Guides
How Much Internet Speed Do You Actually Need? A Reality Check for US Homes
How much internet speed you need depends on peak-hour load, not a marketing number. Use the FCC 100/20 Mbps benchmark and measure what you actually get.
What to take away
- The FCC's current broadband benchmark is 100 Mbps download and 20 Mbps upload, raised from 25/3 in March 2024.
- Most US homes with three or fewer people streaming in 1080p or 4K are served by 100 to 300 Mbps down; heavier households want 500 Mbps or a gigabit tier.
- The number that matters is measured throughput at 8 to 10 p.m., not the speed printed on your bill.
- A wired speed test is the only fair reading. Wi-Fi adds its own limits on top of the line.
- If measured download sits below 70 percent of your plan's advertised speed for three nights running, that is your threshold to call the provider.
What to measure
Throughput, meaning the bits per second actually delivered to a device, is the metric. Advertised speed is a ceiling on the access line, not a promise about your evening.
The FCC sets the benchmark for its own reports: 100 Mbps down and 20 Mbps up for fixed broadband. That replaced the 25/3 standard, which had stood since 2015.
Its consumer guidance covers what common activities demand, from video calls to multiple 4K streams. Read the FCC's framing on broadband speed requirements for common activities.
Measure with a wired connection first. Plug a laptop into the router with Ethernet, close other tabs, and run a test. Then repeat over Wi-Fi from the room where you actually watch TV. The gap between those two numbers tells you whether you have a line problem or a Wi-Fi problem.
How to read it
Run the test three times on three separate evenings, between 8 and 10 p.m. Peak hours are when congestion shows up. A single midday test flatters every provider.
- Test wired at the router, three evenings, same time window.
- Record download, upload and latency for each run.
- Test wirelessly from the room with the most devices.
- Compare the wired average against your plan's advertised download figure.
- Note whether latency climbs above roughly 50 ms, which is where video calls start to stutter.
| Household pattern | Typical downstream need | Notes |
|---|---|---|
| One or two people, 1080p streaming, browsing | 100 to 200 Mbps | Meets the FCC benchmark |
| Three or four people, one 4K stream plus gaming | 300 to 500 Mbps | Upload matters for streaming out |
| Remote work with large file transfers | 500 Mbps to 1 Gbps | Check upload, not just download |
| Many devices, smart home, multiple 4K streams | 1 Gbps | Headroom, not necessity |
Upload is the number US cable plans hide. A 1,000 Mbps down plan often ships 20 to 35 Mbps up, which is thin for two people on video calls while a backup runs.
What it cannot tell you
A speed test cannot separate a congested neighborhood node from a bad cable in your wall. It also cannot measure consistency. A line that hits 900 Mbps at noon and 40 Mbps at nine is a different product from one that holds 300 Mbps all evening, even though the first has the better headline number.
The FCC benchmark itself is a policy floor, not a household recommendation. It describes the minimum the agency considers broadband at all. It says nothing about how many people, streams or devices sit behind that connection. Treat it as the line below which a plan is hard to defend, not as a target.
Wi-Fi is the other blind spot. A 2016 router, or one in a closet behind a metal door, can cap a gigabit line at 150 Mbps in the far bedroom.
If your wired test is fine and your wireless test is not, the fix is a router or a mesh node, not a bigger plan. Our guide to turning on accessibility features without losing what already works covers the same principle for device settings: change one thing, then re-test.
Attribution and its limits
When a test comes back low, you need to know whose problem it is. The provider owns the line to the modem. You own everything after it. That boundary is the demarcation point, usually the box where the cable or fiber enters the house.
If wired speed at the modem is low, the provider is the suspect. If wired speed at the modem is fine but a router test is low, the router or its settings are the suspect. Document each reading with a timestamp and a screenshot. Providers ask for this, and without it the call becomes an argument about perception.
Note latency and jitter too. Gaming and video calls care about round-trip time more than raw throughput, and a 200 Mbps line with 15 ms latency beats a 1 Gbps line with 90 ms for a competitive shooter.
The general mechanics of how access networks deliver these figures are covered in the overview of internet access technologies.
When to stop measuring and decide
Stop when the pattern is clear. Three evenings of wired readings below 70 percent of the advertised download speed is enough evidence to open a ticket or switch. Three evenings at or above 90 percent with good latency means your plan is doing its job, and any remaining complaint is a Wi-Fi or device issue.
If the readings are fine but one room is slow, buy a mesh node instead of a faster tier.
If upload is the shortfall and you work from home, look for a fiber or cable plan with symmetrical or high upload. Check whether your address has more than one provider.
Where a provider fails to deliver advertised speeds, state consumer protection offices and the FCC's complaint process are the escalation path. A written log makes that complaint stick.
