By the end of 2026, ~50,000 companies will file CSRD-aligned carbon reports under limited assurance. By 2028 the standard moves to reasonable assurance — the same scrutiny applied to financial audits. California's SB-253 mandates Scope 1 and 2 disclosure for the 2025 fiscal year, with Scope 3 following in 2027. And the data underneath all of these reports — supplier contracts, production volumes, energy consumption by line, fuel burn by asset — is exactly the data most enterprises do not want sitting in a hyperscaler tenant labeled "ESG platform." Sovereign AI changes that calculus. The Scope 1, 2, and 3 calculations run on-premises, the supplier emissions data never leaves the perimeter and accuracy improves by an order of magnitude when activity-based AI replaces spend-based spreadsheet math. Sign up free to explore the sovereign carbon accounting platform.
The Three Scopes · Accuracy Today vs Sovereign AI
The GHG Protocol defines three emission scopes covering everything from a company's combustion fuel to its supply chain's electricity. Each scope has a different accuracy ceiling under traditional spreadsheet-based carbon accounting, and a different accuracy floor under activity-based sovereign AI. Sign up free to model your three-scope accuracy profile.
01
SCOPE 1 · DIRECT EMISSIONS
Fuel combustion, refrigerants, fugitive emissions, process chemistry
~8%
SPREADSHEET · 65%
SOVEREIGN AI · 99%
Direct on-site fuel burn from boilers, furnaces, and fleet vehicles plus refrigerant leaks. Activity data already exists in CMMS, fuel inventory, and refrigerant logs. The accuracy gap is integration, not measurement — sovereign AI pulls activity data directly from the maintenance system and applies GHG Protocol emission factors continuously.
02
SCOPE 2 · PURCHASED ENERGY
Purchased electricity, steam, heating, cooling — by region, by hour
~14%
SPREADSHEET · 70%
SOVEREIGN AI · 97%
Annual utility bills give a 70%-accurate Scope 2 number. Hourly meter data correlated to regional grid mix gives 97%. Sovereign AI pulls meter readings every minute, applies time-stamped grid emission factors per region, and produces market-based and location-based Scope 2 figures continuously. Both methods are auditable.
03
SCOPE 3 · VALUE CHAIN · 15 CATEGORIES
Purchased goods, transport, capital goods, business travel, use-phase, end-of-life
~78%
SPEND-BASED · ±35% error
SOVEREIGN AI · ACTIVITY-BASED · ±4%
Scope 3 is where the audit will fail. Spend-based proxies — total purchase value × industry-average emission factor — produce ±35% error. Activity-based math — actual quantities of materials, supplier-specific factors, shipment-level transport data — produces ±4%. Sovereign AI matches procurement records to verified emission factors from the 70K+ on-prem library and handles all 15 categories. Supplier contracts never leave the perimeter.
100%
Carbon data stays inside your perimeter
10×
Scope 3 accuracy lift vs spend-based
GHG
Protocol + ISO 14064-1 compliant
Audit
Ready · CSRD · SB-253 · SEC · ESRS
The audit trajectory matters as much as the methodology. Limited assurance applies to CSRD reports in 2026; reasonable assurance kicks in by 2028 with verification standards comparable to financial audit. A spend-based Scope 3 number that passes today's bar will not pass the 2028 bar. Sovereign activity-based AI closes the gap before the deadline arrives. Book a free demo of the activity-based Scope 3 path.
Two Real Sovereign-Carbon Scenarios
Two real scenarios from manufacturers who replaced cloud-hosted carbon accounting with sovereign on-prem AI. Each shows the data-residency or accuracy gap the cloud platform created and the hardware integration that closed it. Book a free demo to walk through your scenario in detail.
SCENARIO 01
"We are a defense contractor. Our Scope 3 includes supplier contract data that is ITAR-restricted. The leading cloud ESG platform sends our supplier list to a US-East hyperscaler region. We cannot use it. What is the sovereign path?"
THE PROBLEM
Tier-1 defense manufacturer producing aerospace assemblies for US government programs. ITAR registration requires that supplier identity, part numbers, and procurement quantities for restricted programs cannot be transmitted to or stored on non-US-citizen-controlled infrastructure. The mainstream cloud ESG platforms host data on multi-tenant hyperscalers where access boundaries cannot be guaranteed at the ITAR level. Procurement team has been calculating Scope 3 on spreadsheets — accurate to about ±40%. CSRD limited assurance audit is 14 months away.
HOW THE ON-PREM STACK SOLVES IT
The Asset Telemetry (Jetson)
Connects to plant DCS, energy meters, and CMMS for direct Scope 1 and 2 data capture. ERP procurement integration runs over internal API only — supplier records never leave the perimeter. ITAR boundary preserved by design.
The Carbon Engine (RTX)
Runs GHG Protocol-compliant calculations for all 15 Scope 3 categories. Maps procurement line items to the 70K+ on-prem emission factor library. Activity-based methodology produces ±4% accuracy without sending any supplier data off-site.
CSRD Audit Pack (DGX)
Generates auditor-ready Scope 1/2/3 disclosure aligned with CSRD, ESRS E1, and GHG Protocol. Methodology documentation, emission factor versioning, and data lineage stored on-prem for the assurance auditor. ITAR programs remain compartmentalized.
THE RESULT
Scope 3 accuracy ±40% → ±4%. CSRD limited assurance passed clean. ITAR compliance preserved.
SCENARIO 02
"We are a CPG manufacturer with 14 plants. Our cloud ESG platform shows Scope 3 at 2.4M tCO₂e. The third-party assurance auditor flagged it as unverifiable. How do we get to a defensible number?"
THE PROBLEM
Top-100 consumer packaged goods company. 14 plants across 6 countries. Existing cloud carbon platform produced a Scope 3 figure of 2.4M tCO₂e built mostly from spend-based industry-average factors. Third-party assurance auditor reviewed the methodology and concluded the underlying data did not support a verifiable opinion — the spend-based proxy for purchased goods (Category 1) was 78% of the total Scope 3 number and had no supplier-specific corroboration. CSRD reasonable assurance deadline is 2028; the current path will fail it.
HOW THE ON-PREM STACK SOLVES IT
The Asset Telemetry (Jetson)
Each plant gets an edge unit that pulls hourly Scope 1 fuel burn, Scope 2 grid mix at the regional level, plus procurement records from the local ERP. 14 plants stream summarized footprint data into the corporate Carbon Engine.
The Carbon Engine (RTX)
Replaces spend-based Category 1 calculation with activity-based — actual material quantities (kg of glass, kg of PET, liters of palm oil) × supplier-specific or country-specific emission factors. Spend-based fallback only where activity data is unavailable, transparently flagged in the audit pack.
Scope 3 Models (DGX)
Trains supplier-specific factors as suppliers provide actual data. Models residual emissions and beyond-value-chain mitigation per the latest SBTi Corporate Net-Zero Standard v2. Auditor sees activity-based math with documented uncertainty bounds.
THE RESULT
Scope 3 number revised from 2.4M to 2.1M tCO₂e on activity data. Auditor opinion: verifiable. Reasonable assurance trajectory secured.
Frequently Asked Questions
The most common questions sustainability officers, ESG controllers, and CIOs ask when evaluating sovereign AI for carbon footprint tracking. Sign up free to start the sovereign-carbon evaluation.
What makes sovereign AI different from cloud-hosted ESG platforms?
Cloud ESG platforms send activity data, supplier records, and emission calculations to multi-tenant hyperscaler infrastructure. Sovereign AI runs the same GHG Protocol-compliant calculations on NVIDIA hardware inside your own data center or plant control room. The emission factor library, the supplier-specific factors, the Scope 3 calculation engine — all on-prem. Zero egress. Critical for ITAR programs, defense contracts with data residency clauses, EU member-state sovereignty requirements, and any organization treating supplier contracts as competitive intelligence.
Does sovereign AI sacrifice accuracy compared to cloud platforms?
The opposite — sovereign AI typically improves accuracy because activity data integration is tighter. The Jetson Asset Telemetry pulls live readings from energy meters, CMMS, and ERP procurement directly. The 70,000+ emission factor library from DEFRA, US EPA, Exiobase, and IEA is maintained on-prem with versioned updates. Scope 3 moves from spend-based proxy (±35% error) to activity-based calculation (±4% error). The audit trail is complete and reproducible.
Does the platform align with CSRD, SB-253, SEC, and SBTi requirements?
Yes. Underlying methodology is GHG Protocol Corporate Standard for Scope 1 and 2, GHG Protocol Scope 3 Value Chain Standard for all 15 categories, and ISO 14064-1 for organizational inventory boundaries. Output formats map directly to CSRD ESRS E1 disclosures, California SB-253 schedules, SEC climate disclosure rules, and SBTi Corporate Net-Zero Standard v2 target-setting. The audit pack includes methodology documentation, emission factor versioning, data lineage, and uncertainty bounds — everything a limited or reasonable assurance auditor needs.
How does Scope 3 actually improve from spend-based to activity-based?
Spend-based: total purchase value in dollars × industry-average emission factor per dollar. Highly approximate. Activity-based: actual quantities of materials, weights of shipments, kWh of supplier energy, ton-miles of freight × supplier-specific or product-specific emission factor. Sovereign AI matches each procurement line item to the most specific available factor — supplier-provided where supplier has disclosed, country-specific where the supplier location is known, industry-average only as last fallback. Each line item gets a quality score; the auditor sees exactly where activity data exists vs where proxies were used.
How fast can we get a sovereign-carbon pilot live?
Eight to twelve weeks from contract signature. Weeks 1-2 — site survey, ERP/CMMS/meter inventory, scope boundary definition. Weeks 3-4 — Jetson Asset Telemetry deployed, Scope 1 and 2 data flowing into the Carbon Engine, baseline established. Weeks 5-6 — Scope 3 procurement integration live, top 5 categories switched from spend-based to activity-based. Weeks 7-8 — supplier-specific factor onboarding for top 20 suppliers, CSRD audit pack template loaded, sustainability team trained. Weeks 9-12 — DGX Scope 3 model training on historical data, remaining categories migrated, second-plant rollout planning.
Sovereign Edition · Scope 1/2/3 · 8-Week Pilot
Keep the Carbon Data Inside Your Perimeter. Make the Numbers Audit-Defensible.
Book a 30-minute call with our sovereign-carbon deployment engineers. Walk through your three-scope footprint, your CSRD or SB-253 timeline, and your data sovereignty constraints. See AI-native carbon accounting live against your own activity data. Pilot in 8 to 12 weeks. Buy it once, own it forever — no monthly fees, source code included.
