What Would an AWS International Restriction Cost Amazon?

Key Takeaways

• A 40% non-U.S. AWS exposure implies about $59 billion of annual revenue at risk.
• AWS supplies about 58% of Amazon operating income, making lost cloud sales disproportionately painful.
• Foreign data centers, contracts, AI demand, and partner relationships would deepen the direct revenue loss.

AWS International Restriction Exposure Starts With a $148 Billion Business

Amazon Web Services generated $42.2 billion of sales and $16.6 billion of operating income during the quarter ended June 30, 2026. For the trailing 12 months, AWS produced $148.4 billion of revenue and $54.7 billion of operating income, according to Amazon’s Q2 2026 results. That made AWS far more financially significant to Amazon than its share of corporate revenue might suggest. AWS accounted for about 19% of Amazon’s trailing-12-month sales but approximately 58% of consolidated operating income.

Those figures make an AWS international restriction primarily a profit problem rather than a revenue problem. Losing $1 of AWS sales has a much larger effect on Amazon’s operating earnings than losing $1 of lower-margin retail revenue. AWS’s trailing-12-month operating margin through June 2026 was approximately 36.8%. Its quarterly margin in Q2 2026 was approximately 39.4%.

The scale has also been rising quickly. AWS sales increased 37% year over year in Q2 2026, which Amazon described as its fastest growth in 18 quarters. The quarterly result implied an annualized revenue run rate of approximately $169 billion. Amazon also reported that its AWS artificial intelligence business and its chips business had each surpassed annual revenue run rates of $25 billion. Restricting AWS at this stage would remove part of a growing business rather than a mature operation with flat demand.

The geographic exposure cannot be read directly from Amazon’s financial statements. Amazon describes AWS as generating sales globally from compute, storage, database, and related services, but its 2025 annual filing states that country-level AWS sales are attributed according to the selling entity. That accounting treatment does not provide a direct measure of customer nationality or the geographic location in which AWS services are ultimately consumed.

That disclosure prevents a defensible statement that a particular percentage of AWS revenue comes from outside the United States. Any financial estimate must use a range of explicit assumptions. The appropriate issue is how much revenue would become unavailable under the exact definition of a restriction and how much of AWS’s cost base Amazon could remove after that revenue disappeared.

The global nature of AWS also means that geography matters beyond billing. As of August 19, 2026, AWS Global Infrastructure spans 123 Availability Zones across 39 geographic regions, with additional regions announced for Saudi Arabia and Chile. Its North American footprint includes four conventional U.S. commercial regions and two AWS GovCloud regions, alongside infrastructure in Canada and Mexico. Most of the AWS geographic footprint therefore exists outside conventional U.S. commercial regions.

Region count is not a revenue proxy. A Virginia region can generate more revenue than several smaller foreign regions combined. The footprint nevertheless demonstrates how extensively AWS has designed its service around international delivery, local data processing, latency requirements, regulatory compliance, and proximity to customers.

Defining the Restriction Determines the Financial Result

A restriction on AWS offering services outside the United States can describe two materially different policies.

A geographic infrastructure restriction would prevent AWS from operating cloud regions or data centers outside U.S. territory but could still permit foreign companies to buy services delivered from U.S. regions. Some workloads could migrate to Virginia, Ohio, Oregon, or Northern California. Revenue would decline because latency, data-residency rules, security requirements, local procurement rules, and customer preferences would make U.S.-hosted service unacceptable for some workloads. AWS could nevertheless retain part of its international customer base.

A customer-based restriction would be much more severe. Under that interpretation, AWS could provide commercial cloud services only to U.S. customers or eligible U.S. entities, regardless of whether a foreign customer’s workload could technically run inside an American data center. Foreign businesses, governments, universities, and other prohibited customers would have to migrate to another provider or build alternative infrastructure.

The financial modeling below assumes the broader customer-based restriction because it most closely matches a prohibition on offering AWS services to customers in other nations.

Policy scope would also determine treatment of multinational companies. A German subsidiary of an American corporation might be considered foreign under one rule and eligible under another. A U.S. corporation processing European consumer data could remain an eligible customer but still face foreign data-residency requirements. Governments might receive different treatment from commercial enterprises, and allied countries could receive exemptions. Each exception would reduce the revenue exposed.

A blanket international prohibition would also move in the opposite direction from the U.S. policy established under the July 23, 2025 American AI Exports Program. That executive order calls for promoting deployment of U.S.-origin artificial intelligence technologies abroad and explicitly includes cloud services, data-center storage, networking, chips, servers, accelerators, and AI systems in export packages.

That does not mean American cloud services are unrestricted. The U.S. Department of Commerce’s Bureau of Industry and Security continues to apply destination, end-user, end-use, and advanced-computing controls. May 2026 BIS guidance, for example, clarified licensing requirements for advanced-computing items involving entities headquartered in Country Group D:5 countries or Macau. The existing framework therefore differs materially from a hypothetical prohibition covering all foreign AWS customers.

The distinction matters financially because physically relocating workloads is costly but possible. Losing legal permission to serve the customer eliminates the commercial relationship.

An infrastructure-only prohibition could produce migration costs, foreign asset impairments, reduced customer retention, and weaker future growth without eliminating every dollar associated with international customers. A customer-based prohibition would place the affected customer stream itself at risk.

The timing would matter as much as scope. A multi-year phaseout would allow AWS to stop construction, move portable equipment, renegotiate contracts, redirect customers, sell infrastructure, and reduce staffing. An abrupt prohibition could leave data centers operating with little associated revenue even as depreciation, leases, energy commitments, financing costs, and employment obligations continued.

The same policy could consequently create substantially different accounting outcomes depending on whether implementation took months or several years.

How Much AWS Revenue Could Disappear?

The most defensible approach is to calculate scenarios rather than claim that Amazon discloses a foreign AWS percentage that it does not provide.

The baseline uses AWS’s trailing-12-month results through June 30, 2026: $148.404 billion of sales and $54.681 billion of operating income. The implied operating margin is approximately 36.8%. Amazon generated $93.712 billion of consolidated trailing-12-month operating income and $775.680 billion of trailing-12-month sales over the same period.

The table applies hypothetical non-U.S. AWS revenue shares of 30%, 40%, and 50%. These percentages are scenario assumptions, not geographic revenue estimates published by Amazon.

The 40% case illustrates why the restriction would matter so much to Amazon. Losing approximately $59.4 billion of AWS revenue would equal less than 8% of Amazon’s trailing-12-month corporate sales, yet the proportional $21.9 billion of AWS operating income exposure would equal more than 23% of Amazon’s consolidated operating income. AWS’s high margin creates that imbalance.

Those figures could understate the near-term earnings effect. The calculation assumes costs fall proportionally with revenue. Data-center economics do not work that way during an abrupt contraction. Buildings, leases, networking infrastructure, depreciation, personnel, and contracted energy expenses can continue after customer revenue disappears.

The Q2 2026 quarterly run rate gives another perspective. Annualizing $42.232 billion of quarterly AWS sales produces approximately $168.9 billion. Applying the same 40% assumption yields approximately $67.6 billion of annualized revenue exposure.

That figure should not be treated as a forecast because annualizing one quarter assumes its sales level persists for a full year. It does show that a restriction imposed on a growing AWS could have a larger dollar effect than a calculation based on calendar-year 2025 revenue.

The actual international percentage could sit below or above the illustrative range. Amazon’s filings do not resolve that uncertainty. AWS operates internationally at substantial scale, and the worldwide cloud market itself is geographically diverse.

The most recent quarterly market estimate available as of August 19, 2026 comes from Synergy Research Group. Synergy estimated Q2 2026 worldwide cloud infrastructure service revenue at $143.4 billion, with Amazon holding 28%, Microsoft 20%, and Google 15%. The United States remained the largest national market by a wide margin, but international demand represented a substantial part of the global total. New Space Economy’s broader coverage of AI and cloud market structure provides additional context for the competitive position of those providers.

Why Profit Damage Would Be Larger Than the Revenue Percentage Suggests

AWS’s contribution to Amazon earnings makes the profit effect more severe than a comparison of total corporate sales would indicate.

Amazon’s retail operations process enormous volumes of merchandise but operate on much lower margins than AWS. AWS converted approximately 36.8% of trailing-12-month sales into operating income through June 2026. During Q2 alone, the margin reached approximately 39.4%. Amazon’s North America segment produced a Q2 operating margin of approximately 7.9%, and its International segment produced approximately 4.1%, based on Amazon’s segment results.

A dollar removed from AWS revenue consequently removes much more potential operating income than a dollar removed from Amazon’s retail businesses.

The proportional scenario calculations also assume cost flexibility that Amazon would not possess immediately. AWS operates capital-intensive computing infrastructure. Amazon’s 2025 Form 10-K reported $190.1 billion of property and equipment allocated to the AWS segment at December 31, 2025, up from $110.7 billion at the end of 2024. AWS segment net additions to property and equipment reached $96.5 billion during 2025. Amazon does not break those AWS asset totals into U.S. and non-U.S. components.

Servers can sometimes be transferred or redeployed. Buildings, electrical connections, substations, cooling equipment, fiber routes, and long-term leases are much less portable. Assets designed around a regional demand forecast can become economically impaired when regulation removes access to that market.

An abrupt international prohibition could consequently create an accounting problem separate from lost revenue. Amazon might need to evaluate affected AWS facilities for impairment if expected future cash flows declined sufficiently. The accounting value of an asset can fall even when the physical facility remains operational.

Potential buyers would also know that Amazon faced regulatory pressure to dispose of foreign assets, which could weaken negotiating economics. Facilities optimized for AWS’s proprietary infrastructure might require modification before another operator could use them efficiently.

Cash flow would face another strain. Amazon reported trailing-12-month purchases of property and equipment, net of proceeds and incentives, of approximately $169 billion through June 2026. The company said the increase primarily reflected artificial intelligence investment. Consolidated free cash flow moved to an outflow of $7.6 billion over that period, compared with an inflow of $18.2 billion for the trailing 12 months ended June 30, 2025.

A restriction imposed during such a heavy investment cycle could force Amazon to cancel planned construction, renegotiate supplier commitments, absorb cancellation expenses, and redesign capacity plans. Capital spending would eventually fall, which could improve later cash flow, but the transition could be expensive.

The longer-term arithmetic becomes less severe if Amazon can reduce capacity in line with the smaller business. Power consumption, network traffic, equipment purchases, sales expenses, and some staffing costs would decline. A smaller AWS could eventually restore a healthy margin on its U.S. operations. The period between losing revenue and removing fixed costs would pose the greater earnings risk.

Foreign Infrastructure and Contracts Would Create Another Financial Hit

AWS has spent years turning geographic proximity into a product feature. Customers can select regions near users, address data-residency requirements, reduce latency, and construct redundant systems across multiple locations. A U.S.-only AWS would surrender much of that architecture.

AWS’s investment commitments demonstrate the size of this exposure. The AWS European Sovereign Cloud became generally available in January 2026 with its initial region in Brandenburg, Germany. AWS describes it as physically and logically separate infrastructure located entirely within the European Union and operated through dedicated European legal entities. The initiative is backed by €7.8 billion of planned investment in infrastructure, employment, and skills development.

Canada provides another example. AWS states that it plans nearly CA$21 billion of Canadian data-center infrastructure investment through 2037 and operates infrastructure regions in Montréal and Calgary, with six Availability Zones between them. Such investments make commercial sense when AWS expects decades of regional demand. A prohibition on foreign AWS service would change the economics of those commitments.

Some projects could be canceled before capital is spent. Existing facilities would present a harder problem. AWS could attempt to sell foreign data centers, lease them to local providers, place them in legally separate companies, or license technology to locally controlled operators if the applicable law permitted those arrangements. None of those approaches guarantees recovery of Amazon’s original investment.

Customer contracts introduce another dimension. Amazon’s June 2026 Form 10-Q reported approximately $496 billion of commitments associated with long-term customer performance obligations, primarily related to AWS. Their weighted-average remaining contractual life was 6.4 years.

That $496 billion should not be interpreted as foreign revenue at risk. Amazon disclosed that the total includes very large commitments involving U.S.-based AI companies. In Q1 2026, AWS and OpenAI expanded their existing $38 billion multi-year arrangement by another $100 billion over eight years. In Q2 2026, AWS and Anthropic expanded their existing commitment by more than $100 billion over 10 years. Those agreements demonstrate how much AWS revenue is increasingly tied to multi-year commitments, but Amazon does not publish their complete geographic composition.

A government-mandated prohibition could trigger contract termination provisions, regulatory clauses, refunds, customer migration assistance, disputes, or litigation depending on individual agreements and governing law. Public filings do not provide enough contract detail to quantify those costs.

Customer behavior could create losses beyond legally prohibited accounts. Multinational companies often want common platforms across countries. A corporation using AWS in the United States, Germany, and Singapore might reconsider its U.S. AWS deployment if Amazon could no longer provide the same platform elsewhere. Global purchasing agreements could shift toward another provider even when a U.S. component remained legally eligible.

That indirect churn is one reason a revenue model based strictly on prohibited foreign accounts could understate the commercial damage.

Competitors Would Gain More Than the Revenue Amazon Lost

Cloud workloads do not disappear when a provider becomes unavailable. Customers still need computing, storage, databases, cybersecurity services, analytics, networking, and artificial intelligence infrastructure. Spending would migrate.

If AWS alone faced the restriction, Microsoft Azure and Google Cloud would be positioned to absorb a substantial share of displaced demand. Microsoft’s global infrastructure spans more than 80 Azure regions and more than 500 data centers according to its U.S. infrastructure page as of August 2026. Google Cloud lists 43 global regions and 130 zones across six continents.

The competitive data already show substantial alternatives to AWS. Synergy Research Group estimated that Q2 2026 worldwide market share was 28% for Amazon, 20% for Microsoft, and 15% for Google. AWS remained the largest provider, but its two largest rivals already operated at enough scale to compete for multinational enterprises.

Migration would not happen instantly. Large organizations frequently build applications around proprietary AWS databases, serverless services, identity systems, data pipelines, networking configurations, and management tools. Moving thousands of workloads can require months or years. Those switching costs would slow AWS revenue loss if customers received a transition period.

A legal prohibition changes normal switching economics. Customers that must leave have no long-term commercial reason to tolerate indefinite migration costs. Consulting firms and competing cloud providers would have strong incentives to subsidize migrations, automate conversions, offer credits, and sign replacement agreements.

The competitive effect could last much longer than the restriction itself. A multinational enterprise that spends two years moving from AWS to Azure is unlikely to reverse that migration immediately if U.S. policy later changes. Cloud platforms create operational familiarity, staff skills, software dependencies, purchasing agreements, security certifications, and internal standards. Once those shift, winning the customer back becomes expensive.

AWS would also lose part of its scale advantage. Cloud economics reward high utilization of shared data centers and network infrastructure. Fewer customers can mean lower utilization, reducing the ability to spread fixed expenses over a large revenue base. Amazon could compensate by closing capacity and concentrating workloads in fewer facilities, but downsizing itself carries costs.

If a restriction applied to all U.S. cloud providers rather than AWS alone, the competitive destination would change. Foreign governments and enterprises would have stronger incentives to support domestic providers, sovereign cloud services, private-cloud deployments, and non-U.S. platforms. Such an outcome could permanently reduce the international reach of American cloud companies instead of transferring AWS customers to Microsoft or Google.

The Damage Would Spread Into AI, Space, and Amazon’s Broader Strategy

AWS increasingly functions as more than rented server capacity. It is Amazon’s distribution platform for artificial intelligence models, custom processors, data services, enterprise software infrastructure, and industry-specific computing.

Amazon’s July 30, 2026 earnings announcement said the AWS AI business had surpassed a $25 billion annual revenue run rate. Its chips business had also passed $25 billion. Amazon Bedrock provides managed access to foundation models and related AI infrastructure. Foreign restrictions would remove part of the addressable customer base for those services at a time when AI is driving exceptionally strong growth in cloud demand.

That creates a future-growth cost that does not appear in a trailing-12-month revenue calculation. If foreign AI demand continues expanding and AWS is legally excluded from those customers, the difference between unrestricted and restricted revenue could widen. The financial loss would consist of foregone growth as well as removed existing sales.

The effect reaches the space economy because AWS has become infrastructure for satellite operators, Earth observation companies, government programs, and data-processing services. Ground Stations as a Service increasingly connects geographically distributed antennas with cloud storage, processing, networking, and customer delivery.

AWS Ground Station is built around this model. AWS describes the service as a global network that lets satellite operators command spacecraft, downlink data, process information, and move that data directly into AWS infrastructure without building their own ground-station networks.

International AWS restrictions could weaken the value of such a distributed service if customers could no longer process data through AWS in jurisdictions where antennas or end users are located. Satellite companies that need global contact opportunities could shift processing toward competing clouds, independent ground networks, or hybrid architectures.

Amazon’s own satellite network creates another connection. Amazon Leo is designed in part around integration with AWS and enterprise connectivity. Amazon’s official private networking service allows organizations to connect Amazon Leo traffic directly to network and cloud infrastructure without routing it across the public internet.

The commercial appeal of combining satellite communications with AWS depends partly on international reach. A geographically constrained AWS would weaken that proposition for customers seeking multinational connectivity and computing under a common architecture.

The same principle appears across broader satellite communications markets, where cloud platforms increasingly interact with connectivity, data processing, enterprise networking, mobility, and remote operations. Restricting AWS internationally would separate pieces of a commercial model that Amazon has spent heavily to connect.

Amazon’s international retail business could also face indirect complications. Amazon’s financial filings explain that technology infrastructure supports AWS and other Amazon businesses. A policy written narrowly around the external sale of AWS services might leave Amazon free to operate internal systems supporting its foreign retail businesses. A broader restriction on deploying U.S. cloud technology abroad could impose costs on those operations as well. The legal language would determine whether that spillover occurred.

Partner economics matter too. AWS states that the AWS Partner Network has more than 140,000 partners from more than 200 countries and that 70% are headquartered outside the United States. Software vendors, consultants, managed-service providers, cybersecurity companies, and systems integrators generate business around AWS deployments. Removing international AWS sales would weaken this distribution network and make competing platforms more attractive to partners.

These effects would not all appear as immediate accounting losses. Some would emerge later as slower AWS growth, weaker pricing power, reduced customer acquisition, lower infrastructure utilization, or declining strategic value.

Amazon Could Reduce the Damage but Not Replace the Global Business Quickly

Amazon would have several ways to respond, but none could replace a large foreign customer base on short notice.

Domestic AWS demand could absorb some unused computing capacity. U.S. artificial intelligence companies are signing enormous infrastructure commitments. Amazon’s June 2026 filing shows how commitments from OpenAI and Anthropic contributed to approximately $496 billion of long-term customer performance obligations, primarily associated with AWS. Strong U.S. AI demand could allow Amazon to redirect servers, accelerators, and other portable equipment toward domestic workloads.

Physical capacity outside the United States would be harder to redirect. Servers and networking equipment can sometimes move. Data-center buildings cannot. Amazon could seek buyers for foreign facilities or transfer them into independently controlled companies if regulators permitted such transactions. Local telecommunications companies, infrastructure funds, sovereign cloud operators, or competing data-center businesses might value the facilities even if AWS could no longer operate them.

A licensing model could preserve some economics if the restriction allowed Amazon to license software or intellectual property without directly providing the cloud service. Local operators could potentially run AWS-compatible technology under independent ownership. Such an arrangement would convert some service economics into licensing or technology revenue, almost certainly at a different margin and with less control over customer relationships.

Amazon could also accelerate U.S. capacity utilization by lowering prices, offering larger committed-spend discounts, or pursuing more government and enterprise workloads. Those measures would help fill infrastructure but could pressure margins. Replacing $50 billion or more of foreign sales requires substantial incremental domestic demand even for a company operating at AWS’s scale.

Capital spending would eventually adjust. Projects intended primarily for prohibited markets could be canceled. Server purchases could fall. Hiring could slow. Construction commitments could be reduced. Those actions would protect cash over time.

They would not reverse the strategic cost of becoming a geographically constrained cloud provider.

AWS competes partly through scale, service breadth, customer familiarity, and the ability to deploy workloads near users in many countries. An international prohibition would remove the geographic component and give competitors years to deepen relationships with customers AWS could no longer serve.

The eventual financial outcome could consequently exceed the simple annual revenue loss. A lower growth rate would influence expectations about AWS’s future cash generation. AWS is Amazon’s largest source of segment operating income, so reduced growth in that business could change how investors value Amazon even if the retail, advertising, subscription, logistics, and other businesses continued expanding.

The midpoint scenario offers a useful reference rather than a prediction. If 40% of AWS trailing-12-month revenue were tied to customers prohibited under the hypothetical rule, approximately $59 billion of revenue would be exposed. Applying AWS’s recent operating economics gives approximately $22 billion of proportional operating income exposure before considering fixed-cost disruption. Using the Q2 2026 annualized run rate raises the corresponding revenue exposure to approximately $68 billion.

A restriction limited to foreign infrastructure, with foreign customers still allowed to consume U.S.-hosted AWS services, would produce a smaller result. A broad prohibition on serving foreign customers would move the outcome toward the larger scenario. Restrictions covering only designated countries could be far smaller.

That range is the financially responsible way to frame the issue because Amazon does not publish the customer-geography data required to calculate a single authoritative amount.

Summary

AWS has reached a size at which an international restriction could materially change Amazon’s financial profile. Through June 30, 2026, trailing-12-month AWS revenue stood at $148.4 billion, and AWS generated approximately $54.7 billion of operating income, representing about 58% of Amazon’s consolidated operating income.

Under an illustrative 40% foreign-revenue assumption, approximately $59.4 billion of trailing-12-month AWS revenue would be exposed. A proportional application of AWS’s recent operating margin produces approximately $21.9 billion of operating income exposure, equal to about 23% of Amazon’s trailing-12-month consolidated operating income. An abrupt restriction could produce a larger near-term earnings effect because foreign data centers, depreciation, leases, contractual commitments, staffing, and other fixed expenses would not disappear with revenue.

The financial effect could extend beyond the accounts legally prohibited from buying AWS. Foreign infrastructure could lose value, multinational customers could consolidate with competing cloud providers, international partners could redirect investment, and Amazon could surrender future AI demand that has not yet entered its financial statements. Connections between AWS and businesses such as Amazon Leo could also lose part of their international commercial reach.

Domestic AI demand and long-term U.S. customer commitments could absorb some displaced computing capacity. Capital expenditures could be reduced, foreign assets could potentially be sold, and Amazon could restructure the remaining AWS business around the United States. Those responses would reduce losses over time but would not recreate the economics of a global cloud platform.

The largest uncertainty remains the figure Amazon does not disclose: AWS revenue by actual customer location or nationality. Without that information, a precise dollar estimate would imply more certainty than the public data supports. A practical scenario range puts tens of billions of dollars of annual AWS revenue at risk under a broad foreign-customer prohibition, with the effect on operating profit disproportionately large because AWS remains Amazon’s dominant source of segment earnings.