Amazon’s carbon footprint spans tens of millions of metric tons of CO₂ equivalent each year, driven by a sprawling network of fulfillment centers, delivery vehicles, data centers, and a third-party marketplace that ships billions of packages annually. Understanding where those emissions actually come from requires looking beyond the obvious image of brown boxes on doorsteps. The bulk of the company’s climate impact sits in categories that most customers never think about, and the way emissions are counted matters almost as much as the emissions themselves.
How Amazon Counts Its Carbon
Like most large corporations reporting climate data, Amazon organizes its emissions into three “scopes.” Scope 1 covers direct emissions from sources the company owns or controls, such as diesel burned in its branded delivery vans and natural gas used to heat fulfillment centers. Scope 2 covers indirect emissions from purchased electricity, mainly the power running its massive data centers and warehouse operations. Scope 3 is the catch-all for everything else in the value chain: the factories that manufacture products sold on the platform, the emissions generated when customers drive to pick up packages, the lifecycle of packaging materials, and the carbon cost of products after they leave the customer’s door.
Scope 3 is by far the largest category for any retailer or e-commerce platform, and Amazon is no exception. For most companies in the retail space, Scope 3 can represent well over three-quarters of total reported emissions. That ratio matters because Scope 3 is also the hardest to measure accurately. Companies rely on estimates, supplier-reported data, and industry averages rather than direct measurement. A growing wave of mandatory corporate climate disclosure rules is pushing Scope 3 from a voluntary best-practice exercise into something companies will increasingly be required to report, which raises real questions about what the numbers are actually telling us and how comparable they are across firms.1Frontiers in Sustainable Energy Policy. Scope 3: what question are we trying to answer?
Transportation and Delivery
Transportation is one of the most visible slices of Amazon’s emissions. The company operates one of the largest delivery networks on the planet, encompassing long-haul trucking between distribution hubs, air cargo flights, and the last-mile vans and drivers that bring packages to your door. Each of those legs carries a different emissions profile.
Middle-mile freight, the long-distance trucking and air transport that moves goods between Amazon’s fulfillment and sortation centers, relies heavily on diesel. Diesel trucks remain the most cost-competitive option for heavy-duty freight in the United States, largely because the refueling infrastructure is so well established and fuel-cost volatility is lower compared to newer alternatives.2Environmental Research: Infrastructure and Sustainability. A multi-criteria assessment of decarbonization pathways for heavy-duty trucks Battery-electric trucks show promise in certain regional settings, particularly in Europe where electricity grids tend to be cleaner and policy incentives are stronger, but hydrogen fuel-cell trucks generally rank as the least competitive option across regions right now. For a company operating tens of thousands of tractor-trailers across North America, that means the middle-mile segment is likely to stay carbon-intensive for years, even as pilot programs with electric and alternative-fuel trucks expand.
Air cargo is an even more carbon-intensive part of the middle mile. Amazon has built its own air freight network, Amazon Air, with dozens of aircraft operating out of dedicated hubs. Jet fuel produces roughly three to four times more CO₂ per ton-mile than trucking, and viable zero-emission alternatives for aviation remain far off. Sustainable aviation fuel exists but is produced in small quantities and at a premium price, making fleet-wide adoption slow.
Last-mile delivery, the final leg from a local delivery station to your home, is the segment Amazon has made the most visible progress on. The company has ordered tens of thousands of custom electric delivery vans and has begun deploying them in cities across the United States and Europe. Electrifying last-mile fleets is one of the most impactful steps any delivery company can take, because those vehicles make frequent stops in dense areas and are well-suited to battery power. Fleet managers evaluating the switch must weigh several powertrain options and renewal timelines, but numerical modeling shows that well-planned transitions to electric vans can deliver meaningful greenhouse gas reductions while also cutting costs over time.3Preprints. Electrification of Last-Mile Delivery: A Fleet Management Approach with a Sustainability Perspective The challenge is that last-mile delivery, while visible, represents only one slice of the total transportation pie. Even a fully electric last-mile fleet would leave the carbon-heavy middle-mile and air cargo segments largely untouched.
Data Centers and Cloud Computing
Amazon Web Services is the world’s largest cloud computing platform, and its data centers are enormous consumers of electricity. Servers generate heat, and cooling those servers requires even more energy. The electricity powering these facilities falls under Scope 2 emissions when Amazon buys it from the grid, or reduces toward zero on paper when it comes from renewable sources the company has contracted.
Amazon has invested heavily in wind and solar projects and claims to be the world’s largest corporate purchaser of renewable energy. The company has stated a goal of powering its operations with 100 percent renewable energy. However, renewable energy procurement for data centers is a complicated accounting exercise. Much of it works through power purchase agreements and renewable energy certificates, which means Amazon funds the construction of wind and solar farms that feed clean electricity into the grid somewhere, but the data center itself may still draw power from a regional grid that runs partly on fossil fuels. Whether that arrangement truly offsets the emissions at the point of consumption is a matter of ongoing debate among climate analysts.
The rapid growth of artificial intelligence workloads is adding another layer of concern. Training and running large AI models is far more energy-intensive per query than traditional cloud computing tasks. As demand for AI services surges, the electricity consumption of data centers is climbing faster than efficiency improvements and renewable procurement can keep pace in some regions. This dynamic could make AWS’s energy footprint a growing share of Amazon’s total emissions profile in the years ahead, even as other segments improve.
The Upstream Supply Chain
The single largest share of Amazon’s carbon footprint almost certainly comes from its upstream supply chain: the factories, raw material extraction, and manufacturing processes that create the hundreds of millions of products sold on the platform. Most of those products are made by third-party sellers, which means Amazon does not directly control the factories or their energy sources. This is classic Scope 3 territory, and it is where the numbers get fuzziest.
Amazon requires its suppliers to comply with environmental principles that include reducing energy use, greenhouse gas emissions, waste generation, and pollution. Suppliers are expected to track, record, and report their greenhouse gas emissions back to Amazon, and they are encouraged to integrate environmentally sustainable practices throughout their own operations and supply chains.4Advances in Economics Management and Political Sciences. Sustainable Supply Chain Strategies in E-Commerce: Case Studies of Amazon and Cainiao On paper, this is a robust framework. In practice, enforcing emissions reductions across millions of third-party sellers and their own tiers of sub-suppliers is an enormous challenge. Many smaller manufacturers in regions with coal-heavy electricity grids have limited ability or incentive to decarbonize their operations, and Amazon’s leverage over them varies widely depending on how much of their output flows through the platform.
The sheer diversity of products complicates any aggregate number. A lightweight phone case shipped from Shenzhen has a radically different manufacturing carbon footprint than a cast-iron skillet made in Tennessee. Averaging across millions of product categories produces a headline number that is more of a statistical exercise than a precise measurement. When Amazon reports improvements in its upstream supply chain emissions, it is worth asking whether those improvements reflect genuine decarbonization by suppliers or simply changes in the product mix, shifts in estimation methodology, or growth in lower-carbon product categories outpacing higher-carbon ones.
Packaging
Amazon ships billions of packages a year, and the cardboard, plastic mailers, air pillows, and other materials that protect those items carry their own carbon cost. The emissions from packaging come from multiple stages: manufacturing the material, transporting it to fulfillment centers, and then dealing with it after the customer opens the box. Amazon has made high-profile efforts to reduce packaging, including its “Ships in Own Container” program for products that do not need an additional Amazon box and a shift toward lighter-weight paper mailers for certain items.
Packaging is a real but relatively modest contributor to Amazon’s total footprint compared to transportation and manufacturing. The carbon intensity of packaging depends heavily on the material. Corrugated cardboard, Amazon’s most common packaging, has a lower manufacturing carbon footprint per unit than rigid plastics but a higher one than thin film mailers. The tradeoff is that lighter packaging can mean more product damage and more returns, which creates its own emissions loop.
The Hidden Cost of Returns
Product returns are a growing and underappreciated source of emissions for e-commerce companies. When a customer sends something back, the item travels in reverse through the logistics network, gets inspected, repackaged, or disposed of, and sometimes makes the journey again to a new buyer. Each of those steps burns fuel and generates waste.
Research quantifying the carbon footprint of apparel returns in the United States found that transportation dominates the emissions profile of reverse logistics, accounting for over 90 percent of the total carbon footprint in some cases. For a retailer with a centralized return system, items traveling over 1,000 miles generated roughly 29,000 metric tons of CO₂ in a single year, representing about 91 percent of that retailer’s total return-related transport emissions. By contrast, a retailer using a decentralized network, where returns are processed at regional facilities closer to the customer, saw long-distance returns contribute a lower but still substantial share. Packaging for returns added another 1,100 to 1,750 metric tons annually per retailer on top of that.5Sustainable Futures. Hidden footprints in reverse logistics: The environmental impact of apparel returns and carbon emission assessment
Amazon’s return volume is staggering given the scale of its operations, and the company processes returns through a mix of centralized and regional facilities. The research suggests that how a company designs its reverse logistics network, centralized versus decentralized, is one of the most important levers for reducing return-related emissions. Amazon has an advantage here in that its dense network of fulfillment and sortation centers across the country allows returns to be absorbed at relatively nearby facilities, potentially cutting the long-distance transport emissions that dominate the return footprint. Still, with millions of items returned daily, even marginal inefficiencies add up to a meaningful emissions category.
What “Net Zero by 2040” Actually Requires
Amazon co-founded The Climate Pledge in 2019, committing to reach net-zero carbon emissions by 2040, a decade ahead of the Paris Agreement’s 2050 target. The pledge covers all three scopes of emissions. Meeting that target would require simultaneous progress on every front discussed above: decarbonizing transportation across all legs, shifting data centers to genuinely clean power rather than certificate-based accounting, persuading or pressuring millions of suppliers to cut their manufacturing emissions, reducing packaging waste, and curbing the carbon cost of returns.
The “net” in net zero is doing a lot of work. Amazon has invested in nature-based carbon offsets, such as reforestation projects, and in carbon removal technologies. Critics point out that offsets can be unreliable: trees can burn down, offset projects can fail to deliver the promised sequestration, and the accounting for avoided emissions is inherently uncertain. The scientific consensus is that offsets should complement direct emissions reductions, not substitute for them. How much of Amazon’s path to 2040 depends on offsets versus genuine reductions in its own operations and supply chain will determine whether the pledge is met in substance or primarily on paper.
There is also the growth question. Amazon’s business has expanded rapidly, and each new fulfillment center, each new Prime delivery, each new AWS data center adds to the baseline that needs to be decarbonized. The company has at times reported that its carbon intensity (emissions per dollar of revenue or per package shipped) has improved even as absolute emissions have risen. That framing is common across the tech industry but can obscure the fact that what matters for the climate is the total amount of CO₂ entering the atmosphere, not the amount per unit of economic activity.
Why Scope 3 Numbers Are Hard to Trust
The difficulty of Scope 3 accounting deserves its own discussion because it affects how you should interpret any headline number Amazon or any other large company publishes. Scope 3 emissions are estimated, not measured. A company like Amazon has to decide which estimation methodology to use, which product categories to include, what emission factors to assign to each material and process, and how to handle gaps in supplier data. Different methodological choices can swing the total by tens of millions of tons without anything changing in the physical world.
As mandatory climate disclosure rules expand in the EU, California, and other jurisdictions, companies are under increasing pressure to report Scope 3 numbers in standardized ways.1Frontiers in Sustainable Energy Policy. Scope 3: what question are we trying to answer? Standardization should, in theory, make year-over-year comparisons more meaningful and cross-company comparisons possible. But researchers have raised the question of whether Scope 3 reporting, as currently practiced, is actually answering the right question. Is the goal to create a precise inventory of every gram of CO₂ in the value chain, which may be impossible at scale? Or is it to identify the biggest levers for reduction and push companies to pull them? The answer matters because it changes what counts as a good Scope 3 report and what counts as greenwashing.
For a consumer trying to evaluate Amazon’s environmental claims, the practical takeaway is that the total carbon number the company publishes is a rough estimate, not a lab measurement. Trends over time are more informative than any single year’s figure, and the composition of the number, which categories went up and which went down, tells you more than the headline total.
How Amazon Compares to Physical Retail
A common question is whether ordering from Amazon is better or worse for the climate than driving to a store. The answer depends on so many variables that no single comparison holds universally. E-commerce delivery can be more efficient per item when a single delivery van drops off dozens of packages on one route, replacing dozens of individual car trips. But that advantage erodes when customers choose fast shipping options that prevent consolidation, when items are shipped individually from distant warehouses rather than bundled, or when high return rates send products bouncing back and forth through the network.
Physical retail has its own emissions: the energy to light, heat, and cool a store; the customer’s drive to get there; and the fact that unsold inventory eventually gets discarded. Studies on this comparison tend to find that e-commerce has a slight edge in carbon efficiency for many product types, but the margin narrows or disappears when delivery speed is prioritized or when the customer lives close to a physical store and would have made the trip anyway.
Amazon’s scale introduces a wrinkle that smaller retailers do not face. The company’s logistics network is so vast that it effectively creates demand for transportation infrastructure, warehouse construction, and energy consumption that would not otherwise exist at the same scale. Whether the net effect is positive, because consolidation reduces per-unit emissions, or negative, because the convenience of fast delivery encourages more consumption overall, is a question that current emissions accounting frameworks are not designed to answer. The carbon footprint of a single order can be measured, at least roughly. The carbon footprint of a system that changes how much people buy in the first place is a much harder thing to quantify.
What Supplier Pressure Can and Cannot Achieve
Amazon’s supplier compliance framework asks manufacturers to track and report their own emissions and to pursue continuous improvement in energy efficiency and greenhouse gas reductions.4Advances in Economics Management and Political Sciences. Sustainable Supply Chain Strategies in E-Commerce: Case Studies of Amazon and Cainiao In principle, a company with Amazon’s purchasing power could reshape manufacturing practices across entire industries. In practice, the influence is uneven. Amazon’s own private-label products and devices, like Kindle readers and Echo speakers, flow through supply chains the company has more direct control over, making emissions reduction more tractable. But the vast majority of items sold on Amazon come from independent merchants who may sell through multiple platforms and respond to cost pressures more than sustainability mandates.
The effectiveness of supplier requirements also depends on enforcement. Tracking whether a factory in Vietnam or Bangladesh is actually reducing its energy consumption requires auditing capacity that scales poorly. Self-reported supplier data is better than no data, but it introduces obvious reliability concerns. The companies that have made the most credible progress on supply chain decarbonization tend to be those with relatively concentrated supplier bases, like automotive manufacturers working with a few hundred key parts suppliers. Amazon’s marketplace model, with millions of sellers, sits at the opposite end of that spectrum.
Still, the signals matter. When a company as large as Amazon tells its suppliers that emissions data is a condition of doing business, it creates a market incentive for suppliers to invest in cleaner operations, or at least to start measuring. That measurement infrastructure, imperfect as it is, lays the groundwork for future reductions if regulatory or commercial pressure increases. The question is whether the pace of improvement can match the pace of growth in the volume of goods flowing through the platform.