- •Carrier-grade has a specific standard: five nines of availability, or 99.999% uptime, roughly five minutes of downtime per year.
- •It requires geo-redundant data centers, direct carrier relationships, and carrier-layer signaling visibility, not a marketing label on someone else's network.
- •CPaaS routes through carrier wholesale agreements but is optimized for developer access, not carrier-grade reliability or unified routing.
- •Evaluate a claim by asking about direct carrier agreements, where numbers live, the uptime SLO, and 487 visibility.
"Carrier-grade" gets used as a marketing term so often it has nearly lost meaning. Every SaaS phone system, every UCaaS vendor, and every AI voice startup claims it. Most of them are wrong.
Here's what the term actually means, why it matters, and how to tell the difference between real carrier-grade infrastructure and a marketing claim layered on top of someone else's network.
Where does the term "carrier-grade" come from?
Carrier-grade has a specific technical definition rooted in how public telephone networks were built. The standard is five nines of availability: 99.999% uptime. That translates to roughly five minutes and fifteen seconds of downtime per year. Not per month. Per year.
Achieving that number requires redundant architecture at every layer - multiple data centers, automated failover, no single points of failure in call routing or signaling, and operations teams monitoring the network continuously. It also requires direct carrier relationships, not resold capacity sitting on top of someone else's infrastructure.
The term was meaningful when telephone networks were built this way by default and consumer-grade alternatives clearly were not. Now that the distinction has been blurred by vendors applying the label broadly, the only way to evaluate a "carrier-grade" claim is to look at the architecture behind it.
What does real carrier-grade voice infrastructure require?
Geo-redundant physical infrastructure. This means multiple geographically distributed data centers with live-active or active-standby configurations. If a data center goes offline - power failure, connectivity issue, physical event - calls route automatically to another location. Our infrastructure runs across four data centers. The uptime SLO is 99.999%.
Direct carrier relationships. Carrier-grade infrastructure connects directly to Tier 1 and Tier 2 carriers under wholesale agreements. The voice traffic traverses the public network under those agreements, not through a cloud telephony API or a reseller arrangement. Direct peering gives you quality of service commitments, priority routing, and visibility into what the carrier is doing with your traffic. A company routing through a CPaaS API layer does not have this.
Carrier-layer signaling visibility. Real carrier infrastructure gives you SIP-level instrumentation - the ability to see response codes, call setup times, post-dial delay, and failure modes at the network layer. This is how you see a 487 rate, how you detect early media anomalies, how you monitor STIR/SHAKEN attestation in real time. Without this layer of visibility, you're operating on aggregated call data from inside your application stack, which is not the same thing.
Number portability and routing control. In true carrier-grade infrastructure, numbers live in the network, not in an application. This means numbers can be pointed at any application - Teams, a contact center, an AI voice agent, a SIP endpoint - without porting. Routing changes are made at the carrier layer. Applications come and go. The numbers and the network stay stable.
What is CPaaS, and is it carrier-grade?
CPaaS - Communications Platform as a Service - is a category of cloud APIs that let developers add voice and messaging to applications. Twilio is the largest example. These platforms are genuinely useful for building applications that need to make or receive phone calls.
| CPaaS | Carrier-Grade Infrastructure | |
|---|---|---|
| Uptime standard | Varies; typically 99.9% | 99.999% (five nines) |
| Carrier relationships | Wholesale agreements via API layer | Direct Tier 1 & Tier 2 peering |
| Number portability | Numbers live in the application | Numbers live in the network |
| Signaling visibility | Aggregated app-layer data | SIP-level: 487 rates, STIR/SHAKEN, post-dial delay |
| Cross-product failover | Requires custom development | Unified routing layer across all voice applications |
CPaaS is not carrier-grade infrastructure.
CPaaS providers route traffic through the public network using carrier wholesale agreements - the underlying network is there - but the architecture is optimized for developer accessibility, not carrier-grade reliability or performance. Programmable voice and elastic SIP trunking are separate products on most CPaaS platforms. Cross-product failover - routing a failed AI agent call to a live agent on a different system - requires custom development. There is no unified routing layer that treats all voice applications as part of the same network.
The difference matters when you're running mission-critical voice at scale. A collections operation processing millions of outbound calls per month cannot afford the visibility gaps that come with a CPaaS architecture. A franchise network running AI-assisted order taking across dozens of locations needs the reliability that five-nines infrastructure provides.
How do you evaluate a carrier-grade claim?
Ask these questions:
Does this provider have direct carrier agreements, or are they routing through a CPaaS layer? The answer tells you whether you're getting true carrier infrastructure or an abstraction on top of it.
Where do your numbers live? If numbers live in the application, they don't live in the network. This has real consequences for portability and routing flexibility.
What is the uptime SLO and how is it backed? A 99.9% SLO is nine hours of downtime per year. A 99.999% SLO is five minutes. These are not the same thing, and the architecture required to achieve them is fundamentally different.
Do they have visibility into carrier-layer signaling? Can they show you your 487 rate? Can they show you STIR/SHAKEN attestation rates by carrier? If the answer is no, they don't have the instrumentation layer that real carrier infrastructure provides.
The Bottom Line
Carrier-grade voice infrastructure means five nines of availability, geo-redundant architecture, direct carrier relationships, and carrier-layer visibility into what the network is doing with your traffic. It is a specific technical standard, not a marketing category.
Most of what's sold as carrier-grade doesn't meet it. The ones that do operate differently - and the difference shows up in performance, reliability, and how much control you actually have over your voice environment.
Frequently Asked Questions
What does carrier-grade mean for voice infrastructure?
Carrier-grade means five nines of availability: 99.999% uptime, which translates to roughly five minutes and fifteen seconds of downtime per year. Achieving that standard requires geo-redundant data centers, automated failover, no single points of failure in call routing or signaling, and direct carrier relationships.
How is CPaaS different from carrier-grade infrastructure?
CPaaS platforms route traffic through carrier wholesale agreements, but the architecture is optimized for developer accessibility rather than carrier-grade reliability. Programmable voice and elastic SIP trunking are separate products on most CPaaS platforms, and cross-product failover requires custom development. There is no unified routing layer that treats all voice applications as part of the same network.
Why does it matter where phone numbers live?
In true carrier-grade infrastructure, numbers live in the network rather than in an application. This means numbers can be pointed at any application, such as Teams, a contact center, or an AI voice agent, without porting. Routing changes are made at the carrier layer, so the numbers and the network stay stable as applications change.
What is carrier-layer signaling visibility and why does it matter?
Carrier-layer signaling visibility means SIP-level instrumentation: the ability to see response codes, call setup times, post-dial delay, and failure modes at the network layer. This is how you see a 487 rate, detect early media anomalies, and monitor STIR/SHAKEN attestation in real time. Without it, you are operating only on aggregated call data from inside your application stack.
What is the difference between a 99.9% uptime SLO and a 99.999% uptime SLO?
A 99.9% uptime SLO allows for roughly nine hours of downtime per year, while a 99.999% SLO allows for roughly five minutes. The architecture required to achieve five nines is fundamentally different from what is needed for three nines, requiring redundancy at every layer of the network.
How do you tell if a provider actually has carrier-grade infrastructure?
Ask whether the provider has direct carrier agreements or routes through a CPaaS layer, where your numbers live, what the uptime SLO is and how it is backed, and whether they have visibility into carrier-layer signaling such as 487 rates and STIR/SHAKEN attestation. If they cannot answer those questions with specifics, they do not have the instrumentation layer that real carrier infrastructure provides.
