How the New 400% AI Tax Deduction Changes the Economics of SME AI Projects
Artificial intelligence is becoming easier for SMEs to access, but deciding whether an AI project is financially worthwhile remains more complicated than choosing a software subscription.
Businesses need to consider implementation costs, consulting fees, data preparation, staff training, ongoing subscriptions and whether the technology will actually reduce costs or increase revenue.
From Year of Assessment (YA) 2027, Singapore businesses will also have a new tax incentive to include in that calculation.
Under the Enterprise Innovation Scheme (EIS), qualifying businesses can claim a 400% tax deduction on the first S$50,000 of qualifying AI expenditure per year for YA 2027 and YA 2028.
This can materially improve the economics of an eligible AI investment.
However, SMEs should understand an important distinction:
A 400% tax deduction does not mean the Government pays 400% of the AI project cost.
The benefit works through the company’s taxable income, and the EIS cash payout option does not apply to this new AI adoption category.
What Has Changed Under the Enterprise Innovation Scheme?
The Enterprise Innovation Scheme supports several types of innovation expenditure through enhanced tax deductions and allowances.
For YA 2027 and YA 2028, AI adoption becomes a new qualifying activity.
Businesses can claim a 400% tax deduction on the first S$50,000 of qualifying AI expenditure incurred for the purposes of their trade or business in each YA.
For qualifying expenditure that would ordinarily be deductible under Section 14 of the Income Tax Act, the 400% deduction generally consists of:
- the normal 100% base deduction; and
- an additional 300% enhanced deduction.
Qualifying expenditure above the first S$50,000 may continue to receive the normal 100% deduction where it satisfies the ordinary deduction rules.
IRAS also provides for the 400% deduction on the first S$50,000 of qualifying AI expenditure that would otherwise not be deductible under Section 14, subject to the prevailing qualifying conditions.
What AI Expenditure Can Qualify?
IRAS defines qualifying AI expenditure broadly around two areas:
- subscription to or licensing of an AI system from another person; and
- subscription to, acquisition of or licensing of qualifying AI business services.
An AI system can include capabilities such as:
- content generation;
- reasoning and problem solving;
- knowledge retrieval and representation;
- natural language processing;
- automated planning and optimisation; and
- multimodal processing such as image recognition or audio-to-text processing.
Qualifying AI business services can include services relating to the development, deployment, operation or maintenance of an AI system.
Examples can include:
- online AI platforms;
- system development;
- AI consultancy and strategy;
- data and analytics services;
- AI-related research and development;
- system engineering and compliance services; and
- system-related training.
This means the incentive can potentially cover more than simply purchasing access to an AI chatbot.
Depending on the project, implementation and professional-service expenditure may also form part of qualifying AI adoption costs.
What Does Not Qualify?
The incentive does not cover every cost associated with an AI project.
IRAS specifically excludes expenditure on physical infrastructure and hardware such as:
- servers;
- storage devices;
- computing equipment; and
- other physical infrastructure.
Qualifying AI expenditure also excludes expenditure that has been subsidised by a Government grant or subsidy.
This distinction is important when SMEs build the project budget.
A S$100,000 AI project does not automatically mean S$100,000 qualifies for the enhanced deduction.
Separate the AI Project Into Cost Categories
Before estimating the tax benefit, management should divide the project into its individual components.
For example, suppose an SME plans an AI implementation costing S$90,000:
- AI software licensing: S$25,000
- AI implementation consultancy: S$18,000
- Data analytics and integration services: S$12,000
- AI-related staff training: S$5,000
- New servers and computing hardware: S$30,000
Total project cost:
S$90,000
Based purely on these categories, the physical hardware portion would not qualify as AI expenditure under the new EIS category.
The remaining expenditure would need to be reviewed against the prevailing IRAS requirements to determine what is actually qualifying expenditure.
This is why SMEs should avoid applying the 400% figure to the entire project budget without first identifying eligible costs.
What Does a 400% Tax Deduction Actually Mean?
Suppose an SME incurs S$40,000 of qualifying AI expenditure that would ordinarily receive a 100% tax deduction.
Under the enhanced EIS treatment:
S$40,000 × 400% = S$160,000 total tax deduction
The company does not receive S$160,000 in cash.
Instead, S$160,000 may be deducted when determining taxable income, subject to the prevailing tax rules.
Without the enhancement, a normal 100% deduction on the same S$40,000 would produce a deduction of only S$40,000.
The additional deduction created by the EIS enhancement is therefore:
S$160,000 – S$40,000 = S$120,000
The actual tax benefit depends on the company’s taxable position and applicable tax treatment.
Example: Potential Tax Impact for a Profitable Company
Singapore’s prevailing corporate income tax rate is 17%.
For illustration, assume a company has sufficient taxable income to use the full deduction and ignore other tax exemptions, rebates and adjustments.
The company spends S$50,000 on qualifying AI expenditure that would normally be deductible.
With the 400% deduction:
S$50,000 × 400% = S$200,000 deduction
Without the enhancement, the ordinary deduction would be:
S$50,000 × 100% = S$50,000 deduction
Additional deduction:
S$200,000 – S$50,000 = S$150,000
At a 17% corporate income tax rate, the simplified additional gross tax reduction could be:
S$150,000 × 17% = S$25,500
This does not mean every SME spending S$50,000 will automatically save S$25,500 in tax.
The actual benefit depends on taxable income, applicable exemptions, tax losses, other deductions and the company’s overall tax position.
But the example demonstrates why the enhanced deduction can materially change the after-tax economics of an AI project.
The Tax Benefit Does Not Reduce the Supplier Invoice
This is an important cash-flow distinction.
If an SME purchases S$50,000 of qualifying AI services, it still needs to pay the vendor according to the agreed payment terms.
The enhanced deduction affects income tax computation.
It does not reduce the amount payable to the supplier at the time the project is implemented.
Management should therefore distinguish between:
- project cash cost;
- eventual tax benefit; and
- timing of that tax benefit.
An attractive after-tax project can still create immediate cash-flow pressure.
There Is No EIS Cash Payout for AI Adoption
The wider Enterprise Innovation Scheme includes an option for eligible businesses to convert certain qualifying expenditure into a cash payout.
However, IRAS specifically states that the cash payout option does not apply to the new AI adoption category.
This matters particularly for smaller or early-stage companies.
A profitable SME paying corporate income tax may be able to benefit relatively directly from an enhanced deduction.
A business with little or no taxable income may not receive the same immediate cash-flow benefit.
Management should therefore avoid valuing the incentive as though it were a cash grant.
Do Not Confuse a Tax Deduction With a Grant
A grant and a tax deduction improve project economics in different ways.
A grant may reimburse or subsidise part of an eligible project cost.
A tax deduction reduces taxable income.
The new AI incentive also specifically excludes expenditure subsidised by Government grants or subsidies from qualifying AI expenditure.
This means an SME should carefully separate:
- costs covered by Government support;
- qualifying AI expenditure potentially eligible for the enhanced deduction;
- non-qualifying expenditure; and
- the company’s own remaining project cost.
This prevents management from accidentally assuming that the same expenditure can receive overlapping support.
Build the Business Case Before the Tax Case
The enhanced deduction makes qualifying AI expenditure more attractive, but an SME should still ask whether the AI system creates measurable business value.
Potential benefits may include:
- reducing manual administrative work;
- improving customer response times;
- automating document processing;
- reducing repetitive data entry;
- improving forecasting;
- supporting sales or customer-service teams;
- optimising scheduling or inventory;
- improving fraud or anomaly detection; and
- increasing employee productivity.
Management should try to convert these benefits into financial estimates rather than relying only on statements such as “AI will improve productivity.”
Example: Calculate Labour Time Saved
Suppose an SME has five employees who each spend eight hours per week preparing repetitive reports.
Total staff time:
5 employees × 8 hours = 40 hours per week
An AI system reduces the work by 60%.
Time potentially released:
40 hours × 60% = 24 hours per week
If the estimated employment cost of that time is S$30 per hour:
24 hours × S$30 = S$720 per week
Across 48 working weeks:
S$720 × 48 = S$34,560 per year
This does not automatically mean the company will reduce payroll by S$34,560.
If employees remain employed, the benefit may instead appear through additional capacity, faster work or more time spent on higher-value activities.
Management should therefore distinguish between:
- actual cash savings;
- capacity released;
- productivity improvement; and
- potential revenue gains.
Calculate the AI Project Payback Before Tax Incentives
Suppose an SME expects an AI project to cost S$50,000 and produce S$20,000 of measurable annual operating benefits.
Simple payback before considering tax:
S$50,000 ÷ S$20,000 = 2.5 years
This tells management whether the project already has a reasonable commercial case.
The tax incentive can then be considered as an additional benefit rather than the reason for making the investment.
Then Consider the After-Tax Economics
Assume the full S$50,000 is qualifying expenditure that would ordinarily be deductible and the company can fully utilise the enhanced deduction.
Using the simplified earlier example, the additional 300% deduction could potentially create up to S$25,500 of additional gross tax savings at a 17% tax rate before considering other tax treatments.
The project would still require the company to pay its S$50,000 project cost.
But the eventual tax benefit improves the effective economic return.
This can make projects with moderate pre-tax returns more attractive.
However, it should not rescue a project that produces little or no operational value.
Watch Recurring Subscription Costs
Many AI systems use subscription pricing.
An implementation may therefore have:
- an initial setup cost;
- monthly or annual licensing;
- usage-based charges;
- API costs;
- data-storage charges;
- consulting support; and
- ongoing maintenance or monitoring.
SMEs should calculate total annual cost rather than only the initial implementation bill.
For example:
- Initial implementation: S$25,000
- Annual licence: S$18,000
- Support and maintenance: S$6,000
- Usage charges: S$5,000
First-year cost:
S$25,000 + S$18,000 + S$6,000 + S$5,000 = S$54,000
Ongoing annual cost after implementation may still be:
S$18,000 + S$6,000 + S$5,000 = S$29,000
If the AI system only produces S$20,000 of annual benefits after the first year, the project may not be financially sustainable despite an attractive first-year tax deduction.
Tax Benefits Are Temporary, Operating Costs May Continue
The enhanced AI deduction is currently available for YA 2027 and YA 2028.
An SME should therefore avoid evaluating a multi-year AI subscription as though the 400% deduction will continue indefinitely.
Suppose an AI system is expected to remain in use for five years.
Management should calculate:
- cost during the enhanced deduction period;
- cost after the enhanced period;
- expected annual benefits throughout the project life; and
- whether the system remains worthwhile without the temporary incentive.
A strong AI project should still make commercial sense when the enhanced deduction eventually ends.
Do Not Underestimate Data Preparation Costs
AI projects often depend on business data.
An SME may discover that its information is:
- stored across multiple spreadsheets;
- inconsistent;
- duplicated;
- poorly labelled;
- incomplete;
- held in incompatible systems; or
- not suitable for the intended AI use case.
Cleaning and restructuring data can materially increase implementation costs.
Where qualifying AI business services include data and analytics work related to the development, deployment, operation or maintenance of an AI system, some of these professional-service costs may potentially qualify, subject to IRAS requirements.
However, SMEs should still budget the full cost before determining the tax treatment.
Include Governance and Compliance Costs
AI adoption can also create governance requirements.
Depending on the use case, management may need to consider:
- data protection;
- access controls;
- cybersecurity;
- model monitoring;
- human review;
- accuracy testing;
- record keeping;
- vendor risk; and
- regulatory requirements.
IRAS includes qualifying system engineering and compliance services among examples of AI business services when they relate to the development, deployment, operation or maintenance of the AI system.
These costs should nevertheless be incorporated into the commercial project budget rather than treated as an afterthought.
Start With a Narrow Use Case
For many SMEs, a smaller AI project can provide more useful financial information than a large enterprise-wide rollout.
A business might begin with one process such as:
- invoice extraction;
- customer enquiry classification;
- internal document search;
- sales lead qualification;
- inventory forecasting;
- meeting or call transcription; or
- automated report drafting.
The company can then measure actual:
- time savings;
- error reduction;
- usage levels;
- employee adoption;
- customer impact; and
- financial benefits.
This evidence can help management decide whether a wider rollout is justified.
Beware of Paying for AI That Employees Do Not Use
Software licences create little value if employees continue using the old process.
Suppose an SME buys 30 AI-user licences at S$100 per employee per month.
Annual licence cost:
30 × S$100 × 12 = S$36,000
If only 10 employees actively use the system, two-thirds of the licence capacity may be producing little value.
Management should therefore track:
- active users;
- frequency of use;
- workflows actually improved;
- time saved;
- output quality; and
- whether unused licences can be reduced.
AI adoption should be measured after implementation, not merely approved at the purchasing stage.
Compare AI With the Existing Process
The correct financial comparison is usually:
Cost of Current Process versus Cost of AI-Enabled Process
For example, suppose an SME currently spends:
- S$70,000 per year on outsourced document processing;
- S$10,000 on corrections and rework; and
- S$5,000 on temporary staff during peak periods.
Total current cost:
S$85,000 per year
An AI-enabled process costs:
- S$30,000 annual software and usage fees;
- S$15,000 annual human review and support;
- S$5,000 of other operating costs.
Total AI-enabled cost:
S$50,000 per year
Potential recurring cash saving:
S$85,000 – S$50,000 = S$35,000 per year
This creates a clear financial basis for evaluating the project.
Stress-Test the Benefit
AI projects should also be tested under weaker assumptions.
For example:
- What if adoption is only 50% of expectations?
- What if licence fees increase?
- What if implementation takes six months longer?
- What if human review remains necessary?
- What if the expected productivity benefit is only half of the forecast?
- What if the vendor changes its pricing model?
If the project remains financially attractive under a reasonable downside scenario, management can have greater confidence in the investment.
Do Not Buy AI Because Everyone Else Is Buying AI
AI can create genuine productivity improvements, but it can also become a technology expense without a clear return.
An SME should be cautious where:
- there is no clearly defined business problem;
- management cannot identify who will use the system;
- there is no baseline against which improvement can be measured;
- the project relies primarily on vague productivity claims;
- existing data is unsuitable;
- integration costs are unknown; or
- the only strong reason for investing is the tax incentive.
A 400% deduction improves the cost of a useful project.
It does not make an unnecessary AI system useful.
A Practical AI Investment Checklist
Before committing to an AI project, an SME can ask:
- What specific business problem are we trying to solve?
- What is the full first-year and ongoing cost?
- Which expenditure may qualify for the enhanced AI deduction?
- Are any project costs excluded because they relate to hardware or Government-subsidised expenditure?
- What measurable cash savings, capacity improvements or revenue benefits should the system create?
- What does the project payback look like before the tax incentive?
- How does the tax deduction change the after-tax economics?
- Does the business have sufficient taxable income to obtain an immediate benefit from the deduction?
- What happens when the enhanced deduction ends after YA 2028?
- Does the project remain worthwhile if benefits are materially lower than forecast?
Final Thoughts
The new 400% tax deduction for qualifying AI expenditure can materially improve the economics of AI adoption for Singapore businesses in YA 2027 and YA 2028.
But SMEs should interpret the incentive correctly.
It is an enhanced tax deduction, not a 400% reimbursement and not an EIS cash payout.
The first S$50,000 of qualifying AI expenditure per YA can receive the enhanced deduction, while hardware and Government-subsidised expenditure are excluded from the new AI category.
For management, the strongest approach is therefore to build the commercial case first.
Identify the business problem, calculate the full project cost, quantify realistic operating benefits, measure adoption and stress-test the expected return.
Then calculate how the enhanced tax deduction changes the after-tax economics.
AI can be a valuable productivity investment when it solves a genuine operational problem and produces measurable benefits.
The tax incentive should make a good AI investment better.
It should not be the reason an SME buys technology it does not need.
Note: The availability and calculation of tax deductions depend on prevailing tax legislation, qualifying expenditure and the taxpayer’s circumstances. Businesses should refer to current IRAS guidance and obtain appropriate tax advice where necessary before relying on the EIS AI adoption deduction in an investment decision.