MIT survey finds AI sustainability gains do not always mean lower emissions
Companies report benefits from AI, yet warehouse energy use and total emissions tell a more complicated story. Governance and measurement remain obstacles.
Loading page…
Companies report benefits from AI, yet warehouse energy use and total emissions tell a more complicated story. Governance and measurement remain obstacles.
Listen to this story
MIT’s 2026 State of Supply Chain Sustainability report finds that AI-related efficiency gains do not reliably correspond to lower overall emissions. In a survey of 1,810 supply-chain professionals across 91 countries, 53% of mature agentic-AI adopters reported lower total greenhouse gas emissions, while 29% reported increases; the responses do not establish that AI caused either outcome. For companies deploying AI, the results underscore the need to track total environmental impact—not just local efficiency—and to monitor the energy and waste costs of automation.
Among North American organizations using AI for sustainability, 87% reported a positive impact on sustainability goals, while 63% used AI to estimate missing supplier emissions data.
At advanced or fully automated warehouses using AI, 78% reported higher energy use and 76% higher waste; only 24% used AI for real-time environmental monitoring.
Although 83% rated sustainability important to long-term business success, only 33% of companies with public goals had board-level oversight.
AI is helping companies pursue sustainability goals, but those gains do not consistently translate into lower total emissions. MIT’s 2026 State of Supply Chain Sustainability report, published October 15, finds widespread reported benefits alongside rising energy use and waste in advanced warehouses using AI. The findings expose a gap between improving individual operations and reducing their overall environmental burden.
The MIT Sustainable Supply Chain Lab’s report draws on responses from 1,810 supply chain professionals across 91 countries. It examines sustainability commitments, organizational oversight, technology adoption and trade disruption. Its AI findings describe respondents’ experiences, rather than independently measured environmental results.
Among North American organizations, 78% reported using AI in sustainability initiatives, and 87% of adopters reported a positive impact on sustainability goals. The report also records reductions in material waste at 86% and energy use at 81%. Another application is filling information gaps: 63% use AI to estimate missing supplier emissions data.
The picture changes when the measure is total greenhouse gas emissions. Among users the report classifies as mature adopters of agentic AI, 53% reported lower total greenhouse gas emissions, while 29% reported an increase. Those results distinguish progress on particular sustainability goals from a reduction in the organization’s total emissions. They do not, by themselves, establish that AI caused the changes.
Warehousing makes that tension especially visible. Among advanced and fully automated warehouses using AI, large majorities reported higher energy use and waste after implementation. Yet AI-driven, real-time monitoring of environmental consequences was much less common within that group.
Reported increased energy use since implementing AI.
Reported increased waste since implementing AI.
Used AI-driven real-time monitoring to track environmental consequences.
The uneven technology outcomes sit within a broader execution problem. Sustainability was rated important or very important to long-term business success by 83% of respondents—the highest share in the report’s seven-year history. But among companies with public sustainability goals, only 51% had a formal team spanning business functions, and 33% had board-level sustainability oversight.
That structure was associated with a substantial difference in daily decisions. At companies with formal cross-functional teams, 57% reported that sustainability was frequently or always built into decisions. The share was 13% where responsibility rested with a single function. The comparison shows an association with shared responsibility, not proof that creating a team alone delivers better outcomes.
Sustainability was most deeply embedded in operational choices: transportation and logistics at 39%, procurement at 35%, warehousing at 34%, and product design at 33%. It was less embedded in capital investment, network design and supply-chain risk planning—the longer-term decisions that shape future emissions.
External pressures also constrained execution. While 59% reported becoming more committed to sustainability following recent U.S. policy shifts that deprioritize environmental concerns, 68% said tariff uncertainty had forced delays, reductions or abandonment of sustainability strategies. Nearly one-third reported that tariff-related disruption had increased operational emissions.
Freight transportation illustrates the difference between operational improvements and replacing underlying technology. Nearly every freight company was pursuing emissions reductions through route optimization, load consolidation and driver-behavior programs. Nevertheless, 43% of fleets remained mostly or entirely dependent on conventional diesel, while only 8% had fully transitioned to low-emission energy.
The regional split was pronounced. In Europe, 84% of freight companies said recent policy changes had helped their decarbonization strategy, compared with 44% in North America. Battery-electric vehicles were operating at scale at 60% of European companies, versus 36% in North America.
Supplier collaboration offered a more encouraging finding. Participation in joint supplier engagement and decarbonization programs rose from 31% to 58% in one year. Partnerships addressing shared suppliers’ emissions increased from 23% to 55%, while the share working entirely alone fell from 18% to 6%.
Loading discussion...
Join the conversation
Explain how you would weigh narrower environmental gains against the overall footprint.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.