Skip to content

#22 of 30 · Wafer-scale chipmaker & cloud

Cerebras

A dinner-plate chip drawing 23 kW; six data centres; no environmental report.

Sunnyvale, California·cerebras.ai ·Verified Oct 17, 2026
DLittle data
Index score
Little data

In plain words

How scores work

Cerebras scores 37 out of 100 — laggard. Little or no disclosure, fossil-backed capacity, or active disputes with regulators and communities. Confidence in this score is little data (D).

What helps

  • Efficiency · Vendor claim of ≈3× compute per watt versus an eight-GPU DGX; very high tokens per second per system.

What hurts

  • Efficiency · Each CS-3 draws up to 23 kW; no independent efficiency measurement.
  • Carbon · Sites on Texas, Oklahoma and California grids plus G42's UAE capacity; no renewable procurement disclosed.
  • Transparency · No environmental disclosure in press releases, funding announcements or IPO materials.

What nobody knows

  • Per-prompt energy disclosed
  • Per-prompt water disclosed
  • Training emissions in model cards
  • Annual environmental report
  • Location-based Scope 2 reported
  • +3 more

Cerebras' CS-3 packs a wafer-sized chip that draws up to 23 kW, cooled by water at 100 litres a minute, and claims about three times the compute per watt of an eight-GPU DGX. It runs inference for Perplexity, Mistral and others from Santa Clara, Stockton, Dallas, Oklahoma City (300-plus CS-3s) and Montreal, with G42's Condor Galaxy systems in the UAE and United States. It raised $1.1 billion at an $8.1 billion valuation in 2025 ahead of an IPO and publishes no emissions, energy or water data.

Products
Cerebras Inference · CS-3 systems · Condor Galaxy (with G42)
Hosting
Own hardware in partner data centres: Santa Clara, Stockton, Dallas, Oklahoma City (Scale Datacenter), Montreal (Enovum); G42 sites in the UAE and US.

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.
0.15
Wh / typical prompt
≈ 12.5 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.CerebrasEpoch AI

ecogpt.co estimate; no company figure.

How we got this number

Cerebras serves 70B-class open models (GPU baseline ≈0.25 Wh) and claims about three times the compute per watt of an eight-GPU DGX; with no independent measurement we credit 1.7×. 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).
0.05
mL / typical prompt
≈ 0.09 500 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.CerebrasEpoch AI

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

How we got this number

0.15 Wh per prompt × 0.3 L/kWh (estimated WUE for closed-loop water-cooled systems in partner sites). Counting the water evaporated to generate the electricity adds ≈0.27 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.
0.058
gCO₂e / typical prompt
≈ 0.34 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.CerebrasEpoch AI

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

How we got this number

0.15 Wh × 390 gCO₂/kWh (US 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.
0.16
MtCO₂e / year (AI compute only)
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.CerebrasCerebras

Location-based estimate for the AI fleet only; not comparable with a full corporate inventory.

How we got this number

60 MW × 8,760 h × 80% utilisation × 390 gCO₂/kWh (US grid). Compute only — nothing else the company does is counted.

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.
+200
% (six new data centres in 2025)
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.Cerebras
How we got this number

Three sites were live before 2025 and six more were announced for the year, tripling capacity.

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.
0.30
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.CerebrasKoomey Analytics
How we got this number

Closed-loop water cooling at system level inside partner data centres whose heat rejection is undisclosed; 0.3 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.
40.0
% (US grid average non-fossil, no procurement)
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.Cerebras
How we got this number

The US grid is about 40% non-fossil; Cerebras discloses no renewable purchases.

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
120,000
H100-equivalents (≈2,000 CS-3 wafer-scale systems)
≈ 1,667 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.CerebrasCerebras

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

How we got this number

More than 300 CS-3 systems at Oklahoma City alone, six inference sites, and G42's Condor Galaxy supercomputers; one CS-3 does roughly the work of 60 H100s. ±2×.

Power draw
60.0
MW (2026)
≈ 45,000 US homes' worth of power
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.CerebrasCerebras

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

How we got this number

≈2,000 CS-3 systems at up to 23 kW each plus cooling and networking.

Electricity a year
0.42
TWh / year
≈ 40,000 US homes for a year
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.CerebrasCerebrasSemiAnalysis

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

How we got this number

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

Carbon baked into the hardware
61.7
ktCO₂e embodied in chips and servers
≈ 13,413 petrol cars driven for a year
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.CerebrasCerebrasarXiv 2607.05408NVIDIA developer blog

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

How we got this number

120,000 accelerators × 514 kg CO₂e each (chip plus share of the host server). Emitted in manufacturing, before the first prompt is served.

Future e-waste
2,040
tonnes of servers to retire within ~5 years
≈ 11 empty Boeing 747s by weight
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.CerebrasCerebrasNVIDIAThe Register

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

How we got this number

120,000 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
1,620
tonnes of copper in the build
≈ copper for 19,518 electric cars
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.CerebrasCerebrasBHPS&P Global Market Intelligence

Tonnes of copper in cabling, busbars and transformers.

How we got this number

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

Fossil generators on site
23.0
backup diesel generators (implied)
≈ 60 MW of backup capacity
DerivedDerived. Arithmetic on a disclosed or independent input using a published conversion factor.CerebrasCerebrasLatitude MediaLatitude Media / Floodlight

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

How we got this number

60 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 C
62
  • Vendor claim of ≈3× compute per watt versus an eight-GPU DGX; very high tokens per second per system. Cerebras
  • Each CS-3 draws up to 23 kW; no independent efficiency measurement. Cerebras
Carbon intensity — How clean that electricity is, and whether total emissions are rising.weight 25% · confidence C
38
  • Sites on Texas, Oklahoma and California grids plus G42's UAE capacity; no renewable procurement disclosed. Cerebras
Water — How much water the data centres consume, and whether it is replaced.weight 15% · confidence C
45
  • Water-cooled at about 100 L/min per system in closed loops; no WUE or consumption published. CerebrasCerebras
Transparency — How much the company actually publishes.weight 20% · confidence D
8
  • No environmental disclosure in press releases, funding announcements or IPO materials. HPCwire
Commitments — Whether it has dated, measurable climate and water targets.weight 10% · confidence C
10
Conduct — How it behaves: permits, lawsuits, gas turbines, community water.weight 10% · confidence C
55
  • No enforcement actions on record. Cerebras

Models

What this company ships, and what each prompt costs

No model-level entries — this company hosts or supplies others rather than serving its own models.

Where the compute lives

Named sites from company disclosures and press reporting

OperationalConstructionPlanned
  • Sunnyvale HQCalifornia, USA
  • Santa ClaraCalifornia, USA
  • DallasTexas, USA
  • Oklahoma City (Scale Datacenter)Oklahoma, USA · 300+ CS-3 systems.
  • Montreal (Enovum)Quebec, Canada
  • Condor Galaxy (G42)Abu Dhabi, UAE