TL;DR
- ▸Public cloud VMs run on shared hypervisors with variable virtualisation overhead, making precision timestamping much harder than on bare metal.
- ▸Microsoft Azure's PTP-enabled VM types currently deliver the tightest cloud-native synchronisation guarantees. AWS and GCP are looser.
- ▸For genuine single-digit microsecond requirements, bare-metal compute with hardware-timestamping NICs in your own racks is still the right answer.
Why cloud is hard
Public cloud VMs run on shared hypervisors. Every system call from a guest VM passes through the hypervisor, every interrupt is delivered through a hypervisor-mediated path, and every clock read can be intercepted and rewritten by the hypervisor for various legitimate reasons (live migration, time skew correction, multi-tenant isolation). This is fine for the millisecond-class precision that NTP delivers — the variability is small relative to the precision floor — but it's a problem for anything tighter.
The result is that most public cloud environments cannot deliver single-digit microsecond accuracy on general-purpose VMs because the virtualisation overhead exceeds the precision budget. Achieving sub-microsecond precision on a cloud VM requires the hypervisor to expose hardware-grade timestamps to the guest, which requires deliberate engineering work in the hypervisor layer — work that not every cloud has done yet.
What works
Microsoft Azure has done the engineering work and exposes PTP-enabled VM types with documented sub-microsecond accuracy. For applications that need cloud-native deployment plus microsecond-class timing, Azure's PTP-enabled VMs are currently the right choice. AWS and GCP have looser guarantees on general-purpose VMs and are the wrong choice if precision is the binding constraint.
For applications that genuinely need single-digit microsecond precision — financial timestamping, low-latency trading, broadcast IP video, 5G fronthaul — the practical answer is still bare-metal compute with hardware-timestamping NICs in your own racks, with cloud workloads running adjacent to that bare-metal core for the parts of the application that don't need the same precision tier. Hybrid deployment models like this are increasingly common.
The hybrid pattern
Most applications that need precision timing don't need it everywhere — only in the specific subsystems that interact with regulated timestamps, hardware control, or coordinated events. Those subsystems run on bare metal with PTP grandmasters; the rest of the application runs cloud-native and locks to the bare-metal core for time. This is the pattern we see most often in production.
Where TimeBeat fits
TimeBeat builds the hardware grandmasters and the TimeBeat Cloud managed timing service that customers use to bridge the gap between bare-metal precision requirements and cloud-native deployment models. For customers running hybrid architectures, TimeBeat hardware in the precision-critical racks delivers the source of truth that the cloud workloads lock to. For customers who want cloud-native precision time as a service, TimeBeat Cloud provides it managed.
Frequently asked questions
Can I achieve single-digit microsecond sync on AWS or GCP?+
Why does cloud timing have a precision ceiling?+
What's the right architecture if I need precision time and cloud deployment?+
Related reading
Blog · Cloud
Microsoft Azure: The Most Accurate Cloud for Clock Sync (For Now)
Microsoft Azure publishes the tightest clock synchronisation guarantees of any major public cloud, with documented PTP support across selected VM types. What Azure delivers, where the limits are, and how it compares to AWS and GCP.
Blog · Protocols
Precision Time Protocol vs NTP: When Each Belongs in Production
The honest engineering comparison between Precision Time Protocol and NTP — what each protocol can actually deliver, where the boundary lives, and how to choose between them without falling for either side's marketing.
Blog · Partners
cloudSwXtch Integrates TimeBeat PTP for Hybrid Cloud Timing
cloudSwXtch integrated TimeBeat's PTP implementation to extend hybrid cloud networking with precision time. The integration and what it enables for hybrid cloud customers.