For most of this industry's history, Section 280E foreclosed the federal R&D credit. In April 2026 state-licensed medical cannabis moved to Schedule III and left 280E behind. Medical licensees, ancillary businesses, and hemp operators can claim the federal credit now. Adult-use plant-touching operators remain on Schedule I, and for them the question is which states run a credit that 280E does not reach.
Every claim on this page is cited and dated. Verified as of August 29, 2026.
Nearly every article on this topic asks whether a state decoupled from Section 280E. That is not the question that decides it. Section 280E is a federal income tax provision. It disallows deductions and credits in computing federal taxable income, and it has no independent operation on a credit created by state statute. What decides the answer is how the state built its own credit.
Vermont decoupled from 280E and still delivers nothing to an adult-use operator, because its credit is a fixed percentage of the federal credit and that figure is zero. Ohio never decoupled at all and holds one of the cleanest positions available, because its credit runs against a gross receipts tax where 280E has nothing to operate on. Two states, opposite decoupling status, opposite outcomes, and decoupling explained neither one.
Holding both a medical and an adult-use license is the common case, not the exception. It is an allocation question rather than a choice between the two, and it is worth getting the structure right before a claim is built rather than after.
Federal positions verified against the DEA regulatory actions record and the Federal Register. Verified August 29, 2026.
Section 280E disallows any deduction or credit for amounts paid or incurred in carrying on a trade or business that consists of trafficking in controlled substances within the meaning of Schedule I or II of the Controlled Substances Act.
Credits are named in the statute, not implied. This is why the R&D credit was foreclosed rather than merely reduced, and why a plant-touching operator on Schedule I gets nothing federally regardless of how strong the research is. Cost of goods sold remains available, which is the only reason these businesses can file at all.
The practical result has been effective federal rates commonly exceeding 70 percent for retail-heavy operators.
The order moved FDA-approved marijuana products and products regulated under a qualifying state medical marijuana license into Schedule III. Because 280E reaches Schedule I and II only, it stopped applying to those operators on the effective date.
Everything else stayed where it was. Adult-use cannabis, and medical cannabis outside a licensed state program, remain in Schedule I and remain fully subject to 280E. The broader rescheduling question is still open and is being litigated on two tracks at once.
Ancillary businesses and compliant hemp were never in scope and are unaffected by any of this.
The April order encourages the Treasury Department to consider relief from Section 280E for prior years in which an operator held a state medical license. That language is precatory. It does not grant relief, no guidance has issued, and anyone promising a refund for closed years is selling something we would not sign.
The defensible posture is to preserve the position rather than to claim it: identify which prior years would be in scope if relief arrives, document the qualifying research for those years while the records still exist, and be ready to file if and when Treasury acts. That costs very little now and is impossible to reconstruct later.
Cultivation, extraction, and formulation work in this industry tends to pass the test more cleanly than operators expect, because the work is genuinely experimental and the results are genuinely uncertain. What fails is production running on an established protocol, however skilled the operator.
Running a proven cultivar through a proven environmental recipe is production. Running four cultivars against a controlled baseline to resolve why one of them keeps expressing inconsistent terpene profiles is research. The plants look identical. The tax treatment is not.
The work must aim to develop a new or improved business component: a product, process, formula, technique, invention, or software. Improvement can be in function, performance, reliability, or quality.
A processor developing a new extraction method to raise cannabinoid recovery while lowering residual solvent has a permitted purpose on two counts: the process itself is the business component, and the improvement is in both performance and quality. A cultivator selecting for a stable phenotype with a target cannabinoid ratio is improving a product.
This is the least restrictive of the four parts. Cost reduction, throughput improvement, and yield improvement all count as improvements in performance.
The work must fundamentally rely on principles of the physical or biological sciences, engineering, or computer science. It does not require a laboratory, a scientist, or a formal research department.
Phenotype selection relies on plant genetics. Extraction relies on chemistry and separation science. Emulsion stability in a beverage relies on colloid science and physical chemistry. Analytical method development relies on chromatography. This part of the test is rarely the obstacle in this industry, because the underlying work is unavoidably scientific.
Market research, consumer preference testing, and brand development do not rely on the hard sciences and fail here.
At the outset, the operator must have been uncertain about capability, method, or appropriate design. Uncertainty about whether the market will accept the result does not count. Uncertainty about whether it can be achieved does.
A manufacturer that cannot hold a nano-emulsion stable through a twelve-month shelf life faces method uncertainty: it is not known which surfactant system, particle size, or process order will achieve it. A cultivator seeing unexplained variance in cannabinoid expression across identical rooms faces design uncertainty.
This is where most cannabis claims are won or lost. The uncertainty must be documented at the start, not reconstructed after a good result.
Substantially all of the activity must be a systematic process of evaluating alternatives: forming a hypothesis, testing it, analyzing the result, and refining. Trial and error qualifies when it is designed rather than incidental.
Running three solvent ratios at four temperatures across replicated batches, measuring recovery and purity for each, and narrowing on the basis of the results is a process of experimentation. Adjusting a parameter once because the last run was disappointing is not.
This is the part that seed-to-sale tracking systems document almost perfectly, and almost entirely by accident. See the documentation section below.
For the general framework applied across every industry, see our R&D tax credit overview.
A vertically integrated medical operator ran a two-year phenotype selection program across nine genetics, hunting for a stable high-CBD chemovar with a consistent terpene profile for a patient population that was not responding well to available products. Rooms were matched, environmental setpoints held constant, and every cutting was tagged and tracked through harvest and lab result. Seven of the nine were discarded.
The operator had never considered the credit, because their CPA correctly told them in 2023 that 280E foreclosed it. After April 2026 that answer changed, and the work is squarely within the four-part test: an unknown at the outset, a systematic comparison of alternatives, plant genetics as the technological basis, and an improved product as the purpose. The seven discarded phenotypes are not a failure in the record. They are the evidence that alternatives were evaluated.
An adult-use extractor could not get residual solvent below the action limit on a live resin line without stripping the terpene profile that made the product worth selling. Over eight months they ran staged trials across three solvent systems, four purge temperature profiles, and two vacuum schedules, measuring recovery, residual solvent, and terpene retention on every batch.
Federally, 280E disallows the credit and will continue to while adult-use remains Schedule I. The state analysis is a different question entirely. In a state whose credit is computed on state qualified research expenses rather than as a percentage of the federal credit, the federal disallowance does not reach it. Several such states exist. In two of them the operator would have had to file an application by a fixed date to claim anything, and that date does not move.
An accredited laboratory developed a sample preparation and chromatographic method for cannabinoid quantification in high-fat edibles, where the matrix was causing interference that existing published methods did not resolve. The development ran across four extraction approaches and two column chemistries before landing on a method that held through validation.
The laboratory is ancillary, so 280E has never applied and the federal credit was available throughout. The difficulty is not eligibility, it is segregation. The same analysts, in the same room, on the same instruments, spend most of their time running compliance panels that are explicitly excluded as quality control. A defensible claim depends on separating method development hours from production testing hours with a contemporaneous basis, and a claim that does not do that carefully is a liability rather than an asset.
A manufacturer of closed-loop extraction systems spent a year developing a control system that holds pressure and temperature within a much tighter band during a run, which required new sensor placement, a rewritten control algorithm, and several failed thermal management approaches before one held.
The company had never claimed the credit because their accountant assumed that selling into cannabis carried the cannabis tax problem with it. It does not. Section 280E reaches businesses that traffic in a controlled substance. An equipment manufacturer does not, has never been in scope, and was eligible for the federal credit for every year they did not claim it. The look-back on the three open years was the largest component of the engagement.
Every state-licensed operator runs a seed-to-sale tracking system because the state requires it. METRC, BioTrack, and their equivalents produce a timestamped, tamper-evident, per-plant and per-batch record of what was grown, under what conditions, from what genetics, processed how, and with what measured result. Nobody built that for tax purposes. It is, structurally, exactly what an examiner asks for.
The weakest point in most R&D credit claims is that documentation was assembled after the fact. Seed-to-sale entries are made at the time of the event because the state requires it, and they cannot be backdated. That is the standard an examiner applies, met without anyone trying.
A process of experimentation has to show which alternative was tried, under what conditions, and what resulted. Seed-to-sale ties a specific plant tag or batch to its genetics, its room, its inputs, and its lab results. The experimental structure is already in the data.
These records are subject to state regulatory inspection and reconciliation. A record that a state cannabis regulator has already examined carries a weight that an internal spreadsheet does not, and it is difficult to characterize as reconstructed.
The gap is not evidence. It is framing. Seed-to-sale data records what happened and a credit claim needs to show what was being tested and why. That layer is thin: a trial design note at the outset, a statement of the uncertainty, and a mapping from plant tags or batch identifiers to the question being asked. We build that layer on top of records you already keep, rather than asking you to start keeping new ones. Operators who begin recording the intent alongside the compliance data are in a materially stronger position within one cycle.
Section 280E is a federal income tax provision with no operation on a state-created credit. Vermont decoupled and delivers nothing, because its credit is a share of a federal credit worth zero. Ohio never decoupled and is one of the cleanest, because its credit runs against a gross receipts tax.
New Jersey, Massachusetts, and Connecticut all run their research credits against a corporate tax. A cannabis LLC with individual owners cannot use those credits at all, regardless of how strong the research is, and in New Jersey that is true even though the state decoupled from Section 280E for both its corporate and personal taxes. Most operators in this industry are pass-throughs, so this is not an edge case.
Seven jurisdictions gate the credit on an application filed by a fixed date, and New York gates it on program admission before any filing exists. In five of the seven a missed date cannot be recovered by amended return. Those windows recur annually and do not track your tax return, so a timely return does not protect the credit.
| Jurisdiction | Grade | Credit runs against | Reaches pass-throughs | Filing deadline |
|---|---|---|---|---|
| Pennsylvania | Confirmed | Corporate net income tax and personal income tax | Yes | December 1 |
|
The Department of Revenue states that the R&D credit is available to businesses and individuals subject to either corporate net income tax or personal income tax, and its published guidance confirms that a medical marijuana company may apply where the activity qualifies under IRC Section 41. Keystone Innovation Zone credits are separately available and may be sold.
On the deduction side, Pennsylvania splits three ways and the Department says so plainly. For personal income tax, deductions are allowable for all ordinary, reasonable and necessary business expenses, because Pennsylvania computes business income under its own standard rather than starting from federal taxable income. For corporate net income tax, the calculation does begin with federal taxable income, and Act 56 of 2024 allows the deduction only for businesses holding an active grower or processor permit. The Department states there are currently no adjustments for medical marijuana dispensaries, so a dispensary-only corporation carries the federal treatment through to its Pennsylvania return. Credits are sellable. A first application requires two prior years of Pennsylvania research expenses, the application window runs August 1 to December 1 with no extensions, and the Department screens tax compliance for the entity and every 20 percent owner before awarding. PA Department of Revenue published FAQ, updated June 2025. Article XVII-B, Tax Reform Code. 61 Pa. Code §9.17. Verified August 29, 2026.
| ||||
| New Jersey | Confirmed | Corporation business tax only | No, not to individuals | None |
|
The most direct statement any state tax authority has made on this question. The Division of Taxation's technical bulletin provides that a registered cannabis licensee claiming a qualified research expense as a deduction may also claim that expense for the New Jersey R&D credit on Form 306, even though those expenses were disallowed for the federal research and development credit. A detailed rider must accompany the computation. Note the entity limitation: the 2023 decoupling covers both the corporation business tax and the gross income tax, but the R&D credit itself runs against the corporation business tax only. An LLC or partnership with individual owners gets the deduction relief and cannot claim the credit. Prospective from January 2023, with no look-back.
P.L. 2023 c. 50 (A3946). Division of Taxation TB-106, July 2023. Verified August 29, 2026.
| ||||
| California | Confirmed | Corporate franchise tax and personal income tax | Yes | None |
|
The Franchise Tax Board told the Legislature that cannabis taxpayers have been able to claim deductions and credits under the Corporation Tax Law since Proposition 64. That statement has been public since 2022 and is largely unremarked. Separate relief covering pass-throughs was extended in 2024 and now runs through tax year 2029. Two cannabis-specific state credits also exist alongside the research credit and are worth reviewing on the same return.
FTB reports to the Legislature, December 2022 and July 2024. R&TC §17209 as amended by SB 167, Stats. 2024 ch. 34. R&TC §§23609 and 17052.12. Verified August 29, 2026.
| ||||
| Colorado | Confirmed | State income tax, enterprise zones only | Yes | Before research begins |
|
Colorado barred any business not legal under both state and federal law from claiming enterprise zone credits, which excluded cannabis entirely and had nothing to do with 280E. A 2024 act removed that bar for state-licensed marijuana businesses for tax years beginning in 2025. The credit is available only for research conducted inside a designated enterprise zone, requires the same zone for three years, and pre-certification must be approved before the research begins. See the risk note below this table.
Colo. Sess. Laws 2024 ch. 410 (SB24-076), amending C.R.S. §39-30-103(8). C.R.S. §39-30-105.5. Verified August 29, 2026.
| ||||
| New Mexico | Strong | Gross receipts, compensating, or withholding tax | Yes | Within 1 year of period end |
|
The cleanest structural position of the eighteen. The basic credit is a flat percentage of qualified expenditures with no base amount to exceed, doubled in rural areas, and it runs against gross receipts, compensating, or withholding tax rather than income tax. New Mexico also supplies its own statutory definition of qualified research instead of importing IRC Section 41, so the credit is independent of federal law at every level. A separate additional credit does run against income tax and requires payroll growth.
NMSA §§7-9F-1 to 7-9F-13, particularly §7-9F-5 and §7-9F-3(D). Verified August 29, 2026.
| ||||
| Michigan | Strong | Corporate income tax, or employer withholding tax | Yes, at entity level | March 15 tentative claim |
|
A new credit effective for tax years beginning in 2025, and it is refundable, which is rare. Flow-through entities do not claim it against income tax at all. They claim it against employer withholding tax on the annual withholding return, at the entity level rather than passing it to owners. Computed on Michigan research expenses against a three-year base, with an enhanced rate for employers under 250 people and an additional component for work with a Michigan research university. The tentative claim deadline is hard and Treasury does not accept late filings. One structural limit: a disregarded entity is not an eligible flow-through entity, so a single-member LLC that has not elected corporate treatment is outside this credit, and the entity must have Michigan withholding obligations to have anything to claim it against.
Public Acts 186 and 187 of 2024 (HB 5100, HB 5101). MCL 206.677 and 206.717. Treasury Revenue Administrative Bulletin 2026-4. Verified August 29, 2026.
| ||||
| Ohio | Strong | Commercial Activity Tax, a gross receipts tax | CAT payers only | None, self-certifying |
|
Ohio has no corporate income tax and never decoupled from 280E, because there was nothing to decouple. The statute contains no requirement that a federal credit be allowed; it imports the Section 41 definition of qualified research expenses and nothing else. No application, no cap, claimed directly on the return. The best legal position on this page and the smallest dollars: the credit is nonrefundable and only useful against actual CAT liability, and the CAT exclusion threshold means smaller operators owe none. Best suited to operators with substantial Ohio gross receipts.
ORC §5751.51, effective October 3, 2023 (HB 33). Verified August 29, 2026.
| ||||
| Minnesota | Strong | Income and franchise tax | Yes | None |
|
The subtraction statute names cannabis and hemp license holders explicitly, which is unusually direct and appears as a line item on the state's own return instructions. The credit is tiered on Minnesota research expenses over a base and reaches corporations, partnerships, S corporations, estates, trusts, and individuals. A portion became refundable by election starting in 2025, with the election required on a timely filed return.
Minn. Stat. §290.068 and §290.0134 subd. 19. Minnesota Revenue 2025 M4 instructions. Verified August 29, 2026.
| ||||
| Delaware | Strong | State income tax, refundable | Yes | September 15 |
|
Two elective methods, and the election is independent of the method used federally. The incremental method computes on Delaware expenses over a base at an enhanced rate for smaller businesses, and unused credit is paid out as a refund. A separate provision preserves the Delaware credit even if the federal Section 41 credit is terminated outright, which is strong support that it stands on its own. The alternative method is a share of the federal credit and is worth nothing to an adult-use operator, so the election matters.
30 Del. C. §§2070 to 2075, particularly §2070(c) and §2073. Form BUS-RDC. Verified August 29, 2026.
| ||||
| Rhode Island | Strong | Corporate, insurance, and personal income tax | Yes | None |
|
Decoupling took effect for tax years beginning in 2025 on both the corporate and the personal income tax sides, and it is implemented rather than merely enacted: it appears as a named line on Rhode Island's own modification schedule for individuals, corporations, and fiduciaries. The credit computes on Rhode Island expenses over a base period and runs against three separate tax chapters, so entity structure is not an obstacle here.
R.I.G.L. §44-32-3, §44-11-11(a)(1)(vii), §44-30-12(c)(13). 2025 RI Schedule M. Verified August 29, 2026.
| ||||
| Missouri | Strong | State income tax, transferable | Yes | Aug 1 to Sept 30 window |
|
Revived for tax years beginning in 2023 and sunsetting at the end of 2028. Computed on additional qualified research expenses over a three-year average, at an enhanced rate for work with a Missouri college or university, with a portion of the program cap set aside for small, minority, and women-owned businesses. Credits carry forward twelve years and are transferable. The Department of Economic Development's published ineligible list contains no federal-legality bar of the kind Colorado had until 2024.
RSMo §620.1039. Missouri DED Qualified Research Expense program guidelines. Verified August 29, 2026.
| ||||
| Maryland | Strong | State income tax, refundable for small business | Yes | November 15 |
|
Computed on Maryland expenses over a base and administered by the Department of Commerce rather than the Comptroller, with a statewide cap and proration if oversubscribed. Refundable for small businesses, defined by net book value of assets, a threshold most operators in this industry fall under. Reaches every entity type, including partnerships, LLCs, S corporations and C corporations. The program is authorized through January 2031, subject to extension.
Md. Code Tax-General. Maryland Department of Commerce R&D tax credit program guidance. Verified August 29, 2026.
| ||||
| Maine | Strong | State income tax | Yes | None |
|
Computed on Maine expenses over a three-year base with an additional component for basic research payments to universities and qualified research organizations. Reaches corporations and individuals, so pass-throughs work, and there is no application gate. Nonrefundable and capped relative to liability, with a long carryforward and no carryback, so the practical value is modest relative to the other Strong jurisdictions.
36 M.R.S. §5219-K. Verified August 29, 2026.
| ||||
| Illinois | Open | State income tax | Yes | None |
|
Architecture and entity reach are both good: computed on Illinois expenses over a three-year base, reaching corporations, S corporations, partnerships and LLCs with pro rata allocation to owners, and extended by Public Act 103-0595 through tax years ending on or before December 31, 2031, up from a prior 2027 sunset. Legislation to remove the sunset entirely is pending. It lands on Open for a drafting reason. The administrative rule defines qualifying expenditures as those "allowable under" IRC Section 41, which gives an examiner a textual hook that New Jersey and Massachusetts do not. We read that as importing definitions rather than outcomes. We would rather flag the ambiguity than count Illinois and hope nobody looks.
35 ILCS 5/201(k). 86 Ill. Adm. Code 100.2160. Verified August 29, 2026.
| ||||
| Massachusetts | Entity split | Corporate excise only | No, not to individuals | None |
|
Decoupled from 280E for licensed marijuana businesses on both the personal and corporate sides, confirmed by a Department of Revenue release. But the credit itself is available only to corporations taxable under the corporate excise. An S corporation can use it at entity level, including against the non-income measure, though it does not pass to shareholders. An LLC or partnership with individual owners cannot use it at all, because the regulation attributes the expenses to owners who are not subject to the corporate excise, and there is no personal income tax version of the credit.
M.G.L. c.63 §38M. 830 CMR 63.38M.1(3)(b), (3)(e), (3)(f), (7)(i). St. 2022 c.180. DOR TIR 23-6. Verified August 29, 2026.
| ||||
| Connecticut | No | Corporation business tax only | No | None |
|
Two separate research credits, both computed on Connecticut expenses, and both sitting in the corporation business tax chapter. Available to C corporations only. Connecticut operates a pass-through entity tax and has extended other credits to it, so the omission looks deliberate rather than accidental. Nothing here turns on 280E or on decoupling. For a C corporation the ordinary analysis applies and Connecticut is likely workable.
C.G.S. §§12-217j and 12-217n, Chapter 208. Verified August 29, 2026.
| ||||
| Vermont | No | Income tax, as a share of the federal credit | Irrelevant, credit is zero | None |
|
The single clearest demonstration that decoupling is the wrong question. Vermont appears on every published list of states that decoupled from Section 280E. Its credit statute sets the amount at a fixed percentage of the federal credit allowed in the taxable year, and the Department of Taxes states plainly that taking the federal credit is the predicate. For an adult-use operator the federal credit is zero, so the state credit is a percentage of zero. The decoupling is real and entirely irrelevant to the outcome.
32 V.S.A. §5930ii. Vermont Department of Taxes business tax credit guidance. Verified August 29, 2026.
| ||||
| New York | No | Franchise tax, as a share of the federal credit | Program-gated | Certificate required first |
|
New York has no broad-based research credit. The Excelsior program credit is a share of the federal credit attributable to New York, capped as a percentage of state research expenditures, so it fails on the same arithmetic as Vermont. It is also discretionary: it requires a certificate of eligibility from Empire State Development, operation in a designated strategic industry, and job creation or investment commitments. Cannabis is not among the listed strategic industries. A pending bill would change the computation for participants without a federal credit, and we are tracking it.
Excelsior Jobs Program statute and Empire State Development program materials. NY Assembly Bill A10247, status unconfirmed. Verified August 29, 2026.
| ||||
Colorado's administering agency has not updated its guidance. The legislature removed the federal-legality bar in 2024, but the economic development office that approves pre-certification still publishes the repealed rule on its program page. Pre-certification must be approved before research begins and cannot be backdated, so an operator who reads that page, concludes they are ineligible, and does not apply loses the year permanently. Confirm current administrative practice before relying on this.
Pennsylvania's Department of Revenue has recommended that the legislature require a federal credit as a condition of the state credit. The recommendation appears in the department's own annual report to the General Assembly. If it is enacted, Pennsylvania moves from the strongest position in this table to no position at all for plant-touching adult-use operators. Nothing has been introduced yet. This is the item on this page most likely to change.
Every jurisdiction above researched to statute, session law, regulation, or published agency guidance. No claim in this table rests on a secondary source. Verified August 29, 2026.
The standard R&D engagement is a look-back study: amend the open years and recover what was missed. In seven of the jurisdictions we track, that engagement does not exist. The credit is gated on a separate application filed by a fixed date, and missing it forfeits the year outright.
These windows recur every year, and they do not track your tax return. That is the part operators get wrong. Pennsylvania's window runs August 1 to December 1 for expenses in the tax year that ended the previous calendar year, with awards issued the following May. A Pennsylvania operator can file a perfectly timely return in April and have already forfeited the credit for the year that return covers, because the application gate closed on a different calendar entirely.
New York gates differently, and harder. There is no annual filing date, because there is no filing to make until Empire State Development admits you to the Excelsior program through the Consolidated Funding Application. That requires operating in a designated strategic industry, meeting job creation or investment commitments, and annual performance reporting to keep the certificate. Cannabis is not a listed strategic industry. Prior years cannot be captured retroactively, because the certificate is a precondition rather than a filing.
The remaining ten are claim-on-the-return jurisdictions with no separate application: Ohio, Minnesota, Maine, Rhode Island, Connecticut, Illinois, Massachusetts, New Jersey, California and Vermont. For those, ordinary amended return rules apply and a look-back is available. Seven gated on a date, one gated on program admission, ten open on the return.
A different question from the filing window, and it changes how far ahead a program is worth planning around. Some of these credits are permanent. One sunsets inside three years.
Two states went further and wrote in permanence deliberately. New Jersey and Delaware each provide that their state credit survives even if Congress repeals the federal credit outright. That is worth knowing for planning, and it also reinforces that both credits stand on their own rather than borrowing federal availability.
We have confirmed the expiration question to a primary source for eleven of the eighteen jurisdictions we track. Michigan, Rhode Island, Maine, Connecticut, Vermont, Massachusetts and Colorado are not among them. We would rather tell you that than assume a program is permanent because we did not find an end date.
A free assessment covers which credits your license type actually reaches, whether any filing deadline is close enough to change your sequencing, and whether the work qualifies at all. If it does not, we will tell you that and there is no engagement.
Book a Free AssessmentNo fee to find out whether you qualify.