← back to blog

Cloud Android phones for Telegram: userbot fleets on real SG mobile IPs in 2026

If you are running Telegram at scale in 2026, you already know the shape of the problem. Accounts flagged as spam_bot within hours of first action. Phone number verification loops that never resolve. Entire batches wiped in a single sweep because they shared the same ASN. The platform has not gotten more lenient, and the detection surface has grown considerably since Telegram integrated more aggressive client-side checks after its 2023 monetization push. Cloud phones are not new, but real-hardware hosted phones with genuine carrier SIMs sit in a different category from what most operators have tried. This post covers exactly why that difference matters for Telegram specifically, how to set up a working userbot fleet on cloudf.one, and where operators typically go wrong.

why Telegram hits walls without real hardware in 2026

Telegram's client-side detection has quietly become one of the more thorough on any messaging platform. The Android client calls Play Integrity API on login and at irregular intervals during active sessions. Play Integrity replaced SafetyNet in 2023, and unlike SafetyNet, it is not trivially bypassable on emulators. An emulator running a stock AOSP image will fail the MEETS_DEVICE_INTEGRITY verdict, which Telegram's backend receives and ties to that session. That verdict failure does not always mean an immediate ban. But it is a permanent signal attached to the account fingerprint, and when combined with other low-trust signals, it is enough to trigger a spam_bot classification or a forced re-verification loop that never completes.

The second layer is IP reputation and ASN classification. Telegram maintains a blocklist of datacenter ASNs and buys feeds to keep it updated. Any IP from a cloud hosting provider, a residential proxy reseller, or a known VPN exit node lands in a higher-scrutiny tier. A Singapore residential proxy sounds like a fix, but proxy pools share IPs across hundreds or thousands of simultaneous users. Telegram's session correlation logic picks up multiple distinct device fingerprints appearing from the same IP block within short windows, and that pattern matches proxy pool behavior, not organic mobile use. The account trust score drops accordingly. Real carrier IP assignment from SingTel, StarHub, M1, or Vivifi does not carry that pattern because the IP is assigned to one physical device on one mobile session.

The third layer is behavioral biometrics at the gesture and timing level. Telegram's client samples touch event cadence, scroll velocity distributions, and inter-action timing. Emulated input events from Android emulators, even those driven by real ADB commands, have statistical signatures that differ from human-generated touch input on physical glass. Automated userbots need to operate within windows that look like human-paced interaction, which is easier when the underlying device is a physical Samsung Galaxy with a real touchscreen rather than a virtualized input layer. These three detection vectors together (Play Integrity verdict, IP reputation, and behavioral sampling) explain why operators who have tried emulator farms or antidetect browser approaches keep burning accounts in 2026. For a detailed comparison of what changes at the hardware layer, see real cloud Android phone vs emulator.

what a cloudf.one phone gives Telegram operators specifically

A cloudf.one phone is a physical Samsung Galaxy S20, S21, or S22 unit sitting in a Singapore data facility, assigned exclusively to one renter for the duration of the rental. Not a virtual machine. Not a container running a modified Android kernel. The Play Integrity API call goes to Google's attestation server and comes back with a MEETS_STRONG_INTEGRITY or MEETS_DEVICE_INTEGRITY verdict, the same result you would get from that handset sitting on a desk in an office. The Telegram client receives that verdict and has no basis to treat it differently from an organic user's device. That is the foundational layer, and no software-only solution can replicate it.

Each phone has a physical SIM from one of Singapore's four carriers (SingTel, StarHub, M1, or Vivifi) inserted in the device. The IP assigned to that session is a real mobile carrier IP from that carrier's Singapore mobile ASN. Not a proxy. Not a relay. The IP changes when the mobile session re-registers, the same as on any consumer mobile device, which means the IP history on that account looks like organic mobile use rather than a static datacenter assignment. This matters because Telegram does run ASN reputation checks, and mobile carrier ASNs in Singapore carry significantly lower suspicion weight than any datacenter or residential proxy ASN. For more on why proxy-based IP workarounds break down for platform-level checks, the analysis in why VPNs don't work for TikTok applies equally to Telegram's IP trust model.

Access to the phone is through two channels: the STF (Smartphone Test Farm) browser UI for live screen view and touch interaction, and ADB shell access over the network. For Telegram operators, that combination covers most workflows. STF handles account setup, app installation from the Play Store, and initial login verification. ADB handles automation hooks, logcat monitoring for session health, and scripted interaction sequences via adb shell input or higher-level tools like uiautomator. The phone is customer-isolated, meaning no other renter shares its session state, its SIM assignment, or its IP history. Telegram's session fingerprint for an account on that phone will consistently show the same device identifiers, the same carrier, and the same geographic region across logins. That consistency registers as a positive trust signal on the platform.

step-by-step setup for userbot fleets on real SG mobile IPs to avoid spam_bot flags

1. Provision a phone from cloudf.one plans. Start with hourly rental for initial evaluation and account seasoning tests. Move to monthly rental once you have confirmed account retention rates and established your automation workflow. Monthly rates cut the per-hour cost substantially for sustained fleet operation.

2. Open the STF browser UI, claim the phone to your session, and go to the Play Store. Search for Telegram and install directly from the Store. Do not sideload an APK or XAPK. Play Protect scans sideloaded installs and the non-Store provenance is a flag independent of what the app actually does. Installing from the Play Store on a device with a verified Google account gives the Telegram install a clean provenance record from day one.

3. Complete phone number registration inside the STF browser UI using a real SIM-registered number, or a number from a provider whose numbers Telegram accepts for your target account tier. On first launch, Telegram will send an SMS OTP. Because the device has a real SG SIM, you can also use that SIM's number directly if you are registering SG accounts. A carrier number verified over the SG SIM gives the account a mobile-verified status that Telegram treats as higher trust than VOIP-verified registrations. After OTP entry, complete the profile setup fully (name, optional profile photo) before taking any other action. Accounts that skip profile completion and immediately jump into joining channels or sending messages have a higher early flag rate.

4. Set up your userbot automation layer. For Python-based userbot frameworks (Pyrogram, Telethon), the preferred approach on a cloud phone is to run the userbot client on the phone itself via Termux, or to run it on a separate server while using the cloud phone as the authenticated session source. To use the phone as the session source, log in to Telegram on the device via the native Android app, export the session string using a session-string generator script run via ADB shell, then import that string into your server-side userbot instance. The session string carries the device fingerprint established on the physical Samsung hardware with the SG carrier IP, meaning subsequent API calls from your server use a session that was legitimately created on real mobile hardware.

5. For persistent login across rental periods, export and store the session string before the rental window closes. A Telegram session established on a real device and exported as a Pyrogram or Telethon session string stays valid as long as you do not trigger a forced re-authentication (which happens on IP class changes that Telegram deems suspicious, or after 30 or more days of no activity). Re-import the session string to a new phone instance when you provision the next rental period. If the account requires re-verification, complete it again via the physical phone in STF rather than programmatically, so the verification event is attached to a real device attestation context.

three real workflows this fits

community and channel management across multiple SG groups

A Telegram operator managing several Singapore-focused community groups typically runs moderation bots, welcome message automation, and periodic broadcast accounts alongside their main admin account. Running all of these from a single IP or a shared proxy pool means that any single account getting flagged can cascade to the others through IP and session correlation. With dedicated cloud phones, each moderation account lives on its own phone with its own SG carrier IP. When Telegram's systems look at the account graph, they see independent device fingerprints, independent carrier assignments, and independent behavioral histories. The broadcast account sending high message volume in a short window does not touch the trust score of the moderation bot on a separate phone. Fleet isolation is the operational gain here, not just IP diversity.

account seasoning and warming for new numbers

New Telegram accounts need a warming period before they can reliably join large groups, send messages to non-contacts, or run inline bot queries without hitting limits. The warming sequence (organic-looking activity over several days, joining small groups, exchanging messages, reading channels) is tedious but cannot be skipped. On a cloud phone with a real SG SIM, the warming activity is indistinguishable from a real user's behavior because the device signals, location signals, and network signals are all genuine. Operators who warm accounts on emulators find the warming period does not take correctly, because the underlying device fingerprint is low-trust throughout. A properly warmed account on a real Samsung with a real SIM can handle userbot actions at rates that would immediately flag an emulator-warmed account.

Telegram-based customer interaction and sales automation for SG businesses

Businesses using Telegram for customer service, order management, or sales pipelines in the Singapore market often need multiple operator accounts to handle volume across shifts or product lines. Each operator account needs to look like a genuine user to avoid being restricted from initiating conversations or being muted by recipients who mark the account as spam. Running these accounts on cloudf.one phones means each account has a persistent device identity with a Singapore carrier IP, and the interaction patterns (typing delays, read receipts, response timing) can be managed via STF manual operation or light ADB scripting rather than aggressive automation. The platform sees mobile users in Singapore because that is exactly what the infrastructure delivers. For more on what the browser-based antidetect alternative looks like in comparison, see cloud phone vs antidetect browser.

cost math at three realistic scales

The honest cost comparison for Telegram fleet operation has to include account replacement cost, not just infrastructure cost. An operator running emulators at $30-60/month per emulator slot plus $20-40/month per proxy bundle loses accounts regularly. Each lost account takes with it the warming investment (days of activity, group memberships, contact history), and replacement numbers from quality providers are not free. When account churn is high, the effective per-account cost over a quarter is multiples of the headline infrastructure cost.

At one phone, the use case is evaluation or a single critical account that cannot afford to be flagged. Check current rates at cloudf.one plans for exact figures, but monthly dedicated rental for one phone is roughly the cost of a mid-tier residential proxy subscription, with the difference being that you get a real attestable device rather than a shared IP pool. At five phones, you are running a small fleet, likely one phone per account category (broadcast, moderation, community, customer service, reserve). The per-phone cost drops on monthly plans, and the operational benefit is full isolation between account roles. At twenty phones, you are running a production fleet and the numbers shift clearly: the monthly cost of twenty cloud phones is comparable to what a mid-size operation spends on antidetect browser seats plus proxy subscriptions, and the account retention rate difference makes the cloud phone approach the lower total cost option over any six-month window. Maintaining physical devices on-premises at this scale adds hardware refresh costs, Singapore office or co-location costs, and the operational overhead of physical device management, none of which apply to cloudf.one rentals.

common pitfalls for Telegram operators

• treating the cloud phone like a browser session. Telegram on Android holds persistent state including cache, notification history, and contact sync data. Clearing this state between uses, or factory resetting a phone between account assignments, destroys the device continuity that makes the account trustworthy. Assign one phone to one account identity and leave the Telegram installation in place for the duration of that account's active life. The phone is not a stateless tool.
• rotating SIMs too aggressively. Changing the SIM on a phone that has an active Telegram account attached to a number verified on a different carrier will trigger Telegram's device-number mismatch detection in some account states. SIM rotation should be planned around account lifecycle, not IP freshness. The carrier IP on a single SIM already changes naturally as the mobile session re-registers. There is no benefit to SIM swapping for IP diversity on a device that already has a real carrier assignment.
• not pinning a phone per account long-term. Telegram builds a device fingerprint history for each account. An account that has consistently logged in from a Samsung Galaxy S21 on a SingTel IP for three months has a much stronger trust profile than one that bounces between device fingerprints. Moving accounts between phones resets part of that history and is a detectable pattern. Pin each account to one phone and keep it there. See the discussion on Android sandbox isolation for why device identity persistence matters at the OS level.
• installing Telegram via XAPK or sideload. Play Protect runs a safety check on sideloaded apps and flags the install provenance in the device's safety status. Telegram itself may not care about the provenance directly, but Play Integrity API results include device safety state, and a device with a Play Protect warning active scores lower on attestation checks. Always install from the Play Store on a device with a clean Google account.
• logging in across multiple phones in 24 hours. Telegram's account-linking detection watches for the same account authenticating from multiple distinct device fingerprints in a short window. If you import a session string from phone A and then also log in via phone B on the same day, Telegram sees two active sessions with different hardware fingerprints appearing at the same time. That pattern looks like session sharing or account theft, and it triggers elevated scrutiny. Establish one device as the canonical hardware for each account and do not authenticate the same account from multiple phones within the same session window.

frequently asked questions

can Telegram detect that this is a cloud phone

The detection vectors Telegram uses are Play Integrity API attestation, IP reputation by ASN, device fingerprint consistency, and behavioral sampling. A cloudf.one phone returns a legitimate Play Integrity verdict because it is a real Samsung Galaxy device with an unmodified bootloader. The IP comes from a Singapore mobile carrier ASN, not a datacenter or proxy ASN. The device fingerprint is persistent across sessions because it is the same physical hardware. There is no signal available to Telegram that identifies a cloudf.one phone as not a user's own phone. The platform cannot distinguish it from an organic user running the same model handset on the same carrier in Singapore, because at every detectable layer, that is exactly what it is.

how many Telegram accounts per phone

Telegram's Android client supports up to five accounts in the native multi-account UI. In practice, for account health and behavioral isolation, most operators run one primary account per phone and use the second or third account slot for a companion bot or reserve account only. Running five fully active accounts from a single device creates inter-account behavioral correlation (all five share the same device fingerprint and IP) and increases the blast radius if that phone's session gets flagged. For userbot fleets where account independence matters, one active account per phone is the right ratio, with secondary slots used sparingly.

does the SIM rotation cause Telegram account flags

SIM rotation means changing the physical SIM in the phone, which changes the carrier IP and potentially the registered phone number context for the device. Telegram does not flag accounts solely because the IP changed, since mobile IPs change regularly on consumer devices. What matters is the device fingerprint staying consistent and the new IP still being a clean mobile carrier IP rather than a datacenter IP. If you rotate SIMs between carrier-assigned SG SIMs (for example, from a SingTel SIM to an M1 SIM), the account continues to see a clean Singapore mobile IP. Rotating to a SIM from a low-reputation provider or a number pool associated with prior spam activity is a different matter.

can I use ADB to automate Telegram actions

Yes. ADB shell access on cloudf.one phones supports adb shell input tap, adb shell input swipe, adb shell input text, and UIAutomator-based interaction scripts. For Telegram specifically, ADB automation works for actions like going to a chat, typing and sending a message, scrolling a channel, and triggering inline bot commands. The input events generated via ADB on a physical device go through the real Android input pipeline and are processed by the same touch event handlers as actual touch input. Timing and cadence still matter: scripted actions at machine speed with zero variance look different from human input. Add realistic delay distributions to your automation scripts. For latency expectations when driving the phone remotely, see cloud phone latency explained.

what about Singapore-specific Telegram features

Telegram has run SG-specific features and promotions, including regional sticker packs, SG-local bot directories, and in some periods, regional verification tiers for business accounts. A real SG SIM gives the account a Singapore phone number in the +65 country code, placing it in Telegram's Singapore regional tier for features gated by registration country. Accounts registered on VOIP numbers or foreign numbers with SG proxies do not get the +65 registration status. For operators building SG-market audiences, group discovery, and regional channel credibility, having genuine +65 registered accounts matters both for feature access and for the trust signal a local number sends to group members reviewing account profiles.

how does this compare to running emulators

Emulators fail Play Integrity's MEETS_DEVICE_INTEGRITY check because they run a virtualized Android environment without a verified bootloader or hardware-backed keystore. Telegram receives that attestation result. Beyond attestation, emulators typically run in datacenter infrastructure, so their IPs land in datacenter ASNs that Telegram's reputation feeds flag. Emulators also produce behavioral signals (input event timing, sensor data, GPU rendering patterns) that differ from physical hardware. Cloud phones address all three of these failure modes at once. The tradeoff is cost per unit: emulators are cheaper per slot. The question is whether the lower cost per slot offsets the account churn rate, and for most operators running Telegram at scale in 2026, the math favors real hardware.

getting started for Telegram

The practical starting point is one or two phones from cloudf.one plans on hourly rental to run through the full account setup and session export workflow before committing to a monthly fleet. Pick the S21 or S22 tier if you expect to run sessions continuously, since the newer hardware handles thermal management better for sustained operation. Establish your phone-per-account ratio based on how many concurrent active accounts your workflow requires, then move to monthly billing once you have confirmed your setup works end to end. The cloudf.one blog has additional posts on related infrastructure decisions. For most Telegram operators, the shift from emulator or proxy-based setups to real cloud phones comes after the first account retention comparison: accounts on real hardware with real SG SIMs simply do not burn at the same rate, and that difference compounds over any meaningful operating period.