Internet speed reportUnited States
Aggregated internet performance from the source named below. IP-based geographic annotations are approximate.
Internet performance in United States
Available aggregates represent 19,490 tests completed during September 2025. No raw-test distribution or time series is available here. Download performance was 0.3% better than August 2025; this comparison describes two user-test samples, not every connection in the area.
Measured download percentiles are unavailable. We do not infer them from the average. A pair of percentiles cannot reconstruct a full distribution, so no percentage-of-tests curve is displayed.
How reports are calculated
Historical user speed-test results are normalized and aggregated by period and geography.
Dataset: bandwidthplace.monthly_report
Data methodology
Historical reports
United States Internet speed report: methodology and answers
During September 2025, this report represents 19,490 aggregated tests for United States. The central download value is 346.46 Mbps, upload is 128.43 Mbps and latency is 41.5 ms. These figures describe the available measurement sample, not every household or address.
| Measure | Value | Context |
|---|---|---|
| Download speed | 346.46 Mbps | 0.3% better than August 2025 |
| Upload speed | 128.43 Mbps | 0.2% better than August 2025 |
| Latency | 41.5 ms | 1.7% worse than August 2025 |
| Packet loss | —% | Share of packets not successfully delivered |
Methodology
Users initiate speed tests, which are assigned to an approximate country or city using IP-based geography and grouped into the selected period. We aggregate the named historical speed-test source after excluding invalid or incomplete rows where identifiable. The resulting user-test sample is used to calculate the ranking, summary and report metrics; a country with no tests in a period receives no calculated result. Displayed aggregates come from the imported measurement sample.
Full methodology →Data freshness
- Page updated
- Data period
- September 2025
- Responsible editor
- Internet Analysis Research Team
Primary sources
- Measurement Lab NDTPrimary measurement protocol and dataset documentation
- M-Lab data documentationAccess, schemas and limitations
- Internet Analysis methodology
Limitations
- IP-based geographic annotations can be approximate.
- People who run a test are not a statistically representative sample of every resident.
- Aggregated results do not guarantee performance for a specific ISP, plan, address or time of day.
- Only imported percentiles are shown; missing bounds remain unavailable.
Related questions
What was the internet speed in United States during September 2025?
The available sample reports 346.46 Mbps download and 128.43 Mbps upload, with 41.5 ms latency.
How are the ranking, summary and reports calculated?
They are calculated from speed tests initiated by users, grouped by period and approximate IP-based geography. Valid test measurements form the sample used for aggregate performance metrics and ranking; no result is invented when the sample is empty.
Does this result represent every internet connection?
No. It summarizes tests present in the source dataset and should not be interpreted as a census of all connections.
Why can my own speed test differ?
Device, Wi-Fi, plan, server path, congestion and time of day can all make an individual result differ from an aggregate report.