Photo by Austin Distel on Unsplash
- As of 2025, AI tool spending among Asian founders rose 20%, with Anthropic's Claude growing 3x and coding tool Cursor growing 4.2x in regional startup usage, per Aspire's study of 37,000+ SMEs across 16 countries.
- Singapore SME AI adoption tripled from 4.2% to 14.5% in 2024, with government-supported adopters reporting an average of 52% in cost savings, per IMDA Singapore official statistics.
- Singapore offers 400% tax deductions on qualifying AI spend — capped at S$50,000 annually for YA2027-2028 — a subsidy many founders are not actively claiming.
- MIT research found 95% of AI projects fail to deliver ROI, and Uber reportedly exhausted its entire 2026 AI budget in four months with no proven link to customer value — a cautionary counterweight to the growth numbers.
What Happened
4.2x. That is how fast Cursor usage grew among Asian startups in 2025 — not a rounding error, but a signal that something structural is shifting in how founders build software. Originally reported by Google News and The Edge Singapore, drawing on financial platform Aspire's analysis of more than 37,000 small and medium-sized businesses (SMEs) across 16 countries, the picture is one of a region accelerating hard into AI — with both the upside and the warning signs fully in frame.
According to Fortune's coverage of Aspire CEO Andrea Baronchelli's findings, AI tool spending among Asian founders rose 20% in 2025. Anthropic's Claude saw usage triple (3x), while Cursor — an AI-powered code editor — surged 4.2x. Baronchelli's read: "This signals a reallocation of capital towards efficiency," with founders deploying AI to write and ship code, not just tidy up inboxes.
The macro context matters. As of June 16, 2026, according to IMDA Singapore's official statistics, SME AI adoption in Singapore tripled from 4.2% to 14.5% in 2024 alone, while non-SME adoption climbed from 44% to 62.5%. Microsoft pledged $5.5 billion into Singapore's AI infrastructure through 2029 (April 2026). OpenAI committed S$300 million to establish its first applied AI lab outside the US in Singapore (May 2026). And the Singapore government's S$1 billion public AI research commitment — paired with a National AI Council chaired by PM Lawrence Wong — signals this is policy, not hype.
The Job These Tools Are Actually Hired To Do
The jobs-to-be-done frame is useful here. Founders are not buying Cursor because they read a LinkedIn thread about it. They are hiring it to ship faster with a smaller headcount. They are hiring Claude to absorb the cognitive overhead of debugging, generating test suites, and drafting technical documentation — the tasks that drain senior engineers who should be making architecture decisions instead.
This is a meaningful departure from the first wave of business AI, which was predominantly administrative: drafting emails, summarizing documents, booking meetings. TechNode Global's reporting on Aspire's newly launched "Aspire for Startups" program (May 2026) reveals cash bonuses tied specifically to AI-related spending in the first three months of membership — an incentive structure that implicitly acknowledges productive AI use requires graduating past the admin tier and embedding into core product workflows.
As of June 16, 2026, 73.8% of Singapore workers report using AI tools at work — most of them several times weekly or daily, per IMDA Singapore data. That is no longer an early-adopter statistic. That is a new baseline for how competitive teams operate.
Photo by Sang Kwak on Unsplash
Which Tools Win — and Where the Budget Bleeds
Chart: Singapore AI adoption rates among SMEs and non-SMEs, 2023 vs. 2024. Source: IMDA Singapore official statistics, as of June 16, 2026.
The growth numbers for Cursor (4.2x) and Claude (3x) are striking, but they exist alongside a harder story. SmartToolbox AI's detailed breakdown of Cursor vs. Copilot vs. Tabnine covers the specific feature and pricing deltas coding teams need before committing to a per-seat subscription — useful context before the stack decision is made.
The cautionary data is equally important. Uber reportedly exhausted its entire 2026 AI budget in just four months, primarily driven by Claude Code usage — with the company's COO noting there was no proven link between high AI adoption rates and useful customer products. Apollo Global Management's John Zito was blunter in Bloomberg's reporting: "too many companies are deploying AI tools for relatively mundane tasks that don't justify the significant costs and computing power."
Asia-Pacific companies that execute AI well expect $2.85 in return for every dollar invested in 2026, representing a 2.8x ROI, per Aspire's research. But a recent MIT report found 95% of AI projects fail to deliver ROI. The distance between those two statistics is where most small businesses will actually operate. One CEO with 25 years of experience, quoted in coverage of the sector, observed that the way AI is discussed in boardrooms and investor decks "has stopped being thoughtful and started being hysterical." That is not an argument against AI tools. It is an argument for using them with specificity.
The Switching Cost Nobody Discusses Before Signing Up
In my analysis, the real lock-in risk for small teams is not the tool itself — it is the workflows crystallized around it. The moment you outgrow a free tier and start embedding Claude into your engineering codebase or Cursor into your daily build cycle, migration costs compound fast. Engineers develop muscle memory. Prompt libraries accumulate. Context and institutional knowledge end up stored inside a vendor's platform rather than your own systems.
The 400% Singapore tax deduction on qualifying AI spend (YA2027-2028, capped at S$50,000 annually) is a meaningful offset — but it applies to qualifying categories, and most founders are not aware they need to document usage against the Productivity Solutions Grant (PSG) framework to capture the 52% average cost savings that PSG-supported adopters reported in 2024. Data export reality: if you are evaluating tools now, ask explicitly whether your AI-generated outputs — code, trained workflows, document libraries — are portable before you are twelve months deep.
The team-size cliff is also real. A two-person startup can absorb overconsumption on a $20/month plan. A 15-person team deploying Claude Code across every engineer at $100-plus per seat per month hits a different arithmetic within a quarter. Aspire's data showing a 20% year-on-year spending rise reflects this stacking dynamic — more tools, more seats, fewer hard conversations about overlap.
Three Actions for Small Business Owners
List every AI subscription your team currently pays for. For each one, identify the specific workflow it is embedded in — not the use case you bought it for, but what it actually does on a weekly basis. Cut anything where the honest answer is mostly administrative tasks we could consolidate or batch. The Aspire data shows the ROI gap emerges precisely when AI spending stays in low-value territory without discipline.
As of June 16, 2026, Singapore businesses can access 400% tax deductions on qualifying AI-related expenditure capped at S$50,000 annually for YA2027-2028, plus PSG co-funding for approved AI-enabled solutions. Aspire's revamped Aspire for Startups program (launched May 2026) also offers S$500,000-plus in founder benefits including cash bonuses tied to AI-related spending in the first three months. These are live programs, not proposals.
Before any team AI subscription renews, run a simple output audit: hours saved per week, code shipped per engineer, support tickets resolved per hour. If you cannot produce a number, you are running on assumption. The 52% average cost savings reported by PSG-supported SMEs in 2024 came from structured adoption — not from subscribing and hoping the efficiency would materialize.
Frequently Asked Questions
What are the best AI tools for startups in Singapore right now?
As of June 16, 2026, the tools with the fastest verified adoption growth among Asian startups — per Aspire's study of 37,000+ SMEs — are Cursor (AI coding assistant, 4.2x usage growth in 2025) and Anthropic's Claude (3x usage growth). For workflow automation and document-heavy tasks, ChatGPT remains broadly deployed. The right answer depends entirely on the job being hired: code-heavy teams should evaluate Cursor against GitHub Copilot directly; reasoning and writing tasks favor Claude or GPT-4o. There is no single best tool without knowing the specific workflow pain it needs to solve.
How much do AI tools cost for small businesses on a monthly basis?
Individual subscriptions typically run $20–$30 per seat per month for tools like Claude Pro, ChatGPT Plus, or Cursor Pro. The risk for small businesses is not the individual line item but cumulative stack cost — multiple overlapping tools across a 10-person team can reach $500–$2,000 per month before anyone audits the overlap. Aspire's data showing a 20% spending rise in 2025 among Asian founders reflects this stacking dynamic more than any single tool's price increase.
Is Singapore's 400% AI tax deduction worth pursuing for small businesses?
Yes, if your spend qualifies. As of June 16, 2026, Singapore businesses can claim 400% tax deductions on qualifying AI-related expenditure capped at S$50,000 annually for YA2027-2028 — effectively a S$200,000 deduction on S$50,000 of spend. Combined with PSG co-funding for approved solutions, the effective net cost of AI adoption for eligible Singapore SMEs is substantially lower than list price. The requirement: spend must fall under specific qualifying categories, so verify against the official IMDA and IRAS criteria before assuming your current subscriptions are covered.
Which AI coding tool is growing fastest among Asian startups?
Cursor is growing fastest in this dataset — 4.2x usage growth among startups in Asia during 2025, per Aspire's research of 37,000+ regional SMEs. Anthropic's Claude follows at 3x growth, though Claude covers a wider range of tasks beyond pure code generation. Growth rates reflect adoption momentum, not necessarily the best fit for every team's stack. Feature-level comparisons against GitHub Copilot and Tabnine are more actionable than usage statistics when making a purchase decision for your team.
How do you actually measure whether your startup's AI tool spending is generating ROI?
Track hours saved per week per tool — ask your team directly and log the responses. Measure output metrics before and after adoption (code commits, documents processed, support tickets closed). Compare against total subscription cost per month. The 52% average cost savings reported by Singapore SMEs using PSG-supported AI solutions in 2024 were documented outcomes, not estimates. MIT research finding that 95% of AI projects fail to deliver ROI consistently points to the same root cause: measurement was not built into the adoption process from day one. If you cannot define the metric before you subscribe, define it before the first renewal.
Disclaimer: This article is editorial commentary based on publicly reported data and is intended for informational purposes only. Tool features, pricing, and government program eligibility may change — verify current details on official websites before making business decisions. Research based on publicly available sources current as of June 16, 2026.