Ably vs Pusher vs PubNub: Real Realtime Messaging Cost per 100 Million Messages in 2026

The short answer: Ably and Pusher don't even agree on what "a message" costs, because Ably's meter is fundamentally different in kind, and PubNub's pricing is widely described as too opaque to budget confidently at all. Ably counts a message per delivery, not per publish — one published update fanned out to 50 subscribers bills as 50 messages — and at 100 million delivered messages, plus a modest connection and channel footprint, Ably's consumption-based model comes to roughly $715/month ($250 for messages, $432 for connection-minutes, the rest base and channel fees). Pusher Channels bills on flat tiers defined by two dimensions — peak concurrent connections and messages per day — with no per-minute connection charge at all, meaning an app with many quiet, long-lived connections can be structurally cheaper on Pusher than on Ably, while a message-heavy, fan-out-driven app (like realtime cursor tracking or chat) tends to favor Ably's lack of a hard daily cap. PubNub's exact cost at this volume is QUOTE-ONLY / UNKNOWN — multiple independent sources describe its usage-based pricing as "complex" and "hard to predict," without a public rate card specific enough to model here.

The billing-unit mismatch, up front

Ably bills messages per delivery: a single published update to a channel with 50 active subscribers counts as 50 billable messages, not one — this is confirmed directly in Ably's own pricing documentation, which lists message delivery, presence events, history retrieval, and even push notification delivery as each generating their own billable outbound message per recipient. Pusher Channels bills on two flat-tier dimensions simultaneously — peak concurrent connections and total messages per day — without a documented per-delivery fan-out mechanic as explicit as Ably's, and without any connection-duration meter at all (a connection open for one minute or one month costs the platform the same, tier-wise, as long as peak concurrency stays under the plan's cap). PubNub's billing unit and exact rate structure are not publicly confirmed, beyond its general reputation for usage-based pricing that scales unpredictably at volume.

What each vendor bills

Ably (OFFICIAL, ably.com/pricing and ably.com/docs/pub-sub/pricing). Free: 200 concurrent connections, 500 messages/second, 6 million messages/month, 1-day message storage. Standard: $29/month base, 10,000 concurrent connections and channels included, 2,500 messages/second, 30-day storage. Pro: $399/month base, 50,000 concurrent connections/channels, 10,000 messages/second, 365-day storage. Usage on top of the base, on both paid tiers: messages at $2.50 per million (with volume discounts at higher scale), connection-minutes at $1.00 per million (dropping to $0.20 per million at volume), channel-minutes at $1.00 per million, and data transfer at $0.25/GiB. Every minute a client maintains an open connection counts toward connection-minutes, regardless of whether it sends or receives anything during that minute — a quiet, long-lived connection still accrues cost. Enterprise: custom, unlimited connections/channels, HIPAA BAA available.

Pusher Channels (OFFICIAL, pusher.com/channels/pricing). Sandbox (free): 100 concurrent connections, 200,000 messages/day. Startup: $49/month. Pro: $99/month. Business: $299/month, scaling up to $1,199/month at the top self-serve tier — each tier defined by a specific combination of peak concurrent connections and a daily message cap, not a usage-metered rate. No connection-minute billing of any kind — Pusher's model rewards steady, predictable connection counts over Ably's consumption-based approach, but a documented "peak-connection trap" means briefly exceeding a tier's concurrent-connection ceiling (even for a short traffic spike) can force an upgrade to a tier several times the price, independent of whether sustained usage would otherwise fit the lower tier.

PubNub (REPORTED, characterization only; no confirmed rate card). Positioned for "enterprise-scale" realtime messaging with support for both one-to-one and one-to-many delivery patterns. Multiple independent sources describe PubNub's usage-based pricing as complex and difficult to predict for rapidly growing applications, and as "prohibitively expensive" at large scale relative to Ably or Pusher — but no specific, current per-message or per-transaction rate precise enough to model a confident cost at 100 million messages is publicly available.

Cost at 100 million messages/month

Formula: Ably = base plan fee + (delivered messages ÷ 1M × $2.50) + (connection-minutes ÷ 1M × $1.00) + (channel-minutes ÷ 1M × $1.00). This article's ILLUSTRATIVE assumption: 100 million delivered messages, 10,000 concurrent connections maintained continuously for the month (43,200 minutes each), and 100 active channels maintained continuously.

Ably's connection-minute charge is not a rounding error at this scale — at 10,000 continuously-open connections, it's nearly double the message charge itself ($432 versus $250). This is the same dynamic the short answer describes from the other direction: Pusher has no connection-minute meter at all, so an app with many idle-but-open connections and comparatively few messages shifts the cost balance toward Pusher's flat, connection-count-based tiers, since Ably keeps billing those open connections every minute regardless of how little they're used. A chatty, high-fan-out app with fewer concurrent connections shifts the balance the other way, toward Ably, since Pusher's tiers don't reward low message volume the way Ably's lack of a connection-minute charge rewards low connection counts.

Why fan-out is the single most consequential mechanic here

Formula: Ably billed messages = published messages × average subscribers per channel. A chat application publishing 10 million actual messages, each delivered to an average of 10 subscribers, bills as 100 million messages on Ably — ten times the "messages sent" figure a naive estimate might use. This fan-out multiplication applies not just to ordinary publishes but to presence events, history retrieval, and push notification delivery as well, each generating its own per-recipient outbound message on Ably's meter. Pusher's own published documentation does not describe an equivalently explicit per-delivery fan-out mechanic for its daily message cap, which remains a genuine reader-facing uncertainty: a direct apples-to-apples comparison between the two platforms' "message" figures stays approximate without confirming exactly how Pusher counts a fanned-out publish against its daily cap.

The peak-connection trap on Pusher

Because Pusher's tiers are defined by peak concurrent connections, not average or sustained connections, a brief traffic spike — a product launch, a viral moment, a scheduled event drawing a crowd — can push an app's peak concurrency over its current tier's ceiling even if typical daily usage would comfortably fit a lower tier. One independent teardown describes this dynamic turning a $99/month Pro-tier bill into a $499/month bill purely from a short-lived connection spike, without any corresponding change in average daily message volume or baseline connection count.

Sensitivity

1. Fan-out ratio (subscribers per published message), for Ably specifically. The single largest lever on Ably's bill; a 2x change in average fan-out roughly doubles or halves the message-cost component.
2. Connection duration and count. On Ably, this is a direct, continuous-time cost; on Pusher, it is a step-function (which flat tier your peak concurrency falls into), not a continuous meter.
3. Traffic spikiness, for Pusher specifically. A brief peak-connection spike can force a tier upgrade that a sustained-average view of usage would not predict.
4. Whether PubNub's actual rate card, once obtained via direct quote, turns out comparable to Ably's or Pusher's structure. Not publicly confirmed; treat any third-party estimate of PubNub cost with skepticism until confirmed directly.

Budgeting traps

• Estimating Ably cost from "messages published" rather than "messages delivered." Fan-out can multiply the real billable message count several times over for any channel with multiple subscribers.
• Sizing a Pusher tier off average daily traffic without accounting for peak concurrency spikes. A short burst above the tier's connection ceiling can force an expensive tier upgrade independent of typical usage.
• Treating PubNub's cost as roughly comparable to Ably's or Pusher's without a direct quote. Multiple independent sources flag PubNub's pricing specifically as hard to predict from public information alone.
• Ignoring Ably's connection-minute charge because it looks small per-unit. At meaningful concurrent-connection counts sustained over a full month, it can exceed the message charge itself.

What to ask before you buy

Measure your application's actual fan-out ratio (subscribers per published message) before estimating Ably's cost, since this article's $250 message-cost component scales directly and linearly with that ratio. Ask Pusher directly about your expected peak concurrent connections, not just average daily message volume, since the tier-defining dimension that most often causes bill shock is a brief connection spike, not sustained throughput.