Digital & IT · Sustainable Web Design
Video Streaming & Conferencing Emissions Calculator
Audit-grade CO₂e for video streaming and video calls, broken into the three segments that actually consume the energy: the data centre that serves the stream, the network that carries it, and the device that plays it. Streaming runs from resolution and hours; conferencing from per-participant bandwidth and meeting length. The result is an attributional, allocation-based estimate carrying an explicit uncertainty band — because in this field, a single point figure without a band is a red flag, not a number.
Three segments, one data-driven chain. Every streaming hour or conferencing minute is resolved into the energy drawn by three segments — network transport, data-centre serving, and the end-user device — each converted to CO₂e through a location-based grid factor and reported as a separate line plus a total. An optional embodied-infrastructure line can be added to the network and data-centre segments.
The energy chain. Network and data-centre energy follow data volume (GB) × energy intensity (kWh/GB) × grid factor. Device energy follows device power (W) × hours ÷ 1000 × grid factor — real measured device watts, not a per-gigabyte device proxy. Streaming data volume comes from resolution (SD 0.7, Full HD 3.0, UHD/4K 7.0 GB per hour). Conferencing data volume is per participant, summed across the meeting, from the selected bandwidth mode.
Attributional, not marginal. The network and data-centre intensities are allocation-based figures from the Sustainable Web Design model (v4, 2023) — they apportion total infrastructure energy across total data transferred. This is deliberately not a marginal model: the energy a single extra stream adds to a network running largely on fixed, always-on equipment is close to zero, so a marginal figure would understate a service’s fair-share footprint. The discredited high-end per-gigabyte estimates that circulated in earlier reporting used a marginal-times-total conflation this calculator avoids.
Network intensity and the uncertainty band. Fixed-broadband delivery uses network 0.059 + data-centre 0.055 kWh/GB (SWD v4). Mobile delivery uses network 0.14 kWh/GB — the midpoint of the IEA’s 2020 4G range (0.1–0.2) — plus the same 0.055 data-centre intensity, and the calculator discloses that the mobile basis differs from SWD. Because published estimates for network energy span an order of magnitude across sources, the headline carries an explicit band (roughly 0.114–0.127 kWh/GB net + data centre for fixed broadband). Report the basis, not just the point value.
Grid basis. Grid factors are location-based, live from Master Brain (Ember globally; the US uses EPA eGRID, Great Britain uses DEFRA), with an optional transmission-and-distribution gross-up. Device power and embodied figures are also live Master Brain reads, shared with the end-user devices calculator so the two tools stay mutually consistent.
Two factor tiers, fully disclosed. Grid, device power, device embodied carbon, and T&D losses are live Master Brain values. The network and data-centre intensities, the resolution-to-data-volume table, and the television power and embodied figures are page-pinned constants — fixed with a cited source, pending promotion to Master Brain once the methodology page settles the allocation method and the television-size basis. Each factor’s tier is stated where it is used; this is an estimate built on disclosed assumptions, not a black box.
Reporting basis. This calculator estimates a third-party digital service from the user’s consumption perspective — Scope 3 for an organisation whose staff stream or attend calls. The reporting period is whatever you set through hours and session count (single mode) or the estate multi-row aggregation (for an organisation-wide total). There is no own-versus-sold toggle: the consumer perspective is the only one modelled.
Excluded: the manufacture of network and data-centre infrastructure beyond the optional embodied line, content production and encoding upstream of delivery, and the device’s full embodied footprint where you are accounting it separately (model that in the end-user devices calculator).
Netflix bitrates via IEA 2020. Higher resolution moves more data (network + data centre), not more device power.
Grid intensity loads live from MasterBrain. UK/US are CO₂e; international (Ember) are CO₂-only.
Default SWD v4: network 0.059 + data centre 0.055 kWh/GB (fixed).
Blank = use the quality preset above.
Pick a mode, quality, device and country above to calculate
Results appear instantly with a range band, the network / data-centre / device split, a full breakdown and the audit trail. Each factor is tagged MasterBrain-live or page-pinned. Add embodied (Scope 3) with one toggle.
Results are indicative use-phase emissions for a third-party video service, computed with a three-segment attributional model (Sustainable Web Design v4 lineage): network + data-centre energy for the data transferred, plus end-user device energy, multiplied by the selected country’s grid carbon intensity. Grid factors, T&D losses and non-TV device power are read live from MasterBrain; the network / data-centre kWh-per-GB (SWD v4 2023), resolution GB-per-hour (Netflix bitrates via IEA 2020) and TV power and embodied carbon (EU energy label; Samsung EPD / JRC Ecolabel) are page-pinned with citations, pending MasterBrain — each factor is tagged in the audit trail. The headline carries a range band because published network-energy estimates span an order of magnitude; always report the allocation basis, not just the point value. International grids (Ember 2025) are CO₂-only; UK (DEFRA) and US (eGRID) are full CO₂e. The optional embodied line amortises device manufacturing over its life and adds network/data-centre hardware embodied — reported as Scope 3, separate from the use-phase electricity. This is a transparent attributional estimate, not metered telemetry; content production, CDN edge specifics and end-of-life are out of boundary. If you operate the infrastructure yourself, use the cloud-compute and data-centre calculators instead. Confirm data rates and grid factors against your own records, and where material proceed to third-party verification.
You have probably seen the headline: an hour of streaming emits as much as driving a few miles, or a year of video calls equals a transatlantic flight. Most of those figures are wrong — not by a little, but by an order of magnitude — because they rest on a single seductive shortcut: that data equals energy, that each gigabyte carries a fixed carbon cost you can multiply out. Networks do not work that way.
This calculator throws out the per-gigabyte myth and models what actually consumes power — the data centre, the network, and your device — as three separate, auditable segments.
An hour of full-HD streaming emits roughly 50–60 gCO₂e on a typical low-carbon grid — about 0.05 kg, not the inflated figures often quoted. The energy splits across data centre, network, and device; no single segment universally dominates. Resolution, device, and grid drive the number.
Accounting basis
Scope 3 — a third-party digital service from the consumer’s perspective. Per session, scaled by hours × sessions.
Three-segment model
Network · Data centre · Device, each a separate line (+ optional embodied). Composition shown, not a single black-box total.
Energy chain
GB × kWh/GB × grid (network + DC) · W × hours ÷ 1000 × grid (device). Attributional, not marginal.
Grid basis
Location-based · Ember · EPA eGRID (US) · DEFRA (GB) · optional T&D gross-up.
Factor sources
SWD v4 (2023) net+DC intensity · IEA 2020 (mobile range, resolution data) · Lövehagen / DIMPACT / Digst (device).
Uncertainty
Headline carries an explicit band (~0.114–0.127 kWh/GB fixed). Published estimates span an order of magnitude.
It is tempting to crown one villain — “it’s the data centres” or “it’s the network.” The truth is that the dominant segment changes with the scenario. Stream 4K to a power-hungry television on a clean grid and the network and data centre lead; watch on a phone over a coal-heavy mobile network and the device and grid take over; sit in an audio-only call and the device dominates outright. This calculator shows the composition for your specific inputs rather than asserting a universal split, because there isn’t one.
Account for the Device Itself Next →
What “Streaming & Conferencing Emissions” Covers — The Three-Segment Chain
Streaming and conferencing emissions are the greenhouse gases from the electricity that delivers video to and from a screen. Unlike a fuel or a flight, that electricity is not drawn in one place — it is spread along a delivery chain, and accounting it honestly means following the chain rather than reaching for a single per-gigabyte multiplier.
The Delivery Chain: Data Centre → Network → Device
- Data centre. The servers that store, encode, and serve the video stream draw power, scaled up by facility overhead. For an attributional estimate, a share of that total is allocated to each unit of data served.
- Network. The fixed or mobile infrastructure — fibre, routers, base stations — that carries the data between the data centre and the viewer. Largely always-on, which is exactly why a marginal per-stream figure misleads.
- End-user device. The screen the video plays on — television, laptop, tablet, or phone — drawing real measured watts for the duration of viewing.
Streaming vs Conferencing — Two Different Workloads
The two look similar but model differently. Streaming is one-to-many and download-dominated: a server sends a high-bitrate stream to a passive viewer, and resolution is the main lever on data volume. Conferencing is many-to-many and symmetric: every participant both sends and receives, so the footprint scales with the number of participants and with whether cameras are on. A one-hour film and a one-hour all-hands with forty cameras are not remotely the same number, and the calculator handles them as separate modes.
Scope Mapping — Where Video Lands in Your Inventory
For an organisation whose staff stream training content or attend video calls, the footprint is a purchased third-party digital service — Scope 3. The device electricity, where staff use company devices in an office you operate, is Scope 2 on the same logic as any other plugged-in equipment; the network and data-centre segments remain your provider’s operational emissions, reported by you as Scope 3. A platform that hosts and sells streaming or conferencing reports the viewer-side energy as Scope 3 Category 11 (use of sold products). This calculator takes the consumer’s perspective throughout.
Included vs Excluded
| Included in this calculator | Excluded — report separately |
|---|---|
| Network transport energy (fixed or mobile), allocation-based | Manufacture of network and data-centre infrastructure beyond the optional embodied line |
| Data-centre serving energy, facility overhead included in the intensity | Content production, filming, and encoding upstream of delivery |
| End-user device power for the viewing or call duration | The device’s full embodied footprint (model in the end-user devices calculator) |
| Optional embodied line for network + data-centre infrastructure | Storage of content not part of the streamed session |
| Location-based grid conversion, with optional T&D gross-up | Software, subscription, and platform overhead unrelated to delivery energy |
The Energy Chain — Data Volume × Intensity × Grid
Two of the three segments — network and data centre — are driven by data volume; the third — device — is driven by power and time. Understanding the chain is what separates a defensible estimate from a viral statistic.
The Core Model
Network + DC: GB × kWh/GB × grid · Device: W × hours ÷ 1000 × grid → sum the three segments
The data-driven segments multiply the data transferred by an energy intensity expressed in kilowatt-hours per gigabyte, then by the grid factor. The device segment ignores data entirely — a screen draws roughly the same power whether the picture is simple or complex — and instead multiplies measured device watts by viewing hours. Summing the three (plus an optional embodied line) gives the total. The intensities are fixed constants with cited sources, while the grid factor and device watts are live Master Brain reads; both tiers are disclosed.
Why Resolution Drives the Data-Driven Segments
Higher resolution means a higher bitrate, which means more gigabytes per hour, which scales the network and data-centre segments linearly. The jump is steep: standard definition moves around 0.7 GB per hour, full HD around 3.0, and 4K around 7.0 — so 4K transfers roughly ten times the data of SD and carries roughly ten times the network-and-data-centre energy. Resolution is the single biggest lever a viewer controls on those two segments.
Device Draw — The Segment Data Can’t See
The device segment behaves differently and is easy to forget precisely because it does not scale with data. A 57-watt television draws its 57 watts whether it is showing SD or 4K; a one-watt phone draws its watt either way. This is why the device’s share of the total swings so widely: as a fraction of a low-data SD stream on a power-hungry TV it can be small, but for an audio-only call on that same TV it can be almost the whole footprint. The device watts are the same live Master Brain rows the end-user devices calculator reads, so the two tools never disagree.
Streaming Emissions — Per-Hour by Resolution
Streaming is the simpler of the two modes: pick a resolution, a device, a network type, and a number of hours, and the calculator returns the three-segment breakdown. Resolution is the dominant input on the data-driven segments.
Per-Hour Data Volume by Resolution
The resolution-to-data-volume figures are fixed constants, drawn from published streaming bitrates via the IEA’s 2020 analysis. They will move to our live data once the methodology page settles the basis.
| Resolution | Data per hour | Relative to SD | Basis |
|---|---|---|---|
| SD (480p) | 0.7 GB | 1× | Page-pinned · IEA 2020 |
| Full HD (1080p) | 3.0 GB | ~4× | Page-pinned · IEA 2020 |
| UHD / 4K (2160p) | 7.0 GB | 10× | Page-pinned · IEA 2020 |
A custom gigabytes-per-hour override is available where you know your platform’s actual bitrate. There is no 8K tier: it remains a negligible share of real-world viewing, and adding it would imply a precision the underlying data does not support.
The instinct is to treat streaming carbon as a data problem — cut the gigabytes and you cut the carbon. But the device you watch on and the grid powering everything are frequently the larger levers. Dropping from 4K to HD on a phone barely moves the device line; switching from a 200-watt television to a laptop, or from a coal grid to a clean one, can halve the total. Optimise the segment that dominates your scenario, which the calculator shows you — not the one the headlines fixate on.
Video Conferencing Emissions — Calls, Cameras, and Participants
Conferencing scales with people, not just hours. Every participant sends and receives video, so a meeting’s footprint is the per-participant figure multiplied across everyone in the room — which makes participant count and camera state the levers that matter.
Per-Participant Bandwidth by Mode
The calculator computes conferencing data volume per participant-hour as bandwidth in megabits per second times a fixed conversion (GB per hour = Mbps × 0.45), summed across the meeting. The bandwidth figures below are representative of typical Zoom, Teams, and Meet usage — they are stated assumptions; the gigabytes-per-hour arithmetic from them is exact.
| Mode | Bandwidth (assumed) | Data per participant-hour |
|---|---|---|
| Audio only (camera off) | 0.10 Mbps | ~0.045 GB |
| SD video (360p) | 0.60 Mbps | ~0.27 GB |
| HD video (720p) | 1.50 Mbps | ~0.675 GB |
| Full HD video (1080p) | 3.00 Mbps | ~1.35 GB |
The Camera-Off Lever
Turning a camera off is the most-discussed conferencing lever, and the data explains why: audio-only moves roughly a thirtieth of the data of a full-HD video feed. Across the data-driven segments, that is a large cut — though the device, which keeps drawing its watts whether the camera is on or off, is unaffected. So camera-off meaningfully reduces the network and data-centre lines but leaves the device line intact, which is precisely the kind of segment-level nuance a single-number estimate hides.
Camera on, Full HD
~1.35 GB per participant-hour on the data-driven segments. The high-bandwidth default for a large all-cameras meeting.
Camera on, SD
~0.27 GB per participant-hour — a five-fold cut from Full HD with little perceptible difference for most calls.
Camera off, audio only
~0.045 GB per participant-hour — the floor for the data-driven segments. The device line is unchanged.
The “Data ≠ Energy” Problem — Why Most Streaming Numbers Are Wrong
The single biggest source of error in streaming-carbon reporting is the assumption that energy scales linearly and steeply with data, captured in a fixed kilowatt-hours-per-gigabyte figure applied marginally. The figures it produces are often an order of magnitude too high, and they have driven a decade of misleading headlines.
An influential 2019 estimate put network energy at around 0.9 kWh/GB and multiplied it straight through streaming data volumes, producing the “one hour of video equals driving a car” claims that spread widely. The figure conflated a marginal calculation with a total-energy allocation and used rapidly outdated efficiency data. Network energy intensity has fallen sharply with each hardware generation, and the always-on nature of the infrastructure means the marginal energy of one extra stream is far below any per-gigabyte average. This calculator does not use that figure in any result — it appears only here, as the cautionary outlier it is.
Allocation vs Marginal — and Why It Matters
There are two honest ways to count network energy and one dishonest one. The marginal question — how much extra energy does one more stream cause — yields a very small number, because the network runs whether you stream or not. The attributional question — what fair share of total network energy does this stream represent — yields a larger, defensible number used for footprinting, and is what the Sustainable Web Design model provides. The error is to take a marginal-style per-gigabyte figure and multiply it as though it were attributional, inflating the result. This calculator uses the attributional SWD intensities throughout and states the basis explicitly, with a band rather than a false-precision point value.
Worked Example — An Hour of Full-HD Streaming, Step by Step
This example reproduces the calculator’s default boot state: one hour of full-HD streaming to a television on UK fixed broadband. The figures are the engine’s exact default output and bracket the Carbon Trust’s widely-cited ~55 gCO₂e-per-hour video-on-demand estimate — the deliberate validation anchor.
One hour of Full HD (3.0 GB) on a 57-watt LED television, on UK fixed broadband, draws an estimated 0.399 kWh across the three segments. Converted at the UK location-based grid factor, the total is 52.25 gCO₂e — about 0.0523 kg, with an uncertainty band of 52.3–57.4 g. The split is network 44.4%, data centre 41.4%, device 14.3%.
| Segment | Calculation basis | CO₂e | Share |
|---|---|---|---|
| Network | 3.0 GB × 0.059 kWh/GB × UK grid (page-pinned intensity) | 23.18 g | 44.4% |
| Data centre | 3.0 GB × 0.055 kWh/GB × UK grid (page-pinned intensity) | 21.61 g | 41.4% |
| Device | 57 W × 1 h ÷ 1000 × UK grid (page-pinned TV power, live grid) | 7.46 g | 14.3% |
| Total | 0.399 kWh × UK grid 0.13096 kgCO₂e/kWh | 52.25 g | 100% |
The Segment Split, Visualised
For this default scenario — a TV on a relatively clean grid — the two data-driven segments lead and the device trails. Change the device to a phone or the grid to a coal-heavy one and these bars shift substantially:
Audit trail note: Factor tiers — page-pinned (cited, pending Master Brain promotion): network 0.059 kWh/GB and data-centre 0.055 kWh/GB (SWD v4, 2023), Full HD 3.0 GB/hour (IEA 2020), TV power 57 W LED. Live Master Brain reads: UK grid 0.13096 kgCO₂e/kWh (location-based, DEFRA 2026). T&D gross-up off; use-phase only (embodied line disabled). Consumer Scope 3 perspective, single session. The 52.25 gCO₂e figure is the engine’s default-boot output recorded on the review date; it brackets the Carbon Trust ~55 g/h VoD reference, which validates the method.
Standards and Methodology — How These Numbers Are Grounded
The calculator’s credibility rests on a published method and disclosed factor tiers. The relevant references fall into three groups.
Accounting Standard
The footprint is reported under the GHG Protocol Scope 3 Standard as a purchased digital service, with device electricity falling under Scope 2 where staff use company devices in operated premises. The consumer perspective is the only one modelled; a platform reporting use of sold products would map the same viewer-side energy to Scope 3 Category 11.
Digital Methodology Basis
Network and data-centre intensities follow the Sustainable Web Design model (v4, 2023), an allocation-based framework for apportioning digital-infrastructure energy across data transferred. The mobile-network range is drawn from the IEA’s 2020 work, whose basis differs from SWD and is disclosed as such. The worked example is reconciled against the Carbon Trust and DIMPACT video-on-demand white-paper figure of roughly 55 gCO₂e per hour, which the default output brackets. The full derivation, including the allocation-versus-marginal reasoning and the television-size basis, is documented in the video streaming methodology page; the website digital-carbon methodology covers the adjacent per-page model.
Grid and Energy Basis
Use-phase electricity is converted with location-based grid factors from Ember Yearly Electricity globally, with the US drawing on EPA eGRID and Great Britain on DEFRA. Device power and embodied figures share the Master Brain rows used by the end-user devices calculator.
Relationship to the Other Digital Models
Streaming sits inside the wider digital-delivery chain. The website and digital carbon calculator applies the same data-centre, network and device model to general web pages, the end-user devices calculator models the playback-device segment in depth, and the data-centre PUE calculator resolves the facility overhead behind the origin and CDN. They share the same SWD factor lineage, so the estimates stay mutually consistent. See also the AI compute calculator, the cloud compute calculator and the IT asset and e-waste calculator for adjacent digital workloads.
Reducing Streaming & Conferencing Emissions — Levers That Move the Number
The levers are not equal, and which one leads depends on the scenario the calculator shows you. Ranked by typical leverage for an office workforce:
1 · Right-size the device
Watching on a laptop instead of a large television can cut the device line several-fold — often the largest single move.
2 · Cap resolution
SD or HD instead of 4K cuts the network and data-centre segments proportionally. Biggest lever on the data-driven lines.
3 · Cameras off when not needed
Audio-only moves a fraction of the data of full-HD video. Trims the data-driven segments sharply on long calls.
4 · Prefer fixed over mobile
Fixed broadband carries a lower network intensity than mobile (0.059 vs 0.14 kWh/GB). Wi-Fi beats cellular for long sessions.
5 · Grid and timing
A cleaner grid scales every segment down at once. Matters most for high-draw devices and dirty grids.
6 · Avoid redundant streaming
Background autoplay and idle video draw full energy for no value. Pausing genuinely unwatched streams is a free cut.
The practical implication: there is no universal “do this one thing.” A team streaming 4K training videos on conference-room televisions should right-size the screen and cap resolution; a remote team on back-to-back HD calls should turn cameras off and prefer Wi-Fi. The calculator’s segment breakdown tells you which lever is yours.
Audit Checklist — Common Streaming-Emissions Reporting Errors
The errors below are the recurring sources of overstated or unverifiable streaming and conferencing figures. The calculator guards against most of them; viral statistics rarely do.
- Using a fixed marginal per-gigabyte figure. The most common and most damaging error. Multiplying data by an inflated, outdated kWh/GB figure produces results an order of magnitude too high. Use an attributional intensity with a disclosed basis and band.
- Forgetting the device segment. Network-and-data-centre-only estimates omit the screen, which for many scenarios — phones, audio calls — is the largest segment. Count all three.
- Treating the device as data-driven. A screen draws its watts regardless of resolution. Modelling device energy from gigabytes rather than watts × hours misallocates the device line.
- Ignoring participant count in conferencing. A meeting’s footprint scales with people, not just minutes. Counting a forty-person call as one stream understates it by the participant count.
- Quoting a point value with no uncertainty band. Network intensity estimates span an order of magnitude across sources. A single figure without a band overstates the precision available.
- Mixing fixed and mobile intensities silently. Mobile network energy is materially higher than fixed, and on a different basis. State which delivery path you assumed.
- Double-counting the device with a separate device inventory. If you account device electricity here and again in a device calculator, you count it twice. Pick one boundary and state it.
- Conflating allocation tiers. Page-pinned intensities and live grid factors are different provenance tiers. An auditor will ask which numbers are measured and which are assumed — state the tier per factor.
Data Sources, Factor Provenance, and Uncertainty
Factor Provenance — Two Tiers
The calculator’s factors fall into two disclosed tiers. Live Master Brain v2026.284 reads:
- Grid factors — Ember Yearly Electricity globally; US via EPA eGRID; GB via DEFRA.
- Device power and embodied carbon — the same end-user-device rows the devices calculator reads (DIMPACT, Digst, Lövehagen).
- Transmission & distribution losses — per-country loss rates, applied as an optional gross-up.
Page-pinned constants (cited, pending Master Brain promotion):
- Network and data-centre intensity — Sustainable Web Design model v4 (2023), attributional; mobile range from IEA (2020).
- Resolution-to-data-volume — streaming bitrates via IEA (2020).
- Television power and embodied carbon — page-pinned because the underlying device dataset carries no television row.
Uncertainty
| Input | Nature of uncertainty |
|---|---|
| Network intensity | Published estimates span an order of magnitude; the headline carries a band. The most contested input. |
| Data-centre intensity | Allocation-based class average; a specific provider’s efficiency varies. |
| Resolution data volume | Representative bitrates; actual platform encoding differs. Custom override available. |
| Conferencing bandwidth | Stated assumption per mode; the GB/h arithmetic from it is exact. |
| Device power | Live class-average watts; actual draw varies by model and brightness. |
| Grid factor | Annual location-based average; real intraday intensity varies. |
Update Schedule
Grid and device factors refresh with each Master Brain release; the page-pinned intensities update when their source models do and on promotion to Master Brain. The data version badge in the calculator footer always reflects the live MasterBrain version in use.
Frequently Asked Questions
Roughly 50–60 gCO₂e for an hour of full-HD video on a television on a relatively clean grid — about 0.05 kg, far below the inflated “equals driving a car” figures that circulated in earlier reporting. The exact number depends on resolution, the device you watch on, the network type, and the grid. The energy splits across data centre, network, and device, and no single segment universally dominates. The calculator returns the breakdown for your specific inputs.
Per hour, a video call is a small footprint — comparable to streaming and far below the travel it usually replaces. It scales with participants and camera use: a forty-person all-cameras call carries roughly forty times the per-participant data of one feed. The largest reductions come from turning cameras off when they add no value and lowering video resolution. Against the alternative of physical travel, conferencing is almost always the lower-carbon choice.
Yes, for the network and data-centre segments — audio-only moves roughly a thirtieth of the data of a full-HD video feed, so those two lines drop sharply on a long call. The device line does not change, because your screen draws the same power whether the camera is on or off. So camera-off is a real cut to the data-driven segments but not a total elimination; its impact is largest on long, many-participant meetings.
Because they multiply data volume by a fixed, inflated energy-per-gigabyte figure — often a 2019 estimate of around 0.9 kWh/GB — as though network energy scaled linearly and marginally with bytes. It does not: the network is largely always-on, so one extra stream adds almost no energy, and intensity has fallen with each hardware generation. The honest approach is attributional allocation with a disclosed basis and uncertainty band, which is what this calculator uses.
It matters for the data-driven segments. 4K moves around 7 GB per hour against 3 GB for full HD, so the network and data-centre lines roughly double. It does not change the device line, which depends on the screen’s power, not the resolution. On a large television the device line is already significant, so capping resolution helps most when paired with right-sizing the screen. The calculator shows the trade-off for your setup.
Fixed broadband over Wi-Fi is lower than mobile for the network segment — the calculator uses 0.059 kWh/GB for fixed against 0.14 for mobile, the latter from the IEA’s 2020 4G range. For long sessions on a phone, switching from cellular to Wi-Fi cuts the network line meaningfully. The data-centre and device segments are unaffected by the network choice. Note the mobile figure sits on a different basis from the fixed one, which the calculator discloses.
For an organisation consuming the service, it is Scope 3 — a purchased third-party digital service. The device electricity, where staff use company devices in premises you operate, is Scope 2 like any plugged-in equipment, while the network and data-centre energy stays your provider’s operational emissions that you report as Scope 3. A platform that hosts and sells streaming reports the viewer-side energy as Scope 3 Category 11. This calculator takes the consumer’s perspective.
Because the honest uncertainty in network energy intensity is wide — published estimates span an order of magnitude across the IEA, the Sustainable Web Design model, and others. A single point value would imply a precision that does not exist. The calculator reports a central estimate with a band (around 0.114–0.127 kWh/GB for fixed broadband, network plus data centre) so you can see the range your figure sits within rather than a false-precision single number.
Methodology Notes and Limitations
Three segments, reported separately. Network, data centre, and device are distinct lines summed into a total, with no single segment universally dominant — the composition shifts with resolution, device, and grid.
Attributional, not marginal. Network and data-centre intensities allocate total infrastructure energy across total data transferred. The marginal energy of one extra stream is far lower, but a fair-share footprint uses allocation. The discredited high-end per-gigabyte figures are not used in any result.
Two factor tiers, disclosed. Grid, device power, device embodied carbon, and T&D losses are live Master Brain reads. Network and data-centre intensity, resolution data volumes, and television power and embodied carbon are fixed constants with cited sources, until they can move to our live data once the methodology page settles the allocation method and television-size basis.
Device, not data, drives the device line. A screen draws its watts regardless of resolution, so the device segment is modelled from measured watts and hours, sharing the end-user-device rows for consistency. Account the device in one calculator, not two.
An uncertainty band, by design. Network intensity estimates span an order of magnitude; the headline carries a band rather than a false-precision point value. Report the basis alongside the figure.
Consumer perspective, user-set period. The calculator models a third-party service from the consumer’s Scope 3 viewpoint. The period is whatever you set through hours and session count, or the estate multi-row aggregation for an organisation-wide total — there is no own-versus-sold toggle.