Introduction
In May this year, the Economic Strategy Review presented 32 recommendations across eight thrusts, the product of five committees and consultations with more than 7,700 stakeholders, with the full report released in June. It is a serious piece of work and I commend those involved. But as I noted in this House last September, Singapore has never lacked task forces and committees: the Economic Strategies Committee in 2009, the Committee on the Future Economy in 2016, the Future Economy Council, the Industry Transformation Maps, the Emerging Stronger Taskforce in 2020, and now the ESR and the Singapore Economic Resilience Taskforce. At this rate, reviewing the economy risks becoming a growth industry in its own right.
The ESR asks searching questions of our businesses and our workers. Sharpen our value proposition. Be agile. Be resilient. These are fair propositions. But there is one economic actor of whom the report asks remarkably little: the Government itself. And specifically, the Government in its most lucrative role as our landlord. The State owns some 90 per cent of the land in Singapore and it is, by a wide margin, the biggest landlord in the country. My contention today is a simple one. The ESR's ambitions for enterprise and innovation will not be realised so long as the Government continues to see itself as a landlord, and as long as Singaporeans continue to share in the Government’s rentier mindset.
In fact, even as economists often identify the key factors of production being land, labour, capital and entrepreneurship, there is not a single mention of land in the 32 recommendations of the ESR report, even as the report itself has acknowledged that land, energy and demographic constraints are expected to tighten.
Land Costs and The Cost of Everything
Mr Speaker, rent-seeking behaviour permeates our society, and the tone is set at the top. As I have shared in my speech at the opening of this term of Parliament, policy should not inadvertently reward the highest bidder at the expense of wider societal outcomes. And it is in upstream land costs that we must go further. State land forms part of our reserves, but excessive land prices push up development costs and rents, impacting tenants and companies and eventually consumers. As a local C-suite member of one of Asia’s largest real estate groups has shared in the past, in Singapore, land costs now form roughly 70% of total project development expenditure, up from just 4% in the 1980s for Raffles City. These costs do not vanish. They cascade from developer to tenant, from tenant to hawker, clinic and shopkeeper, and finally to every consumer.
Now, I anticipate the familiar reply that rental is not the largest cost businesses face. Indeed, when this House debated hawker culture, the Senior Minister of State cited an NEA survey showing that rental forms less than 10 per cent of stallholders' operating costs, against 56 per cent for raw materials and 20 per cent for manpower. But this argument proves too little. Where does the hawker's supplier operate from? A rented industrial unit on state-priced land. Where does the wholesaler store his goods, the logistics firm park its lorries, the central kitchen prepares its ingredients? All on land whose price the State sets and extracts. High land prices feed into rentals, and rentals feed into every line of the cost structure, including the 56 per cent we are told to look at instead. Rent is not one cost among many. It is the cost embedded in all the others, and an important fixed overhead that cannot be avoided.
Members will recall the $52,000 monthly rental bid for a GP clinic in an HDB estate. Minister Ong Ye Kung himself acknowledged that this must feed through to healthcare costs one way or another, and that the highest rental bid does not deliver the best care for the community. I agreed with his view then and I continue to hold this view. During this year’s Committee of Supply, I appreciate that the HDB is expanding the Price-Quality Method for commercial tenders of neighbourhood shops. But if we accept this logic downstream, at the level of a single shop tender, we must have the intellectual honesty to apply it upstream, to land policy itself.
Land Prices That Can Only Go Up
Mr Speaker, there is a structural reason why these costs keep rising: our land sales system is built so that prices can only ratchet upwards. Sites are awarded to the highest bidder, subject to a reserve price that is not published for any individual site. When the market bids above the reserve, the land is sold and a new benchmark is set. When the market bids below it, the tender is simply not awarded barring other specific circumstances, and the State waits for a better day. Heads, land prices rise. Tails, land prices are not allowed to fall.
In a reply to my parliamentary question in July 2021, the Minister for Law disclosed that the reserve price is pegged to 85 per cent of the estimated market value assessed by the Chief Valuer's Office, and that of the 133 sites launched for tender between 2016 and 2021, 17 were not awarded because the highest bids were deemed not reflective of the fair market value of the land. That is roughly one in eight tenders aborted rather than allowed to clear at what the market was actually prepared to pay. The pattern continues. In February 2024, URA declined to award the Marina Gardens Crescent white site because the sole bid of $770 million, or $984 per square foot per plot ratio, was assessed to be “too low”.
The history of Paya Lebar Quarter, one of my favourite mixed use sites in the east, tells the same story. In November 2011, URA rejected the sole bid from a consortium of UOL Group and Singapore Land for a mixed-use site in Paya Lebar Central, at $529.5 million or $566 per square foot per plot ratio, on the grounds that it was too low. The consortium asked for reserve prices to be made public, given the costs and effort of submissions of such a scale. URA declined, saying disclosure would not be in the public interest. When the site was eventually tendered and awarded in April 2015, it was sold at $1.67 billion, or $943 per square foot per plot ratio, at a rate some two-thirds higher than the bid the Government had earlier refused. The land waited four years, the price went up by two-thirds, the state coffers are higher by a billion, even as the site area is now bigger, and we saw private residential prices in the area reach new highs after Park Place Residences was eventually launched. Is this a healthy development?
The irony is that the Government understands this perfectly well in certain cases: NEA proudly sets no reserve rent at all for hawker stall tenders since 2012 and does not set a minimum bid price for hawker stall tenders.
My first call, therefore, is that maximising revenues from land and buildings should no longer be a primary consideration of land policy. As I have argued in this House before, we should bring back two-envelope, concept-and-price tenders as the default for strategic and community-facing sites, so that land is awarded on the quality and value of what is built, and not simply to whoever bids the most. The Government as the biggest landlord of all benefits, but it does so at everyone else’s expense.
Shackled Households Cannot Be Entrepreneurial Households
The same logic applies to households. Home loans make up at least 70 per cent of total household liabilities in Singapore, and mortgage borrowing grew 5.8 per cent year on year in the first quarter of this year. Our own regulatory framework, the Total Debt Servicing Ratio, contemplates households committing up to 55 per cent of gross monthly income to debt obligations. A household servicing a 30-year mortgage at anywhere near that level does not leave the perceived security of a monthly pay check to start a company with no visibility over cash inflows for years. It optimises for the next pay cheque, because it must. Moreover, according to ADP Research, 60 per cent of workers in Singapore were living pay cheque to pay cheque as of 2024. What capital is there for entrepreneurship? It is no wonder that risk aversion in Singapore is not some cultural mystery, but a lived balance sheet reality.
This is why, in the last term of Parliament, I moved an adjournment motion calling for rental housing to be developed as a genuine housing model, and not merely a residual safety net for those with no other options. The build-to-own model has served asset accumulation well, but it front-loads a lifetime of leverage onto Singaporeans in precisely the decades when they are most able to take entrepreneurial risk. In many other global cities, young professionals rent for years without stigma, preserving the flexibility to move, retrain and build.
My second call is therefore for a meaningful public rental option for young and middle-income Singaporeans, with tenure security and rents set with reference to income rather than market value. We already have a variant of this in the Parenthood Provisional Housing Scheme (PPHS) for married or engaged couples with more modest incomes and who wish to rent a temporary flat while waiting for their new home to be completed. If we want a generation of founders, we should stop shackling them to mortgages in their twenties and thirties.
Compute Is the New Land
Mr Speaker, since the Government seems to relish its position as being the largest landlord in the country, let me offer a new estate to develop. In the AI economy, compute is to value creation what land was to the industrial economy: the scarce foundational input on which everything else is built.
We have made a start, and I acknowledge that an Enterprise Compute Initiative was announced at Budget 2025 with up to $150 million of funding support. This is however, only meant to be a short-term programme that will only run for one year for the respective launch dates of each cloud service provider. By October 2025 it had reportedly paired some 1,000 firms with cloud partners, but this is a very modest figure as compared to the 358,000 or so firms in Singapore’s enterprise landscape in 2024 as identified by the ESR final report. I also recognise that NTUC's AI-Ready SG subsidises half the subscription cost of eligible AI tools for members, if you first go for an approved course. In other words, the principle of subsidised access to compute has been accepted by the Government. The questions that remain are of scale, of duration, and of ambition.
Some examples from overseas may be instructive. Hong Kong built its own Artificial Intelligence Supercomputing Centre at Cyberport and then allocated HK$3 billion to a three-year AI Subsidy Scheme under which eligible users receive subsidies of up to 70 per cent of the Centre's service list price. That is a state-owned facility with a published rate card and a discount. Canada's Sovereign AI Compute Strategy pairs public supercomputing investment with a C$300 million Access Fund that covers two-thirds of eligible costs for local cloud-based AI computing services, on projects with compute costs of up to C$5 million. Note the word local, and that these are targeted at SMEs. The United Kingdom, having found itself with only 1.3 per cent of global compute capacity in late 2022, committed up to £2 billion by 2030 under its Compute Roadmap, and its AI Research Resource now allocates compute directly: 10,000 GPU hours for first-time users and 20,000 GPU hours for small and medium enterprises, awarded as capacity rather than cash.
Consider also the asymmetry closer to home. We extract some $20 billion in a single year in land sales proceeds, and yet we are only setting aside $150 million for the digital land on which our own companies must build, and $1 billion over five years in overall National AI research. And for all our first-place finishes in AI readiness, only 4 per cent of firms had embedded AI into core business processes by 2026, with adoption at 15 per cent among SMEs against 63 per cent for larger firms. AI diffusion is a big concern.
I could not have summed it up better than a 2025 World Bank report on Digital Progress and Trends report, that governments should create an enabling environment to facilitate access to compute resources, catalysing AI adoption, adaptation and innovation. This can include targeted interventions such as compute subsidies for small and medium enterprises and researchers, regional data centres, and public-private partnership. More recently, NVIDIA CEO Jensen Huang also remarked that a strong AI ecosystem is not a foregone conclusion. Policymakers have an important opportunity to act, such as through expanding access to compute for startups and researchers and investing in shared training assets.
This is not a one-off or time-limited effort, but instead we should spare no effort until a significant majority of our local enterprises have truly embedded AI into their core business processes.
My third call is therefore a so-called “HDB model” for compute. The HDB public housing model made the government’s objective of home ownership for the people programme during our early years attainable, through subsidised access to state land. Similarly, a national compute programme should offer subsidised, allocated capacity to local companies and particularly SMEs and entrepreneurs. This will draw on capacity in which the State has an ownership interest, priced below market with a published rate card with tenure certainty, and eligibility weighted towards local ownership and local value capture.
As what Liang Wenfeng, CEO of DeepSeek said recently, 人才的差距，本质上也是因为算力的差距。Let us not have a talent gap in Singapore, simply because of a lack of access to compute.
The Reserves Belong to Singaporeans, Not the Government
Mr Speaker, I can already anticipate the Government’s response to the proposals I have raised. That pricing land or compute below the highest market bid amounts to raiding the reserves. Allow me three responses. First, the reserves belong to all Singaporeans, not to the Government of the day. Foregone rent extraction that flows into the productive capacity of Singaporean households and firms is not a raid but an investment in our collective future. It is the owners of the reserves investing in themselves.
Second, on the Government's own explanation to this House, leasehold land reverts to the State at the end of its term and is protected as past reserves once again. The debate is therefore about the pricing philosophy for the use of land over 99 years, not about giving the freehold away. Third, we do not apply this stinginess to foreign investment promotion. We extend generous low tax incentives and charge as low as 5% corporate income tax instead of 17% to attract multinational corporations via a series of incentive schemes. Why is a discount for a foreign MNC prudent economic strategy, but a discount for a Singaporean entrepreneur a raid on the reserves?
Conclusion: What Is the Real Risk?
Allow me to conclude in Mandarin Mr Speaker, on what the real risks are for Singapore.
议长先生，我们的国家，我们的经济真正面临的风险究竟是什么？
是我们将少数地段、少数租赁组屋或少数GPU集群的定价定在低于市场可承受的最高水平吗？还是一代又一代的国人会悄然得出这样的结论：在自己的国家，最吃香的生意就是当房东，最稳妥的生活就是当半辈子的房奴偿还房贷？
我欢迎经济策略检讨委员会的《最终建议报告》，更支持张文杰和林志蔚议员所提出的这项动议。但只要国家依然是我们经济体中最为成功的租金最大化者，这一战略的宏图就只能停留在纸面上。政府应当停止将自己视为新加坡最大的房东，而应该更积极地将自己视为人民最大的助力者。
议长先生，我支持这项动议。
