Best VPS for Algo Trading in India: What to Check in 2026

The best VPS for algo trading in India depends on how you trade. For a Python bot on a broker API, look for a Linux server in India with a dedicated IPv4 address that only you use and that stays the same across reboots, because since 1 April 2026 brokers accept API orders only from an IP you have registered. For AmiBroker or MetaTrader you need a Windows VPS. If you don't write code, a broker-hosted platform needs no server at all.

Below is the checklist, followed by a comparison of real options with prices taken from each provider's own pricing page as of October 2026.

First, check whether you need a server at all

NSE's implementation standards (circular NSE/INVG/67858) make a static IP mandatory for API access when you wrote the algo yourself. NSE's retail algo FAQ of November 2025 adds that a client static IP is needed only by a "tech savvy" investor placing orders through an API. When an empanelled vendor's platform is hosted on the broker's infrastructure, the IP that counts is the broker's server.

So if you trade through your broker's app, its own algo product or a platform hosted on its servers, you don't need a VPS. Our static IP guide for algo trading explains the IP rule in full, and SEBI algo trading rules covers the rest of the framework. If you run your own code, read on.

The IP address: dedicated, stable and IPv4-first

Dedicated to you. Under NSE's standards, a static IP can be mapped to only one client at a time, although family members as SEBI defines them may share one if the broker allows it. That rules out shared hosting, shared VPNs and proxies where many customers leave from one address.

Stable across reboots and reinstalls. You register a primary IP and an optional secondary one, and you can change them no more than once a calendar week. An address that changes without warning can stop your bot for days, and providers differ a lot:

• AWS EC2 releases the automatic public IPv4 when you stop an instance and assigns a new one on start. You need an Elastic IP to keep it.
• AWS Lightsail changes the default IP on every stop and restart unless you attach a static IP.
• Oracle Cloud keeps an ephemeral public IP when you stop the instance but deletes it when the instance is terminated. A reserved public IP survives that.
• DigitalOcean keeps a Droplet's addresses for the life of the Droplet, including a rebuild.
• HostEthical Algo VPS keeps its dedicated IPv4 across reboots and operating-system reinstalls.

Moving to a new server almost always means a new IP and another registration, so pick a provider you are happy to stay with.

IPv6 behaviour. Many servers have IPv6 as well as IPv4, and when the broker's API host has both, Linux usually tries IPv6 first. Your order then leaves from an address you never registered and is rejected. Some brokers accept only IPv4 for now, including Angel One, according to its SmartAPI notice. Ask how the server handles IPv6, then test which address your code connects from. The static IP guide has the test commands and the one-line /etc/gai.conf fix.

Location, power and uptime

Choose an Indian data centre. Your bot talks to an Indian broker, so a server in Mumbai, Hyderabad or Bengaluru keeps each request inside India. OpenAlgo's hosting guide recommends Mumbai for Indian exchanges. Keep expectations realistic: a retail VPS is not exchange colocation, and no location makes a strategy profitable.

Don't rely on anyone's latency claims, ours included. Measure from a trial server during market hours, using your broker's API host:

curl -o /dev/null -s -w 'connect %{time_connect}s  total %{time_total}s\n' https://api.kite.trade

Why not the PC at home? It can work with a static IP from your ISP, but your bot stops whenever the power, the router or the broadband does, and an automatic update can restart the machine mid-session. A data-centre server has backup power and keeps running when your laptop is shut.

Linux or Windows: match the server to your software

Python bots belong on Linux. The broker SDKs (kiteconnect, dhanhq, fyers-apiv3, upstox-python-sdk, smartapi-python) are ordinary Python packages, systemd runs bots as services, and Linux plans cost less because there is no Windows licence. Amazon Lightsail's price list shows the gap: a 2 GB bundle is $12 a month on Linux and $22 on Windows, as of October 2026.

AmiBroker and MetaTrader want Windows. AmiBroker is Windows software: its guide lists Windows 11, 10, 8 and 7 and recommends 4 GB of RAM or more. MetaQuotes runs MetaTrader 5 on Linux through Wine, a compatibility layer, which adds one more moving part to a live trading machine. If you depend on Windows software, buy a Windows VPS. HostEthical's Algo VPS is Linux only, so it isn't the right choice for that.

One middle path: self-hosted OpenAlgo runs on Linux, takes signals from AmiBroker or TradingView and places orders from the server's IP. Your signal source still has to be running. See how to install OpenAlgo on a VPS.

How much CPU and RAM a Python bot needs

Usually less than you'd expect. A typical bot spends most of its time waiting for the market to open, a candle to close or the broker to reply. CPU matters when you compute indicators across many symbols, backtest or run several bots. Memory grows with the data you hold: a short watch-list in pandas is small, a day of ticks for hundreds of instruments is not.

Reference points from the software makers: OpenAlgo's guide calls 2 GB of RAM the minimum for OpenAlgo and 4 GB comfortable, and AmiBroker recommends 4 GB or more. The smallest cloud plans have 512 MB, which is fine for experiments, but check your bot's real usage before relying on one. Run it on a test server for a full trading day, then look:

free -h                                  # RAM used and free on the whole server
ps -eo rss,args --sort=-rss | head -5    # the biggest processes, memory in KB
systemctl status algo@mybot              # for a systemd service, see the "Memory:" line

Leave headroom for the morning login and a second strategy. On AWS, T3 and T4g instances launch in "unlimited" mode by default: if average CPU stays above the baseline over 24 hours, surplus credits cost $0.05 per vCPU-hour on T3 and $0.04 on T4g for Linux.

Price, billing and payment in India

• Currency. AWS prices are in US dollars, but an account with an Indian billing address is with AWS India, and the invoice total is listed in rupees. DigitalOcean says USD is the only currency it accepts. HostEthical bills in rupees.
• Payment methods. AWS India accepts Visa, Mastercard, American Express and RuPay cards, net banking and UPI, with automatic card payments only for bills up to ₹15,000. DigitalOcean accepts cards, PayPal and Google Pay, among others, and Indian cards must be enabled for international transactions; that page doesn't mention UPI. Oracle's free tier needs a credit card, or a debit card that works like one, for identity checks, and its FAQ rules out PIN debit, virtual and prepaid cards. HostEthical takes UPI, cards, netbanking and crypto.
• Separate charges. On AWS EC2, the IPv4 address, disk and extra data transfer are billed on top of the instance. Most VPS hosts bundle them into one price.
• Introductory prices. Some hosts advertise a low first term that renews higher. Compare the renewal price too.

For a wider look at budget servers, see cheap VPS hosting in India.

Root access, support and backups

Root access lets you install packages, set the time zone, control IPv6 and run your bot as a service. Every server option below gives it.

Managed or unmanaged. Almost every algo-trading server is unmanaged: the host keeps the hardware and network running, and you look after the operating system and your code. If that is new to you, read managed vs unmanaged VPS, and see how to run a Python trading bot 24/7 for systemd, timers, alerts and restarts.

Support. Check how you reach support and when, because a problem at 9:10 IST can't wait for an overnight reply. No host will debug your strategy.

Backups. Keep your code in a private git repository and your API secrets out of it. Server backups usually cost extra: DigitalOcean charges 20% (weekly) or 30% (daily) of the Droplet price, AWS charges $0.05 per GB-month for snapshot storage in Mumbai, and Oracle's Always Free tier includes five volume backups.

VPS options for algo trading in India compared (October 2026)

Prices are from each provider's own pricing page, as of October 2026, in the currency they quote and before tax. Monthly cloud figures assume 730 hours.

AWS. In Mumbai (ap-south-1), a t4g.small (2 Arm vCPUs, 2 GiB) and a t3.micro (2 vCPUs, 1 GiB) both cost $0.0112 an hour. Add the public IPv4 charge of about $3.65 a month, charged whether the address is in use or idle, and about $1.82 for 20 GB of gp3 disk. That is roughly $13.65 a month before tax. Check that your packages install on Arm before choosing t4g. Lightsail bundles everything into one price, but Mumbai plans get half the data-transfer allowance of other regions. New accounts get $100 in Free plan credits, with up to $100 more, but a Free plan account closes after six months or when the credits run out, unless you upgrade.

Oracle Cloud Always Free. Oracle has regions in Mumbai and Hyderabad, and free servers must run in your home region. The Always Free tier covers two small AMD servers (1/8 OCPU and 1 GB each) or Arm servers worth 1,500 OCPU hours and 9,000 GB hours a month, which Oracle equates to 2 OCPUs and 12 GB. Its docs warn of "out of host capacity" errors when free shapes run short. Oracle may also reclaim an instance as idle if, over 7 days, its 95th-percentile CPU use and its network use both stay below 20% (and memory too, on Arm). A bot that waits most of the day can easily look idle, so this is a good place to learn and a risky one for live orders.

DigitalOcean Bangalore. BLR1 is still open to new Droplets, including Basic plans, which start at $4 a month for 1 vCPU, 512 MiB and 10 GiB of SSD. It is a simple, cheap Linux option if you are happy paying in dollars.

Windows VPS. If you need AmiBroker or MT5, this is the better fit. Several Indian hosts sell Windows VPS plans with RDP access. Before you buy, ask which city the server is in, whether the IP is dedicated to you, whether the Windows licence is included and what the plan renews at. Lightsail Windows in Mumbai is an alternative with a published price.

ISP static IP and broker platforms. Few ISPs publish a home static IP price; the figures above come from one trader's reports in a TradingQnA thread in March 2026, so ask your own provider. A broker-hosted platform needs no IP from you, but you are limited to what the platform supports.

HostEthical Algo VPS Mumbai. It has 4 vCPUs, 8 GB of RAM, 100 GB of SSD and a 200 Mbit/s port in Mumbai, with a dedicated IPv4 address that nobody else uses. Ubuntu 24.04 is set up at first boot: IST time with the clock kept in sync, a Python environment with the Zerodha, Dhan, FYERS, Upstox and Angel One SDKs, and IPv4 preferred for outgoing connections. The algo-ip command prints the IP to register and warns you if Python connects from another address. A systemd template (algo@<name>) runs each bot as a service, and the firewall leaves only SSH open. Your dashboard shows the IP beside links to each broker's API settings. It is unmanaged and Linux only, billed monthly or yearly.

To be fair about it, 4 vCPUs and 8 GB is more than one small bot needs; the headroom is for several strategies, OpenAlgo or a database. If you would rather set things up yourself, our plain Cloud VPS uses the same hardware class from ₹860 a month, or about ₹774.17 a month when billed yearly, and Mumbai adds a monthly location fee.

Which option fits which trader

• Learning or testing a strategy: Oracle Always Free or AWS Free plan credits. Move to a paid server before you trade live.
• You don't write code: a broker-hosted platform. No server, no IP.
• You use AmiBroker, MT4 or MT5: a Windows VPS from an Indian host, or Lightsail Windows in Mumbai. Not our Algo VPS.
• Your desktop is on all day with reliable power: an ISP static IP is the cheapest choice.
• You already work in AWS: EC2 or Lightsail in Mumbai, with an Elastic or static IP attached before you register it.
• You want the cheapest Linux server and can pay in dollars: DigitalOcean in Bangalore.
• You run Python bots and want a server that's ready when you log in, billed in rupees: Algo VPS Mumbai. Our Kite Connect and Dhan API guides cover the broker side.

Whichever you pick, check three things before you register the IP: that it is yours alone, that it survives a reboot, and that your code really connects from it.

Frequently asked questions

Do I need a VPS for algo trading in India?

Only if your own code places orders through a broker API. Since 1 April 2026 those orders must come from a static IP you have registered with your broker, and a VPS is the usual way to get one that runs all day. If you trade through your broker's app, its own algo product or a platform hosted on the broker's servers, you don't need a server or an IP of your own.

Is a Linux or Windows VPS better for algo trading?

Linux is the better fit for Python bots, because the broker SDKs and systemd run natively on Ubuntu, and Linux plans cost less without a Windows licence. Choose a Windows VPS if you rely on Windows software such as AmiBroker or a MetaTrader terminal. HostEthical's Algo VPS is Linux only, so Windows users should look at a Windows VPS from another provider.

Will my VPS IP address change after a reboot?

It depends on the provider, so check before you register the IP with your broker. On AWS EC2 the automatic public IPv4 changes when you stop and start the instance unless you attach an Elastic IP. A DigitalOcean Droplet keeps its IP for its lifetime, including rebuilds. HostEthical's Algo VPS keeps the same IPv4 across reboots and reinstalls. Brokers let you change the registered IP only once a calendar week.

Can I use a free cloud server for algo trading?

Yes for learning and testing, but it is risky for live orders. Oracle Cloud's Always Free tier offers small servers with a public IP in Mumbai or Hyderabad, but Oracle may reclaim instances that look idle over seven days, and free capacity can run out. AWS's Free plan gives credits, but the account closes after six months or when the credits run out unless you upgrade.