Microsoft Sets the Bar for US Forest Management Credits

At a glance

The news

What Microsoft and Relae published

On 8 October 2026, Relae (formerly Carbon Direct), working with Microsoft, published quality criteria for improved forest management (IFM) credits on US industrial timberland. They go beyond what any registry requires today, and because Microsoft is the largest retiree of ACR IFM credits in our Canopy data, at 1.08 million, they work in practice as a procurement standard.

• Baselines from the landowner's own planning models, benchmarked against comparable forests and re-assessed periodically.
• Additionality shown by practice changes beyond normal year-to-year harvest variation.
• Leakage deductions above registry minimums where the harvest cut is large, screened by comparing leakable carbon with total benefits.
• Benefit-sharing scaled to a project's impact on timber-dependent workers and towns.

The reset was overdue. Older IFM methods credited the gap between a forest and an imagined worst case: profit-maximising harvest baselines, coarse regional averages and baselines frozen at year zero. CarbonPlan found about 29% of the California forest credits it analysed were over-credited, and a 2023 study found enrolled forests held 127% more carbon than regional averages, which points to land that was never going to be logged hard.

The rules

How the registries and regulators compare

Most major programmes are moving away from frozen baselines, by different routes. Verra uses matched control plots, ACR re-checks a modelled baseline before each issuance, and Canada's federal protocol takes whichever of two baselines stores more carbon.

Even the newest methods are unproven. ICVCM recorded that VM0045 has not been tested empirically and that satellite data can saturate under full canopy, and ACR's comparable-properties tool still lets the developer choose which matched property sets the harvest cap. Relae asks for that comparable to be fully independent of the landowner.

Canopy data

288 Mt of IFM supply, little of it to the new standard

IFM has more unretired credits than any other project type in the voluntary market. In Canopy Market Watch data, 329.9 Mt of IFM credits have been issued and only 41.6 Mt retired, leaving 288 Mt unretired: 23% of the market's overhang from a type that is 13% of issuance.

Our supply and demand model shows why that stock matters. In the base case, demand for nature-based removals rises from 13.5 Mt in 2026 to 23.3 Mt in 2030, while new nature-based supply grows only from 6 Mt to 8 Mt a year. The gap, about 154 Mt by 2034, has to come from credits already issued, and IFM's overhang could cover it almost twice over.

So the constraint is quality, not volume. ACR IFM issuance has risen every year since 2018, to 14.7 million credits in 2025, and three in four ACR IFM credits are still unretired. Most of that stock was issued before the first dynamic-baseline credits reached the market in 2025. If buyers apply the new criteria, the shortfall lands on a small pool of qualifying credits, and the legacy overhang becomes a pricing problem rather than a supply cushion.

Scope: voluntary market only; California and Washington compliance offsets are excluded. Registry data to May 2026; model base July 2026.

How we look at it

How Canopy screens an IFM project

The new criteria tell buyers what good looks like. Canopy's due-diligence engine is how we test a specific project against it. Every IFM project is scored on eight concerns, and the result is a Canopy integrity indicator, not a rating.

Three rules that change the answer

• A 2021 vintage floor. We only assess credits from 2021 vintages onward, and older inventory is kept out of what we recommend. On its own that screens out the oldest part of the IFM overhang.
• Leakage is a project number, not a typology. We compare the project's applied deduction with the band for its type and say whether it sits inside, below or above it. A general paragraph about leakage without the project's own percentage does not pass.
• The satellite check has to agree with the documents. For one Malaysian logged-to-protected IFM project, forest loss since 2015 was 1.46% inside the boundary against 4.71% in a 10 km ring around it, about three times lower. That independently supported the additionality case in its documents. We can run the same check on about 84% of Verra IFM projects.

Seven of the eight concerns set the result: Solid needs at least four Solid scores and no Watch on the concerns that matter most for that project type; a single Weak score drops a project to Watch at best. Concern 8, the rater view, is shown but never counted, so third-party ratings inform the read without deciding it.

Mapped against the Microsoft–Relae criteria, the overlap is on baselines, additionality, leakage and benefit-sharing. Our framework adds the checks a buyer is exposed to after purchase: permanence and country risk, Article 6 status, and reputation.

Beyond the baseline

Leakage, permanence and fire

Leakage: size it to the project

The right deduction depends on how much of a project's benefit comes from harvest it did not do. In Relae's examples, a 60% harvest cut on Pacific Northwest Douglas-fir puts 78% of credits at leakage risk, where a 30% deduction is not conservative; a two-year rotation extension on Southeast loblolly puts only 29% at risk. Standard-setters and regulators agree: ICVCM told CAR to lift its 20% Mexico default, and Washington doubled its US default to 40%.

Permanence: 100 years or tonne-years

Every protocol here sells a tonne as permanent and backs it with 100-year monitoring or a buffer. Tonne-year accounting instead credits carbon for how long it is stored, but the result swings with the method: over 100 years, the stricter Lashof approach needs about 128 tonnes stored for a year to equal one tonne emitted. It is honest about temporary storage only if the credit is sold as temporary, never as a permanent tonne.

Fire: manage for the forest that survives

Fuel treatment changes how much carbon survives a fire. Across 216 thinning projects in California's Sierra Nevada, treatments cut high-severity fire by 88%, and treated stands held more live carbon through the 2020–21 megafires. Static baselines penalise that thinning. Washington’s watershed-level fire pricing, and the buffer discounts for fuel work in Washington and Canada’s federal protocol, reward it.

What to watch

What comes next

• Whether Microsoft writes the criteria into its contracts and RFPs, and whether other large tech buyers follow.
• Whether older ACR and CAR IFM methodologies win CCP approval, and on what baseline basis.
• How much the leakage screen cuts volumes in practice. The paper gives a method, not a number.
• Price. No premium for credits that meet the criteria has been disclosed yet.

If you hold or are being offered IFM credits, we can run them through the eight Canopy questions. See how Canopy works →

Sources