Power, Energy, Capacity & Efficiency
The vocabulary that separates a nameplate spec sheet from a bankable energy asset — MW vs MWh, capacity factor, C-rate, round-trip efficiency, and why “how big is it?” is never a complete question.
Musk Practical Energy Guide · Part 1 — Energy Fundamentals · Chapter 3 of 80 · 26 min read
A 500 MW solar plant. A 100 MW / 400 MWh battery. A cell with a 2C discharge rate. A power station running at a 90% capacity factor. These numbers all sound like they measure “how much energy something can handle” — and they all measure something different.
Confusing them is one of the most common and most expensive mistakes in energy engineering, finance and procurement. This chapter is the vocabulary that separates a nameplate spec sheet from a bankable energy asset.
3.1 — Power vs energy — the water-tank model
Chapter 1 introduced the distinction; this chapter builds every later concept on top of it, so it is worth restating in its most useful form.
Table 3.1 — The water-tank analogy
| Water system | Energy-system quantity | Unit |
|---|---|---|
| How much water is in the tank | Energy | kWh, MWh, GWh |
| How fast water flows out | Power | kW, MW, GW |
| How much the tank can hold | Capacity | kWh or MWh |
| How much water reaches the destination | Efficiency | % |
E = P × t
A 10 MW generator running for 5 hours delivers 10 × 5 = 50 MWh
3.2 — MW vs MWh — reading a BESS nameplate
A 100 MW / 400 MWh battery energy storage system tells you two independent things. The first number is the maximum instantaneous power it can charge or discharge at. The second is how much energy it holds. Dividing one by the other gives the figure the industry actually trades on:
400 MWh ÷ 100 MW = 4 hours
Which is why the industry calls this a “4-hour battery”
Table 3.2 — Same power rating, very different assets
| Configuration | Duration at full power | Typical duty |
|---|---|---|
| 100 MW / 100 MWh | ~1 hour | Frequency response, fast reserve |
| 100 MW / 400 MWh | ~4 hours | Evening peak shifting, capacity market |
| 100 MW / 800 MWh | ~8 hours | Overnight arbitrage, long-duration shifting |
In plain English
Two BESS projects can both advertise “100 MW” while one sustains that output four times longer than the other. “How big is the battery?” is not a complete question. The complete question is: what is the MW/MWh configuration, and what duration does that imply?
3.3 — Nameplate capacity is not actual generation
A wind farm’s 200 MW nameplate does not mean it generates 200 MW continuously; wind varies, so generation varies. A 100 MW solar plant does not produce 100 MW × 24h × 365d. That is only its theoretical ceiling under continuous full-power operation, which never actually happens. The ratio between the two has a name:
CF = Actual energy generated ÷ (Rated capacity × 8,760 h)
8,760 is the number of hours in a year
Worked example 3.1 — Reading a capacity factor
A 100 MW plant generates 262.8 GWh in a year. Its theoretical maximum would be:
100 MW × 8,760 h = 876,000 MWh = 876 GWh
CF = 262.8 ÷ 876 = 30%
Nothing has gone wrong. 30% is an entirely healthy figure for a well-sited onshore wind farm — it simply reflects that the wind does not blow at rated speed all year.
Table 3.3 — Typical capacity factor ranges
| Technology | Typical capacity factor |
|---|---|
| Utility-scale solar PV | 18–28% |
| Onshore wind | 28–42% |
| Offshore wind | 42–58% |
| Run-of-river / reservoir hydro | 35–55% |
| Combined-cycle gas, baseload duty | 50–70% |
| Nuclear | 85–93% |
Why this matters later
This is why comparing generation technologies purely on $/MW is misleading. The relevant unit is almost always $/MWh delivered. A megawatt of solar and a megawatt of nuclear are not the same product, and no amount of capital-cost comparison will make them so.
3.4 — Efficiency is not capacity factor
Efficiency = Useful output ÷ Energy input × 100
These are independent axes, and conflating them produces nonsense. A plant can be 95% efficient — excellent conversion whenever it runs — and have a 20% capacity factor because it simply runs infrequently. Both statements are true at once, with no contradiction. Efficiency measures quality of conversion; capacity factor measures frequency of operation.
3.5 — Thermal efficiency and the Carnot limit
ηCarnot = 1 − (Tc / Th)
T_h and T_c are the hot- and cold-side absolute temperatures, in kelvin
No heat engine converts 100% of heat into useful work. This is a thermodynamic limit, not an engineering shortfall — no amount of better manufacturing removes it. Real machines fall below even the Carnot ceiling because of friction, heat leakage, pressure losses and turbine and generator inefficiencies.
Technical framing
This is exactly why combined-cycle gas plants outperform simple-cycle ones. A simple-cycle plant discards its turbine exhaust heat; a combined-cycle plant captures that waste heat to raise steam and drive a second turbine, pushing the overall plant closer to — though never reaching — the theoretical limit.
3.6 — Battery round-trip efficiency
RTE = Energy delivered ÷ Energy charged × 100
Worked example 3.2 — Where the missing 10% goes
Charge a BESS with 100 MWh and recover 90 MWh on discharge:
RTE = 90 ÷ 100 = 90%
The missing 10 MWh is lost to cell internal resistance, BMS electronics, the PCS/inverter conversion stage, thermal management and auxiliary loads. Our Battery 101 chapter on the BMS and the cell walks through exactly where those losses arise inside the pack.
Table 3.4 — Typical round-trip efficiency by storage technology
| Storage technology | Typical round-trip efficiency |
|---|---|
| Lithium-ion BESS | 85–95% |
| Pumped hydro | 70–85% |
| Vanadium flow battery | 65–80% |
| Green hydrogen (electrolysis → storage → fuel cell or turbine) | 30–45% |
| Thermal storage (molten salt) | 40–90%, highly design-dependent |
3.7 — State of charge, C-rate and discharge duration
State of charge describes how much energy remains relative to usable capacity: a 100 kWh battery at 80% SoC holds roughly 80 kWh. Only roughly, though — terminal voltage depends on chemistry, current, temperature and internal resistance, not on SoC alone.
Current = C-rate × Rated capacity (Ah)
A 100 Ah cell at 1C draws 100 A; at 2C, 200 A; at 0.5C, 50 A
3.8 — Energy density vs power density — the Ragone trade-off
Energy density (Wh/kg or Wh/L) tells you how much energy a system carries per unit of mass or volume. Power density (W/kg) tells you how fast it can deliver that energy. No storage technology maximises both at once, and the classic way to show that trade-off is a Ragone plot.
Technical framing
There is no universally best battery. An EV pack prioritises energy density, because range per kilogram of pack mass is the product. A grid frequency-regulation asset prioritises power density and cycle life, because it may perform thousands of shallow cycles a year rather than one deep cycle a day. Chemistry and pack architecture follow directly from which side of the Ragone curve the application sits on. Note too that pumped hydro’s dreadful energy density per kilogram is irrelevant at grid scale — the “mass” in question is a mountain reservoir nobody has to carry.
3.9 — Utilisation and its financial consequences
A factory with 10 MW of installed electrical capacity but only 5 MW of average load runs at 50% utilisation. That unused headroom — transformers, switchgear, contracted grid capacity — still costs money every month whether it is used or not. The same logic applies to EV chargers, transmission lines, generation assets and batteries alike. Idle capacity is not free capacity.
3.10 — Data centres: capacity, load and consumption
Worked example 3.3 — Maximum capacity vs actual draw
A data centre with a 500 MW maximum IT load, running flat out continuously, would consume:
500 MW × 8,760 h = 4,380 GWh/year
If its actual average load is 400 MW, annual consumption is instead:
400 MW × 8,760 h ≈ 3,504 GWh/year
That ~876 GWh gap between nameplate capacity and actual draw drives power procurement contracts, backup generation sizing, BESS sizing, cooling design and grid-connection agreements. It is why hyperscale AI campuses negotiate multi-hundred-megawatt PPAs years before a single GPU rack is installed.
The companion metric is power usage effectiveness, which captures everything the facility burns that is not compute:
PUE = Total facility energy ÷ IT equipment energy
Worked example 3.4 — Reading a PUE figure
A data centre draws 600 MW in total, of which 500 MW is IT load:
PUE = 600 ÷ 500 = 1.2
The extra 100 MW covers cooling, pumps, fans, power conversion losses and lighting. Best-in-class hyperscale facilities push PUE toward 1.1 or below; legacy enterprise data centres often sit at 1.6–2.0. As AI training clusters drive up rack densities and push operators toward liquid cooling, PUE becomes a direct lever on both operating cost and grid-connection sizing.
3.11 — DC/AC ratio and clipping
A solar plant might carry 150 MWp DC of array behind only 100 MW AC of inverter — an inverter loading ratio, or DC/AC ratio, of 1.5. Modules rarely hit peak output simultaneously across an entire array, so oversizing DC relative to AC raises inverter utilisation across far more hours of the day. Total annual energy goes up, even though peak output gets clipped at midday.
3.12 — Degradation, depth of discharge and usable energy
Batteries, solar panels, wind turbines, transformers and power electronics all degrade over their operating life. For batteries, degradation accelerates with calendar age, cycle count, high temperature, high state of charge, high C-rate and deep discharge.
100% SoC → 20% SoC = 80% depth of discharge
Depth of discharge is the size of the swing, not the ending state of charge
Worked example 3.5 — Usable energy vs nameplate energy
A BESS rated at 100 MWh, operated within a 90% usable DoD window, delivers at most:
100 MWh × 0.90 = 90 MWh
— and that is before round-trip efficiency losses are applied at all. Layer in a 90% RTE and the round trip returns roughly 81 MWh of the 100 MWh headline.
Why this matters later
This is why “100 MWh” on its own is not a specification. A serious BESS spec has to state AC versus DC capacity, usable energy, the SoC operating window, the efficiency measurement boundary, auxiliary consumption and the end-of-life capacity guarantee. Our Battery 101 chapter on SoC, DoD, SoH and degradation derives each of those terms from first principles and shows how a DoD limit is actually configured on a real pack.
3.13 — LCOE and LCOS — the bridge to economics
LCOE = Lifetime costs ÷ Lifetime electricity generated
LCOS = Lifetime storage costs ÷ Lifetime energy delivered
LCOE lets different generation technologies be compared on a common economic basis. What it does not capture is when the electricity arrives, what transmission it needs, or what integrating it costs the rest of the system. A cheap megawatt-hour generated at noon is not worth the same as a megawatt-hour delivered into the 8 PM peak.
Why this matters later
That single insight — that the value of electricity depends on when and where it is available — is the entire reason storage has economic value at all. It is also the bridge into the next chapter, which takes up tariffs, PPAs and wholesale markets.
3.14 — The vocabulary, assembled
Table 3.5 — Each term, and the question it answers
| Concept | The question it answers |
|---|---|
| Power | How fast? |
| Energy | How much? |
| Capacity | What is the maximum capability? |
| Capacity factor | How much is actually produced over time? |
| Efficiency | How much of the input becomes useful output? |
| Availability | Can the asset operate when required? |
| Utilisation | How much of its capability is actually used? |
| Energy density | How much energy per kg or litre? |
| Power density | How much power per kg or litre? |
| C-rate | How fast can it charge or discharge, relative to capacity? |
| Depth of discharge | How deeply is the battery cycled? |
| Degradation | How does performance decline over time? |
| LCOE | What does generated electricity cost over its life? |
| LCOS | What does delivered stored electricity cost over its life? |
3.15 — The deeper lesson: no single number is enough
A 1 GW power plant is not automatically better than a 500 MW one. A 500 MWh battery is not automatically better than a 250 MWh one. A 95%-efficient system is not automatically worth more than a 90%-efficient one, and a $20/MWh generation source is not necessarily more valuable than a $40/MWh source once timing and reliability are priced in. Evaluating energy infrastructure means holding size, time, utilisation, efficiency, reliability, cost and location together — never any one of them alone.
In plain English
When someone says “we’re building a 500 MW renewable project”, the useful follow-up questions are:
- •500 MW of what — solar, wind, hybrid?
- •What capacity factor is assumed, and what GWh/year does that imply?
- •What transmission capacity is available at the connection point?
- •What is the CAPEX, and what LCOE does it produce?
- •Is storage required, and at what MW/MWh configuration?
- •What does the PPA structure look like?
- •What happens during curtailment?
- •What degradation curve is assumed over the asset life?
Asking these is the difference between reading a spec sheet and evaluating an asset.
Quick check: test yourself
1.A BESS is rated 150 MW / 600 MWh. What is its duration, and how does that compare with a 150 MW / 150 MWh system?
Show answer
2.A 300 MW wind farm runs at a 38% capacity factor. What is its expected annual generation?
Show answer
3.How can a plant be 95% efficient and yet run at only a 20% capacity factor, without contradiction?
Show answer
4.A solar plant is built with a 1.5 DC/AC ratio — 150 MWp DC behind 100 MW AC. Why would a developer choose that over matching DC and AC one to one?
Show answer
Chapter 3 recap — cheat sheet
Core relationship
E = P × t
Everything else builds on this
Storage duration
Duration (h) = Energy (MWh) ÷ Power (MW)
100 MW / 400 MWh = a 4-hour battery
Capacity factor
CF = Actual ÷ (Rated × 8,760 h)
Solar ~20%, wind ~35%, nuclear ~90%
Round-trip efficiency
RTE = Delivered ÷ Charged × 100
Li-ion BESS ~85–95%
C-rate
Duration (h) ≈ 1 ÷ C-rate
2C ≈ 30 min; 0.5C ≈ 2 hours
Power usage effectiveness
PUE = Total facility ÷ IT load
Best hyperscale ~1.1; legacy 1.6–2.0
LCOE and LCOS
Lifetime cost ÷ lifetime energy delivered
The denominator is delivered, never nameplate
The rule
Size + time + utilisation + efficiency + cost
No single number evaluates an asset
Frequently asked questions
What does a “100 MW / 400 MWh” battery rating actually mean?+
The first number is power — the maximum rate the system can charge or discharge at. The second is energy capacity — how much it holds. Dividing energy by power gives duration: 400 MWh ÷ 100 MW is 4 hours, which is why the industry calls it a 4-hour battery. Two systems can both advertise 100 MW while one sustains that output four times longer than the other.
What is capacity factor and why does it matter?+
Capacity factor is actual energy generated divided by the theoretical maximum if a plant ran at full rated power for every hour of the year. A 100 MW plant delivering 262.8 GWh against a theoretical 876 GWh has a 30% capacity factor. It matters because it makes $/MW comparisons between technologies meaningless — utility solar typically runs 18–28%, onshore wind 28–42% and nuclear 85–93%, so the relevant unit is almost always $/MWh delivered.
Can a power plant be highly efficient and still have a low capacity factor?+
Yes, and there is no contradiction. Efficiency measures how well a plant converts input energy to useful output whenever it runs. Capacity factor measures how often it runs relative to its theoretical maximum. A gas peaker can be 95% efficient on every start while operating at a 20% capacity factor, because it is dispatched only during demand spikes.
What is round-trip efficiency in a battery system?+
Round-trip efficiency is energy delivered on discharge divided by energy put in on charge. Charge a BESS with 100 MWh and recover 90 MWh and the round-trip efficiency is 90%. The missing energy goes to cell internal resistance, BMS electronics, the inverter conversion stage, thermal management and auxiliary loads. Lithium-ion typically achieves 85–95%, pumped hydro 70–85% and green hydrogen only 30–45%.
Why do solar plants install more DC panel capacity than AC inverter capacity?+
Because modules rarely peak simultaneously across a whole array. Putting 150 MWp of DC behind a 100 MW AC inverter — a DC/AC ratio of 1.5 — raises inverter utilisation through the many hours of weaker morning, evening and cloudy irradiance. Some midday output is clipped at the inverter ceiling and genuinely lost, but total annual energy usually rises and LCOE falls.
Why is a “100 MWh” battery rating not a complete specification?+
Because nameplate energy is not usable energy. A 100 MWh system held to a 90% depth-of-discharge window delivers at most 90 MWh before efficiency losses, and roughly 81 MWh after a 90% round trip. A serious specification also has to state AC versus DC capacity, the SoC operating window, where efficiency is measured, auxiliary consumption and the end-of-life capacity guarantee.
Reviewed by
Musk Practical Energy Guide is an original educational series explaining how the modern energy system works, from primary resources through to useful work. Figures and worked examples use representative real-world values for illustration and are not investment advice.