Rent a 24GB GPU in India: INR per hour

A 24 GB GPU runs most 4-bit local models up to about 32B parameters. If you only need one for a few hours or days, renting is usually cheaper than buying. This page lists published rupee prices from providers that run GPUs in India, plus Runpod and Vast.ai converted from USD. Prices checked on 1 November 2026.

Quick answer

On 1 November 2026 the lowest published hourly rates for a 24 GB card at India-based providers were ₹29/hr for an RTX 3090 on GHive^[1] and ₹30/hr on GridShare^[2], and ₹47–48/hr for an RTX 4090 on the same two sites.^[1]^[2] Both take UPI. For datacenter 24 GB cards, Jarvislabs lists the A30 at ₹38.88/hr and the L4 at ₹41.31/hr.^[3] Vast.ai marketplace offers can be lower, but they bill in USD to an international card or crypto.^[7]^[9]

24GB GPU prices in India (INR per hour)

Provider	GPU (24 GB)	Price	Billing	Tax shown	Pay with
GHive	RTX 3090	₹29/hr^[1]	Per second, prepaid wallet	Not stated on price page	UPI, card, netbanking^[1]
GHive	RTX 3090 Ti	₹41/hr^[1]	Per second	Not stated	UPI, card, netbanking
GHive	RTX 4090	₹47/hr^[1]	Per second	Not stated	UPI, card, netbanking
GridShare	RTX 3090	₹30/hr^[2]	Per second	Ex-GST; 18% added at top-up^[2]	UPI, card^[2]
GridShare	RTX 4090	₹48/hr^[2]	Per second	Ex-GST	UPI, card
Jarvislabs	NVIDIA A30	₹38.88/hr^[3]	Per minute	GST invoice issued^[3]	Card via Stripe^[3]
Jarvislabs	NVIDIA L4	₹41.31/hr^[3]	Per minute	GST invoice issued	Card via Stripe
E2E Networks	NVIDIA L4	$0.57/hr (≈ ₹54)^[4]	Per GPU-hour	Excludes taxes^[4]	UPI, net banking, wallets^[4]
E2E Networks	NVIDIA A30	$1.35/hr (≈ ₹128)^[4]	Per GPU-hour	Excludes taxes	UPI, net banking, wallets

GridShare notes that a machine with 32 GB or more of system RAM can add ₹2–4/hr.^[2] GHive and GridShare are marketplaces of hosted machines, so a listed card may not be online when you look.^[1]^[2] E2E's pricing page showed USD to our checker; INR figures are converted at ₹95.05 per USD.^[8]

Bigger GPUs, for context

Provider	GPU	Price
GHive	A100 40GB / A100 80GB / H100 80GB	₹71 / ₹79 / ₹208 per hr^[1]
Jarvislabs	A100 40GB / A100 80GB / H100 SXM	₹84.24 / ₹140.94 / ₹330.48 per hr^[3]
E2E Networks	A100 80GB / H100	$2.10 / $3.99 per hr, ex-tax^[4]

Indian providers vs Runpod and Vast.ai

Provider	GPU	USD/hr	≈ INR/hr
Runpod	RTX A5000 (24 GB)	$0.27^[6]	₹25.7
Runpod	RTX 3090	$0.50^[6]	₹47.5
Runpod	L4	$0.59^[6]	₹56.1
Runpod	RTX 4090	$0.89^[6]	₹84.6
Vast.ai (16 single-GPU offers)	RTX 3090	lowest $0.12, median $0.18^[7]	₹11.6 / ₹16.6
Vast.ai (48 single-GPU offers)	RTX 4090	lowest $0.27, median $0.46^[7]	₹25.6 / ₹44.1

Converted at ₹95.05 per USD.^[8] Vast.ai prices are live marketplace offers that change by the minute. USD prices do not include your bank's foreign-exchange markup or any GST you owe on imported services (see below). Runpod and Vast.ai GPUs are not in India unless a host happens to be.

Where to rent or buy

Run it on a cloud GPU

Vast.ai Affiliate partner Consumer GPU marketplace (4090/3090-class rentals)
Runpod Affiliate partner Pods and serverless GPUs (referral link)
E2E Networks Works with India GPU cloud (official site)

Run it via API instead

Novita AI Affiliate partner Open-model APIs and Sandbox for AI agents (no GPU rental)

Buy this GPU (or a Mac)

B&H Photo Works with GPUs and computing gear (official site)

We may earn from links marked Affiliate partner. Disclosure.

RTX 3090 vs 4090 vs L4 vs A30

All four cards have 24 GB of memory, so they fit the same models. The 4090 is the fastest consumer card here; the L4 and A30 are datacenter cards offered by Jarvislabs and E2E. See our measured speeds before paying more for speed.

Paying with UPI and Indian cards

GHive and GridShare take UPI, and E2E lists UPI, net banking and wallets for Indian customers.^[1]^[2]^[4] Jarvislabs takes cards through Stripe and bills in INR.^[3] Runpod accepts cards through Stripe, crypto, and invoicing for deposits over $5,000.^[9] Vast.ai accepts cards through Stripe and crypto through BitPay and Crypto.com.^[10] Neither lists UPI.

Why your Indian card gets declined, and how to fix it

Since 2020, RBI rules require new and reissued Indian cards to be enabled only for domestic ATM and point-of-sale use. Online and international use stay off until you switch them on, and existing cards that were never used online or abroad were switched off.^[11] Turn on online and international use in your bank's app or net banking first; HDFC, for example, says the change is instant.^[12]

Runpod's help pages also say:^[9]^[13]

Fraud checks often flag first-time or international payments; ask your bank whether it blocked the charge.
Don't try several cards in a row. Repeated failures can block payments for about 24 hours, and Runpod blocks a payment method after a third failed attempt.
Prepaid cards should deposit at least $100 at a time.
Crypto is the fallback, after KYC with the payment processor.

GST on GPU cloud bills in India

Cloud services carry 18% GST.^[14] Indian providers add it to your bill (GridShare adds it at top-up^[2]), and a GST-registered business can claim input tax credit with its GSTIN on the invoice.^[2]^[15] When a GST-registered business buys from a foreign provider like Runpod or Vast.ai, it pays the 18% itself under reverse charge and can then claim credit.^[14]^[15] This is general information, not tax advice.

What models fit in 24 GB VRAM?

From our own model data (Q4_K_M file sizes, before context):

Model	Q4_K_M file	Fits in 24 GB?
Llama 3.1 8B	4.9 GB	Yes, with room for long context
Qwen3 14B	9.0 GB	Yes
Gemma 4 31B	19.6 GB	Weights fit; KV cache not yet sourced
Qwen3 32B	19.8 GB	Yes, with about 8K context (KV cache is about 256 KB per token)

Llama 3.1 70B at Q4_K_M is a 42.5 GB file, so 70B models do not fit on one 24 GB card. See what fits in 24 GB for the full list, or check any model in the calculator.

How to rent for a short job

Check your model's VRAM in the calculator.
Pick a 24 GB card from the table above.
Top up only what the job needs; GHive and GridShare bill by the second from a prepaid wallet.^[1]^[2]
Download your outputs before you stop. GHive erases instance disk when the machine stops.^[1]

IndiaAI compute portal

The government's IndiaAI compute portal lists subsidised GPU rates, but you have to apply and qualify. Eligible groups include DPIIT-recognised startups and MSMEs with AI/ML experience and ₹50 lakh revenue or ₹1 crore funding, researchers meeting publication criteria, students with AI/ML projects, IndiaAI Fellows and government entities.^[5]^[16] It is not a pay-as-you-go option for individuals.

FAQ

What is the cheapest RTX 4090 rental in India?

On 1 November 2026, GHive listed the RTX 4090 at ₹47/hr and GridShare at ₹48/hr ex-GST.^[1]^[2] Vast.ai's cheapest single-4090 offer was about ₹26/hr, billed in USD.^[7]

Can I pay Runpod or Vast.ai with UPI?

No. Neither lists UPI. Runpod takes cards and crypto; Vast.ai takes cards, BitPay and Crypto.com.^[9]^[10]

Why is my Indian card declined on Runpod?

Usually because online or international use is switched off on the card, or the bank's fraud check blocked it. Enable both in your bank app, then retry once.^[11]^[13]

Is there GST on GPU rental?

Yes, 18%. Businesses registered for GST can claim input tax credit.^[14]^[15]

What models can I run on 24 GB?

Most 4-bit models up to about 32B, including Qwen3 32B with modest context. 70B models need more.

Who can use the IndiaAI compute portal?

Approved startups, MSMEs, researchers, students and government bodies that meet IndiaAI's criteria.^[16]

Prices change often. Always confirm on the provider's page before you rent. We earn a commission from Runpod and Vast.ai referral links; links to Indian providers are plain official links (Disclosure).

More India pages

Sources

All sources checked on 2026-11-01.