Solar photovoltaic cells convert sunlight directly into electricity through the photoelectric effect (Einstein, 1905). The first practical cell was made by Bell Labs in 1954 — 6% efficiency, ~$300 per watt in 1954 dollars (~$3,500/W in 2024 dollars). Cells were space technology in the 1960s. By 2024, mass-produced silicon modules are ~22% efficient, cost ~$0.10/W, and produce electricity at ~$30/MWh in optimal sites — cheaper than any source of new electricity in human history. The cost decline of ~99% over fifty years is one of the most extraordinary technology learning curves on record. Wind has followed a similar trajectory; offshore wind is now economical without subsidy in many sites.
What is happening with solar PV is best understood as a manufacturing story rather than a physics one. The underlying photovoltaic effect has been industrially exploited since Bell Labs's 1954 cell, and crystalline silicon — the dominant chemistry then and now — is approaching its theoretical efficiency ceiling. The action is in cost. With every doubling of cumulative production, module prices have fallen by about a fifth, the same learning curve that historically governs semiconductors, and the curve has now run long enough that modules themselves are nearly free at the wholesale level — most system cost lies in mounting, inverters, installation, and transmission. Wind has followed a similar though shallower trajectory, with onshore turbines mature and offshore the major growth area. Both are variable: the sun sets, the wind drops, and a panel produces full output only when the sun is high and unobstructed. The cheapest electricity in human history is also the most schedule-dependent.
Around that central manufacturing-curve story sit distinct bets on what comes next. Perovskite-tandem cells layered on silicon are pushing lab efficiencies past the silicon ceiling and could reset the cost curve again if they prove durable in the field. Hydropower still supplies about a sixth of global electricity but has saturated in developed countries, with most economical sites already developed. Enhanced geothermal — drilling deep enough to access dry hot rock anywhere rather than only in volcanic regions — is at the start of what may be its own Solar-style curve. Nuclear fission is the structural exception: low-carbon, firm, almost incomparably energy-dense, and yet it has not followed a manufacturing learning curve in the West — recent US builds came in at roughly twice their estimates, while South Korea, China, and India still build similar reactors at a fraction of the cost. Whether small modular reactors finally drag fission onto a learning curve, and whether private-sector fusion (Commonwealth Fusion Systems, Helion) reaches commercial demonstration this decade, are the open bets.
By 2024 global solar capacity sits around 1,800 GW with another 600 GW being added each year and accelerating; wind capacity is near 1,100 GW; nuclear has been roughly flat for two decades. Levelized costs in good sites place utility-scale solar and onshore wind cheaper than almost any alternative, while new nuclear in the West remains several times more expensive per megawatt-hour than the same reactors built in China or Korea. The technical-physics curves for clean generation are now broadly favorable; the bottleneck has shifted to absorbing variable output into a reliable grid, treated separately as Grid Integration & Energy Storage.