Skip to content

#16 of 30 · AI lab + cloud

Baidu (ERNIE)

Own Kunlun chips, PUE 1.19, and renewable power stuck in single digits.

Beijing, China·baidu.com ·Verified Oct 16, 2026
BGood confidence
Index score
Good confidence

In plain words

How scores work

Baidu scores 45 out of 100 — middle of the pack. Some disclosure or clean-energy commitments, offset by rising emissions or blind spots. Confidence in this score is good confidence (B).

What helps

  • Efficiency · Average data-centre PUE of 1.19 and custom Kunlun silicon; ERNIE 4.5 is a mixture-of-experts model.
  • Carbon · Yangquan site certified carbon-neutral; 2030 operational neutrality goal.
  • Commitments · Carbon-neutral operations by 2030 across six areas including data centres; PUE target 1.14; offsets for the remainder.

What hurts

  • Efficiency · No per-prompt figures.
  • Carbon · Renewable electricity about 5% (Greenpeace, 2023); 62 GWh of renewables in 2024 against Scope 2 emissions of 777 kt.
  • Water · Water cooling with free cooling described qualitatively; no WUE or consumption published.

What nobody knows

  • Per-prompt energy disclosed
  • Per-prompt water disclosed
  • Training emissions in model cards
  • Water-use effectiveness reported
  • Hardware reuse or e-waste reported

Baidu trains ERNIE on its own Kunlun P800 chips — a 30,000-chip cluster lit in April 2025 — and runs some of China's most efficient data centres (average PUE 1.19). But renewable electricity is a rounding error: 62 GWh in 2024, which Greenpeace put at about 5% of consumption, against Scope 2 emissions of 777 kt location-based. Carbon-neutral operations are promised for 2030, largely via offsets.

Products
ERNIE 4.5 / X1 · ERNIE Bot · Baidu AI Cloud (Qianfan) · Kunlun chips · Apollo
Hosting
Own data centres in China (Yangquan, Baoding, Beijing) running Kunlun P800 clusters.

Key figures

Per-prompt and fleet metrics. Hover a title for what it means; open “How we got this number” for the working.

How to read the labels:Company figureCompany figure. Published by the company itself. We take it as given and record the scope next to it.Measured independentlyMeasured independently. Measured or estimated by researchers or analysts with no stake in the result.ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Not disclosedNot disclosed. The company publishes nothing on this and no defensible estimate exists.— hover any label.
Energy per promptElectricity used to answer one typical request. Under 0.3 Wh is good. A 10 W LED bulb uses 0.3 Wh in about two minutes.
0.25
Wh / typical prompt
≈ 20.8 smartphone charges per 1,000 prompts
ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.Reuters via Yahoo FinanceBaidu via sustainabilityreports.comEpoch AI

ecogpt.co estimate; no company figure.

How we got this number

ERNIE 4.5 is a 424B mixture-of-experts model with 47B active parameters, served on Baidu's own Kunlun P800 chips in data centres with a PUE of 1.19. Uncertainty roughly ±3×.

Water per promptWater evaporated cooling the servers for one request. Scope matters: on-site cooling only, or also the water used to generate the electricity (which adds ≈1.8 mL per Wh).
0.20
mL / typical prompt
≈ 0.40 500 mL bottles per 1,000 prompts
ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.Reuters via Yahoo FinanceBaidu via sustainabilityreports.comEpoch AIUS Securities and Exchange Commission

On-site cooling only, so comparable with Google and OpenAI's figures.

How we got this number

0.25 Wh per prompt × 0.8 L/kWh (estimated WUE for water-cooled sites with free cooling). Counting the water evaporated to generate the electricity adds ≈0.45 mL.

Carbon per promptGreenhouse gas from one request on the grid where it actually runs. Below 0.1 g is clean. A kilometre in a petrol car is about 170 g.
0.145
gCO₂e / typical prompt
≈ 0.85 km driven (petrol car) per 1,000 prompts
ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.Reuters via Yahoo FinanceBaidu via sustainabilityreports.comEpoch AI

Location-based: what the grid actually emitted. Renewable certificates would bring the market-based figure near zero.

How we got this number

0.25 Wh × 580 gCO₂/kWh (Chinese grid, location-based).

Annual emissionsThe company's whole yearly footprint, in million tonnes of CO₂e. Clouds and device makers dwarf pure labs because they also count shops, logistics or products. Compare the trend, not the size.
0.80
MtCO₂e (2024, Scope 1+2 location-based)
Company figureCompany figure. Published by the company itself. We take it as given and record the scope next to it.TracenableBaidu via sustainabilityreports.com

Scope 2 market-based 645 kt; as compiled from the 2024 ESG report.

Emissions trendDirection of travel against last year or the company's own baseline. Negative is good. Positive means the AI build-out is outrunning clean energy.
-13.6
% YoY (2024, Scope 1+2)
Company figureCompany figure. Published by the company itself. We take it as given and record the scope next to it.Tracenable
Water per kWh (WUE)Litres of water consumed per kilowatt-hour of server electricity, across the fleet. 0.15 is excellent, 0.3 is good, 1.8 is the US data-centre average.
0.80
L/kWh
ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.US Securities and Exchange CommissionKoomey Analytics
How we got this number

No disclosure. Baidu describes water cooling with free cooling and a PUE of 1.19, which points to below-average evaporative use; 0.8 L/kWh assumed.

Clean electricityShare of electricity matched with renewable or carbon-free sources. Annual matching counts certificates bought anywhere; hourly matching on the same grid is far stricter.
5
% renewable electricity (Greenpeace Clean Cloud 2024, for 2023)
Measured independentlyMeasured independently. Measured or estimated by researchers or analysts with no stake in the result.Data Center Dynamics

Baidu reports 62 GWh of renewable electricity in 2024 but not the share.

Hardware & materials

The chips, power, copper and future e-waste behind the fleet — the part sustainability reports leave out

Factors we use
AI chips in service
100,000
accelerators (Kunlun and NVIDIA)
≈ 1,389 racks of 72 GPUs
ecogpt.co estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.Reuters via Yahoo Finance

GPUs, TPUs or custom accelerators the company runs. Analyst figures are in H100-equivalents, company figures in physical chips.

How we got this number

One disclosed cluster of 30,000 Kunlun P800 chips, plus earlier Kunlun generations and NVIDIA A800/H20 capacity for Baidu AI Cloud. Uncertainty ±2×.

Power draw
130
MW
≈ 97,500 US homes' worth of power
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinanceSemiAnalysis

Megawatts the AI fleet pulls. One megawatt runs roughly 750 US homes.

How we got this number

100,000 accelerators × 1.3 kW each, all-in. Fleet power implied by the chip count.

Electricity a year
0.91
TWh / year
≈ 86,667 US homes for a year
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinanceSemiAnalysis

Terawatt-hours per year. One TWh is a year of electricity for about 95,000 US homes.

How we got this number

130 MW × 8,760 h × 80% utilisation. Compute fleet only.

Carbon baked into the hardware
51.4
ktCO₂e embodied in chips and servers
≈ 11,174 petrol cars driven for a year
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinancearXiv 2607.05408NVIDIA developer blog

CO₂e emitted making the chips and servers before they answer a single prompt.

How we got this number

100,000 accelerators × 514 kg CO₂e each (chip plus share of the host server). Emitted in manufacturing, before the first prompt is served.

Future e-waste
1,700
tonnes of servers to retire within ~5 years
≈ 9.4 empty Boeing 747s by weight
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinanceNVIDIAThe Register

Tonnes of server hardware that will be retired in three to five years.

How we got this number

100,000 accelerators × 17 kg of server hardware each. Tomorrow's e-waste unless reused; generative AI could produce 2.5 Mt of it a year by 2030.

Copper in the build
3,510
tonnes of copper in the build
≈ copper for 42,289 electric cars
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinanceSemiAnalysisBHPS&P Global Market Intelligence

Tonnes of copper in cabling, busbars and transformers.

How we got this number

130 MW × 27 t of copper per MW (cabling, busbars, transformers).

Fossil generators on site
50.0
backup diesel generators (implied)
≈ 130 MW of backup capacity
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.Reuters via Yahoo FinanceSemiAnalysisLatitude MediaLatitude Media / Floodlight

Diesel backup units or gas turbines at the data centres, permitted or observed.

How we got this number

130 MW ÷ 2.6 MW per generator, the average size permitted at US data centres. Permit counts are not published fleet-wide; this is the number a fleet this size normally carries.

Hardware reused or recycled
—
No hardware reuse or e-waste figure published.
Not disclosedNot disclosed. The company publishes nothing on this and no defensible estimate exists.

Share of retired servers and parts given a second life.

Mineral depletion
—
Only Mistral has published a mineral-depletion figure.
Not disclosedNot disclosed. The company publishes nothing on this and no defensible estimate exists.

Kilograms of antimony-equivalent, the life-cycle unit for scarce minerals consumed.

Score breakdown

Each pillar is 0–100; weights are fixed across the index

Energy efficiency — How much electricity one answer costs.weight 20% · confidence B
62
Carbon intensity — How clean that electricity is, and whether total emissions are rising.weight 25% · confidence B
28
Water — How much water the data centres consume, and whether it is replaced.weight 15% · confidence C
40
Transparency — How much the company actually publishes.weight 20% · confidence B
45
Commitments — Whether it has dated, measurable climate and water targets.weight 10% · confidence B
50
Conduct — How it behaves: permits, lawsuits, gas turbines, community water.weight 10% · confidence C
60

Models

What this company ships, and what each prompt costs

ModelWeightsEnergy / promptWaterTraining CO₂e
ERNIE 4.5
424B total / 47B active (MoE)
Open0.25 Wh Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.0.200 mL Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.4,000 t Estimateecogpt.co estimate. Our own estimate from published factors. The working is shown next to the number; treat it as an order of magnitude, not a measurement.

Where the compute lives

Named sites from company disclosures and press reporting

OperationalConstructionPlanned
  • Beijing HQBeijing, China
  • YangquanShanxi, China · Carbon-neutral certified; Kunlun clusters.
  • BaodingHebei, China