Solar Power: From Photons to Gigawatts
From a single photon striking a silicon wafer to a gigawatt-scale power plant — the physics, manufacturing, engineering and finance that together make solar the fastest-growing source of electricity on Earth.
Musk Practical Energy Guide · Part 2 — Power Generation · Chapter 6 of 80 · 32 min read
“Sunlight falls on a panel and electricity is produced” is technically true and industrially useless. It describes the physics of a single cell and tells you nothing about why one project clears a 12% equity return while an identical-looking one next door cannot raise debt at all.
A modern utility-scale solar plant is millions of cells, thousands of modules, inverters, trackers, transformers, SCADA and transmission infrastructure — a combination of quantum physics, semiconductor manufacturing, power electronics, civil engineering and project finance. The physics is settled. Everything that decides whether a plant is worth building sits in the layers above it.
Chapter 5 ended at the grid’s front door. This chapter walks the chain from a single photon striking a silicon wafer through to a gigawatt of grid-connected capacity earning a price it did not choose — and shows why the three numbers that actually decide a solar project are capacity factor, capture price and cost of capital, not module efficiency.
6.1 — From photon to electron: the photovoltaic effect
E = hf
Photon energy = Planck's constant × frequency; higher-frequency photons carry more energy
When a photon carrying sufficient energy strikes a semiconductor — almost always silicon — it can excite an electron out of its bound state, creating an electron-hole pair. The cell’s internal structure separates these charges before they recombine and routes them through an external circuit, producing current.
That is the whole of the photovoltaic effect. There is no combustion, no working fluid, no rotating mass and no thermodynamic cycle. A solar cell has no moving parts and does not convert heat into work, which is why it sidesteps the Carnot limit that caps every thermal plant in Chapter 1 — and also why, as Chapter 5 set out, it contributes no inertia to the grid it feeds.
This single microscopic interaction, repeated across billions of cells, is what powers factories, electric vehicles, homes and data centres. Everything else in this chapter is engineering and finance built on top of it.
6.2 — Why silicon dominates
Silicon is not the most efficient photovoltaic material available, and it never has been. It won for a combination of reasons that have very little to do with peak laboratory performance: an appropriate bandgap for the solar spectrum, natural abundance, long-term stability under decades of UV and thermal cycling, relatively low toxicity, and — decisively — a mature manufacturing base inherited from the semiconductor industry.
Photovoltaic-grade silicon nonetheless requires extremely high purity, quite distinct from the metallurgical-grade silicon used in steel and aluminium alloys. The purification step is capital-intensive, energy-intensive and one of the most concentrated points in the whole supply chain.
Why this matters later
Solar’s cost collapse was a manufacturing story before it was a physics story. The learning curve in Chapter 4 — roughly 20–24% cost reduction per doubling of cumulative production — ran on silicon precisely because silicon could be industrialised. A superior material that cannot be mass-produced to a 25-year warranty is a laboratory result, not an energy technology.
6.3 — The solar supply chain
Stage 1
Quartz — mined silica, the raw feedstock
Stage 2
Metallurgical silicon — carbothermic reduction in an arc furnace
Stage 3
Polysilicon — purification to photovoltaic grade
Stage 4
Ingot — crystallisation into a monocrystalline boule
Stage 5
Wafer — sawn into thin slices
Stage 6
Solar cell — doping, passivation, metallisation
Stage 7
Module — cells strung, laminated, framed, warranted
Stage 8
Solar plant — mounted, wired, inverted, connected
Each stage is a distinct industrial process with its own capital intensity, yield curve and scale economics. They are not interchangeable and they do not sit in the same places: a country can have module assembly without cell manufacturing, or cell manufacturing without polysilicon, and each gap is a different kind of strategic exposure.
This is why solar economics are as much a manufacturing story as a physics story, and why the phrase “domestic solar manufacturing” means almost nothing until you know which of these eight stages it refers to.
6.4 — Cell architecture: PERC, TOPCon, HJT and IBC
| Architecture | Full name | Key characteristic |
|---|---|---|
| PERC | Passivated Emitter and Rear Cell | Rear-side passivation cuts recombination losses; the dominant architecture through the late 2010s |
| TOPCon | Tunnel Oxide Passivated Contact | Thin tunnelling oxide plus passivated contact; higher efficiency and scales onto existing c-Si lines |
| HJT | Heterojunction Technology | Crystalline silicon with thin dissimilar semiconductor layers; strong temperature performance and bifaciality, more complex to manufacture |
| IBC | Interdigitated Back Contact | All contacts on the rear surface, so no front-side shading; highest efficiency potential, most complex process |
Technical framing
The industry’s broader shift from p-type to n-type silicon wafers — the substrate underneath TOPCon, HJT and IBC — shows how a seemingly small materials decision ripples outward. Dopant type in the base wafer changes efficiency, degradation behaviour, temperature coefficient and ultimately the economics of an entire global industry.
6.5 — Cell, module and efficiency
η = Electrical output ÷ Incident solar energy
Efficiency is the fraction of incident solar energy that leaves as electricity
A cell converting 23 of every 100 units of incident solar energy into electricity has 23% efficiency. The other 77 units are not lost in one place. They are split across reflection, thermalisation, recombination, resistive loss and several other mechanisms, most of them governed by the semiconductor’s bandgap.
Photons carrying less than the bandgap energy pass straight through without generating a charge carrier at all. Photons carrying more than the bandgap generate one carrier and dump the excess as heat. Those two effects alone account for the majority of the theoretical limit, and neither is a manufacturing defect — they are consequences of using a single material with a single bandgap against a broad solar spectrum.
Important
This is why improving cell efficiency is a materials-science problem rather than a process-tuning problem, and why efficiency gains arrive in architecture generations rather than continuously.
6.6 — From cell to module: why glass and encapsulation matter
Individual cells are wired into strings and packaged into a module containing cells, interconnects, glass, encapsulant, a backsheet or rear glass, a junction box and a frame. The cell generates the electricity; the module is what survives.
Modules must endure 25–30 years of rain, humidity, dust, UV, wind loading, hail and daily temperature cycling while losing only a fraction of a percent of output per year. Glass performs both an optical and a structural role, which is why modern bifacial modules increasingly use glass-glass construction, collecting light on the front face and — depending on ground reflectivity and mounting height — on the rear face as well.
In plain English
A 25-year warranty on a module is not a statement about the silicon. It is a statement about the lamination, the encapsulant chemistry, the seals and the frame — the parts that have nothing to do with converting photons and everything to do with whether the thing is still working in 2050.
6.7 — The solar resource: DNI, GHI and DHI
| Term | Meaning |
|---|---|
| DNI — Direct Normal Irradiance | Direct sunlight arriving in a straight line from the sun |
| GHI — Global Horizontal Irradiance | Total irradiance striking a horizontal surface |
| DHI — Diffuse Horizontal Irradiance | Scattered sunlight arriving from the sky rather than directly from the sun |
These vary enormously by geography, and not always in the direction intuition suggests. A plant in Rajasthan and one in Kerala can have very different economics purely from differences in irradiance, cloud cover, temperature and dust — before land, grid access or policy enter the calculation at all.
Concentrating technologies depend almost entirely on DNI, because only direct beam radiation can be focused. Flat-plate photovoltaics use GHI and therefore keep producing under diffuse light, which is why solar works commercially in Germany and the UK despite resource levels that would rule out a concentrating plant.
6.8 — Capacity factor revisited
Worked example 6.1 — Capacity factor of a 100 MW plant
A 100 MW solar plant delivers 175,200 MWh over a year.
Theoretical maximum = 100 MW × 8,760 h = 876,000 MWh
Capacity factor = 175,200 ÷ 876,000 = 20%
That sits squarely in the typical utility-scale solar range from Chapter 3. The plant is not underperforming — it is doing exactly what a fixed-tilt array at a good site does.
Capacity factor is the number that converts a nameplate rating into an energy expectation, and it is the reason MW figures are close to meaningless when comparing technologies. The same 100 MW of nameplate buys roughly 20% capacity factor as solar, 30–45% as onshore wind and 80–90% as nuclear. Anyone comparing installed capacity across technologies without capacity factor is comparing labels rather than energy.
6.9 — DC/AC ratio and clipping
DC/AC ratio = DC capacity (MWp) ÷ AC capacity (MW)
Also called the inverter loading ratio or DC overbuild
A “100 MWp DC / 80 MW AC” plant has a DC/AC ratio of 1.25. The module array is deliberately oversized relative to the inverter, and this is a design choice rather than an oversight.
Worked example 6.2 — Why oversize the array
150 MWp DC ÷ 100 MW AC = 1.50
The array only approaches its peak DC output for a handful of hours around solar noon on clear days. For most generating hours it is well below rating. Oversizing the DC side lifts output through the morning, evening and cloudy hours — when the inverter would otherwise sit half-loaded — at the cost of clipping the midday peak at the inverter’s ceiling.
More annual MWh from the same inverter, transformer and grid connection, and a lower LCOE, in exchange for discarding energy on the best days of the year.
Typical ratios run 1.2–1.4. The optimum depends on the shape of the local resource, the cost of modules relative to inverters, and whether the connection agreement or a co-located battery gives the clipped energy anywhere to go.
6.10 — The inverter does far more than convert DC to AC
P = V × I, maximised continuously
MPPT — maximum power point tracking, run continuously as conditions change
Calling an inverter a DC-to-AC converter is like calling a grid a collection of wires. Beyond conversion, a modern inverter provides maximum power point tracking, voltage regulation, reactive power control, grid synchronisation, fault ride-through, monitoring and communications.
As Chapter 5 argued, this is where an inverter-based resource either does or does not behave like a good grid citizen. Grid-forming control, reactive support and ride-through settings are inverter functions, not module functions — which makes the inverter almost as strategically important to a solar plant as the modules themselves, at a fraction of the capital cost.
| Approach | Advantages | Disadvantages |
|---|---|---|
| Central inverter | High power density, fewer units to manage, lower cost per watt | Larger single point of failure, longer DC collection runs, coarser MPPT |
| String inverter | Granular MPPT, easier fault isolation, better mismatch tolerance, partial failures cost less | More equipment, more distributed maintenance, higher unit count |
6.11 — Tracking systems
Fixed-tilt modules sit at a constant angle. Single-axis trackers rotate through the day to follow the sun, raising annual energy capture — typically by 15–25% depending on latitude and resource — at the cost of added capital, mechanical complexity, maintenance and wind-load engineering.
Trackers also change the generation profile, not just its size, pushing output earlier into the morning and later into the evening. In a market where the midday price has already collapsed, that reshaping can be worth more than the extra megawatt-hours.
Important
As with DC/AC ratio, more energy does not automatically mean better economics. The tracker earns its place only when the incremental energy, valued at the price it actually captures, exceeds its incremental cost over the project life.
6.12 — Balance of system
Everything beyond the module — mounting structures, trackers, cables, combiner boxes, inverters, transformers, switchgear, SCADA, access roads, drainage, fencing, security and the grid connection itself — falls under balance of system. It routinely represents half or more of total project cost.
Plant layout is a genuine multi-variable optimisation across row spacing, inter-row shading, terrain, drainage, cable routing and maintenance access. Tighter spacing lifts energy density per hectare but increases self-shading in winter and at low sun angles; wider spacing costs land and cable.
Why this matters later
Because BOS scales with area rather than with watts, a higher-efficiency module reduces it: fewer structures, less cabling and less land per MW. This is the mechanism by which an expensive module can produce a cheaper plant — and the reason module price alone never settles the question.
6.13 — Where solar energy is actually lost
A solar plant never converts its theoretical resource into AC electricity perfectly. Losses stack multiplicatively across optical effects, temperature, mismatch, soiling, DC wiring resistance, inverter conversion and transformer conversion.
The ratio of delivered AC energy to the energy in the plane of the array is the performance ratio, and a well-built plant lands somewhere around 78–84%. Note what the cascade shows: no single category dominates. Temperature is the largest term, but stripping it out entirely would still leave roughly 12 points of loss spread across seven other mechanisms.
In plain English
There is no single villain. A yield model that applies one blanket derate hides which assumptions are load-bearing — and when a plant underperforms, the diagnosis depends entirely on having modelled the stages separately in the first place.
6.14 — Temperature coefficient
Solar modules lose output as cell temperature rises above the 25°C standard test condition. A typical crystalline-silicon temperature coefficient runs around −0.3% to −0.4% per °C, and cell temperature on a still, high-irradiance afternoon can sit 25–35°C above ambient.
Worked example 6.3 — A 500 W module on a hot afternoon
Coefficient = −0.35%/°C, nameplate 500 W at 25°C
Cell temperature 65°C ⇒ ΔT = 40°C
Derate = 1 − (0.0035 × 40) = 0.86
Output = 500 × 0.86 = 430 W
A 14% shortfall against the datasheet, in normal operating conditions, with nothing faulty.
This is why the hottest sites are not automatically the best sites. High irradiance and high cell temperature arrive together, and the second partly cancels the first. It is also why cool high-altitude deserts such as the Atacama outperform what their irradiance figures alone would suggest, and why HJT’s better temperature coefficient is worth more in Rajasthan than in northern Europe.
6.15 — Soiling and the O&M trade-off
Dust, pollen, bird droppings and industrial deposition accumulate on the glass and reduce the irradiance reaching the cells. In arid and semi-arid regions this is one of the largest controllable losses in the plant, and in some locations soiling can cost several percent of annual yield if left unmanaged.
Cleaning recovers that energy but creates operating cost, and in water-scarce regions it consumes a resource that may itself be contested. Developers therefore optimise cleaning frequency against energy recovered: clean too rarely and generation is lost, clean too often and the marginal wash costs more than the marginal energy it recovers.
In plain English
It is the same arithmetic as deciding how often to service a vehicle. There is an optimum, it is site-specific, and it moves with the season, the dust load, the local water price and the electricity price the plant is capturing that month.
6.16 — Degradation and availability
6.16.1 — Degradation
Modules lose a small fraction of output every year — typically a larger drop in year one, then a steady annual rate thereafter. The exact figures depend on technology, materials and climate, and they compound meaningfully across a 25–30 year life. A plant is a declining asset from the day it is energised, and the financial model has to carry that decline explicitly rather than assuming a flat output.
6.16.2 — Availability
Availability is entirely distinct from resource. It asks whether the equipment is online and able to produce, not whether the sun is shining. An inverter fault, a transformer failure or a tripped breaker reduces production on a perfectly clear day, and neither irradiance data nor degradation curves will predict it.
Both effects must be modelled independently in a bankable energy yield assessment. Collapsing them into a single fudge factor is one of the more common ways an optimistic model gets built.
6.17 — EPC and the project lifecycle
Stage 1
Site and resource assessment — irradiance, terrain, temperature, soiling
Stage 2
Land, interconnection and permits — often the longest lead-time items
Stage 3
PPA or offtake secured, and financing raised against it
Stage 4
EPC and construction — cost, schedule and performance certainty transferred
Stage 5
Commissioning, operations, repowering
An EPC — engineering, procurement and construction — contractor typically bears responsibility for cost, schedule and performance certainty. That transfer is the developer’s core risk-management mechanism at the construction stage, and it is priced accordingly.
Note the ordering. Interconnection and permitting sit before financing for a reason: as Chapter 5 set out, a queue position is frequently the binding constraint, and no lender funds a plant that has nowhere to export.
6.18 — LCOE and the cost of capital
Solar is almost pure capital cost. There is no fuel, operating costs are modest, and the asset runs for decades. That structure makes levelised cost acutely sensitive to the discount rate — more so than for any fuel-burning technology, where fuel price uncertainty dominates instead.
Worked example 6.4 — The same plant at two costs of capital
Take an identical plant, identical resource, identical yield, financed at 6% WACC in one market and 12% in another.
Annuity factor, 25 years at 6% = (1 − 1.06⁻²⁵) ÷ 0.06 = 12.783
Annuity factor, 25 years at 12% = (1 − 1.12⁻²⁵) ÷ 0.12 = 7.843
The same capital cost is spread across 12.783 discounted years in one case and 7.843 in the other. Levelised cost rises by roughly 63% — from nothing more than the financing environment.
Important
Access to low-cost capital is often as competitively decisive for a solar developer as module technology. A 2% efficiency advantage cannot outrun a 600 basis point cost-of- capital disadvantage. This is the same WACC sensitivity introduced in Chapter 4, applied to the most capital-intensive generation technology in the book.
6.19 — Four different costs, four different questions
In plain English
A ₹/W cell cost, a ₹/W module cost, a ₹/W installed system cost and a ₹/MWh electricity cost are four different numbers answering four different questions. They are routinely quoted interchangeably, and almost every confused solar argument traces back to that substitution.
A 24%-efficient module at ₹15/W is not automatically superior to a 22%-efficient module at ₹11/W. The correct question is the lowest lifetime cost per unit of energy delivered, which requires netting the balance-of-system savings that higher efficiency unlocks — less land, fewer structures, less cabling per MW — against the premium on the module itself, then dividing by the energy the plant will actually produce and sell.
Only the last of those four numbers is a decision variable. The first three are inputs to it.
6.20 — Supply-chain concentration and strategic risk
Solar manufacturing has become highly geographically concentrated, and that concentration delivered real scale and cost advantages — it is a substantial part of why modules are as cheap as they are. It also creates exposure to trade restrictions, geopolitical tension, export controls and shipping disruption at several of the eight supply-chain stages simultaneously.
Governments pursuing domestic manufacturing for energy security and industrial policy reasons are making an explicit trade: strategic resilience against the higher production costs that new domestic capacity typically carries relative to established low-cost producers already far down the learning curve.
Technical framing
The learning-curve mathematics of Chapter 4 cut both ways here. Cost falls with cumulative production, not calendar time, so a new entrant does not merely need subsidy — it needs enough sustained volume to descend the curve itself. Protection without volume produces expensive modules indefinitely.
6.21 — Solar + storage: solving the time problem
Solar’s core limitation is timing, not quantity. Generation concentrates in daylight while demand persists well into the evening — the evening ramp of Chapter 5, seen from the generator’s side. Pairing solar with a battery converts a purely variable resource into a substantially flexible one.
The right battery size depends entirely on the commercial objective, and these are genuinely different design targets rather than interchangeable descriptions of “adding storage”:
- •Energy shifting — move midday surplus into the evening peak.
- •Firming — deliver a contracted, predictable output profile.
- •Clipping recovery — capture DC energy the inverter would otherwise discard.
- •Ancillary services — frequency response and reserve, sold separately.
- •Curtailment avoidance — store rather than spill when the network is congested.
A battery sized for the second of these will be wrong for the fourth. Chapter 4’s revenue-stacking arithmetic is what decides which combination pays.
6.22 — Capture price: what solar actually gets paid
Worked example 6.5 — Average price is not the price you receive
Suppose the average market electricity price across the year is ₹6/kWh, but the volume-weighted price during the hours a solar plant actually generates averages ₹4/kWh.
Capture rate = ₹4 ÷ ₹6 = 66.7%
The plant’s realised revenue per unit is ₹4/kWh, not ₹6/kWh. A financial model built on the market average overstates revenue by 50%.
The mechanism is cannibalisation. Every solar plant on a system generates in the same few midday hours as every other solar plant, and their combined output depresses the price in exactly those hours. The physical output of the next megawatt is unchanged; its revenue is not.
Important
This is the single most important idea in solar economics after cost of capital. The marginal value of solar declines as solar penetration rises — which means storage, transmission, flexible generation and demand response become progressively more valuable precisely as solar becomes cheaper. Falling module prices do not solve this. They accelerate it.
6.23 — Advantages and limitations
| Advantages | Limitations |
|---|---|
| No fuel cost, and no fuel price risk | Intermittent — no output at night, at all |
| Modular from watts to gigawatts | Weather-dependent output |
| Fast to build relative to thermal or nuclear | Substantial land requirement at scale |
| Mass-manufacturable, and on a learning curve | Integration challenges at high penetration |
| Deployable from rooftop to utility scale | Often needs storage to shift value in time |
| Low operating complexity, no moving parts in the module | Frequently transmission-constrained, not resource-constrained |
| No inertia requirement to operate | Supplies no inertia to the system either |
6.24 — The complete solar stack
Layer 1
Sun — the resource, measured as DNI, GHI and DHI
Layer 2
Silicon — the semiconductor and its bandgap
Layer 3
Cell and module — architecture, efficiency, encapsulation
Layer 4
Strings, inverters, transformers — conversion and control
Layer 5
Plant, transmission and interconnection — getting it to market
Layer 6
PPA, capture price, CAPEX, debt, equity, IRR
Read it downward and each layer constrains the one below. Read it upward and each layer decides whether the one above it was worth building. The physics sets the ceiling; the bottom two layers decide whether anyone reaches it.
6.25 — The core solar mental model
- •Resource — how much sunlight is available, and in what form?
- •Technology — which cell and module architecture, and why?
- •Efficiency — how much of that sunlight becomes electricity?
- •Design — DC/AC ratio, tracking, layout and spacing?
- •Losses — temperature, soiling, mismatch, wiring, conversion?
- •Grid — can the electricity actually be exported, and when?
- •Revenue — what capture price, not what average market price?
- •Financing — what is the cost of capital?
- •Lifetime — how does degradation shape 25–30 years of output?
- •Risk — what happens if costs, prices, generation or policy move?
This framework moves the conversation from “solar is cheap” to “I understand why this particular solar project is or is not economically attractive” — which is the only version of the statement that survives contact with a term sheet.
Chapter 7 turns to the other great variable resource, and to a technology whose physics is far less forgiving: wind power, where output scales with the cube of wind speed and a 10% error in the resource assessment becomes a 33% error in the energy estimate.
Chapter summary
- ✓The photovoltaic effect converts photons directly into charge carriers, with no thermodynamic cycle — which is why solar sidesteps the Carnot limit and also why it supplies no inertia.
- ✓Silicon dominates because it could be industrialised, not because it is the best photovoltaic material. Solar’s cost collapse was a manufacturing story.
- ✓Cell architecture progresses PERC → TOPCon → HJT → IBC, buying efficiency with manufacturing complexity — and complexity is not free.
- ✓Most of the incident energy is lost to bandgap physics: sub-bandgap photons pass through, above-bandgap photons dump their excess as heat.
- ✓A module’s 25-year warranty is a statement about encapsulation and glass, not about silicon.
- ✓Capacity factor, not nameplate MW, converts a rating into an energy expectation; 20% for utility-scale solar against 80–90% for nuclear.
- ✓DC/AC ratios of 1.2–1.4 deliberately oversize the array, trading midday clipping for higher inverter utilisation and lower LCOE.
- ✓The inverter carries the grid-citizenship functions — MPPT, reactive power, ride-through, grid-forming control — at a fraction of the plant’s capital cost.
- ✓Delivered energy is the product of eight sequential derates. Performance ratio lands around 78–84%, and no single loss category dominates.
- ✓Temperature costs a 500 W module roughly 14% of nameplate at 65°C cell temperature, which is why the hottest sites are not automatically the best sites.
- ✓Balance of system scales with area rather than watts, which is the mechanism by which an expensive module can produce a cheaper plant.
- ✓Solar is almost pure capital cost, so LCOE is acutely sensitive to WACC: 6% versus 12% raises levelised cost by roughly 63% with no physical change.
- ✓Cell cost, module cost, installed cost and electricity cost answer four different questions; only the last is a decision variable.
- ✓Capture price falls as solar penetration rises. The marginal value of solar declines as solar gets cheaper — which is exactly why storage and transmission grow more valuable.
Quick check: test yourself
1.Why doesn’t higher module efficiency automatically mean a better project?
Show answer
2.A 150 MWp DC array sits behind a 100 MW AC inverter. What is the DC/AC ratio, and why would a developer choose this over 1:1 sizing?
Show answer
3.What is capture price, and why does it typically fall as solar penetration rises?
Show answer
4.Why can a project have excellent irradiance and cheap land yet still be uninvestable?
Show answer
5.A 500 W module carries a −0.35%/°C coefficient. What does it deliver at 65°C cell temperature, and what does that imply about site selection?
Show answer
Chapter 6 recap — cheat sheet
Photon energy
E = hf
Higher frequency carries more energy per photon
Cell efficiency
η = Electrical output ÷ Incident solar energy
PERC 20–22%, TOPCon 23–24%, HJT 24–25%, IBC 24–26%
DC/AC ratio
DC capacity (MWp) ÷ AC capacity (MW)
Typically 1.2–1.4; trades clipping for utilisation
MPPT
Continuously maximise P = V × I
One of many inverter functions, not the only one
Temperature derate
≈ −0.3 to −0.4 %/°C above 25°C
500 W at 65°C ⇒ 430 W, a 14% shortfall
Performance ratio
Product of every derate ≈ 78–84%
Eight sequential losses; no single villain
Capacity factor
Actual MWh ÷ (MW × 8,760)
≈ 20% utility-scale solar; nameplate MW means little alone
Capture price
Price during generating hours ≠ market average
Falls as penetration rises — cannibalisation
Cost of capital
6% vs 12% WACC ⇒ ≈ 63% higher LCOE
Solar is almost pure CAPEX, so WACC dominates
The rule
Lowest lifetime cost per unit of energy delivered
Not efficiency, not ₹/W, not nameplate MW
Frequently asked questions
Why does a solar cell not obey the Carnot limit?+
Because it is not a heat engine. The photovoltaic effect converts photons directly into electron-hole pairs through a semiconductor junction, with no working fluid, no combustion and no thermodynamic cycle to extract work from a temperature difference. That is why a solar cell has no moving parts and no efficiency ceiling set by hot and cold reservoir temperatures. The same absence of rotating mass is also why solar contributes no inertia to the grid it feeds.
Why does higher module efficiency not automatically mean a cheaper project?+
Higher-efficiency architectures such as HJT and IBC cost more per watt to manufacture. The decision variable is lifetime cost per unit of energy delivered, not efficiency or ₹/W in isolation. Because balance-of-system cost scales with area rather than watts, an efficiency premium has to be repaid through less land, fewer mounting structures and less cabling per MW, then measured against the energy the plant actually sells. A 22%-efficient module at ₹11/W can beat a 24%-efficient module at ₹15/W.
What is a DC/AC ratio and why is a solar array deliberately oversized?+
It is DC capacity in MWp divided by AC capacity in MW, typically 1.2–1.4. A 150 MWp array behind a 100 MW inverter has a ratio of 1.50. Because the array only approaches peak DC output for a few hours around solar noon on clear days, oversizing lifts output through morning, evening and cloudy hours when the inverter would otherwise sit half-loaded. The cost is clipping the midday peak at the inverter ceiling — more annual energy from the same inverter, transformer and grid connection, and usually a lower LCOE.
What is performance ratio and what causes the losses?+
Performance ratio is delivered AC energy divided by the energy striking the plane of the array, and a well-built plant lands around 78–84%. It is the product of roughly eight sequential derates: soiling, reflection, spectral effects, temperature, mismatch, DC wiring, inverter conversion and transformer losses. Temperature is the largest single term in a hot climate, but removing it entirely would still leave about 12 points spread across the other seven — which is why a bankable yield model derates each stage explicitly rather than applying one blanket figure.
How much output does a solar module lose to heat?+
Crystalline silicon typically loses 0.3–0.4% of output per °C above the 25°C standard test condition, and cell temperature can sit 25–35°C above ambient on a still, high-irradiance afternoon. A 500 W module with a −0.35%/°C coefficient running at 65°C delivers 500 × (1 − 0.0035 × 40) = 430 W, a 14% shortfall with nothing faulty. This is why the hottest sites are not automatically the best sites: high irradiance and high cell temperature arrive together, and the second partly cancels the first.
What is capture price and why does it fall as solar penetration rises?+
Capture price is the volume-weighted average price a plant receives during its own generating hours, which can sit well below the market average — ₹4/kWh against a ₹6/kWh average is a 67% capture rate. It declines with penetration because of cannibalisation: every solar plant generates in the same few midday hours as every other solar plant, so their combined output depresses the price in exactly those hours. The physical output of the next megawatt is unchanged; its revenue is not. This is why storage, transmission and demand flexibility grow more valuable precisely as solar grows cheaper.
Why is the cost of capital so decisive for solar specifically?+
Solar is almost pure capital cost — no fuel, modest operating costs, and decades of output — so nearly its entire lifetime cost is exposed to the discount rate. Over 25 years the annuity factor is 12.783 at 6% and 7.843 at 12%, so the identical plant with identical yield shows roughly 63% higher levelised cost purely from the financing environment. A two-point efficiency advantage cannot outrun a 600 basis point cost-of-capital disadvantage, which is why access to cheap capital is often as competitively decisive as module technology.
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.