Big Tech AI carbon credits have become a central topic in the voluntary market because AI adoption is driving up electricity use and the operational emissions of data centers. When renewables and Scope 2 tools cannot keep pace, many companies increase their focus on offsets and, above all, on higher-quality credits.

Why is AI driving up data center emissions and pushing Big Tech to buy carbon credits?

The key issue is electricity. Google reports that in 2024 data center electricity consumption grew +27% year on year versus 2023, due to business growth and the adoption of products “including AI” (sustainability.google). When loads rise that fast, the “physical” decarbonization of the grid does not always keep up.

Emissions reflect this. In its 2024 environmental report, Google states that 2023 emissions increased +13% YoY, reaching 14.3 MtCO₂e, driven by data center consumption and the supply chain (blog.google). This creates a mismatch between net-zero roadmaps and operational reality: energy demand rises immediately, while infrastructure, grids, and new renewable capacity take time.

The phenomenon is not isolated. A UN/ITU analysis cited by the media points to an average increase in the operational emissions of major tech groups of around +150% between 2020 and 2023, linked to AI and data centers (Al Jazeera). It is a useful figure for understanding why, in the short term, the use of credits in the VCM tends to grow.

Even “system-level” estimates point in the same direction. An academic projection on arXiv estimates data center electricity demand rising from about 415 TWh in 2024 to about 945 TWh in 2030, with AI contributing disproportionately (arXiv). If these orders of magnitude are close to reality, pressure on energy procurement and carbon strategy will remain high for years.

This is where the procurement angle comes in. When PPAs and certificates (RECs) are not enough due to grid constraints, permitting timelines, or where loads are located, many companies turn to carbon credits to “neutralize” a residual share and support claims. Reputational risk increases if offsetting becomes a substitute for real reductions rather than a complement.

What types of credits do Big Tech companies prefer (avoidance vs removal), and how are quality criteria changing?

The distinction matters more than ever. “Avoidance” credits generally represent emissions avoided or reduced versus a baseline (typical VCM examples include REDD+, renewables, cookstoves). “Removal” credits, by contrast, represent measured CO₂ removals, with one central issue: durability.

Demand is pushing toward removals. A clear signal is the use of multi-year offtake agreements: Microsoft, for example, has a 3.7 Mt agreement over 12 years with CO280, linked to removals in pulp & paper facilities (Axios). This type of contract indicates a preference for future supply and for projects that can scale.

For “engineered” removals with higher durability, prices can be much higher. Tom’s Hardware reports deals with costs even around ~$350/t for some solutions, with a focus on multi-million-ton volumes (Tom’s Hardware). This is not a “new market price” for everything, but it shows how far the premium segment can diverge from spot.

Quality, meanwhile, is becoming more formalized. The ICVCM defines the Core Carbon Principles (CCP) as an integrity threshold for credits in the VCM, with criteria on governance, additionality, MRV, leakage, permanence, and avoiding double counting (ICVCM). For buyers, this translates into tougher requirements in contracts: MRV clauses, buffers or insurance, transparency on baseline and vintage, and clarity on rights and claims (offsetting vs contribution).

One important detail for procurement teams: the CCP do not say that all credits with an ICVCM label are equivalent, but they define a shared minimum threshold. Due diligence on the specific methodology, the registry, and the developer is still necessary to distinguish quality within the same standard.

How are prices, contracts (forward/offtake), and credit availability evolving with Big Tech’s large-scale entry?

The dominant mechanism is “near-industrial” procurement. Big Tech companies tend to prefer forward/offtake contracts to lock in future supply (often 5–12 years), reduce the risk of non-availability, and finance CAPEX for removal projects. This is a shift away from spot shopping for generic credits.

Price benchmarks help interpret the market. Sustainable Atlas reports typical VCM ranges: nature-based avoidance ~$2–$25/t, nature-based removals ~$15–$50/t, engineered removals (DAC) ~$200–$600+ (Sustainable Atlas). These are ranges, not price lists, but they explain why portfolio composition can change total spend dramatically.

The market, however, is not moving in one direction. MSCI notes average prices falling in 2024 versus 2023, with wide dispersion and premiums for nature restoration and carbon-engineering, often via forwards (MSCI). In practice: “average down,” but “premium up” for the most sought-after segments.

One example of aggregated demand is Frontier, a coalition of tech buyers using legally binding offtakes. Carbon Pulse reports for 2025 $261M for 688,300 t (Carbon Pulse). This type of structure can accelerate the project pipeline, but it also shifts bargaining power toward those buying large volumes.

The impact on availability is real. A “barbell” market emerges: lots of low-quality supply at low prices and limited high-quality supply at high prices. For SMEs and late-adopting sectors, the risk of crowding-out from top-tier projects increases. This is where strategies such as multi-technology portfolios, phased procurement, delivery options, and make-up clauses in case of under-delivery become sensible.

Greenwashing risk arises when credits cover a structural increase in consumption. If AI increases data center loads and a company uses offsets to maintain net-zero claims, credibility depends on how “residual” the offsetting is compared to real reductions. This is also where the distinction between Scope 2 market-based and location-based matters: accounting tools and certificates may not reflect actual reductions on the local grid where the data center consumes power.

Double counting is the other major issue. A reduction can be claimed both by the company and by the host country in its NDC. In the CORSIA context, IATA highlights the concept of a corresponding adjustment as a practice to avoid counting the same reduction twice (IATA). Even for non-aviation buyers, it is a useful reference for raising the bar.

CORSIA also functions as an integrity stress test. ICAO publishes lists of CORSIA-Eligible Emissions Units and eligibility criteria (ICAO). For many companies, choosing credits that can withstand similar criteria means reducing the risk of challenges and improving traceability.

Practical checklist for solid projects, aligned with ICVCM CCP (ICVCM):

  • Additionality and a credible baseline
  • Independent, verifiable MRV
  • Leakage management
  • Permanence with a buffer pool or insurance
  • Clear rights over environmental attributes and claims
  • Transparency on communities and co-benefits
  • Developer and registry track record
  • Alignment with CCP where applicable

What impact will new rules and regulated markets (ETS2, CORSIA) have on Big Tech strategies in the voluntary market?

ETS2 is a directional signal, even though it does not directly cover data centers. The European Commission describes ETS2 for buildings, road transport, and additional sectors, designed to start in 2027, with a possible delay to 2028 depending on energy price conditions; additionally, in 2026 the EU Council supported measures for a “smoother” start in 2028 (European Commission). For the VCM, the point is the indirect effect: more regulatory attention on prices, impacts, and governance. (For context, ETS2 is an EU policy instrument and applies within the European Union.)

CORSIA raises the bar for credits usable for compliance. ICAO notes that the TAB reviews eligibility for future periods, including a reassessment in 2025 for compliance 2027–2029 (ICAO TAB). Even if a Big Tech company is not buying for CORSIA, the tightening creates a “halo effect”: more demands for MRV, additionality, and no double counting even in the voluntary market.

In Europe, the removals framework also matters. The European Commission notes the adoption of the Carbon Removals and Carbon Farming Certification Framework (CRCF) (December 2024) and the start of transparency and audit rules via Implementing Regulation 2025/2358 (European Commission). This could shift preferences toward certified, standardized removals, especially for those seeking more defensible disclosures. (For context, CRCF is an EU framework intended to standardize how carbon removals are certified across Member States.)

The procurement strategy that emerges is often “dual track”: VCM for voluntary claims today, but with processes and contracts already prepared for future assurance needs. This includes choices of registries and methodologies, and internal governance that involves marketing, legal, and investor relations.

Tokenization and carbon credits: can it improve transparency and traceability, or does it add new risks for buyers and investors?

Tokenization can help if it solves real traceability problems. A blockchain layer can provide an audit trail, fractionalization, and faster settlement. In enterprise contexts, a permissioned blockchain often makes more sense than a public chain, for controls and compliance.

The main risk remains the link to registries. Toucan openly discusses the issue of double issuance and double counting if the bridge between traditional registries and tokens is not governed by strong rules, such as locking and burning, one-way bridging, and controls on unbridging (Toucan). If the token “lives” without rigorous anchoring to the original credit, confusion increases around rights and claims.

Practical due diligence on platforms/tokens for buyers and investors:

  • Proof of lock or retirement on the registry, with verifiable evidence
  • Unique identifiers and clear serial mapping
  • Governance for custody and KYC/AML controls
  • Explicit rules on rights and claims linked to the token
  • Disclosure on fees and operational risks

The EU trend on removals suggests a simple principle. With CRCF and audit rules, tokenization works only if it remains anchored to recognized certifications and robust verification (European Commission). A token does not “create quality”: quality remains in methodology, MRV, and additionality.

In practice, tokenization can be useful in supply chains with many actors, where chain-of-custody and ESG reporting are needed. But for those buying Big Tech AI carbon credits, the rule is always the same: first assess the integrity of the credit, then decide the format.