What GPU hardware does Qube Compute use?

We deploy NVIDIA Vera Rubin R100 NVL72 — the most powerful commercially available GPU system with 1,400+ ExaFLOPS FP4 per rack and NVLink 6.0 fabric. We also offer Groq LPX for sub-10ms real-time inference.

How much does GPU cloud cost at Qube Compute?

Anchor contracts start at $14/GPU-package-hour (6-24 month terms). Cloud On-Demand is $19/hr and Spot/Night is $25/hr. Our energy cost of $0.048/kWh makes us 3x cheaper than AWS/Azure.

Is Qube Compute Sharia-compliant?

Yes. We are the world's only AFSA-certified halal GPU cloud. Our Mudaraba profit-sharing structure has zero debt (riba) and no derivatives (gharar). All payments are held in Sharia-compliant escrow at Al Hilal Bank.

Where is the data center located?

Our 8 MW Tier III TIA-942 facility is located in SEZ PIT Alatau, Almaty, Kazakhstan. The Special Economic Zone provides 0% corporate tax, VAT, and personal income tax until 2029.

How are payments protected?

All prepayments are held in escrow at Al Hilal Bank under AIFC English Common Law. Funds are released only upon verified GPU access delivery. If we fail to deliver — automatic full refund.

Birinshy Rubin R100 NVL72 bulut provaideri

AI Superbulut masshtabqa arnalgan

Janga buyn NVIDIA Rubin R100 NVL72 + Groq LPX naqty uaqyttagy inferens. $0.048/kVt·s. SEZ-de salyq joq.

GPU qol jetimdilik 2027 shilde ayinan

GPU brondau Satylym bolimimen baylanys

MVt quaty

$0.048

/kVt·s energiya

<10

ms keshiguy

salyq SEZ-de

Nege koshbasshylar Qube Compute tandaidy

Janga buyn onimdilik

NVIDIA Vera Rubin R100 NVL72 — H100-den token basyna 5 ese quatty. 1,400+ ExaFLOPS FP4. NVLink 6.0.

Tendesissiz ekonomika

$0.048/kVt·s gaz generatsiyasy — AWS/Azure-den 3 ese arzan. ABHM salyqtyru PUE 1.10. SEZ-de 0% salyq.

Halal-saikestik

Alemdepi jeke AFSA-sertifikattalgan halal GPU bulut. Mudaraba qurylymy. $4.5 trln islamdyq qorlarqa qol jetimdilik.

Developer-First API

Deploy GPUs in Seconds

Full REST API, Python SDK, Node.js SDK, and CLI. Spin up NVIDIA Rubin R100 NVL72 instances or call Groq LPX inference with a single command.

Get API Key

Launch Instance

Choose GPU type, count, and container image

Train or Infer

SSH in for training, or use Groq API for <10ms inference

import qube

client = qube.Client(api_key="your-api-key")

# Launch a GPU instance
instance = client.instances.create(
    gpu_type="rubin-r100-nvl72",
    gpu_count=8,
    image="nvidia/pytorch:24.04",
    region="almaty-sez"
)

print(f"Instance {instance.id} is {instance.status}")
print(f"SSH: ssh root@{instance.ip}")

# Run inference with Groq LPX
response = client.inference.create(
    model="llama-3.1-70b",
    messages=[{"role": "user", "content": "Hello!"}],
    max_tokens=512
)
print(response.choices[0].message.content)
# Latency: 8ms

api.qubecompute.com

SDK Access: Private Beta for LOI Signatories

Bes jensiz artykshylyq

Esh bulut provaideri bu kush tustardy biriktirmeydi

months monopoly

First Rubin R100 NVL72 Cloud Provider

No NVL72 in KZ, UZ, KG. Nearest competitor with Rubin — Finland (Nebius), UAE (G42). 18-24 month exclusive window.

/kWh

3x Cheaper Energy

Gas-powered generation at $0.048/kWh vs $0.09-0.18 at Western competitors. ABHM absorption cooling, PUE 1.10.

addressable market

Only Sharia-Compliant GPU Cloud

AFSA/AIFC certified Mudaraba structure. Access to PIF ($925B), QIA ($475B), ADIA ($993B) sovereign wealth funds.

taxes until 2029

SEZ Tax Exemptions

0% corporate tax, VAT, personal income tax in SEZ PIT Alatau. Government investment contract guarantees.

ms latency

Groq LPX Real-Time Inference

Sub-10ms LLM inference API globally. Unique Rubin + Groq combination for training + real-time serving.

Performance

Benchmark Comparisons

NVIDIA Rubin R100 NVL72 delivers up to 5x more performance per dollar compared to H100. Combined with Groq LPX for inference — unmatched speed and efficiency.

LLaMA 3.1 70B Training

Time to train (1T tokens)

Rubin R100 NVL72~3 days

H100 SXM (8×)~15 days

A100 SXM (8×)~38 days

Inference Throughput

Tokens/sec (LLaMA 70B)

Groq LPX~3,000 tok/s

Rubin R100~800 tok/s

H100 TensorRT~350 tok/s

A100~120 tok/s

Memory Bandwidth

Per rack

Rubin R100 NVL72468 TB/s

GB200 NVL72~380 TB/s

H100 SXM (8×)26.4 TB/s

FP4 Performance

Per rack

Rubin R100 NVL721,400+ ExaFLOPS

GB200 NVL72~720 ExaFLOPS

H100 SXM (8×)~16 ExaFLOPS

* Benchmark estimates based on NVIDIA published specifications and industry testing. Actual performance may vary by workload. Rubin R100 NVL72 specs from NVIDIA GTC 2025 announcements.

70%-ga deiyin ynemdeu

Qansha ynemeytininiyzdi koriniyz

Qube Compute-ti irbi GPU bulut provaiderlermen salyqtyrynyz. Rubin R100 + arzan energiya = tendesissiz ekonomika.

GPU sany8 GPUs

172 (full rack)

Kunine sagat24h/day

1h24h

Provaider	GPU	$/GPU/sagat	Ailyq qun	Zhyldyq qun	Ynemdeuiniyz
Qube ComputeEn zhaqsy baga	Rubin R100	$14	$80,640	$967,680	—
AWS (p5.48xlarge)	H100	$163.85 ($32.77 x 5hrs)	$943,776	$11,325,312	91% $10,357,632/yr
Azure (ND H100 v5)	H100	$136.00 ($27.20 x 5hrs)	$783,360	$9,400,320	90% $8,432,640/yr
GCP (a3-highgpu-8g)	H100	$156.10 ($31.22 x 5hrs)	$899,136	$10,789,632	91% $9,821,952/yr
CoreWeave	H100	$23.80 ($4.76 x 5hrs)	$137,088	$1,645,056	41% $677,376/yr
Lambda Cloud	H100	$17.45 ($3.49 x 5hrs)	$100,512	$1,206,144	20% $238,464/yr

* Qube Compute yakorldyq baga ($14/GPU-paket/s) paidalanady. Baqtalastar bagalary 2025 zh. ashyq prais-tizimderden.

GPU esepteulerden 3 ese ynemdeuge daiynbyz?

Klient agymy

Korporativtik suranysh qalypta

4 sektorda kasiporndarmen belgendi LOI jane kelissozder. 1-Faza syiymylygy (8 stoika) yakorldyq klientterge bolinudi.

LOI agymda

576+

GPU suralgan

$8M+

Kelisimsharttyyn zhyldyq quny

Sala sektorlary

Belgendi klient agymy (attar NDA astynda)

Sektor	Aimaq	Qoldanu	GPU syiymylygy	Statys
Qarazhylyq qyzmetterge	Tauyq Shygysh	Antifraud ML, kredit skoringi	Birneshe stoika	LOI orynedaluda
Energetika men onerkasip	Euraziya	Boldau analitikasy, optimizatsiya	Arnaiy stoika	LOI orynedaluda
Memleketik sektor	Euraziya	Ulttyq AI platformasy, NLP	Kop stoykalyq klaster	Kelissozde
Densaulyq saqtau men bioteh	Tauyq Shygysh	Dari izdeuy, diagnostika	Arnaiy stoika	Kelissozde

Klient attary NDA-men qorghalgan. Egzheilikter investor NDA boiynsha surau salasynda qolzhetimdi.

Arber salaga arnalgan AI sheshimder

En talap etti juktemelerge arnalgan infraqurylymy

Oil & Gas AI

Seismic analysis, predictive maintenance, and well optimization for KMG, TCO, Karachaganak.

Financial AI

Anti-fraud ML, algorithmic trading signals, and risk scoring for banks and fintech.

Government AI

National digitalization projects, NLP for public services, and smart city infrastructure.

Healthcare AI

Real-time diagnostics, drug discovery, and molecular dynamics on Groq LPX under 10ms.

Tehnologiyalyq seriktester men ekojuye

AI infraqurylymy, qarzhy jane komplaens salasyndagy alem koshbasshylarimen qurylan

NVIDIA

GPU Partner

Groq

Inference Partner

AIFC

Legal Framework

Al Hilal Bank

Escrow Partner

ComplyAdvantage

KYC/AML

Astana Hub

Tech Ecosystem

Built for Every Industry

Serving enterprises globally — from AI startups to Fortune 500

AI Startups

Fortune 500

Sovereign Wealth Funds

Fintech

Healthcare AI

Oil & Gas

Research Labs

Government AI

AI-dy masshtabtauga daiynbyz?

1-Faza syiymylygy shekteulik — 8 stoika. Yakorldyq bagamen brondanyz.

GPU qol jetimdilik 2027 shilde ayinan. Yakorldyq baga ushin qazir brondanyz.

GPU brondau Investitsiyalyq materialdardy surau