Singapore's 2026 AI grant stack, explained for SME owners

If you run an SME in Singapore and you’ve been putting off AI work because of the price tag, 2026 is the year the maths changed. Between the Budget 2026 measures, the National AI Impact Programme and the bank-led schemes, the government has stacked more co-funding behind SME AI adoption than any market we operate in.

The catch is that the stack is genuinely confusing. Five different schemes, run by different agencies, with different windows, covering different costs. Owners either assume none of it applies to them, or assume all of it stacks neatly. Neither is true.

Here’s the plain-English version. One disclaimer before we start: we’re practitioners who implement AI systems, not grant consultants. Treat this as a map, not advice, and check the IMDA and EnterpriseSG sites for current terms before you spend anything, because criteria change and windows close.

Updated 6 October 2026. The biggest change since this was first published: EDG, PSG and MRA stopped taking new applications on 29 September 2026 and were replaced by a single scheme, the EDGE Grant, from 30 September. The section below has been rewritten for EDGE; anything you read elsewhere that still says “apply for PSG” is out of date.

The National AI Impact Programme (NAIIP)

This is the umbrella. Launched in March 2026 under NAIS 2.0 (Singapore’s second National AI Strategy), NAIIP sets the headline targets: 10,000 enterprises adopting AI and 100,000 “AI Bilingual” workers over three years. When a programme names a number like 10,000 enterprises, read it as intent: the agencies involved are measured on getting businesses like yours through the door.

Two pieces matter practically. First, the Digital Leaders Accelerator Bootcamp (DLAB), aimed at getting business leaders fluent enough to direct AI work rather than just approve budgets for it. Second, the pre-approved solutions list: IMDA is raising the AI-enabled share of pre-approved solutions from 30% to 50%, and pre-approved solutions come with grant support attached.

Who it’s for: broadly, locally registered enterprises, and the programme explicitly wants SMEs, not just the usual large-company suspects. F&B and hospitality businesses should pay particular attention; the sector keeps appearing by name in these schemes (more on that below).

Budget 2026: the 400% tax deduction

This is the one that makes accountants sit up. Under the Enterprise Innovation Scheme (EIS), Budget 2026 (announced February 2026) introduced a 400% tax deduction on qualifying AI investments, capped at S$50,000 of expenditure per Year of Assessment, for YA2027 and YA2028.

In English: spend S$18,000 on a qualifying AI implementation and you may be able to deduct S$72,000 from your taxable income. At Singapore’s 17% corporate rate, that’s worth up to roughly S$12,200 in tax saved, on an S$18,000 spend. Even allowing for the fine print, that is an unusually generous lever.

Who it’s for: businesses paying Singapore corporate tax, on qualifying expenditure, within the cap and the two-YA window. The qualifying criteria are exactly the sort of thing that changes between Budget announcement and IRAS implementation guidance, so this is the scheme where “talk to your accountant first” is not a throwaway line.

The honest caveat: YA2027–28 is a fixed window. If your AI project is perpetually next quarter’s problem, you can drift straight past the deduction without noticing.

The EDGE Grant (which replaced PSG and EDG)

EDGE is now the workhorse. From 30 September 2026 it folds the old Productivity Solutions Grant, Enterprise Development Grant and Market Readiness Assistance into one scheme covering eight business areas and over 150 activities. For an AI project the relevant area is Automation and Digitalisation, with the Business Strategy and the Innovation areas covering the consultancy-shaped work the EDG used to fund.

The numbers, from EnterpriseSG’s EDGE FAQ: SMEs get up to 50% support on most activities (70% only for internationalisation and sustainability), non-SMEs up to 30%. Each company has an annual grant cap of S$100,000 across all EDGE activities, resetting every 1 April, and within that a S$30,000 sub-cap for single-function digital solutions, integrated enterprise systems and pre-scoped equipment. Grants are paid on reimbursement after the project completes.

Two practical changes from the PSG days. Pre-approved vendors still exist for packaged solutions (single-function tools are listed on IMDA’s SMEs Go Digital platform), but you may now engage other vendors if the functions you need aren’t available pre-approved. And you may pay the vendor after submitting the application rather than waiting for approval, though submission does not guarantee approval, so treat that as a cash-flow convenience, not a green light.

One detail for anyone buying consultancy rather than software: the FAQ makes SS 680 certification a condition for Standards-adoption projects, but the application document lists for Business Strategic Planning and the Financial Management activities also ask for the consultant’s SS 680 certification, so assume a consultancy-only engagement needs a certified consultant. For which EDGE route fits an AI project, and the one gap to watch for, see our EDGE Grant guide for AI projects.

Who it’s for: any Singapore-registered business with at least 30% local shareholding. SMEs (group turnover of not more than S$100 million, or not more than 200 employees) get the higher rate. Applications go through the Business Grants Portal with Corppass. If you submitted a PSG or EDG application before 30 September 2026, it continues under the old scheme’s rules; nothing new can be lodged there.

SkillsFuture Enterprise Credit (SFEC)

Implementation without training is how companies end up with shelf-ware, so it’s sensible that the stack covers the people side too. SFEC offsets up to 90% of out-of-pocket training costs. Unused credits under the current SFEC expire on 30 November 2026; from 1 December 2026 the scheme is redesigned, with eligible employers (those with at least three resident employees) getting a fresh S$10,000 tranche.

Pair this with another Budget 2026 sweetener: employees who complete SkillsFuture AI courses get six months of premium AI tools free. The government is quite literally paying your team to learn the tools and then paying for the tools while they practise.

Who it’s for: eligible employers with SFEC balances. Many SMEs have credits sitting unused without knowing it. Worth a five-minute check before you budget a single training dollar.

The honest caveat: subsidised training is only worth what you do with it. Generic prompt-writing courses are now effectively a free commodity (the AI labs themselves give them away); spend your subsidised hours on training tied to a workflow you’re actually changing.

DBS Spark GenAI

The outlier in the stack: a bank scheme rather than a government one. DBS Spark GenAI, enhanced in April 2026 in partnership with EnterpriseSG and IMDA, offers grant support of up to 50% of eligible costs for SME AI adoption.

Its predecessor, the GenAI Sandbox, subsidised around 300 SMEs across retail, F&B, education and hospitality. That sector list is worth noticing: if you run a restaurant group, a café chain or a hotel F&B operation, you are not a marginal case for these schemes. You’re the named target market. We work primarily with hospitality businesses, so we’d say that, but the sector designation came from IMDA, not us.

Who it’s for: SMEs, with DBS banking relationships being the natural path in. Eligibility terms are the bank’s to set and revise.

What it means in dollars: a worked example

Take a realistic engagement: an S$18,000 implementation: say, a 3–4 week sprint that deploys one supervised agentic workflow (invoice processing, reporting assembly, booking-data hygiene) plus the staff training to run it.

Path one: EDGE route. If the project is approved under Automation and Digitalisation at the SME rate, 50% support takes your out-of-pocket from S$18,000 to S$9,000, reimbursed on completion. Neither the S$30,000 packaged-solution sub-cap nor the S$100,000 annual cap bites at this size.

Path two: EIS route. If the spend qualifies for the 400% deduction instead, you pay the full S$18,000 now but deduct S$72,000, worth up to about S$12,200 off your tax bill at 17%, a net effective cost somewhere near S$5,800, realised at tax time rather than at invoice.

The training layer. If S$3,000 of a project is formal training, SFEC at up to 90% can take your out-of-pocket on that slice to a few hundred dollars.

Can you stack all of this on the same dollar? Assume not until told otherwise. Grant-funded amounts and tax deductions on the same expenditure are precisely where double-claiming rules live, and the answer depends on rulings, not blog posts. The realistic takeaway is more modest and still striking: through one path or another, a well-structured S$18,000 project plausibly costs an eligible Singapore SME somewhere between a third and a half of sticker. The APAC backdrop suggests plenty of businesses have done this maths: IDC (April 2026) projects AI and GenAI spend across Asia/Pacific growing from US$73B in 2024 to US$370B by 2029.

The caveats, all in one place

Because a grants article without caveats is marketing:

• Windows close. The EIS deduction covers YA2027–28. The current SFEC’s credits run only to 30 November 2026, with a redesigned SFEC (a fresh S$10,000) from 1 December 2026. Programmes get revised at every Budget. None of this is permanent furniture.
• Criteria change. Pre-approved lists are living documents; eligibility definitions get tightened. Check IMDA (imda.gov.sg) and EnterpriseSG (enterprisesg.gov.sg) primary sources before committing money, not summaries like this one.
• We’re practitioners, not grant consultants. We build and run agentic systems; we can tell you what a project costs and what it returns. For eligibility rulings, stacking questions and applications, use your accountant or a grant specialist.
• A grant doesn’t make a bad project good. A poorly chosen AI project at 50% off is still a poorly chosen AI project. It just wastes public money alongside yours. The schemes reduce the cost of acting; they don’t reduce the need to pick the right workflow first.

Do this next week

Three checks, maybe ninety minutes total. First, look up your SFEC balance. You may be sitting on training credits that expire on 30 November 2026. Second, run your project through the BizSG grant recommender on the EnterpriseSG site and skim the pre-approved solutions on IMDA’s SMEs Go Digital for your sector, to see what’s already co-fundable under EDGE. Third, send your accountant a one-line email: “If we spend S$15–20K on a qualifying AI implementation this year, what does the EIS 400% deduction do to our YA2027 position?”

Then, and only then, start scoping what you’d actually build. The funding stack is the tailwind, not the destination.