Turn every device on Earth into one programmable supercomputer.
Data centers hold the raw horsepower — but the billions of phones, laptops, tablets and
wearables sold in the last decade are sitting mostly idle. MFA Mesh is the
control plane that tracks that fleet, contracts a slice of each device's compute, and manages
security, energy, redundancy and reporting across all of it — globally, in one pane.
Devices Online
0
▲ 1.92M joined / 24h
Aggregate Compute
0
▲ 3.1% vs last week
Contributing Nodes
0
● live scheduling
Efficiency vs Data Center
0
▲ work per watt
Illustrative concept data — figures are placeholders to show the surface, not live measurements.
Device fleet vs. the world's data centers
the question from the call — who actually has more compute?
Compute under management
Estimated usable FP throughput, ten-year device cohort vs. global data-center build-out.
All data centers (global)~5.2 ZFLOPS
10-yr consumer device fleet~9.4 ZFLOPS
Mesh-addressable today (idle %)~1.6 ZFLOPS
Data centersDevice meshlatency-tiered scheduling
How the work is placed
Each job is routed to the tier that fits its latency + horsepower needs.
Data-center tier
heavy training · <5ms intra-rack
28%
Edge-cluster tier
regional inference · <40ms
34%
Device-mesh tier
embarrassingly-parallel · best-effort
31%
Cold / opportunistic
overnight · charging + Wi-Fi only
7%
Device & compute tracking
every enrolled device, what it's contributing right now
Device class
Enrolled
Contributing
Avg. share
Throughput
Trend
State
📱Smartphones
1.41 B
22.4 M
8%
612 EFLOPS
contributing
💻Laptops
486 M
11.8 M
22%
540 EFLOPS
contributing
🖥️Desktops / workstations
198 M
3.9 M
35%
288 EFLOPS
contributing
📟Tablets
274 M
2.1 M
11%
96 EFLOPS
idle / charging
⌚Wearables
512 M
0.9 M
4%
14 EFLOPS
standby
Security & privacy controls
Per-device consent and isolation — the device owner is always in control.
Sandboxed execution
Each job runs in an isolated VM/enclave; no host filesystem or contacts access.
End-to-end encrypted payloads
Workloads + results sealed; the mesh sees ciphertext, never your data.
Owner consent & revocation
Opt in per device; revoke instantly. Default: charging + Wi-Fi only.
Result attestation
Redundant verification + signed proofs reject poisoned or faulty returns.
Region / data-residency fencing
Pin sensitive jobs to compliant geographies (GDPR / HIPAA).
Energy usage monitoring
Draw across the contributing fleet — and work-per-watt vs. a data center.
38.6 MW
Fleet draw
6.7×
Work / watt vs DC
71%
On renewable grid
Devices reuse power they'd draw anyway while charging — marginal energy cost is a fraction of a dedicated server.
Redundancy & health checks
devices are flaky by nature — the mesh assumes it and routes around failure
Replication
Every shard runs on N independent devices.
Replication factor
live target
3.0×
Quorum reached
last 1h
100%
Liveness
Heartbeat + drop detection across nodes.
Healthy nodes
heartbeat <30s
99.4%
Re-dispatched
dropped mid-job
0.6%
Failover
Time to re-place a job after a device vanishes.
Median re-place
p50
0.8s
Jobs lost
last 24h
0
Customizable reporting
pick the metrics, scope and cadence — export or schedule
Build a report
Tap metrics to include them.
Devices onlineThroughput (FLOPS)Energy drawWork / wattRedundancySecurity eventsCost / jobGeo distribution