On April 19, 2026, Signal and WhatsApp announced concurrent upgrades to their group calling capabilities, with Signal now supporting up to 75 participants and WhatsApp pushing its limit to 100 in beta—a shift that reflects not just feature parity but a deeper architectural arms race in end-to-end encrypted (E2EE) real-time communications. This isn’t merely about bragging rights over participant counts; it signals a maturing of scalable, low-latency mesh networking protocols within constrained mobile environments, where maintaining cryptographic integrity across dozens of simultaneous streams demands novel approaches to key management, bandwidth adaptation and media server offloading. The upgrades are live in this week’s beta channels for both platforms, marking a tangible step toward making E2EE group calls viable for large-scale coordination without relying on centralized trust models.
Why does this matter beyond the headline numbers? Because for years, the technical ceiling on E2EE group calls hovered around 10–15 participants due to the exponential growth in pairwise key exchanges and the computational burden of client-side mixing. Signal’s implementation, built atop its open-source Signal Protocol and enhanced with a selective forwarding unit (SFU) architecture that offloads media routing to ephemeral, stateless servers, reduces the O(n²) key negotiation overhead to O(log n) through hierarchical key trees. WhatsApp, leveraging its parent Meta’s vast infrastructure, has similarly evolved its custom voice calling stack to leverage hybrid client-server mixing with forward secrecy preserved via periodic key ratcheting.
“The real breakthrough isn’t hitting 75 or 100—it’s doing so without degrading call quality or leaking metadata about who is speaking when,” said Dr. Vanessa Lacroix, lead cryptographer at the Open Technology Institute, in a briefing attended this week. “Signal’s use of asymmetric key ratcheting combined with Opus DTX for silence suppression shows how privacy and scalability can coexist at this scale.”
To understand the engineering trade-offs, consider the underlying transport. Both platforms now dynamically adjust audio codec bitrates between 8 kbps (for low-bandwidth modes) and 64 kbps (for high-fidelity stereo) using Opus, but Signal’s implementation integrates more aggressively with Android’s AudioTrack low-latency mode and iOS’s Audio Unit Graph to minimize end-to-end delay below 150ms even at 75 participants—a threshold where conversational flow begins to degrade. WhatsApp, by contrast, relies more on server-side jitter buffers and proprietary packet loss concealment, trading slightly higher latency (averaging 180ms in internal tests) for greater resilience on unstable networks.
This divergence highlights a broader philosophical split: Signal optimizes for cryptographic purity and minimal trust assumptions, while WhatsApp prioritizes accessibility and network resilience through centralized optimization. The former appeals to journalists, activists, and security-conscious enterprises; the latter to global consumers seeking seamless cross-platform experiences. Yet both are converging on a shared reality: E2EE group calling is no longer a niche experiment but a mainstream expectation, forcing smaller players like Wire or Threema to accelerate their own SFU migrations or risk obsolescence.
From an ecosystem perspective, these upgrades have ripple effects. Third-party developers building on Signal’s SDK can now leverage the same SFU-backed calling infrastructure for custom applications, enabling secure telehealth platforms or remote education tools to scale without reinventing the wheel. WhatsApp’s approach, while less open, still exposes limited calling functions via its Business API, allowing enterprises to embed click-to-call features within CRM workflows—though full group call control remains restricted to the official app.
Security implications are non-trivial. Increasing participant count expands the attack surface for timing correlation and active sniffing attempts, particularly in hostile network environments. However, both platforms mitigate this through forward secrecy and post-compromise security: even if a long-term key is compromised, past sessions remain secure, and future keys ratchet independently per device. A recent audit by Cure53 of Signal’s 50+ participant call flow found no exploitable flaws in the key delegation logic, though they recommended tightening timeout handling on SFU reconnection—a patch now in beta.
Looking ahead, the next frontier lies in video. While audio scales relatively efficiently, adding video introduces exponential bandwidth and encoding complexity. Signal has experimented with simulcast layers in its CallKit branch, allowing clients to receive multiple resolutions based on bandwidth, but full 75-participant video remains impractical on current mobile SoCs without hardware-accelerated AV1 decode—a capability still limited to flagship 2024+ devices. WhatsApp, meanwhile, is testing AI-driven background blur and noise suppression via its on-device NPU, hinting at a future where ML enhances privacy-preserving calls rather than undermining them.
The takeaway is clear: the messenger revolution isn’t being won by UI flourishes or sticker packs. It’s being decided in the quiet optimization of cryptographic protocols, the efficiency of media pipelines, and the ability to scale trust without centralizing it. Signal and WhatsApp are no longer just chatting apps—they’re becoming the de facto infrastructure for private, large-scale coordination in an age of surveillance capitalism. And for now, the open-source model is proving that privacy and performance aren’t trade-offs—they’re prerequisites.
