Running Telegram at Scale: Proxies, VPS, Antidetect & Multi-Account Infra

Running one Telegram account on your laptop is trivial. Running 10-50 accounts for OFM automation requires infrastructure: proxies, antidetect browsers, a VPS, and a monitoring setup. Get any of it wrong and the accounts die, not because of bad warmup or junk CTAs, but because Telegram's anti-spam detects that "Sarah from Miami" and "Emma from LA" are coming from the same Amsterdam datacenter IP, same browser canvas fingerprint, same Windows timezone.

This guide is the infrastructure deep-dive. It covers the three-legged stool (phone + IP + fingerprint), proxy choices, VPS setup for Cupid, antidetect configuration, density limits, and what breaks when you scale from 1 → 50.

---

1. The infrastructure triangle, phone + IP + fingerprint

Telegram's account-flag heuristic reads three signals and looks for coherence:

1. Phone number country and operator, "this is a Swedish Telia SIM."
2. IP geography, "this user is connecting from Stockholm, Sweden."
3. Device/browser fingerprint, "this is Chrome on Windows 11 with Swedish timezone, Swedish locale, typical Swedish screen resolution."

If all three agree: Telegram thinks you're a normal user. Score: clean.

If one disagrees (US SIM + US IP + German-locale browser): mild flag. Survivable if warmup is good.

If two disagree (US SIM + Brazilian datacenter IP + Swedish browser): heavy flag. Account likely dies in days.

If all three disagree: dead on arrival.

Common mistake: operators obsess over one leg (say, the proxy) while completely ignoring the others. A $15/mo mobile proxy with a mismatched browser fingerprint is worse than a $3/mo datacenter proxy that matches everything else. Coherence > quality per leg.

---

2. Dedicated proxy per account vs shared

Question from the data: "Do you guys recommend putting each Telegram on a different proxy or as long as you press 'new fingerprint' in adspower you should be fine?"

Clean answer: every account needs its own proxy OR at most 2-3 accounts behind a very clean proxy. "New fingerprint in AdsPower" without a proxy change is not enough, the IP still gives you away.

Proxy-sharing reality:

• 1 account per proxy: safest, most expensive. Standard for high-value accounts.
• 2-3 accounts per proxy: acceptable for warming / burner accounts on a clean ISP proxy. Telegram tolerates some density.
• 5+ accounts per proxy: Telegram's IP heuristics flag this. The accounts may verify fine but fall in days.
• 50+ accounts per proxy (what some bad "bulk" setups do): pure account churn, every account dies on day 2-5.

Proxy types, ranked by OFM survivability:

1. Mobile proxies (4G/5G cellular IPs). Rotating or sticky. Highest quality, $30-80/mo per IP. Telegram sees them as regular mobile users. Best for hero accounts.
2. ISP / static residential proxies. Look like home ISP IPs but are sold for commercial use. $5-15/mo per IP. Good for production accounts.
3. Rotating residential proxies. Different IP each session, pulled from real residential pool. $5-15 per GB or per IP. Good for warming but note: IP change during a session = instant flag.
4. Datacenter proxies. Cheap ($1-3/mo). Telegram's blocklists hit most datacenter ranges. Usable for burners only.

For production OFM: one ISP proxy per account, same country as the phone number. Costs $5-15/mo/account. Worth it.

---

3. VPS setup for running Cupid 24/7

Cupid and similar chatbots need to run continuously. Your laptop isn't viable (sleep, reboots, disconnects). A VPS is the answer.

Spec minimums per Telegram account (running Cupid):

• RAM: 1-2 GB per account with an antidetect browser profile open. 10 accounts = 20 GB RAM minimum (not accounting for OS overhead).
• CPU: 2-4 cores per 10 accounts. Cupid itself is light; the browser automation is heavy.
• Disk: 50 GB per 10 accounts (browser profiles and caches bloat).
• Bandwidth: 500 GB-1 TB/mo for 10 accounts doing DM volume.

Windows Server vs Linux:

• Windows Server VPS (Hyper-V or dedicated): easiest for AdsPower / Dolphin / most antidetect tools which ship as Windows apps. Most OFM operators run Windows VPS.
• Linux VPS + Wine or Docker Chromium: works but painful. Only worth it for specific stacks (playwright, custom automation).

Recommended spec: 8 vCPU / 32 GB RAM / 200 GB SSD Windows Server VPS handles 10-15 Telegram accounts comfortably. Cost: $60-150/mo depending on provider.

Why Cupid runs slower on VPS than local:

Question from the data: "When running AdsPower Cupidbot on Telegram locally it works fast and well, when on the VDS it seems to message the fans way less."

Causes:

• VPS CPU is shared, "8 vCPU" on a cheap VPS can be 4 physical cores heavily oversubscribed. You get bursts, not sustained performance.
• Network latency to Telegram datacenters is higher than from your local ISP.
• Antivirus / security on the VPS throttles browser automation (check if Windows Defender is scanning the browser profile constantly).
• Single shared IP for outbound, if the VPS's IP is rate-limited, your requests queue.

Fixes:

• Run fewer accounts per VPS than the spec sheet suggests.
• Use a provider with guaranteed dedicated cores (OVH, Hetzner dedicated).
• Exclude the antidetect browser folders from antivirus.
• Ensure each Telegram session uses its own proxy (not the VPS's IP).

---

4. Antidetect browser configuration

The options (roughly ranked by OFM adoption):

• AdsPower. Most-used in OFM. Supports 10-50+ profiles easily, integrates with most automation stacks. $9-50/mo depending on profile count.
• Dolphin Anty. Comparable to AdsPower. Some operators prefer it for better Telegram webapp handling.
• Sunbrowser. Less common but cheaper. Used by some small operators.
• GoLogin. More geared toward ecommerce / social but works for Telegram.
• Multilogin. Premium tier, $99+/mo. Used mostly by enterprises.

Per-profile fingerprint settings that matter:

• Canvas / WebGL fingerprint: randomize per profile (all good tools do this).
• User agent: match the "country" you're claiming. Don't use Japanese Chrome UA for a US persona.
• Timezone: match the account country.
• Language / locale: match the country.
• Screen resolution: pick common resolutions for the persona's implied device (1920x1080 desktop, 390x844 for iPhone-emulated).
• WebRTC IP leak: set to "proxy", so the IP from WebRTC matches the proxy, not the VPS.
• DNS leak: route DNS through the proxy.

The most-missed setting is WebRTC. It's what betrays the mismatched IP even when everything else is configured. Check it with browserleaks.com or similar.

---

5. How many accounts per IP

Real question from data: "how many accs per static ISP are you all doing on Telegram?"

Observed tolerance:

• 1 account per static ISP IP = safe.
• 2-3 accounts per static ISP = acceptable if they're warmed separately and messages don't overlap heavily.
• 5+ accounts per static ISP = Telegram's anti-spam starts flagging the IP for "concentrated bot activity."
• 10+ accounts per static ISP = the IP itself gets shadow-banned; accounts verify but fail to send DMs.

For mobile proxies, the tolerance is higher (many real users do share mobile IPs through NAT), up to 5-8 accounts per sticky mobile IP is often fine.

The density ceiling isn't about Telegram's accounting of "accounts per IP" directly, it's about the aggregate behavioral signal from that IP. If the IP produces DM patterns that look coordinated (same hours, same content styles, mass-DM bursts), Telegram flags the IP regardless of how many distinct accounts sit behind it.

---

6. Region matching, phone + IP + device locale

Real question from data: "I have an iPhone with US SIM card, I want to add my Telegram account in that phone but my phone number on Telegram is Swedish, is it gonna affect the phone."

The phone itself doesn't matter, the iPhone is just a device, its SIM is the US one. What Telegram cares about:

• Which number is the Telegram account registered under? Swedish (+46).
• Which IP is the Telegram app connecting from? If you're physically in the US, the US home/cell IP is showing a Swedish account logging in from the US. Fine for a real user on vacation, but for a bot account this is the kind of mismatch that flags.

For OFM production accounts: always pair the phone number country with a proxy in the same country. Swedish number + Swedish proxy = clean. Swedish number + US home WiFi = mild flag. Swedish number + German datacenter = heavy flag.

For a real user (you, traveling with your personal Telegram), Telegram is lenient. For a fresh account that was just created, it's strict.

---

7. What to do when your VPS IP is flagged

Real question from data: "If the Telegram my Cupid bot was running on got frozen, is there a possibility I'd have to buy a new VPS?"

Depends on what "running on" means:

• If Cupid was running on your VPS and each Telegram account had its own proxy (the correct setup): losing one account has zero blast radius. Just rotate that one proxy+account. VPS is fine.
• If Cupid was running on your VPS with all Telegram accounts using the VPS's native IP (the wrong setup): losing one account means the VPS's IP is flagged. Next account you create from that IP is already pre-flagged. You need either a new VPS (new IP) or you need to add proxies correctly.

Rule: never use the VPS's native IP as the Telegram-visible IP. Always route each account through its own proxy, even if the proxy is in the same country as the VPS.

---

8. Scaling from 1 → 10 → 50 accounts

Stage 1 (1-3 accounts): runs on your laptop + one mobile proxy or home connection. No VPS needed. Manual Cupid, daily check-in. Cost: <$30/mo.

Stage 2 (5-10 accounts): cheap VPS ($30-80/mo) + AdsPower ($30/mo) + 10 ISP proxies ($50-150/mo). Manual monitoring, manual account provisioning. Cost: $100-250/mo. Single-operator manageable.

Stage 3 (20-30 accounts): bigger VPS ($100-250/mo), AdsPower higher tier, 30 ISP proxies ($200-600/mo), custom monitoring scripts. Starts to require a part-time operator to keep clean. Cost: $500-1200/mo. Usually 1-2 ops handling it.

Stage 4 (50+ accounts): multiple VPS instances (1 per 15-20 accounts), multiple proxy providers for redundancy, custom dashboards, dedicated ops team (1-3 people). Cost: $2k-5k/mo infrastructure alone. Below 50 accounts, scaling into this tier doesn't pay; above 50 it does.

Diminishing returns signal: if you're adding account #40 and it lives 3 days, your infrastructure isn't clean enough to support that scale. Fix the setup before adding more.

---

9. Mobile device farms, physical phones as an alternative

Some agencies run physical iPhone/Android farms instead of antidetect + proxy. Each phone has its own SIM, each phone runs one Telegram account.

Pros:

• Most authentic device fingerprint possible.
• No antidetect browser simulation to get wrong.
• Mobile app Telegram behavior is cleanest for warmup / chat UX.

Cons:

• $300-800 per phone + $10-30/mo per SIM.
• Operational nightmare: physical devices need charging, reboots, battery management.
• Doesn't scale beyond ~20-30 devices without warehouse space and racking.
• Cupid / chatbots typically don't integrate with mobile Telegram as cleanly as webapp/desktop.

Realistic fit: 1-5 hero accounts on physical devices (the most valuable ones), the rest on antidetect + proxy. The hybrid gives you the best of both.

---

10. Monitoring at scale

At 10+ accounts, manually checking each one daily doesn't scale. Monitoring setup:

• Uptime ping, each account has a monitoring script that tries to send a harmless "heartbeat" message hourly. If it fails, alert.
• DM volume tracking, if Cupid's daily DM count drops by 80% for a specific account, something's wrong (rate limit, session kill, shadowban).
• "Can I DM non-contacts" check, daily automated test: does each account still pass the spam filter?
• Session liveness, is the account still logged in? Has Telegram killed the session remotely?
• Session export schedule, every account's session file backed up daily, so a crashed VPS doesn't mean re-verifying 30 SIMs.

Mature ops build this as a dashboard (Grafana, simple homemade UI, or a dedicated OFM management tool). Starting simple: a Telegram bot that pings you when something's wrong is enough at Stage 2.

---

11. Cost calculator, true per-account monthly

Stage 2 (10 accounts), broken down per account:

• Account purchase (amortized over 2-month lifespan): $5-15.
• ISP proxy: $5-15.
• Antidetect slot: $3-5.
• VPS share: $5-10.
• SIM / phone number rental: $5-20.
• Cupid / chatbot share: $10-30.
• Total: $33-95 per account per month.

If each account generates 5-15 OF subs/mo at $30/sub = $150-450/mo gross. Net after OF's 20% and infrastructure: $90-270 per account per month. Viable margin at this scale.

At Stage 4 (50 accounts), per-account cost drops to $25-60 due to scale efficiencies, but management overhead rises.

If per-account cost is over $100/mo for you, infrastructure is mis-sized. Usually means too-expensive proxies, over-provisioned VPS, or redundant tooling.

---

12. When to stop adding accounts

The "more accounts = more money" mental model breaks down fast:

• Warming bandwidth is finite. You can warm 20 accounts in parallel with a rolling schedule. Warming 100 in parallel is nearly impossible without automation most operators don't have.
• Low-quality fingerprints start to cluster. At 50+ accounts, your tools' "randomized" fingerprints start to share patterns Telegram detects.
• Operational attention per account drops. At 100 accounts, you can't seriously watch for each account's health daily. Dead accounts accumulate silently.
• Revenue per account drops as you dilute the source traffic. Splitting the same Reddit funnel across 10 TG accounts vs 50 TG accounts doesn't 5x revenue; it just splits the same leads over more accounts.

The observed sweet spot for most single-agency OFM ops: 20-40 active accounts. Above that you're running a business; below you're running a side hustle.

---

Related guides

• Guide 02, Preventing bans (the rules infrastructure enforces)
• Guide 03, Warmup (infrastructure dependencies)
• Guide 05, SMS & phone numbers
• Guide 15, Hiring a Cupid chatter (when ops exceeds owner time)