- Sustainability Strategy and Goals
- Progress vs. Target Tracker
- Key Sustainability Innovations and Technologies
- Measurable Impacts
- Challenges and Areas for Improvement
- Future Plans and Long-Term Goals
- Comparisons to Industry Competitors
- Semiconductor ESG Metrics (Latest Available Data)
- What to Watch: 12 to 18 Month Indicators
NVIDIA, the $216 billion revenue AI accelerated computing company headquartered in Santa Clara, California, published its FY25 Sustainability Report in June 2025, covering fiscal year January 29, 2024 to January 26, 2025. The report marks NVIDIA’s most significant sustainability disclosure to date, introducing its first SBTi-validated emissions reduction targets, confirming the achievement of 100% renewable electricity for global offices and data centers, and disclosing a total GHG footprint of 7.15 million tCO2e, an 86.86% increase from FY24 driven by the explosive growth of AI chip manufacturing. A Greenpeace East Asia report published in March 2026, titled “Nvidia’s Green Illusion,” identified what it described as a “decarbonisation deficit” across NVIDIA’s East Asian manufacturing supply chain, where 24.09% of supplier electricity comes from renewables compared to 80.55% among North American suppliers.
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://www.nvidia.com/en-us/sustainability/
https://carboncredits.com/nvidia-hits-almost-216-billion-revenue-as-ai-boom-tests-its-climate-strategy-emissions/
Sustainability Strategy and Goals
NVIDIA’s sustainability strategy rests on the premise that making AI computation more energy efficient per unit of work delivered is its greatest opportunity to reduce absolute environmental impact at a global scale. The strategy is structured around three pillars: Energy, Efficiency, and Climate; People, Diversity, and Inclusion; and Responsible Business. NVIDIA’s first two SBTi-validated targets were disclosed in June 2025, more than five years after SBTi validation became standard practice among technology peers, covering a 50% absolute Scope 1 and 2 reduction by FY30, and a 75% reduction in Scope 3 emissions intensity per PetaFLOP (PFLOP) by FY30, both from a FY23 baseline. The absence of an absolute Scope 3 reduction target and any formal net zero commitment with a specific date represents a material gap in the strategy.
Net Zero and Carbon Emissions
NVIDIA’s total GHG footprint reached 7.15 million tCO2e in FY25, an 86.86% increase from 3.83 million tCO2e in FY24, driven almost entirely by Scope 3 purchased goods and services as AI chip production volumes surged. Total operational emissions (Scope 1 and 2) reached 241,330 tCO2e in FY25, up 27.03% from FY24, despite the achievement of 100% renewable electricity for offices and data centers under NVIDIA’s operational control, which reduced market-based Scope 2 to zero. Scope 3 emissions of 6.91 million tCO2e in FY25 represent 96.6% of the total footprint, with purchased goods and services alone accounting for approximately 6 million tCO2e, or 87% of Scope 3. NVIDIA has not set a net zero target with a specific date as of March 2026.
- Total GHG footprint in FY25: 7.15 million tCO2e, up 86.86% from FY24
- Scope 1 and 2 (market-based) in FY25: 241,330 tCO2e, up 27.03% from FY24; Scope 2 market-based reduced to zero through 100% renewable electricity
- Scope 3 emissions in FY25: 6.91 million tCO2e across 8 of 15 GHG Protocol categories
- Purchased Goods and Services (Scope 3 Cat. 1): approximately 6 million tCO2e; 87% of total Scope 3
- Supply chain emissions more than doubled in three years: from 2.97 million tCO2e in FY23 to 6.03 million tCO2e in FY25
- First SBTi-validated targets: 50% absolute Scope 1 and 2 reduction by FY30, and 75% Scope 3 intensity reduction per PFLOP by FY30, both from FY23 baseline
- No absolute Scope 3 reduction target and no formal net zero target with a specific date published as of March 2026
Water Stewardship
NVIDIA’s most commercially significant water innovation in FY25 is the Blackwell GPU architecture’s liquid cooling system, which the company reports increases data center water efficiency by over 300x compared to air-cooled systems by enabling compute density to increase without proportional increases in water use per computation. Because NVIDIA is fabless, its direct operational water consumption is confined to offices and leased facilities rather than manufacturing plants. The most material water impacts in NVIDIA’s value chain sit with TSMC, Samsung, and SK Hynix, whose semiconductor fabs in Taiwan, South Korea, and Japan are among the most water-intensive manufacturing operations globally, and which NVIDIA’s supplier engagement programme has not yet addressed with formal water reduction targets.
- Blackwell GPU liquid cooling: over 300x improvement in water efficiency per computation vs air-cooled systems at equivalent workloads
- Fabless model: NVIDIA’s direct operational water use confined to offices; manufacturing water impacts sit with TSMC, Samsung, and SK Hynix
- No formal supply chain water reduction target published for manufacturing partners as of FY25
- NVIDIA Earth-2 platform enables climate and hydrological modeling at kilometer resolution, providing tools for other organisations to plan water risk responses
Regenerative Agriculture
As a semiconductor design and software company, NVIDIA does not operate an agricultural supply chain and does not maintain a regenerative agriculture programme. Its indirect contribution to agricultural sustainability is delivered through the Earth-2 AI climate modeling platform, which provides AI-accelerated weather and climate simulation tools that enable farmers, agricultural policymakers, and water resource managers to anticipate and respond to drought, flood, and extreme heat risks.
Deforestation and Biodiversity
NVIDIA does not source forest-risk commodities and does not publish a formal biodiversity or deforestation-free commitment. Its packaging programme uses 100% recycled HDPE-2 for bulk carton packaging, resulting in 97% recyclability for GPU system packaging and 94% overall consumer packaging recyclability in FY25. The company continues to reduce reliance on plastic foam, increasing applications for paper-based cushioning, and requires vendors to maintain FSC certification for paper-based packaging materials.
- 97% recyclability for GPU system packaging cartons in FY25
- 94% overall consumer packaging recyclability in FY25
- 100% recycled HDPE-2 used for bulk carton packaging
- Plastic foam reduction ongoing; paper-based cushioning applications expanding across product packaging lines
- FSC certification required of paper-based packaging vendors
- No formal biodiversity or deforestation target published
Packaging and Circular Economy
NVIDIA’s circular economy programme focuses on three areas: packaging material transformation, hardware reuse and component recovery, and logistics optimization. Older-generation DGX systems are resold in secondary markets, and systems that cannot be resold are broken into components and recycled. The company has implemented a multimodal distribution system and consolidated shipping volumes to reduce freight fuel use and logistics emissions. Packaging redesign has reduced size, weight, and plastic foam usage across GPU, networking, and systems product lines, contributing to lower transportation emissions per unit shipped.
- 97% recyclable GPU system packaging cartons; overall 94% recyclable consumer packaging in FY25
- Older DGX systems resold in secondary markets; non-resold systems broken into components and recycled
- Multimodal distribution and shipping consolidation: reduce freight fuel use and logistics Scope 3 emissions
- Packaging redesign reduces size, weight, and plastic foam; paper-based cushioning replacing expanded foam across product lines
- No formal circular economy target with a 2030 recycled content percentage or recovery rate published
Human Rights and Responsible Sourcing
NVIDIA’s supply chain responsibility is governed by its alignment with the Responsible Business Alliance (RBA) Code of Conduct and the Responsible Minerals Initiative (RMI), covering conflict mineral sourcing governance across tin, tantalum, tungsten, gold, and cobalt. The company expects all suppliers to maintain progressive employment, health, safety, and ethical practices meeting or exceeding applicable laws and the RBA standards. The supplier SBT engagement programme reported that over 80% of Scope 3 Category 1 emissions were covered by supplier engagement discussions in FY25, but the company has not published a formal supplier SBT adoption rate with a specific percentage of suppliers who have formally set or validated targets.
- Over 80% of Scope 3 Category 1 supplier emissions covered by active supplier engagement in FY25
- No formal supplier SBT adoption rate published (e.g., percentage of suppliers who have set or validated targets)
- RBA Code of Conduct applied to supply chain; RMI membership governs conflict minerals sourcing
- No direct renewable energy investment in East Asian manufacturing hubs disclosed; suppliers in Taiwan, South Korea sourcing only 24.09% renewable electricity
- Greenpeace March 2026 report identified a “decarbonisation deficit” across NVIDIA’s 20 largest suppliers, accounting for over 50% of cost of goods sold
Nutrition and Health
As a semiconductor design and software company, NVIDIA does not operate in the food or nutrition sector. Its health-adjacent contribution comes through AI tools and platforms for genomics, drug discovery, and medical imaging, specifically through BioNeMo and Clara, which are used by pharmaceutical and healthcare organisations to accelerate disease research and clinical diagnostics.
Community and Social Impact
NVIDIA’s most publicly visible social programme is its AI for social good platform, which makes NVIDIA Inception accelerator programme membership and GPU computing credits available to nonprofits working in climate, health, education, and humanitarian response. The company’s employee turnover rate was 2.5% in FY25, one of the lowest in the technology sector, reflecting the concentrated talent retention dynamics of a company growing revenues from $26 billion to $130 billion within three fiscal years. NVIDIA’s Women in Technology and HBCU (Historically Black Colleges and Universities) partnership programmes are the primary formal diversity equity and inclusion delivery mechanisms for the company.
- Employee voluntary turnover: 2.5% in FY25, among the lowest in the technology sector
- NVIDIA Inception accelerator: GPU computing credits and programme access for nonprofits in climate, health, and humanitarian response
- Women in Technology and HBCU partnership programmes: primary formal DEI delivery mechanisms
- NVIDIA Foundation: funds AI for science and humanitarian research including climate modeling, pandemic response, and wildfire prediction
Governance and Transparency
NVIDIA’s FY25 Sustainability Report was independently assured by PwC, covering Scope 1, 2, and selected Scope 3 metrics with limited assurance. The most substantive governance concern is the five-year delay in setting SBTi-validated targets, which were not published until June 2025 despite NVIDIA achieving trillion-dollar market capitalisation since 2023. The California Voluntary Carbon Market Disclosures Act (VCMDA) filing published in 2025 confirms the two validated targets and their FY23 baseline, but explicitly notes the absence of any absolute Scope 3 reduction target or a formal net zero commitment date.
Technology and Innovation
NVIDIA’s most important sustainability innovation is not an operational or supply chain programme but its product architecture itself. The Blackwell GB200 Grace Blackwell Superchip delivers 25x energy efficiency over the prior Hopper generation for inference of massive large language models. NVIDIA estimates that switching from CPUs to GPUs for AI and high-performance computing could save 40 trillion watt-hours of energy every year globally. NVIDIA holds the top position on the November 2024 Green500 list, the ranking of the world’s most energy-efficient supercomputers, with 8 of the top 10 systems powered by NVIDIA.
The Earth-2 climate AI platform is NVIDIA’s most distinctive contribution to global sustainability infrastructure. Launched in 2024 and expanded with the cBottle generative AI foundation model in June 2025, Earth-2 enables kilometer-resolution global climate simulation at speeds thousands of times faster than traditional numerical models, providing scientists and policymakers with the tools to model flood risk, drought probability, wildfire spread, and sea level rise at actionable local resolution. NVIDIA’s $5 billion investment in Intel in September 2025, partly motivated by accessing Intel’s established Scope 1 and 2 semiconductor emissions reduction capabilities, is the most commercially significant supply chain sustainability investment NVIDIA has made to date.
- Blackwell GB200: 25x energy efficiency improvement over Hopper generation for LLM inference
- GPU vs CPU substitution for AI workloads: estimated 40 trillion watt-hours of annual energy savings globally
- NVIDIA powers 8 of top 10 Green500 most energy-efficient supercomputers as of November 2024
- Earth-2 cBottle: world’s first generative AI foundation model for kilometer-resolution global climate simulation
- $5 billion Intel investment (September 2025): accesses Intel’s 20+ years of semiconductor emissions reduction experience
- Blackwell liquid cooling: 300x improvement in water efficiency per computation vs air-cooled systems
Global Partnerships and Advocacy
NVIDIA co-operates the Earth-2 climate platform with leading climate institutions including the Max Planck Institute for Meteorology, the Allen Institute for AI (Ai2), and national meteorological agencies across 20+ countries. The company is a signatory of the Paris Agreement Business Ambition campaign and a member of the UN Global Compact. NVIDIA advocates publicly for AI-accelerated climate science and for policy frameworks that value computation efficiency metrics as a core component of sustainability regulation for the semiconductor and data center sectors.
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://tracenable.com/company/nvidia/ghg-emissions
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
https://blogs.nvidia.com/blog/earth2-generative-ai-foundation-model-global-climate-kilometer-scale-resolution/
https://datacentremagazine.com/hyperscale/how-nvidia-is-boosting-water-efficiency-with-blackwell
Progress vs. Target Tracker
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://tracenable.com/company/nvidia/ghg-emissions
https://carboncredits.com/nvidia-hits-almost-216-billion-revenue-as-ai-boom-tests-its-climate-strategy-emissions/
Key Sustainability Innovations and Technologies
The Blackwell architecture is NVIDIA’s most consequential sustainability contribution of FY25, and potentially of the decade. The GB200 Grace Blackwell Superchip delivers 25x energy efficiency over the Hopper generation for LLM inference, meaning that a given AI workload completed on Blackwell consumes 25 times less energy than on the prior generation hardware. At the scale of global AI infrastructure, where data center AI workloads are projected to consume 1,000 TWh annually by 2026, a 25x efficiency improvement per computation is the single most impactful decarbonization lever available to the AI industry, exceeding the impact of any renewable energy procurement, supplier engagement, or carbon offset programme in both magnitude and speed.
Earth-2 is NVIDIA’s most transformative contribution to global sustainability science infrastructure. The cBottle generative AI foundation model, launched in June 2025, simulates global climate at kilometer resolution, thousands of times faster and more energy-efficiently than traditional numerical models. This enables decision-makers at city, watershed, and national grid level to access climate projections previously available only to major research institutions with weeks of supercomputer access. Earth-2 is being used by the Max Planck Institute for Meteorology to pioneer the first kilometer-scale full Earth system simulations and by Ai2 to improve prediction of flooding rains and wildfire-spreading winds.
Blackwell’s liquid cooling architecture delivers a 300x improvement in water use efficiency per computation by enabling ultra-high compute density within a liquid-cooled thermal envelope, where heat is removed by direct liquid contact rather than by air flow through large volumes of evaporated water. This is commercially significant because traditional air-cooled data center expansion at AI workload densities would require dramatically more water for evaporative cooling than liquid-cooled Blackwell deployments at equivalent computational output.
- Blackwell GB200: 25x energy efficiency for LLM inference over Hopper; 50x+ over CPUs for AI inference workloads
- Global GPU-vs-CPU substitution potential: 40 trillion watt-hours per year saved
- Earth-2 cBottle: kilometer-resolution global climate simulation, thousands of times faster than traditional models
- Blackwell liquid cooling: 300x water efficiency improvement per computation vs air-cooled systems
- Green500 top position: NVIDIA powers 8 of 10 most energy-efficient supercomputers globally as of November 2024
- $5 billion Intel investment: access to Intel’s 20+ years of semiconductor manufacturing emissions reduction R&D
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://blogs.nvidia.com/blog/earth2-generative-ai-foundation-model-global-climate-kilometer-scale-resolution/
https://datacentremagazine.com/hyperscale/how-nvidia-is-boosting-water-efficiency-with-blackwell
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
Measurable Impacts
NVIDIA’s FY25 sustainability data presents a company where product-level environmental performance is advancing at an extraordinary rate while supply chain absolute emissions are growing at an equally extraordinary rate in the opposite direction. The 25x Blackwell efficiency gain for LLM inference is commercially verified by benchmark data and confirmed through the Green500 top ranking. The 100% renewable electricity achievement for offices and data centers under operational control is a genuine milestone for a company whose own facilities are powered entirely by clean energy.
The supply chain trajectory is the defining counter-narrative. Total emissions grew 86.86% in a single year from FY24 to FY25, with Scope 3 purchased goods emissions rising from approximately 3.2 million tCO2e in FY23 to 6 million tCO2e in FY25, a figure that the Greenpeace March 2026 report confirms as continuing to rise. NVIDIA’s 80%+ supplier engagement coverage in FY25 demonstrates reach, but the absence of a disclosed adoption rate for supplier SBTs means the engagement has not yet been converted into formal emissions reduction commitments at the supplier level.
- Total GHG footprint: 7.15 million tCO2e in FY25, up 86.86% from 3.83 million tCO2e in FY24
- Scope 1 and 2 (market-based): 241,330 tCO2e in FY25; Scope 2 market-based at zero due to 100% renewable electricity
- Scope 3 purchased goods (Cat. 1): approximately 6 million tCO2e in FY25, up from 2.97 million tCO2e in FY23
- Supply chain emissions more than doubled in three years (FY23 to FY25)
- Blackwell 25x energy efficiency for LLM inference vs Hopper generation
- 8 of 10 Green500 most energy-efficient supercomputers powered by NVIDIA
- 97% recyclable GPU system packaging
- 80%+ Scope 3 Category 1 supplier engagement coverage
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://tracenable.com/company/nvidia/ghg-emissions
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
Challenges and Areas for Improvement
The absence of a net zero target date and an absolute Scope 3 reduction target is the most material governance gap in NVIDIA’s sustainability framework. With total emissions at 7.15 million tCO2e in FY25 and Scope 3 growing by 86.86% in a single year, the intensity-only Scope 3 target (75% reduction per PFLOP by FY30) allows absolute emissions to continue rising indefinitely as long as computational output grows faster than the efficiency improvement rate. Given that NVIDIA’s revenue grew from $26 billion to $130 billion in three fiscal years, and that AI computing demand is projected to continue compounding, an intensity-only framework has the same structural accountability problem as Salesforce’s intensity-based Scope 3 approach, but at a scale that is an order of magnitude larger.
The Greenpeace East Asia “Green Illusion” report, published March 2026, identifies the specific supply chain mechanism through which NVIDIA’s emissions accountability gap is materialising. NVIDIA’s 20 largest suppliers, accounting for over 50% of cost of goods sold, source only 24.09% of their electricity from renewables, compared to 80.55% among North American suppliers. TSMC, Samsung, and SK Hynix have seen electricity consumption rise by up to 16% between 2022 and 2024, while renewable energy adoption increased by less than 4%. NVIDIA has not made direct investments in renewable energy infrastructure near its East Asian manufacturing hubs, and has set no 2030 deadline for major suppliers to transition to 100% renewable energy.
The five-year delay in SBTi target setting means NVIDIA’s science-based targets were validated only after the company had already doubled its emissions footprint. The Scope 1 and 2 target of a 50% absolute reduction by FY30 from a FY23 baseline is set against emissions that had already grown 27.03% in FY25 vs FY24, meaning the reduction pathway requires both reversing recent growth and then delivering the full 50% cut in five years. Scope 3 disclosure covers only 8 of 15 GHG Protocol categories, specifically excluding downstream use-phase emissions from sold products from formal Scope 3 disclosure, which is the category most affected by Blackwell’s 25x efficiency claims and where the most commercially material customer-level emissions reductions are occurring.
- No net zero target date published as of March 2026
- No absolute Scope 3 reduction target; intensity-only framework allows absolute Scope 3 growth at scale
- Total emissions up 86.86% in one year; supply chain emissions doubled in three years (FY23 to FY25)
- Top 20 suppliers: 24.09% renewable electricity in East Asia vs 80.55% in North America
- No direct renewable energy investments in East Asian manufacturing hubs (Taiwan, South Korea)
- TSMC, Samsung, SK Hynix electricity consumption rose up to 16% (2022 to 2024) with renewable adoption up less than 4%
- SBTi targets set five years after peer companies; first validated targets only disclosed June 2025
- Scope 3 Category 11 (use of sold products) excluded from formal Scope 3 disclosure
- No formal supplier SBT adoption rate published despite 80%+ engagement coverage
Source
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
https://carboncredits.com/nvidia-hits-almost-216-billion-revenue-as-ai-boom-tests-its-climate-strategy-emissions/
https://tracenable.com/company/nvidia/ghg-emissions
Future Plans and Long-Term Goals
By FY30, NVIDIA has committed to a 50% absolute Scope 1 and 2 reduction and a 75% Scope 3 emissions intensity reduction per PFLOP, both from the FY23 baseline. The roadmap for achieving the Scope 3 intensity target rests entirely on continued GPU architecture generational efficiency improvements, the Blackwell generation having already delivered 25x inference efficiency improvement in FY25. If successive architecture generations (Rubin, post-Rubin) deliver comparable step-change improvements, the 75% per PFLOP intensity reduction is achievable well before FY30. The challenge is that this represents a relative improvement in a rapidly expanding denominator, not an absolute reduction in emissions.
For supply chain absolute emissions, the $5 billion Intel investment and the Greenpeace March 2026 pressure may accelerate NVIDIA toward direct renewable energy co-investment in East Asian manufacturing hubs, which would be the structural intervention required to bend the supply chain emissions curve. NVIDIA’s future sustainability credibility depends on whether FY26 disclosures include an absolute Scope 3 reduction target, a net zero target date, and a disclosed plan for transitioning its top 20 suppliers toward 100% renewable electricity. The Earth-2 platform expansion, with the cBottle model launched January 2026, signals continued investment in AI climate infrastructure that will have compounding impact through the decade.
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://carboncredits.com/nvidia-hits-almost-216-billion-revenue-as-ai-boom-tests-its-climate-strategy-emissions/
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
Comparisons to Industry Competitors
AMD and Intel are NVIDIA’s most directly comparable semiconductor peer group for sustainability benchmarking. AMD targets a 50% absolute Scope 1 and 2 reduction by 2030 from a 2020 baseline, achieving 24.5% reduction to 46,605 tCO2e by 2023, and requires 80% of suppliers to source renewable energy by 2025, with 83% of supplier manufacturing sites audited under sustainability standards. Intel has been reducing emissions for over two decades, cutting absolute emissions by 70% since its peak year of 2006, and targets net zero Scope 1 and 2 by 2040 and Scope 3 upstream emissions by 2050. Intel estimates that its chips sold in 2024 will contribute 3.2 million metric tonnes of annual downstream emissions, and has pledged a tenfold energy efficiency improvement for its chips between 2019 and 2030.
NVIDIA’s total Scope 3 footprint of 6.91 million tCO2e in FY25 dwarfs AMD’s total disclosed footprint but is partially explained by the extraordinary scale difference in production volumes following the AI boom. Intel’s fabrication facilities generate material Scope 1 and 2 manufacturing emissions that NVIDIA’s fabless model outsources to TSMC, meaning the emissions exist in the same geographic and energy system but appear under different corporate reporting boundaries. NVIDIA’s 25x Blackwell inference efficiency gain leads the semiconductor sector for downstream use-phase efficiency improvement, while Intel’s 70% historical absolute Scope 1 and 2 reduction leads the sector for operational decarbonization over time.
Semiconductor ESG Metrics (Latest Available Data)
Source
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
https://carboncredits.com/amd-stock-hits-all-time-highs-can-esg-and-net-zero-momentum-sustain-the-rally/
https://csrreportbuilder.intel.com/pdfbuilder/pdfs/CSR-2024-25-Full-Report.pdf
https://trellis.net/article/nvidia-benefits-5-billion-intel-investment/
What to Watch: 12 to 18 Month Indicators
FY26 Disclosure of Absolute Scope 3 Target and Net Zero Date
NVIDIA’s FY26 Sustainability Report, expected in June 2026, is the most consequential near-term sustainability disclosure in the semiconductor sector. The critical question is whether NVIDIA adds an absolute Scope 3 reduction target and a formal net zero target date to its currently intensity-only Scope 3 framework. If NVIDIA discloses an absolute Scope 3 reduction commitment aligned with SBTi’s value chain standard, it would represent a structural change in accountability that directly addresses the Greenpeace East Asia “Green Illusion” finding. If the FY26 report maintains only the 75% per PFLOP intensity target without an absolute backstop, it confirms that NVIDIA’s emissions accountability framework permits unlimited absolute Scope 3 growth as long as computational performance scales faster than emissions.
East Asian Supplier Renewable Energy Transition Plan
Greenpeace’s March 2026 analysis identified that NVIDIA’s 20 largest suppliers in East Asia source only 24.09% of electricity from renewables, with TSMC, Samsung, and SK Hynix electricity consumption rising up to 16% between 2022 and 2024. NVIDIA’s FY26 sustainability disclosure will be the first annual report following this publicly released finding. Any formal commitment to direct renewable energy investment in Taiwan or South Korea, a supplier 2030 renewable electricity transition deadline, or a revised supplier engagement programme with binding renewable energy targets would represent a direct response to the structural supply chain risk identified. Absence of any supply chain renewable energy target in FY26 would confirm that the Greenpeace findings have not yet translated into formal governance change.
Rubin Architecture Efficiency Benchmark and Green500 Performance
NVIDIA’s post-Blackwell GPU architecture, codenamed Rubin, is expected for launch in 2026. Its energy efficiency per PFLOP relative to Blackwell will be the most commercially significant indicator of whether NVIDIA’s Scope 3 intensity target trajectory is on track for the FY30 deadline. Blackwell delivered 25x over Hopper; any Rubin efficiency improvement compounding further on that baseline would provide independent validation that the per-PFLOP intensity reduction is real and accelerating. NVIDIA’s Green500 rank in November 2026 will provide a public, independently verified reference point for that efficiency trajectory, separate from company-disclosed benchmarks.
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
https://carboncredits.com/nvidia-hits-almost-216-billion-revenue-as-ai-boom-tests-its-climate-strategy-emissions/
NVIDIA sits at the most consequential intersection in the global sustainability debate: the company whose products are simultaneously the most energy-efficient computing architecture ever built for AI workloads, and the company whose supply chain emissions are growing faster than any disclosed technology company in the world. The 25x Blackwell efficiency improvement and the 40 trillion watt-hour global energy saving potential of GPU substitution for CPU-based AI workloads are genuine, technically verified contributions to global decarbonization that no other company in any sector can match in absolute scale of potential impact. The Earth-2 cBottle climate modeling platform adds a second category of global sustainability infrastructure contribution that extends NVIDIA’s sustainability footprint far beyond its own operational boundaries.
The accountability framework that governs NVIDIA’s own supply chain is structurally inadequate relative to both the scale of the problem and the company’s own resources. Supply chain emissions of 6.91 million tCO2e in FY25, doubled from FY23, with no absolute reduction target, no net zero date, no renewable energy transition deadline for suppliers, and suppliers sourcing only 24.09% renewable electricity in East Asia, describes a company that has set product-level efficiency goals while leaving the manufacturing emissions question entirely to its supplier base and to the energy policy of Taiwan and South Korea. NVIDIA’s $216 billion in revenue in FY25 provides the financial capacity to make direct renewable energy infrastructure investments in East Asia at a scale that would shift the grid mix available to TSMC, Samsung, and SK Hynix.
Three strategic takeaways for practitioners benchmarking or replicating NVIDIA’s approach:
- Intensity targets for Scope 3 are only credible accountability mechanisms when the denominator (computation output) is genuinely bounded. NVIDIA’s 75% per PFLOP target is technically ambitious but structurally unconstrained because there is no ceiling on global AI computing demand. Practitioners designing Scope 3 targets for technology hardware companies should pair intensity targets with disclosed absolute emissions scenarios that show the range of absolute outcomes under different demand growth assumptions, so that stakeholders can assess the realistic absolute emissions trajectory rather than the theoretical intensity improvement in isolation.
- Fabless manufacturing creates a structural supply chain emissions accountability gap that requires a different intervention model than traditional supplier engagement. Because NVIDIA does not own any semiconductor fabrication capacity, it cannot unilaterally decarbonize its supply chain through operational investment. The required intervention is direct co-investment in renewable energy infrastructure near major fab sites in Taiwan and South Korea, either through direct equity in energy projects, through long-term PPAs that create new generation capacity, or through collaborative industry procurement with other fabless companies. Practitioners advising fabless technology companies should treat co-investment in supply chain renewable energy as a prerequisite for any credible absolute Scope 3 reduction commitment.
- Product efficiency improvements should be reported with absolute emissions avoided data, not only relative improvement percentages. NVIDIA’s 25x Blackwell efficiency improvement is its most important sustainability result, but the FY25 report does not disclose the absolute tCO2e avoided through Blackwell deployment vs counterfactual Hopper deployment. Publishing the estimated absolute emissions avoided through product efficiency improvements, independently verified by a third party, would provide the most direct evidence of the net sustainability impact of NVIDIA’s product strategy. For any technology company whose sustainability case rests on product efficiency rather than operational reduction, this kind of impact quantification is the evidence base that sophisticated ESG audiences require.
Source
https://images.nvidia.com/aem-dam/Solutions/documents/NVIDIA-Sustainability-Report-Fiscal-Year-2025.pdf
https://tracenable.com/company/nvidia/ghg-emissions
https://esgpost.com/greenpeace-report-flags-rising-emissions-in-nvidias-east-asia-supply-chain/
https://trellis.net/article/nvidia-top-esg-priority-slash-power-gobbled-by-ai-chips/
https://blogs.nvidia.com/blog/earth2-generative-ai-foundation-model-global-climate-kilometer-scale-resolution/