BloombergNEF data-center demand: wind-PPA and hourly-match screen#
Trigger (11 November 2026): BloombergNEF, Data Center Power Demand Is Set to Surge, Raising New Climate Challenges, reporting the Powering the Data Center Boom study commissioned by Bloomberg Philanthropies.
Source provenance and claims#
Input |
Value |
Status and time boundary |
|---|---|---|
Worldwide data-center electricity demand in 2035 |
2,354 TWh/year |
BNEF forecast, published November 11, 2026; not observed generation, a guaranteed load or a 2026 annual total. |
Share of projected world electricity in 2035 |
about 7% |
BNEF forecast; contextual only. |
Onsite gas generation proposed alongside data centers |
181 GW |
BNEF announced pipeline, project maturity variable; not commissioned or 2035 capacity. |
Wind capacity factor, delivery loss, PPA share |
ranges in script |
Assumed scenarios, not attributed to BNEF. |
168-hour wind profile, site consumption, BESS |
generated deterministically in script |
Synthetic; no geographic, meteorological or power-system calibration. |
The publicly accessible BNEF press announcement is quoted only for three top-line numerical claims and linked to its primary source. No copyrighted underlying analyst dataset, full report, weather records or customer load records are copied into this repository. Values are manually transcribed from the publisher’s November 11 press release; publication date and URL are the reproducibility key. The forecast may change in later updates.
Business question#
An electricity buyer can sign a PPA covering a fraction of annual MWh, but data-center operation also depends on hourly availability and physical grid deliverability. We screen (1) wind capacity required to generate a given annual MWh credit and (2) the difference between energy accounting and an illustrative hourly power balance with optional storage. It does not estimate additionality, contractual rights or emissions avoidance.
Physical method and baselines#
Annual nameplate: (P_{wind,GW} = (E_{DC,TWh} \cdot f \cdot 1000) / (8760 \cdot CF \cdot (1-L))). Here (f) is the purchased annual-energy fraction and (L) is an assumed, constant delivery loss. This is a net annual energy screen under the stated conventional-year hours, not a capacity-adequacy model.
For hourly scenarios, (P_{wind,t}=P_{nameplate} CF_t(1-L)), (E_{wind}=\sum_t P_{wind,t}\Delta t). The aggregate energy match is (\min(E_{wind}/E_{load},1)), regardless of contemporaneous coincidence. Physically served fraction is instead calculated using energy_methods.hybrid_dispatch.simulate_hybrid: prioritize wind supplied directly to demand, then charge the battery from any surplus and discharge under its MW and MWh bounds with symmetric round-trip efficiency losses. Starts at zero state of charge; no grid imports. Zero storage is the hourly physical baseline; 100% aggregate energy accounting does not guarantee hourly matching.
Run and validation#
uv sync --locked --all-packages
uv run --no-sync pytest methods/tests/test_data_center_wind.py
uv run --no-sync python examples/analysis/plot_bnef_data_center_wind_2026.py
The gallery runs without network requests. Figure 1 varies hypothetical wind CF (30–54%) and annual procurement fraction (25%, 50%, 100%) while using BNEF’s 2035 global forecast solely as a demand-scale anchor. Figure 2 varies synthetic storage duration at a 100 MW illustrative site over 168 synthetic hours. It is neither a real 2035 weather trace nor an annual demand forecast. Neither figure has statistically calibrated confidence bands, so ranges must not be described as probabilities or uncertainty intervals.
Scientific and financial limits#
No hourly BNEF data-center forecast, measured wind generation, node-to-node deliverability, wind-resource covariance, battery degradation, PPA price, project cost, curtailment compensation, electrolyser/demand flexibility, construction timetable or regional geographic allocation is provided. A global annual energy projection cannot be attributed to particular data centers or countries. The 181 GW gas proposal figure cannot be converted directly into a wind substitution requirement, because gas firm output and intermittent wind nameplate are not physically equivalent. The synthetic time window cannot establish annual hourly clean-energy share, LOLP, EENS, operational carbon emissions or project ROI.
Next evidence upgrade: obtain licensed country-specific demand forecast slices, real hourly substation load and weather, loss measurements, and time-matched contracted generation before performing 24/7 CFE, grid congestion and cost-of-service studies. Compare with a credible dispatchable baseline and optimize wind, solar, BESS and demand flexibility subject to capacity adequacy and economic constraints.