“How do we pay for this?” is the question that stalls most AI projects. The good news is there are several real levers in 2026; the catch is that the right one depends entirely on what you’re doing. This playbook matches the lever to the situation, with official sources. (dgm is an AI integration partner, not a lender or tax firm — orientation here, claims with your CPA.)

First, decide which kind of AI spend you have

Funding AI splits cleanly into two cases, and confusing them wastes time:

  • You’re building or customizing AI/software — writing code, building agents, integrating systems, training or fine-tuning models. This is where the strongest levers live.
  • You’re buying and rolling out existing AI — subscriptions, off-the-shelf tools, seats. There is no broad grant for this; it comes from operating budget, a loan, or the savings it creates.

Most “where’s the grant?” frustration comes from being in the second case but hunting for first-case money. Be honest about which you’re in.

If you build or customize AI: start with tax

The tax code is the most valuable lever for builders in 2026. Under new Section 174A, domestic research expenses are again immediately deductible for tax years beginning after December 31, 2024, and the Section 41 R&D credit reduces tax dollar for dollar on qualifying work — which, per the IRS four-part test, expressly includes computer science. Even early-stage companies can benefit: qualified small businesses can apply the credit against payroll taxes, with the cap raised to $500,000 per year for tax years beginning after December 31, 2025. This is usually a bigger, more reliable lever than chasing a grant.

If you do genuine R&D: pursue non-dilutive grants

If your work is true research and development, real grants exist:

  • SBIR and STTR (America’s Seed Fund) fund small-business R&D across 11 agencies, with Phase I awards up to about $323,090 and Phase II up to about $2,153,927 (inflation-indexed — verify the current figures).
  • NSF SBIR/STTR funds AI startups specifically, with Phase I awards up to about $305,000 and a dedicated AI topic.

These reward innovation, not adoption, and they’re competitive — budget real effort for the Project Pitch and proposal.

If you’re purchasing: loans and budget

To finance buying AI tools or AI-enabled equipment, SBA 7(a) loans are flexible debt that can cover technology — and the SBA is doubling its cumulative 7(a)+504 limit to $10 million effective July 4, 2026. The 504 program covers fixed assets, generally not software alone. Often, though, the cleanest “funding” for adoption is the savings it generates — which is why a tight ROI case (the kind dgm builds in an assessment) matters more than a grant search.

For guidance, not cash

Some of the most useful help isn’t money at all. NIST’s Manufacturing Extension Partnership helps manufacturers de-risk AI and automation adoption, and Small Business Development Centers provide free or low-cost advising. Where states help, it’s often a matching grant tied to a federal award — for example, New York’s NYSTAR matches federal SBIR/STTR awards up to $200,000.

A simple decision path

  1. Building/customizing AI? Talk to a CPA about Section 174A + the Section 41 credit first.
  2. Doing real AI R&D? Consider SBIR/STTR/NSF.
  3. Buying tools? Fund from budget or an SBA 7(a) loan, and build the ROI case.
  4. Want guidance? Use MEP (manufacturers) or an SBDC.
  5. In NY/CA and winning federal awards? Check state matching programs.

How dgm helps

dgm implements osFoundry and other AI for US businesses. On funding, we help you identify which lever realistically applies, scope and document AI development so it’s better positioned to qualify for R&D treatment, and build the ROI case that often funds adoption on its own. The tax filing belongs with your CPA; the implementation is ours.