Budget Debate Speech: Building an R&D Ecosystem - 26th Feb 2026
(Hongmeiren)
Speaker,
Last week was Lunar New Year. I had the pleasure of sampling a Xiangshan Hongmeiren mandarin orange. The texture was extraordinary — the vesicles of the orange so sweet, the skin almost paper thin. It has been described as “eating jelly”, or “jelly growing on a tree”. Twelve to fifteen percent sugar content, low acidity, no floating skin.
It inspired me to investigate its history. The Hongmeiren was once Ehime No. 28, developed over fifteen years at the Ehime Prefectural Fruit Tree Experiment Station in Japan, prized for its jelly-like texture and ultra-premium positioning.
But due to a lack of IP protection, it was lost. A handful of scion cuttings were carried to Xiangshan county - Zhejiang province in China - in 2001. Today, the Hongmeiren has expanded to over 1 million mu across fifteen Chinese provinces — dwarfing Japan’s 300 hectares by over a hundredfold. Fifteen years of breeding, gone like that. It is one of the most consequential agricultural IP losses in modern Japanese history, alongside the Shine Muscat grape and strawberry leakages.
And that reminded me of something closer to home.
(Mirxes)
“It is well known that our R&D ecosystem over the last thirty years has not produced impressive commercial outcomes. Tens of billions of spending since 1990. More than thirty years. Where are the deep tech commercial successes? The report card of significant commercial outcomes — high-value IPOs, globally competitive companies — is bare.”
That is what I would have said, up to May 23rd, 2025.
On that day, Mirxes, a micro RNA cancer diagnostics biotech company incubated over many years of A*STAR research at Biopolis, IPO’d on the Hong Kong Stock Exchange, raising HK$1.09 billion with China-linked cornerstone investors. Finally, a billion dollar IPO. Coincidentally, also about fifteen years of effort — like the Japanese breeding the Ehime No. 28.
But the story is more complicated. Mirxes listed in Hong Kong, not Singapore, because SGX had no equivalent of HKEX’s pathway for pre-revenue biotech. Its cornerstone investors were Chinese entities holding over forty percent of the IPO. Its manufacturing now is in Zhejiang province. Its growth market is China. The initial science came from Singapore. Commercialisation leaked offshore.
If this is success, what does failure look like?
Now - this is not the fault of the Chinese. They optimised rationally within the realities of the system. This is our failure to capture value. Our responsibility to anticipate, to get it right. To make sure Singaporean taxpayers reap the rewards of tens of billions ploughed into R&D.
Mirxes highlights two facts. First, the system does not produce a strong pipeline — that Mirxes is the only billion dollar IPO or commercially significant R&D company after thirty years of continual investment tells you something. Second, even when a success emerges, the system is naively vulnerable to value-chain leakage.
“To lose one billion-dollar IPO may be regarded as a misfortune. To lose more is carelessness.” I will return to this later.
I welcome the Budget’s commitment to $37 billion under RIE2030 and the expansion of Startup SG Equity into growth capital — these are steps in the right direction. But the question is not whether we are spending enough; it is whether the system converts that spending into commercial outcomes for Singapore.
(System)
Why does the system produce these two outcomes - (a) a limited pipeline of commercially viable R&D companies, and (b) a failure to take advantage of successes when they come? The structural flaw is that agencies like EDB and the statutory boards are structured as grant-givers, not as investors or ecosystem builders.
If you are a grant-giver, your job is to mark the market — to benchmark, to do due diligence against market rates. If you mark the market, you are then subjected to the audit process — every decision scrutinised against whether you followed the benchmark. If you are a civil servant in that position, there is absolutely no reason to stick your neck out. Zero incentive for upside. Only downside risk from audit.
So of course you end up with super-conservative civil servants who say, “I don’t want to do any innovation.” You end up with long meetings where senior officials agonise over five thousand or ten thousand dollars because they are worried about compliance. Worse, annual KPI cycles force officers to judge five-to-seven-year bets on a twelve-month horizon — so promising ventures get culled before they can prove themselves. Meanwhile, you miss the forest for the trees.
To take an ecosystem approach to funding so that the day-to-day operational environment for R&D startups is the best it can be. To design policy to capture R&D upside.
Our agencies are structured to avoid losing money, not to make it. That is why thirty years of spending has not produced thirty years of returns.
I agree with the Prime Minister that R&D is a core imperative. Productivity gains, moving up the value chain — all this matters. But there must be an honest reckoning with thirty years of limited commercial outcomes, and value chain leakage.
The R&D ecosystem needs three things: money, speed, and a market.
First, Money.
I tried to count Singapore’s government-linked startup funds. I stopped at twenty. There are eight sub-schemes under the Startup SG umbrella alone. SGInnovate. Xora Innovation under Temasek. Vertex Holdings with seven sub-funds. NRF, ASTAR, IMDA, MAS run their own programmes. Not an exhaustive list.
Of course, each was created for a reason. Each has a logic. But the aggregate effect is a landscape so fragmented that no single entity has the mandate, the capital concentration, or the institutional authority to make the kind of large, decisive, fast bets that define successful deep tech commercialisation elsewhere.
The SG Growth Capital merger of SEEDS and EDBI is a step in the right direction. But the deeper question is whether we have the institutional courage to truly consolidate — to give one or two entities the capital, the mandate, and critically the permission to fail at scale, rather than spreading accountability so thin that no one is responsible for ecosystem outcomes. The pruning must go deeper.
France has a public investment bank called Bpifrance — a single institution combining innovation investment, SME lending, loan guarantees, export credit insurance, and strategic equity stakes. In 2024, Bpifrance deployed €60 billion. Its officers proactively contact companies with programmes. It is commercially viable, with 2024 net income of €896 million.
Under François Hollande and economic adviser Emmanuel Macron in 2012, France merged four fragmented investment bodies — one for SME loans, guarantees, innovation - one for venture capital - one for strategic equity - and one for regional equity — into a single entity: Bpifrance (Banque Publique d’Investissement). Export credit insurance was added later in 2017. It now employs about 3,500 staff, centralising innovation investment, economy-wide strategic investment, SME financing, loan guarantees, export support, and regional business development.
The diagnosis that prompted Bpifrance’s creation was plain: despite strong science and underlying dynamism, France’s entrepreneurial culture was weak — risk-averse and fearful of failure. Bpifrance also organised learning expeditions and export programmes to help French companies internationalise.
Bpifrance became the anchor investor across French tech — not just deep tech. Its 10-year assessment found that 80% of French startups that raised funds between 2013 and 2021 received Bpifrance support, and two-thirds of French VC funds have Bpifrance as an LP. Bpifrance deliberately accepts below-market returns (targeting ~7% annually rather than 10%+) to take risks that private investors won’t - training a generation of entrepreneurs and building an ecosystem from scratch.
I believe we need an equivalent policy investment bank. An institution outside the civil service, incentivised to take risk, freed from line-item accountability. Staffed with ecosystem builders, not risk-averse civil servants. We need to take a portfolio approach, thinking in ten-year cycles. Ten-year cohorts. Ten-year mandates. They need the right incentives. We had this in the early years with DBS and Temasek — both have since drifted from that mission, and today no institution fills the gap.
The WP has argued before for an EXIM bank; the French example shows a mandate wider still — innovation investment, SME lending, export credit, strategic equity, all under one roof.
It is often said we have a lot of capital in Singapore. But we have the wrong type of capital for R&D commercialisation. Capital tied to government or corporates. Excessive obsession with business metrics way too early. Startups forced to jump through a thousand hoops for small cheques. Family offices largely uninterested in our R&D ecosystem. My colleague Jamus Lim has proposed requiring a modest domestic allocation as a condition for family office tax incentives, which I support.
What does success look like twenty-five years from now? A flywheel, where private capital is sophisticated enough to understand deep tech. Patient capital writing experimental cheques. Less metric-obsessed early on. That is the ecosystem we should be building toward.
Second, Speed.
Funders and regulators need to appreciate that two extra days of approvals or one month of delayed cash flow is life or death for a startup. You do not understand what it takes if you have not tried it yourself.
Fast capital deployment does not just save companies — it trains a generation of entrepreneurs who won’t waste sixty percent of their time fundraising. We should make peace with the fact that roughly eighty percent of bets on startups, made in good faith, will back a losing venture. That is the power law in action. The real gain is often the entrepreneur’s second and third startup. But if the first attempt feels like pulling teeth, they drop out, and the compounding of iterated learning across business formation is lost.
Speed is not just about money. It is about people.
MOM uses income as its proxy for talent — the Tech.Pass at 22500 a month, the ONE Pass at 30000. This may work for hot fields. It seldom works for deep tech which due to their call-option nature, are unfashionable before they are fashionable. A strong biotech engineer in Thailand earns a quarter of Singapore wages. To bring them here, a startup must triple their salary — not because their skills are worth less, but because MOM’s threshold demands it. That is startup cash burnt on regulatory compliance, not R&D. We are asking founders to choose between the talent they need and the runway they cannot afford to lose.
The assumption that salary equals value breaks down where it matters most. Deep tech talent is scarce, specialised, and often transient. An eighteen-month engagement with a materials engineer, or a combustion engine specialist can redefine a company’s trajectory. The visa framework should reflect that reality.
I propose a segmented approach: a company-driven deep tech visa, tied not to income but to the company’s credentials. A monthly review committee of mixed Singaporean and international entrepreneurs evaluating deep tech companies — there are perhaps fifty to eighty serious ones formed here each year. Companies meeting the criteria enter a special segment: flexibility for 3-5 visas in year one, with annual portfolio reviews assessing whether the company is growing. Easy entry, progressively stricter requirements on each renewal, all strictly performance-based. Is the company growing? Are Singaporeans being hired and trained alongside foreign specialists? If yes, renew. If not, exit.
Traditional organisations — banks, established corporates — can afford strict income rules. There is enough local talent. But deep tech startups need radical fluidity, and the visa regime most suited to them should follow the company, not the person.
We should also aim to make visa decisions snappy. It is fine to say no to people. But let’s have the courtesy to do it fast, within a matter of weeks. Let’s not waste our companies’ time.
Third, a market and offtake.
Money and speed are necessary but insufficient if no one is buying. The system must be designed for offtake — structured so that when R&D produces something, there is a buyer on the other end.
Singapore hosts regional headquarters, not global ones. Decision-making centres remain in New York, London, Paris, Shanghai. Regional postings run three to four years. When the regional executive who championed an innovation pilot rotates home, continuity goes with them. Corporate incentive structures often reward launching initiatives, not seeing them through — creating structural discontinuity.
Tax incentives should require backing from both regional and global headquarters. If there is no interlocutor in New York or Shanghai who has signed off, no incentive. This ensures a direct link to decision-making and prevents projects from being forgotten when executives rotate home.
Beyond that, we should attach strategic buyer offtake commitments to grants. Require MNCs to pilot, evaluate, or procure from at least one local startup or SME as part of grant conditions.
And the government itself must practise innovation procurement. Europe and China give SMEs quotas of contracts to help them gain scale. We must do the same.
(IP Protection)
I said I would return to the question of protecting value. Money, speed, and a market can build a pipeline. But pipeline alone is not enough if the value leaks out.
Three countries have dealt with this question, and they are not peripheral economies — they are among the most successful R&D ecosystems in the world.
Israel protects at inception. Under the Law for the Encouragement of Industrial Research and Development, the Israel Innovation Authority attaches binding conditions to all public R&D funding. IP developed with public grants cannot be transferred outside Israel without explicit prior approval. Unauthorised transfer is a criminal offence. No known prosecution has occurred. But the deterrent is powerful.
Taiwan protects through structure. ITRI spun off TSMC in 1987 by transferring fabs, equipment, technologies, and ninety-eight professionals to a new Taiwanese entity — with significant government ownership. The critical feature was not any single restriction but a system: government ownership stakes in spin-offs, personnel who were Taiwanese nationals, incorporation in Taiwan, government funding giving the state structural leverage over deployment.
South Korea designates core protected technologies. Under the Industrial Technology Protection Act, South Korea designates more than seventy technologies across thirteen fields as National Core Technologies — semiconductors, displays, batteries — and requires prior approval for any export or M&A involving these technologies. Maximum prison sentences for overseas technology leakage reach eighteen years, with punitive damages at five times actual losses.
Singapore needs equivalent mechanisms for conditionality.
Some say we should adopt - Golden shares. IP retention conditions, protected technologies. Domestic manufacturing requirements tied to public funding.
My view is: the necessary and plausibly sufficient policy to adopt is an IP conditionality regime, designed for a small open economy. Legislate that IP developed with public R&D funding cannot be permanently transferred offshore without prior approval from a designated authority — enforceable domestically against Singapore-domiciled entities and individuals.
Ultimately we will be judged on the competitiveness of our ecosystem. An SGX that rewards R&D. SEZs that can scale up manufacturing at lower cost. Company-driven R&D manpower policies. Risk capital from a consolidated policy investment bank like BpiFrance. Trade architecture that companies actually use — not headline ATIGA numbers or scarcely used FTAs, but reduced non-tariff barriers and the ability to incorporate once in Singapore and operate across key Southeast Asian markets.
When companies reach a certain size, the siren song of redomicile will prove irresistible for many. The competition shifts — to building a growth-capital ecosystem deep enough and catalysing a Southeast Asian market large enough that staying makes more sense than leaving.
So sunset such an IP conditionality regime once the ecosystem is strong enough.
This will crowd-in private R&D and take the burden of sustaining our headline R&D numbers off the Singaporean taxpayer.
Yes, fewer R&D startups may form under these IP conditions. But I say, better to have fewer startups with real upside capture, than sexy headlines, nice PR pieces in CNA, and nothing to show for it. And it shifts R&D accountability to where it belongs — not on public servants approving a grant, but on the system retaining the value.
You may be very proud of your Ehime No. 28. But unless you take preparatory measures — unless you design for the upside — one day you will find your R&D being sold as a Hongmeiren mandarin. Incubated in Singapore, harvested in Zhejiang and HK.
Thank you Sir.