Why Energy Transition Forecasts Keep Getting the Demand Curve Wrong

Most energy-transition forecasting starts with a quiet assumption that never gets questioned: the shape of today’s market simply gets carried forward. From there, the job becomes finding a low-carbon stand-in for every barrel, kilowatt and tonne the current system delivers. That framing may be inflating the problem well beyond what it actually is.

Analyst Michael Barnard made the case at an Earth Set gathering hosted at Octopus Energy’s London headquarters, in conversation with Amy Rennison. Their talk ranged across shipping, hydrogen, grids, heat pumps, aviation, carbon removal and data centres — and audience questions kept dragging the discussion back to the unexamined assumptions propping up confident-looking forecasts.

Shipping shows how demand quietly shrinks

Maritime freight is the clearest illustration. Around 40% of shipping tonnage is coal, oil and gas, with raw iron ore adding roughly another 15%. As grids electrify, steelmaking shifts and short-sea and inland routes become candidates for batteries, the liquid fuel the sector ultimately needs falls for reasons that have nothing to do with marine fuels themselves.

Barnard’s own projection lands near 70 million tonnes of annual shipping fuel demand by 2100. Other models run into the several hundred million tonnes. That spread reshapes every downstream calculation — how much biomass, synthetic fuel, renewable power and port infrastructure get built. A fuel pathway that looks impossible under one demand curve can look workable under another.

The core argument: Forecasts spend enormous effort on the ingenuity of a proposed solution, and far too little on whether the size of the problem is changing underneath it.

Biological fuels hit the same wall. Cars, trucks and trains are electrifying, much of short-distance shipping can follow, and fossil freight volumes are declining. Aviation, meanwhile, remains the stronger claimant on scarce sustainable liquid fuels — which changes what is realistically left for ships. Barnard says looking at those markets side by side shifted his own numbers for biodiesel and renewable diesel in the maritime pathway.

Hydrogen: a market that shrank while institutions grew

The hydrogen story is a different flavour of the same mistake. The sweeping “hydrogen economy” narrative assumed green hydrogen would get cheap enough to win in transport, heating, storage and a long list of industrial uses. Working through production and delivery economics, Barnard found a far smaller addressable market once hydrogen stayed expensive.

But policies, subsidy schemes, corporate bodies and investment plans had already been built around the bigger expectation. Institutions do not unwind as fast as the assumptions that spawned them. Once money, headcount and political capital pile up around a proposition, reversing course becomes an organisational problem as much as an analytical one.

That partly explains why the least glamorous transition work may matter most:

• Transmission and distribution capacity has to expand, with utilities overhauling how they plan and build without compromising reliability.
• Community consent for wind, solar and battery projects depends on local benefits and genuine engagement, not national polling numbers.
• Deployment over invention — heat pumps and electric cars did not need reinventing, just easier paths for households to adopt them.

Octopus Energy was a fitting venue for that point, since much of its contribution comes from pushing existing hardware through software, customer relationships and new commercial models.

Where engineering cannot settle the argument

The audience pushed into terrain that no spreadsheet resolves. Ocean physicist Helen Czerski challenged Barnard’s framing of aviation demand. He estimates roughly 91% of flying is discretionary — a figure easily misheard as a prediction that almost all aviation vanishes. It is not.

Discretionary travel bends to price in ways commuting rarely can. People fly less often, pick closer destinations, swap some business trips for video calls, or simply pay more for journeys they value. The thornier question is political: showing how demand reacts to pricier sustainable aviation fuel says nothing about how willingly governments will adopt policies that make familiar travel costlier.

Carbon removal carries its own version of the demand trap. If a scenario overstates future fossil fuel use, the modelled removal requirement swells with it — and oversized removal targets can then be used to justify technologies whose real scale and cost remain unproven. Forest, wetland and grassland restoration stand on firmer ground than capturing dilute CO2 and piping it through vast new networks; those cases deserve separate scrutiny rather than a single “carbon capture” label.

Data centres repeat the pattern. Today’s computing loads get extended decades out, while gains in software, chips and system architecture get far less attention. Cost, hardware and power constraints create hard incentives to raise efficiency — demand can climb steeply while electricity per task falls.

The discipline Barnard proposes is unglamorous. Start with the service society needs, ask how the market delivering it will change, and make every assumption visible enough for someone else to break it. Only then estimate the fuel, infrastructure or capital a given solution requires.