Energy, the economy and the future: Growth, forecasts and the future

4.1 Real growth follows S-curves, not endless exponentials

Smil's book Growth starts with the mathematics of how things grow. Linear growth adds the same amount each period: a farm adding 100 tonnes of grain a year. Exponential growth adds the same percentage, so the amount added keeps rising: savings earning 7% a year. A handy rule gives the doubling time of exponential growth: divide 70 by the annual percentage rate. At 7% a year a quantity doubles in about ten years; at 2%, in about 35.

Exponential growth cannot last in any finite system. Sooner or later it runs into limits of resources, space, markets or energy. So most real growth is S-shaped. It starts slowly, speeds up, then slows and levels off near a ceiling. The logistic curve, first described by the Belgian mathematician Pierre-François Verhulst in the 1830s, is the classic form, and related curves describe similar paths. Smil finds the pattern everywhere: in the growth of organisms, the size of machines, the spread of new technologies, and the output of energy sources and industries.

The S-curve has two lessons for energy. The first is Smil's: if every energy conversion follows a logistic path, the shift from fossil fuels will not be exceptionally fast overall, because each new technology must pass through slow early and late phases. The second cuts the other way. In the steep middle of an S-curve, growth looks impossibly fast to anyone used to linear thinking, and forecasters who draw straight lines are repeatedly surprised. Solar power in the 2020s is in that steep phase.

The skill is to ask where on its curve a technology sits, and what sets its ceiling. For solar the ceiling is not the sun but the grid's ability to absorb midday power, storage, land and the share of energy use that can be electrified. For an Indian electricity market growing 6% or more a year, demand itself doubles in about 12 years, which raises the ceiling for every source.

• Smil, Growth: From Microorganisms to Megacities (2019) · via Smil Mastery Course, M15 'Growth trajectories: linear, exponential, logistic'
• Ember, Global Electricity Review 2026 (April 2026) · 2025 data, https://ember-energy.org/latest-insights/global-electricity-review-2026/ and https://ember-energy.org/chapter/2025-in-review/

4.2 Long-range energy forecasts have a poor record

Smil's Energy at the Crossroads has a whole chapter titled 'Against Forecasting'. He reviews decades of long-range energy projections and finds that most missed badly, whatever the authority of their authors. In the 1950s and 1960s leading scientists foresaw a world powered by cheap nuclear electricity, with fast breeder reactors taking over by the end of the century. A celebrated 1976 forecast of 'soft energy' for the United States missed its target for 2000 by around 90%. Oil prices were expected to keep rising after 1980 and instead collapsed. Even population projections, among the most reliable of all, shift: the UN's 1991 projection for 2025 was cut by about 600 million people in its 2004 revision.

The main reason forecasts fail is that they are unrealistically static. They extend current trends, but trends weaken, deepen or reverse, sometimes abruptly, when prices, technologies, politics or wars change. Single numbers make the problem worse: a forecast of so many gigawatts in 2050 hides huge uncertainty behind false precision.

The errors are not all in one direction. Through the 2000s and 2010s, many official projections, including the IEA's, badly underestimated the growth of solar power, because they assumed cost declines would slow and policies would stay modest. The IEA itself now shows how much its baseline has moved. Its 2025 Current Policies Scenario, which assumes no new policies at all, still has cumulative solar and wind generation to 2040 almost four times what its 2015 current policies case projected, and coal supply about 23% lower. The world changed faster in solar than the forecasters assumed, and more slowly in oil than many advocates hoped.

This record does not mean projections are useless. It means they should be read as conditional stories rather than predictions, and judged by how well they expose the assumptions that matter.

• Smil, Energy at the Crossroads: Global Perspectives and Uncertainties (2003) · via Smil Mastery Course, M16 'Against forecasting'
• Smil, Energy Myths and Realities: Bringing Science to the Energy Policy Debate (2010) · via Smil Mastery Course, M16 'Against forecasting'
• Smil, Global Catastrophes and Trends: The Next Fifty Years (2008) · via Smil Mastery Course, M16 'Against forecasting'
• ACCR, Investor bulletin: IEA World Energy Outlook 2025 (13 November 2025) · comparison of WEO 2025 scenarios with the 2015 and 2019 editions, storage capacity by scenario, https://www.accr.org.au/downloads/13112025-investor-bulletin-weo-2025.pdf

4.3 Use scenarios to map constraints and choices, not to predict

Smil's own practice is to avoid point forecasts. Instead he sets out the constraints that bound what is realistically possible: the physics of energy conversion, the scale of the existing system, the lifetimes of its assets, the materials required and the costs involved. Within those bounds he is willing to argue for normative goals, what we ought to aim for, while refusing to pretend he knows the date and size of the outcome.

For an investor this suggests a practical method. First, use several scenarios rather than one, and look for investments that make sense in most of them. Grid expansion, energy efficiency, storage, efficient cooling and clean industrial heat look sound in almost every IEA case. In its 2025 Outlook, battery storage grows from about 77 GW today to about 1,400 GW by 2035 even in the Current Policies case, about 1,700 GW in the Stated Policies case and about 2,900 GW in the net-zero case. An eighteen-fold rise in the most conservative case is a strong signal.

Second, identify the signposts that tell you which path the world is on. For India these include the share of electric two-wheelers and cars in sales, the start and price of the carbon market, the pace of grid and storage tenders, coal plant running hours and the growth rate of electricity demand.

Third, translate every claim into physical quantities and check them against history. How many gigawatts a year, how many tonnes, how many kilometres of line? If an investment depends on a build rate many times anything seen before, it needs a very good reason.

Fourth, respect discontinuities. Wars, pandemics, price shocks and sudden policy shifts have changed energy paths more than steady trends have. They cannot be predicted, but portfolios can be built to survive them. That combination of humility about prediction and rigour about constraints is the most useful thing Smil's work offers anyone who has to commit capital to the energy future.

• Smil, Energy at the Crossroads: Global Perspectives and Uncertainties (2003) · via Smil Mastery Course, M16 'Against forecasting'
• Smil, Global Catastrophes and Trends: The Next Fifty Years (2008) · via Smil Mastery Course, M16 'Fatal discontinuities'
• ACCR, Investor bulletin: IEA World Energy Outlook 2025 (13 November 2025) · comparison of WEO 2025 scenarios with the 2015 and 2019 editions, storage capacity by scenario, https://www.accr.org.au/downloads/13112025-investor-bulletin-weo-2025.pdf
• IEA, World Energy Outlook 2025 (2025) · Overview and key findings, https://www.iea.org/reports/world-energy-outlook-2025/overview-and-key-findings

Check your understanding

Using the rule of 70, roughly how long does a quantity growing at 7% a year take to double?

What does Smil see as the main reason long-range energy forecasts fail?

In the IEA's 2025 Outlook, battery storage capacity by 2035 ranges from about 1,400 GW to 2,900 GW across scenarios, from about 77 GW today. What is the most sensible reading for an investor?