The $0.043 Floor: What Solar and Storage Costs Actually Make Possible

597 GW in One Year, and the Price Still Fell

The world installed 597 GW of solar photovoltaic capacity in 2024, a 33% increase over 2023, bringing the cumulative global total past 2.2 TW.¹ China alone added 329 GW — 55% of the global figure. India added 25.2 GW, a 204% year-on-year rise, its highest annual total ever.² The global weighted-average levelised cost of electricity (LCOE) for utility-scale solar stood at $0.043/kWh in 2024, per IRENA's latest cost report.³ That figure is 56% below combined-cycle gas ($0.077/kWh) and 64% below coal ($0.119/kWh). In China, the LCOE reached $0.033/kWh; in India, $0.038/kWh.³

These are not modelled projections. They are audited weighted averages from projects commissioned in 2024. The underlying module prices bottomed at $0.07–$0.09 per watt in late 2024 and early 2025, a level at which the four largest Chinese manufacturers collectively lost $1.54 billion in the first half of 2025.⁴ Beijing intervened in September 2025, mandating capacity cuts and eliminating the 13% VAT export rebate on modules and storage systems.⁵ Module prices rose roughly 9% in Q4 2025.⁵ Even after this correction, solar electricity remains the cheapest bulk generation source in recorded history.

Storage Costs Fell 93% in Fourteen Years, and the Gap to Useful Remains

Utility-scale battery storage cost $192/kWh (installed) in 2024, a 93% decline since 2010, according to IRENA.³ BloombergNEF's separate survey of lithium-ion pack prices reported $108/kWh globally in 2025, an 8% fall from 2024, with stationary-storage packs specifically at $70/kWh.⁶ Full turnkey battery energy storage system (BESS) costs averaged $117/kWh globally in 2025 — $73/kWh in China, $219/kWh in the United States.⁷

The deployment curve is correspondingly steep. Global battery storage additions reached 112 GW (307 GWh) in 2025, a 48% increase over 2024.⁸ The ratio of solar capacity installed to battery capacity installed narrowed from 56:1 in 2016 to 6:1 in 2025, and BloombergNEF expects 4:1 by 2026.⁸ The system is beginning to co-locate generation and storage as a default rather than an afterthought.

Yet a critical distinction separates "cheap electricity for some hours" from "cheap electricity always." The levelised cost of storage (LCOS) — the all-in cost of storing a unit of energy and dispatching it — stood at $65/MWh in the Ember analysis of late 2025.⁷ Added to solar's $43/MWh, a solar-plus-storage system delivers electricity at roughly $108/MWh, or $0.108/kWh. That is competitive with gas peaking plants. It is not yet competitive with baseload gas or coal in markets where those fuels are cheap and carbon is unpriced. The abundance thesis requires that second condition — always-available cheap power — to open new product categories. The evidence says "approaching, not arrived."

The Categories That Cheap Energy Supposedly Creates

The energy-abundance argument, articulated most forcefully by Casey Handmer of Terraform Industries, holds that once solar-plus-storage electricity falls below roughly $20/MWh on a sustained basis, a set of energy-intensive processes become economically viable for the first time.⁹ The most frequently cited candidates are green hydrogen, synthetic fuels, direct air capture of CO₂, and desalination.

Each of these is real technology. Each faces binding constraints that have little to do with the price of electricity.

Green hydrogen requires approximately 55 kWh per kilogram of output.¹⁰ At $20/MWh electricity, the energy input alone costs $1.10/kg — attractively close to grey hydrogen at $1.50–$2.50/kg. But the installed cost of electrolysers remains $1,500–$2,200/kW for PEM systems, of which 55–75% is balance of plant, not the stack.¹⁰ Stack degradation adds $0.20–$0.40/kg over a three-year operating cycle.¹⁰ Compression to 700 bar for transport adds $1.50–$2.50/kg.¹⁰ Of the 520 GW of green hydrogen projects announced globally, only 4–7% have reached final investment decision; over 33 GW have been cancelled or deferred, including BP's 26 GW AREH project in Australia.¹⁰ The EU Hydrogen Bank's first two auction rounds saw seven projects representing 1.88 GW of the 2.33 GW winning capacity subsequently withdraw.¹⁰ Cheap electricity is necessary for green hydrogen. It is not sufficient.

Direct air capture requires 1.5–2.5 MWh of electricity and approximately 6 GJ of thermal energy per tonne of CO₂ removed.¹¹ First-of-a-kind plant costs run $400–$600/tCO₂, with transport and storage adding $50–$150.¹¹ Climeworks's Mammoth facility in Iceland, the largest operational DAC plant, started in 2024.¹² The $100/tonne figure frequently cited as a target is, per ESI's 2026 analysis, "structurally improbable with current chemistries."¹¹ Frontier's 2026 offtake contract cleared at $315/tonne, the lowest credible commercial price published to date.¹¹

Solar desalination is the closest to commercial viability. Reverse osmosis consumes 2.7–3 kWh per cubic metre with energy recovery, and energy accounts for 60–70% of total production cost.¹³ Solar-powered desalination systems in 2026 produce water at €1–€3 per cubic metre over a 15-year life, compared to €3–€6 for diesel-powered equivalents — a 50–70% cost reduction.¹³ This is already happening. Elemental Water Makers, a Dutch firm, deploys solar-RO units across the Caribbean, East Africa, and South-East Asia. The constraint here is not cost but scale: the world's large desalination plants (Ras Al Khair at 1.025 million m³/day, Sorek B at 627,000 m³/day) run on baseload grid power, and converting them to intermittent solar requires either oversized arrays or battery buffers that raise capital costs.

Synthetic fuels sit at the intersection of all three. Terraform Industries demonstrated a combined electrolyser–DAC–Sabatier reactor producing pipeline-grade methane (>97% CH₄) from solar electricity and atmospheric CO₂ in March 2024.¹⁴ The company's current production cost is approximately $5/mmBTU, roughly 2.5 times the U.S. Henry Hub natural gas benchmark.¹⁴ At scale, with solar at $20/MWh, Handmer projects hydrogen under $1/kg and methane under $0.75/kg.⁹ Those projections require electrolyser capital costs below $100/kW — roughly one-fifteenth of current installed PEM costs.

565 TWh of Data-Centre Demand Is the Nearer Opportunity

While the abundance thesis focuses on chemistry, the most immediate large-scale buyer of cheap solar electricity is computation. Gartner estimates global data-centre electricity consumption at 565 TWh in 2026, a 26% increase over 2025's 447 TWh.¹⁵ The IEA projects this figure exceeding 945 TWh by 2030.¹⁶ AI-optimised servers already account for 31% of data-centre power demand in 2026, and Gartner expects AI server consumption to surpass conventional servers in 2027.¹⁵

The tech sector signed roughly 40% of all corporate renewable power purchase agreements in 2025.¹⁶ This is a buyer that values the marginal cost of electricity ($0.04/kWh solar) but can also pay for the reliability premium ($0.11/kWh solar-plus-storage). A hyperscale data centre at $50,000/kW of compute capital expenditure demands 98% uptime; it will pay for batteries.⁹ The category "cheap solar enables" most immediately is not a chemical product but a computational one — and it is already here, at scale, buying.

What Would Falsify the Thesis

The energy-abundance argument rests on the continuation of two learning curves: solar modules at roughly 24% cost reduction per doubling of cumulative capacity (the observed rate since the 1970s),¹⁷ and lithium-ion batteries at roughly 18–20% per doubling.⁸ Three developments could break this trajectory.

First, critical-mineral bottlenecks. PEM electrolysers require 300–500 kg of iridium per GW of capacity; global annual iridium production is 7–8 tonnes.¹⁰ Reaching 51–109 GW of PEM capacity by 2030 would require 2–6 years of the entire global supply — a binding physical constraint regardless of learning-curve economics.

Second, policy reversal. The Chinese capacity interventions of September 2025 demonstrate that the learning curve is not autonomous; it depends on manufacturing investment subsidised by governments willing to tolerate below-cost pricing. The 13% VAT rebate cancellation alone shifted module prices 9% in a single quarter.⁵

Third, grid-integration costs. The LCOE measures generation cost at the busbar. It excludes transmission upgrades, curtailment losses, and the system-balancing costs that rise as variable renewable penetration increases. In grids above 40–50% solar share, these costs have historically offset a substantial fraction of the generation savings, though quantifying this precisely remains contested.

One Sentence for the Product Builder

The cost curves are real, the chemistry is real, and the demand is real — but the product categories that cheap energy creates are, for now, those whose capital equipment costs less than $1,000/kW and whose processes tolerate intermittency, which means desalination and data-centre power purchase agreements rather than the hydrogen and synthetic-fuel visions that dominate the pitch decks.