← Back to the map

Net Zero Cost Benefit

What does net zero cost?

Scroll down ↓

Going net zero means buying things like solar panels, heat pumps and electric cars.

They are often expensive at first, but they can save money over time.

But net zero also makes energy more expensive.

And huge green investments crowd out other investments, so economies grow more slowly.

These effects pull in different directions, so economy-wide models estimate the total.

The UN Climate Panel collected 540 scenarios from such models.

Cutting a little costs little.

Cutting a lot costs much more.

And almost no scenario reaches a 100% cut by 2050 – net zero.

The UN Climate Panel famously estimated in 2022 that keeping warming to around 1.5°C would cost 2.6–4.2% of global GDP in 2050.

But only two of its scenarios actually reached net zero by 2050.

And all of them assumed deep cuts by now that never happened.

Other studies add 421 more scenarios, for 961 in total.

Different models give different answers, so let's split by modeling team.

Adjusting the scenarios that started cutting too early, and extending each team's curve to 100%, gives costs from about 4% to 13% of GDP in 2050.

Six scenarios actually reach or nearly reach net zero by 2050.

Three assume deep cuts that should already have happened by now, so they are outdated.

That leaves three.

With these, the cost of net zero ranges from 3% to 19% of GDP in 2050. Across the six modeling teams, the median is 9.8% and the average 9.4%.

To estimate cost, this site uses the middle of the three net-zero scenarios, AIM, at 7.8% of GDP. That is measured from AIM’s own 2020 policies, where emissions stay high. Current policies already cut emissions somewhat, so from there net zero costs 7.2%.

(If anything, this is a low-ball estimate, as explained in the notes.)

Notes
  1. UN Climate Panel range. IPCC AR6 Working Group III, Summary for Policymakers, C.12.2 (IPCC 2022): scenarios that limit warming to 1.5°C (>50%) with no or limited overshoot, GDP in 2050 against continued current policies. In these scenarios warming typically peaks around 1.6°C before falling below 1.5°C by 2100. Replicated from the AR6 Scenarios Database (Byers et al. 2022): 57 runs, interquartile range 2.6–4.1%; median cut in 2050, 86%; two of the 57 reach net zero. In all 57, 2025 emissions are 17–46% below actual 2025 emissions as recorded by the European Commission (JRC 2026); counted the way the scenarios count, the gap is even larger.
  2. What the cost measures. The cost is how much smaller the economy is in a given year than under current policies: 7.2% in 2050 means world GDP is 7.2% lower that year, not that growth is 7.2 points slower. Until about 2045, net zero makes economies grow more slowly; after that they grow slightly faster, but remain smaller in absolute terms through 2100. That is the "lost growth" in the tagline: economies still grow, just less than they otherwise would.
  3. What about the years other than 2050? The charts compare all scenarios in one year, 2050, the target year for net zero, because that is where the cost of reaching it shows. But the models estimate much more than that: they report results for many years, usually every ten or every five years, up to 2100. AIM reports every five years (2025, 2030, 2035 and so on), and this site uses AIM's whole path to calculate the cost across the century. Measured from current policies (next note), the cost of world net zero is about 0.2% of world GDP in 2030, rises to 5.6% in 2040 and peaks at 8.3% in 2045. It is 7.2% in 2050, then eases to about 5% in 2060–70 and 4.3% by 2100. Costs for the years in between are filled in on a straight line between AIM's five-year values.
  4. Current policies and cost. AIM measures its 7.8% against its own scenario with the policies in place in 2020 (here called "2020 policies"), in which emissions keep rising (about 20% by 2050). AIM has no scenario similar to CMIP7 Medium (Van Vuuren et al. 2026; see Temperature). But it has one with generally more emissions, "2020 policies", and one with fewer, called "pledges continued" (national pledges kept after 2030). So for each of AIM's ten world regions, the site interpolates between the two with one fixed proportion for the whole century, chosen so that the interpolated scenario emits as much over 2025–2100 as CMIP7 Medium, or as close as the two allow (for example, 95% "pledges continued" and 5% "2020 policies" for Europe; half and half for the world; in South Asia and Latin America CMIP7 emits more than both AIM scenarios, and in the Middle East less, so there the closer one is used in full). Every year, both emissions and GDP are taken in that same proportion, so the cost of going from the interpolated scenario to net zero (the GDP lost) is measured from the same starting point as its effect on temperature. On this site, "current policies" means this interpolated scenario. Measured from it, world net zero costs 7.2% of GDP in 2050. The interpolated scenario starts from CMIP7's 2025 greenhouse gas emissions, which are fitted to historical data and somewhat higher than AIM expected (world 59 instead of 56 billion tons, including land use). Net zero must also cut these extra emissions, about 29 billion tons in all over 2027–2050, falling from about 2.5 billion tons a year now to none by 2050. The site includes these cuts in its temperatures, so they add to the benefits of net zero, but it counts them as free in the costs. If anything, that makes net zero look a little better. (It changes the result very little: if cutting the 29 billion tons cost $300 a ton, it would add about $6 trillion in today's money, under 3% of the $255 trillion that world net zero costs.)
  5. The net-zero path. The cost is AIM's net-zero scenario, in which countries meet their 2030 pledges and then cut faster to reach about zero by 2050. This site's temperature effect assumes a straight line from today to zero in 2050, which is simpler and cuts a little sooner. The site also ran a path like AIM's through the climate model: pledges to 2030, then a straight line to zero in 2050. It changes little: the world avoids 1.33–1.34°C of warming in 2100 instead of 1.38°C, and gets 9¢ back per dollar either way. For the US, the EU and Germany the difference is tiny; China, whose pledge lets emissions grow to 2030, avoids 0.12°C instead of 0.135°C. So for the world as a whole, if anything, the straight line is slightly generous to net zero. Both versions are in the download.
  6. In today's money. All amounts on this site are in 2026 prices, so adjusted for inflation. For instance, the cost or benefit in a single year, such as 2040, for a country or group is in 2026 prices. Totals over a period, such as 2027–2100, are also discounted to today. Discounting is needed because costs and benefits come at different times: more of the cost falls early in the century, and more of the benefit late. To compare them, every cost-benefit analysis turns all future costs and benefits into today's money with a discount rate. A dollar today is worth more than a dollar in 2080, not least because it can be invested and grow in the meantime, and because people in the future will be richer. This site uses the discount rate from William Nordhaus, the only climate economist to win the Nobel Prize in economics, as updated with Lint Barrage, in a linear version of their DICE-2023 rates: about 4.5% a year now, falling linearly to 3.4% by 2100, which slightly favors net zero (Barrage and Nordhaus 2024). So, added up over 2027–2100 and discounted to today, world net zero costs $255 trillion.
  7. Scenarios. 540 scenarios with GDP costs from the IPCC AR6 database (Byers et al. 2022). 421 from other studies: 81 from the AIM scenario archive (Fujimori 2021), 59 from the ENGAGE database, where the AR6 version lacked a matching reference scenario (Riahi et al. 2021; IIASA n.d.a), 17 from later literature, and 264 from later IIASA scenario databases (ENGAGE, POSTPARIS, SENSES, COMMIT, NGFS, CD-LINKS) (IIASA n.d.b), duplicates removed. Each dot is one scenario's GDP loss in 2050 against its own reference scenario. Left out: 11 "no policy" reference runs and one frozen-methane variant, which emit more than current policies.
  8. Team curves. Middle-of-the-road (SSP2) scenarios only: the same assumptions about population and economic growth as everything else on this site, and those of almost all the scenarios. Scenarios set in other futures, such as a greener world with less energy demand (SSP1) or a fossil-fueled boom (SSP5), cost less or more for reasons that have nothing to do with net zero. This leaves out 80 of 972 scenarios, 67 of them from one MESSAGE project (Rogelj et al. 2019). Scenarios that cut too early (2025 or 2030 emissions more than 10% below current policies) are scaled up by each model's cost of delay, which this site calculated from matched ENGAGE scenario pairs (Riahi et al. 2021). For each scenario design with enough deep runs, power and exponential curves are fitted to runs above a 30% cut; a team's value is the midpoint of the two, median across its designs: GEM-E3 3.8%, AIM 5.5%, MESSAGE 11.0%, WITCH 11.3%, REMIND 12.7%.
  9. COFFEE (COPPE/UFRJ, Brazil), one of the larger contributors of scenarios, is left out of the team comparison. It is an energy and land-use model, not an economic model: GDP is set from outside the model, so its GDP losses are not driven by an economic model (Harmsen et al. 2021; IAMC n.d.). Its runs appear among the dots.
  10. Net-zero scenarios. Kept: REMIND ADVANCE_2030_1.5C-2100, 3.0% (Luderer et al. 2018); AIM/CGE EN_INDCi2030_500f (ENGAGE), 7.8% (Riahi et al. 2021); 8.6% after this site's adjustment for its pledge-based start; MIT EPPA global net zero, 18.8% (Morris et al. 2023). Morris et al. report consumption only; the GDP figure is from a chart provided by J. Morris, MIT Joint Program, personal communication, Oct 26, 2023. No longer describing a path from today: REMIND SMP_1p5C_Sust (Bertram et al. 2018), REMIND PEP_1p5C_red_eff (Kriegler et al. 2018) and AIM CD-LINKS_NPi2020_400 (McCollum et al. 2018); in this site's comparison, their 2025 emissions are 48–67% of current policies.
  11. Why AIM's bar (8.6%) is above its 7.8% dot. AIM's net-zero scenario was made around 2020 and assumes the world would start cutting emissions right away, to 55 billion tons in 2025. The world didn't: counted the same way, CMIP7 estimates 2025 emissions at about 59 billion tons. Starting later means cutting faster later, which costs more. AIM itself ran pairs of scenarios that start cutting early or late (Riahi et al. 2021): at the deepest cuts they compare, about 77% by 2050, the late start costs about 10% more in 2050, and that extra cost was still rising. Adjusted by 10%, as for every other scenario that starts cutting too early, AIM's net-zero cost rises to 8.6%. The chart shows the published 7.8%. The site's own calculations start from current policies, which are slightly different, so there net zero costs 7.2% (see Current policies and cost).
  12. Why 7.8% (7.2% from current policies) is likely a low-ball estimate. (a) It is below both the median (9.8%) and the average (9.4%) of the modeling teams. (b) Even AIM's scenario assumes the world started cutting emissions right away, which it did not; adjusted for the later start with AIM's own runs (Riahi et al. 2021), it costs 8.6%, and probably more, since the extra cost of a late start grows with deeper cuts. (c) It measures lost GDP; the loss in what households can consume is larger, 10.8% in 2050 in AIM's own results (ENGAGE database, Riahi et al. 2021). (d) Nine in ten of these scenarios, including AIM's, assume the cheapest possible policy: one carbon price everywhere, from now or after the 2030 pledges. Real climate policies can cost much more: the Energy Modeling Forum's EU study found the EU's 2020 climate targets would cost an average of 1.03% of welfare with the cheapest policy and 2.19% as designed, two to four times the European Commission's own estimate of 0.5% (Böhringer, Rutherford, and Tol 2009). Looking at the very different carbon prices countries actually used in 2019, Tol (2023) estimates that this alone made climate policy 2.6 times as expensive as it needed to be.
  13. Breakthroughs. The models already assume clean technology keeps getting cheaper, as experts now expect, and the costs on this site are based on that. If the world manages big breakthroughs beyond that, green energy could become much cheaper, maybe even cheaper than fossil fuels. That would cut the cost dramatically, and everyone would switch.
  14. Countries. AIM reports these scenarios for ten world regions only, although the model itself has 17 (it separates, for example, the USA and Canada, the EU and Turkey, Japan and Australia/New Zealand). No public source has the 17-region results of these scenarios, but six other public AIM studies report all 17 regions for a deep cut against a reference, and they agree closely on how a big region's cost splits inside it (for example, the medians across the studies in 2050, as a share of the big region's average loss: USA 0.97, Canada 1.23, EU 0.88, Turkey 2.45, Japan 0.85, Australia and New Zealand 1.31; data: Fujimori 2021 and the AIM datasets at zenodo.org/records/7471205, 22802585, 17903582, 8313380, and doi.org/10.1038/s41467-025-67276-5). Each country gets its own AIM region's loss (the big region's loss times the median ratio across the studies for each decade, for example the USA from 0.97 in 2030 to 0.92 in 2100 and Canada from 1.25 to 1.5), as a share of its GDP (SSP2 GDP at market exchange rates from Moyer et al. 2026 for 188 countries; regional PPP shares for 21 small states and territories), and all countries are scaled together to AIM's world cost. So within each of AIM's 17 regions every country loses the same share of its GDP, and a group's cost is the sum of its countries. 32 very small places have no GDP projections and so no cost estimate of their own, for example Monaco, Andorra, Liechtenstein, Greenland, Bermuda and the Cayman Islands (the full list is in the download). Together they are less than 0.1% of the world economy. They are not left out of the World total, because all countries are scaled together to AIM's world cost. Each country's cost is its share of a world where everyone goes net zero, so the site assumes a country going net zero on its own pays the same as when everyone acts. In AIM's scenario every region cuts 90–134% by 2050, so a country's share is close to the cost of its own net zero. National studies of the EU and the US going it alone find costs of 1–12% of GDP in 2050, though they measure cost in different ways: lost GDP, lost consumption or spending on emission cuts (Boitier et al. 2023; European Commission 2018; Bistline et al. 2024; Bistline et al. 2025; Peng et al. 2021); the shares used here, 1.6% for the EU and 2.6% for the US, sit inside that range, toward its low end. For Japan, an economy-wide model in the Energy Modeling Forum's Japan study finds an 80% cut costs about 3% of GDP in 2050, measured as lost consumption rather than lost GDP (Sugiyama et al. 2021), against a GDP loss of 1.4% used here for full net zero.
References