe, reserved for you from boot until you stop it. Shown here: H200.

B300 · €48.00/hrB200 · €40.00/hrH200 · €32.00/hrH100 · €24.00/hr
Workers · llm-eu
min 0 · max 12
+6 workers
Serverless

Endpoints that follow traffic

Workers start when requests queue and stop when the queue drains. Billed per second, down to zero.

Pricing

Pricing without a calculator.

Pick a family and read one number. On demand or reserved, per GPU-hour, metered per second, invoiced in euros.

See all pricing
GPU
Nexpod
Oracle Cloud
B300
288 GB HBM3e
€6.00/GPU·hr
€48.00 /node
B200
180 GB HBM3e
€5.00/GPU·hr
€40.00 /node
H200
141 GB HBM3e
€4.00/GPU·hr
€32.00 /node
H100
80 GB HBM3
€3.00/GPU·hr
€24.00 /node

Rates differ by region, instance size and term, so there is no single number to put here.

→ check Oracle Cloud's own price list for your region

Indicative. Nexpod on-demand rates, excl. VAT, metered per second.
Placeholder

Placeholder quote. A customer who moved GPU work from Oracle Cloud describes, in two plain sentences, what changed on the invoice and in the day-to-day.

Customer nameRole, Company
8GPUs
in every node
1s
billing granularity
€0
egress on network storage
billing / invoice NX-2026-1213
Sample
Billed to
Customer name
Currency
EUR
Instant Cluster · 4× 8 H200
38h 15m 00s · EU-AMS-1
€4,896.00
Egress from network storage
5.6 TB out
€0.00
VATreverse charge
Total due€4,896.00

Comparison

Oracle Cloud and Nexpod, at a glance.

Our answer on the left. On the right, what to confirm for your Oracle Cloud tenancy, where shape, region and contract set the terms.

Nexpod
Check on Oracle Cloud
What you rent
Whole 8-GPU server
?Which shapes include eight GPUs?
GPU families
B300 · B200 · H200 · H100
?Which families are offered near your data?
Billing granularity
Per second
?How is GPU time metered for your shape?
Invoice currency
EUR
?Which currency does your tenancy bill in?
Egress from network storage
None
?What does outbound transfer cost from your region?
Serverless GPU endpoints
?Which service serves GPU models as endpoints?
Multi-node clusters
?How do cluster nodes connect, and up to what size?
Root over SSH on the node
?Bare metal, or a virtual machine?
Rate shown before you deploy
?Which rate applies under your contract?
Qualitative. Cloud offerings change often; answer these with Oracle Cloud's current documentation and pricing for your region.

Workload fit

Choose where each job runs best.

Plenty of teams keep databases on one cloud and send GPU work to another. Here is how to tell which job goes where.

Nexpod fits when

  • You want to start a GPU server today, without a sales cycle.
  • You want per-second billing for jobs that stop and start.
  • You want one euro invoice for servers, endpoints and clusters.
  • You want the same container to run on one node or eight.

Oracle Cloud may fit better when

  • Your GPU work sits next to Oracle databases you already run there.
  • Existing Oracle contracts already cover your cloud spend.
  • You need other services from the same provider in the same region.

Talk to an engineer

A second opinion in 15 minutes.

Show us the cluster you are planning. We tell you how it maps to Nexpod servers and what it costs in euros, with no obligation.

Book a call
nexpod.ai/contact · 15 min call
GE
GPU engineer
video call · CET
15 min
Mon
Tue
Wed
Thu
Fri
09:0009:0009:0009:0009:0009:3009:3009:3009:3009:3010:0010:0010:0010:0010:0010:3010:3010:3010:3010:3011:0011:0011:0011:0011:0014:0014:0014:0014:0014:00
Fri · 11:00 CET held for you

Try it on one workload

Start small. Scale when it proves out.

Run one node for a day, check throughput and cost, then add nodes or move to an Instant Cluster.

Get startedTalk to sales
On-demand ratesEUR · excl. VAT
  • B300€6.00/GPU·hr
  • B200€5.00/GPU·hr
  • H200€4.00/GPU·hr
  • H100€3.00/GPU·hr