From DayOne to Orbital Compute: How China-Origin Deep Tech Internationalisation Is Being Structurally Rewritten

When "Singapore-registered" is no longer sufficient, and "Chinese parent holds a non-controlling stake" becomes the new minimum threshold — the diverging international fates of data centres, AI, LEO communications, and orbital compute.

Executive Summary

DayOne's reported consideration of a simultaneous Singapore and US listing looks, on the surface, like a data centre IPO story. In substance, it may be the opening signal of a new era: China-origin infrastructure companies seeking access to global capital markets can no longer rely on a Cayman structure, a VIE arrangement, or an overseas registration address. What is now required is real overseas assets, real overseas customers, real overseas governance, real capital-structure restructuring — and a Chinese parent that does not control the entity.^[1]^[2]^[3]

This is not the first time a Chinese company has attempted internationalisation. But this time is different. The tools of the previous era — VIE contracts, Cayman holdings, US ADRs, Hong Kong secondary listings — answered a specific question: how can foreign capital connect to the economic upside of a Chinese company? The question has changed. The new question is: who controls the data? Who controls the algorithm? Who controls the board? Who can unilaterally shut down the service under regulatory pressure from any single jurisdiction? VIE structures cannot answer these questions, because VIE is a capital instrument, not a trust instrument.

DayOne is the cleanest current version of the new form. GDS Holdings has carved out its international business as an independent entity, retained approximately 35.6% as a non-controlling major shareholder, no longer consolidates DayOne on its balance sheet, admitted international capital, placed operating assets outside China, and is pursuing a dual listing in Singapore and the US.^[4]^[5]^[6]
TikTok is the coerced, extreme version: ByteDance's stake has been reduced to approximately 19.9%, with US and global investors collectively holding 80.1%, and Oracle, Silver Lake, and MGX among the reported significant shareholders; the algorithmic IP remains with ByteDance, but operational control has been forcibly restructured.^[9]^[10]
Manus is the most important counterexample: the company had relocated to Singapore, but Chinese regulators determined that its core AI technology constituted a sensitive asset and blocked Meta's reported $2 billion acquisition offer.^[11]^[12]^[13]

These three cases together define a structural pattern:

The Chinese parent retains economic upside; the overseas entity earns governance credibility.
The new compromise is not "no Chinese shareholders" — it is "no Chinese control."

For Singapore, this is a value upgrade, not a value loss. But the upgrade is conditional: Singapore must move from being a "holding company jurisdiction" to a verifiable Singapore Trust Layer — with real teams, real operations, real data boundaries, real governance structures, real capital market access, real regional customers, and a commercial language that capital markets and sovereign buyers on both sides of the Pacific can read.

For orbital compute, LEO communications, space-AI, remote sensing, launch services, and sovereign AI companies, this shift will arrive earlier, sharper, and with higher stakes than it has for data centre companies. The closer an industry sits to AI, space, communications, remote sensing, semiconductors, data, and dual-use defence attributes, the less it can rely on a registration-address story. What is required instead is real control restructuring, data segregation, use-case boundaries, audit capability, and deployment architectures that non-Chinese customers can accept.

The core judgements in this report are:

DayOne demonstrates that Singapore architecture still has value; Manus demonstrates that Singapore washing is no longer sufficient; TikTok demonstrates that politically-forced de-control is becoming a template; the next generation of sensitive infrastructure — orbital compute and its peers — must be designed from day one for auditability, controllability, localisation, and multi-jurisdictional operability.

The relevance of this framework to orbital compute and LEO communications is direct: those categories represent the hardest case of all. Among all sensitive technology classes, orbital compute combines the greatest difficulty in restructuring control, the highest segregation costs, and the least clearly defined data boundaries. DayOne's data centre model provides the starting point. The Manus AI lesson provides an upper bound. Orbital compute — layering on-board processing, remote sensing, communications, dual-use attributes, and inter-satellite links — is the extreme scenario in which every variable of this framework appears simultaneously. Understanding Singapore architecture at the data centre layer is a necessary prerequisite for understanding why the orbital compute layer is harder, and why it must be designed that way from the beginning.

1. DayOne Is Not an Ordinary IPO Story

The news that DayOne is considering a dual listing in Singapore and the United States should not be read merely as the Singapore Exchange competing for a large IPO. It is better read as a structural signal: China-origin AI infrastructure assets are searching for a new form that can be, at minimum, tolerated simultaneously by a Chinese parent, US capital markets, Singapore, and international customers.

According to media reports, DayOne is considering a simultaneous listing in Singapore and the US, potentially raising $5 billion. The Financial Times reported that this could value DayOne at close to $20 billion and make it one of Singapore's most significant IPOs in years. Business Times reported that DayOne initially considered a sole New York listing, but SGX persuaded it to consider a co-listing.^[1]^[2] This sequence matters: it indicates that DayOne's core valuation anchor remains the US capital market, and the Singapore listing is an additive rather than a substitutive element.

The significance of this news lies in the fact that DayOne's origins are not "cleanly" Singaporean. It traces back to GDS Holdings — a Shanghai-founded data centre company that operates high-performance facilities in China for hyperscale cloud service providers, large internet companies, financial institutions, and telecoms carriers.^[4] GDS subsequently carved out its international business, initially as GDS International; in January 2025, that business was renamed DayOne.

The renaming is not the point. The structural change is. GDS's public filings indicate that it holds approximately 35.6% non-controlling equity interest in DayOne; in its 2025 annual documents, GDS stated that following DayOne's completion of a Series B round in December 2024, GDS's stake stood at approximately 35.6% and it no longer consolidates DayOne.^[4]^[5] In January 2026, GDS announced the sale of a portion of its DayOne shares, stating that the transaction allowed it to recover approximately 95% of its original investment and realise close to a 6.5x return; meanwhile, GDS's remaining DayOne stake implied a value of over $2.2 billion at the Series C issuance price.^[6]

The essence of the DayOne case: the Chinese parent has not disappeared, and has not retained control; the overseas business is not a shell; international capital is not investing in a registration address; Singapore is not a fig leaf, but the load-bearing layer for governance, fundraising, customers, regional infrastructure, and legal order. This is more complex, and more real, than traditional internationalisation.

DayOne completed a more than $2 billion Series C equity financing in January 2026, and has hired JPMorgan and Morgan Stanley as US IPO underwriters.^[39]^[40] The underwriter lineup is itself a signal: JPMorgan is the traditional gatekeeper for Chinese company US IPOs, and its participation is an indirect endorsement of the investability of this type of asset in the American market.

On the Singapore side, MAS announced regulatory framework enhancements in April 2026 to facilitate dual listings on SGX alongside Nasdaq; Singapore's parliament passed amendments to the Securities and Futures Act on 7 May 2026 to support the establishment of a Global Listing Board (GLB).^[28]^[29]^[41] DayOne has become the most representative candidate under the GLB framework — its "Singapore headquarters + US listing" dual-venue model aligns precisely with the GLB's "SGX-Nasdaq bridge" positioning.

The long-term success of the GLB depends on one condition: whether it can consistently attract "DayOne-class" companies — assets with real international operations, non-China-controlling structures, and regional market appeal. If the GLB becomes a fallback option only for small or mid-sized Chinese companies, its liquidity depth and valuation levels will not be competitive with the Nasdaq main board.

2. From GDS to DayOne: Engineering in China, Business in Singapore

To understand DayOne, you must first understand GDS.

GDS's core business is data centre infrastructure. It does not build foundation models, and it is not a SaaS company. It provides data centre capacity, connectivity, and operations to hyperscale cloud, internet, financial, telecoms, and enterprise customers. In the AI era, data centres have transformed from "real estate plus power plus networking" into bottleneck assets for AI compute growth. Land, power, cooling, redundant supply, submarine cable access, customer pre-leasing, capital expenditure, and green energy pathways have all become valuation variables.^[4]^[20]^[21]

GDS accumulated development and operating experience in China, but the Chinese market itself has come under pressure from valuation discounts, policy uncertainty, and capital market restrictions. At the same time, demand for data centres in Southeast Asia, Japan, Hong Kong, Europe, and the broader Singapore region has grown rapidly. DayOne was born in this context.

The most noteworthy aspect of DayOne's website is not "headquarters in Singapore" — it is the emphasis on SIJORI: Singapore, Johor, and Batam forming a complementary digital infrastructure region. Singapore provides finance, customers, connectivity, and trusted governance; Johor and Batam provide land, power, cost, and expansion capacity.^[7] This triangular architecture has a strong structural parallel to the three-tier topology common in LEO satellite communications — "core gateway plus regional site plus edge node" — with Singapore as the core control layer and Johor and Batam as the capacity-bearing layer.

This gives Singapore a cleaner future positioning: Singapore does not need to absorb all heavy assets, but it can absorb the control layer, capital layer, customer layer, compliance layer, and regional architecture layer. This logic maps directly to AI data centres, sovereign AI, orbital edge compute, and satellite ground segments.

DayOne's formula compressed to a single line:

China-origin engineering + overseas infrastructure assets + Singapore governance + global capital + hyperscaler demand = investable international platform.

Whether this formula can be replicated in more sensitive industries requires layer-by-layer analysis.

Looking at the equity evolution trajectory, GDS's move from controlling to non-controlling shareholder is an active, commercially-driven "de-China-control" process, not a forced adjustment. GDS holding approximately 35.6% with no consolidation means that DayOne's P&L, governance, and strategic decisions are already independent of the Chinese parent. GDS's subsequent share sale to realise a 6.5x return further reduced its holding and confirmed that this is a commercially-motivated reduction, not a symbolic gesture.^[6]

GDS Holdings' own Nasdaq background — listed since 2016 — also provides a reference point for this separation. The split between its China and international businesses represents, in part, a proactive adaptation to the CSRC's overseas listing registration regime introduced in 2023.

3. The Old World: VIE, Cayman, ADRs, and China's Era of Overseas Capital Access

The history of VIE structures traces back to Sina's Nasdaq listing in 2000. As the pioneering architecture for Chinese internet companies accessing overseas capital markets, VIE helped hundreds of Chinese companies reach US investors over more than two decades. The fundamental logic of a VIE is that the overseas listed entity controls the economic interests of the onshore operating entity through contractual arrangements, rather than direct equity ownership. Alibaba's IPO prospectus explicitly disclosed its VIE structure.^[22]^[23]

VIE resolved a core tension of its time: Chinese companies needed dollar capital and US valuations; certain industries did not permit direct foreign equity ownership; so contractual control substituted for equity control. Nankai University finance professor Tian Lihui has described the essence of red-chip structures as "cross-border legal-entity isolation" and "regulatory arbitrage."^[42]

But VIE resolved the questions of foreign access and listing economics — not the questions that are most critical today: who controls the data? Who controls the algorithm? Who controls the board? Who controls security updates? Does Chinese regulation accept this? Does the US national security apparatus accept it? Will customers believe this is not a black box? In short, VIE is a capital structure instrument. It is not a trust structure instrument.

The Ant Group lesson: Ant Group originally planned approximately $37 billion in simultaneous Hong Kong and Shanghai listings. The listings were halted on the eve of trading. On the evening of 2 November 2020, four financial regulatory bodies summoned Ant's actual controller Jack Ma, chairman Eric Jing, and president Simon Hu for regulatory interviews; the Shanghai Stock Exchange announced the pause of the STAR Market listing that night, followed thirty minutes later by Ant announcing a pause of its Hong Kong listing.^[18]^[46]^[47] The CSRC's statement was extremely measured but profoundly consequential: "significant changes in the regulatory policy environment" — this was not a decision about Ant alone, but a prelude to the restructuring of the entire fintech regulatory framework.

The Didi lesson: Didi listed on the NYSE on 30 June 2021. The Cyberspace Administration of China initiated a cybersecurity review on 2 July 2021; "Didi Chuxing" was removed from app stores on 4 July; seven government agencies conducted a joint investigation.^[17]^[48]^[49] On 21 July 2022, the CAC imposed a RMB 8.026 billion fine for violations of the Cybersecurity Law, Data Security Law, and Personal Information Protection Law, along with personal fines of RMB 1 million each on CEO Cheng Wei and President Jean Liu.^[19]^[50] Didi ultimately delisted from the NYSE.^[38]

These two cases together transmitted a clear signal: overseas listings involving financial infrastructure, platform data, and large user bases are no longer purely commercial decisions. They have become matters of state governance.

In 2023, the CSRC implemented the Trial Administrative Measures of Overseas Securities Offering and Listing by Domestic Companies, establishing a registration regime for overseas listings by domestic enterprises and codifying a "substance over form" principle.^[24]^[25]^[43]^[44]^[45] The measures specified that an indirect overseas listing by a domestic company must register with the CSRC if any single indicator — principal revenue, profit, total assets, or net assets — accounts for more than 50% of the consolidated statements and principal operations are conducted onshore. Xinhua characterised this shift as "red-chip structures moving from 'preferred channel' to 'cautious exception.'"^[42]

This phrase is decisive: substance over form. It means that registration address, listing venue, VIE documents, Cayman company, and Singapore entity are none of them the final answer. Both regulators and capital markets will ask: where is the substance?

4. The New World: Capital Is No Longer Neutral, Control Is the Core Variable

The old logic of global capital assumed that as long as a company could grow, its financials could be audited, and its structure was compliant, capital could flow across borders. The new logic is different:

Capital itself has been securitised for national security purposes.

The US Treasury's outbound investment program explicitly covers three national security technology domains: semiconductors and microelectronics, quantum information technology, and artificial intelligence. Related rules took effect on 2 January 2025, aimed at restricting US capital and associated intangible benefits from flowing to entities in countries of concern, including mainland China, Hong Kong, and Macau.^[26] CFIUS represents the other end, reviewing foreign investments in the US for national security implications.^[27]

China has simultaneously strengthened its oversight of overseas listings, data, cybersecurity, technology exports, and sensitive transactions. The result is a pincer movement in both directions: the US restricting capital inflows to Chinese sensitive technology; China restricting sensitive technology outflows. The middle ground — Singapore, Hong Kong, Cayman, and similar jurisdictions — was once a grey channel for navigating bilateral regulation, but the Manus case demonstrates that this channel is narrowing. Looking through registration addresses to examine substance has become the shared tendency of investment review in both directions.^[87]

The Manus case is the sharpest new example. Timeline:

Note: Meta's acquisition offer was never completed. The December 2025 "announcement" was a statement of intent, which was ultimately blocked.^[55]^[56]^[57]

The regulatory basis for the prohibition had three dimensions:^[56]^[57]^[58]

First, core technology outflow. Manus's agent orchestration algorithms, task scheduling engines, and other core technologies were all developed in China and fall within the scope of China's Catalogue of Technologies Prohibited or Restricted from Export. DeHeng Law Offices partner Liu Anbang noted: "if Manus's core technology had been transferred overseas without authorisation, it would likely constitute a direct violation of China's technology export control regulations."^[58]

Second, data security risk. Massive interaction data generated by Manus's early Chinese users and used in model training entered an overseas system when the company relocated to Singapore, apparently bypassing required data export security assessments. Yingli Law Firm partner Xia Bikang observed that "Manus may have crossed regulatory red lines as early as when it re-domiciled to Singapore."^[58]

Third, the regulatory red line of "shell-washing" internationalisation. The pattern — incubated in China, Singapore re-domicile, sold to an American giant — was characterised by Chinese media as "eating Chinese dividends to earn American money."^[59]^[60] Caixin Weekly cited Zhonglun Law Firm's analysis describing this as "typical extraterritorial jurisdiction" — both the acquirer and target are overseas entities, but China's government still exercised jurisdiction.^[61]

This decision set two precedents: the first publicly disclosed case of a foreign acquisition in the AI sector blocked under the Foreign Investment Security Review Measures since they took effect; and the first "investment prohibited" ruling against an AI M&A transaction.^[55]

The Manus signal is clear for the industry: the overseas sale of a China-origin AI company is no longer a purely commercial transaction. It is a matter of national technological sovereignty. A "Singapore shell" cannot automatically sever Chinese regulatory reach.

5. The Three Forms of the New Compromise: DayOne, TikTok, Manus

DayOne, TikTok USDS, and Manus together outline different versions of a new compromise model — ranging from the most proactive to the most coerced, from the cleanest to the most fraught.

5.1 DayOne: Proactive Capital-Market-Driven De-Control

DayOne is the cleanest current version. Its core logic: the Chinese parent retains meaningful economic exposure but relinquishes control; the overseas entity earns governance credibility and valuation re-rating; Singapore provides platform identity, regulatory order, a regional infrastructure vocabulary, and a capital market bridge.^[1]^[4]^[5]^[6] The key open question: is GDS's 35.6% a stable endpoint? If GDS maintains hidden influence through a golden share or veto rights, capital markets will continue to question its substantive independence.

5.2 TikTok USDS: Politically-Forced De-Control

TikTok is the more extreme, more politicised version. According to media reports, through a new joint venture structure, US and global investors collectively hold approximately 80.1% of TikTok's US operations, with ByteDance retaining approximately 19.9% — a number that precisely falls below the US legal threshold of "less than 20% not considered foreign adversary control." Oracle, Silver Lake, and MGX are reported to be among the significant shareholders.^[9]^[10]^[63]^[64]^[65]

According to TMTPost reporting, five of the seven directors of the new joint venture must be US nationals approved by the White House as security experts; the source code ownership of TikTok's recommendation algorithm remains with ByteDance, but the new company holds "mirror access rights" to the algorithm and must retrain it from scratch on Oracle servers; the reported valuation has shrunk dramatically from earlier expectations of around $50 billion to approximately $14 billion.^[66] This is a highly sophisticated "data island separation": ByteDance retains the intellectual property of the algorithm while losing operational control over it.

TikTok demonstrates that for high-sensitivity platform businesses, merely reducing the Chinese parent to a minority shareholder may not be sufficient — it is also necessary to demonstrate that data, algorithms, operations, and security guarantees are locally controllable.

Note: the specific equity ratios and terms attributed to TikTok USDS above are from media reports, not official disclosure documents, and should be treated accordingly.

5.3 Manus: The Limits of Singapore Washing

Manus illustrates the most important counterexample. A Singapore registration address provides no immunity against Chinese regulatory reach — if the company structure involves only a change of address, without real restructuring of control, data, IP, customers, team, and governance, it remains potentially high-risk in the eyes of both Chinese and overseas regulators simultaneously.^[11]^[12]^[13]

Manus was founded in Wuhan, received angel funding from ZhenFund, and subsequently brought in Sequoia China, Tencent, and Wang Huiwen.^[52] Following a Benchmark-led $75 million funding round in April 2025, it relocated its headquarters to Singapore and substantially reduced its China-based team.^[53]^[54] This trajectory was summarised by Chinese media as a classic pattern of "shell-washing internationalisation" — re-domicile, but without transferring core technology, data, or regulatory attributes.^[59]^[60]

The substantive difference between Singapore washing and Singapore architecture:

5.4 Shein: The Difficulties of Supply-Chain-Type Globalisation

Shein is another intermediate form. Founded by Xu Yangtian in Nanjing in 2008, it relocated its headquarters to Singapore in 2021 and attempted to enter overseas capital markets as a global fashion e-commerce platform; but its core supply chain, founding history, and regulatory controversies remain deeply tied to China.^[67]^[68]^[69]

Shein executive chairman Donald Tang offered a "triple identity" definition at a 2024 public event: "if judged from birthplace and supply chain, it's a Chinese company; if judged from headquarters, it's a Singaporean company; if judged from market and values, it's an American company."^[67] The confusion in that definition is itself the best annotation of Shein's struggle in regulatory arbitrage.

Shein's IPO path has been tortuous: blocked in the US (2023), redirected to London but not cleared by China's CSRC (2024–2025), and — when pivoting to Hong Kong — reportedly considering moving headquarters back from Singapore to China to secure regulatory approval (media reports from August 2025, denied by Shein).^[14]^[15]^[72]^[73]^[74] Its core dilemma: a Singapore headquarters can package the narrative, but it cannot substitute for substantive answers to questions about supply chain transparency, profit attribution, and regulatory compliance.^[70]^[71]

5.5 PDD/Temu: The Limits of Formal-Only Internationalisation

PDD Holdings' filing change — moving principal executive offices from China to Ireland — generated market attention, but PDD's spokesperson explicitly denied that this was a "headquarters relocation," emphasising that Pinduoduo's headquarters remains in Shanghai.^[16]^[75]^[76] More analytically relevant is Temu's structural design — Temu is directly incorporated as a Delaware company headquartered in Boston, Massachusetts, bypassing complex VIE layers.^[77] But structural simplification cannot eliminate the policy risk that comes with "Chinese platform" status — the Trump administration's elimination of the de minimis tariff exemption for small packages directly damaged Temu's business model, which no headquarters relocation can avoid.

6. The Essence of This Structure: Not "No China," But "No Chinese Control"

The new structure is not anti-China, nor is it complete de-Sinicisation. It is more like a complex compromise:

• The Chinese parent retains economic upside;
• The overseas entity earns governance credibility;
• International capital receives an investable explanation;
• Overseas customers receive an auditable boundary;
• Chinese regulators retain a veto over the outflow of strategic assets;
• US regulators retain a veto on national security review grounds.

The simplest formula:

China-origin capability + overseas operating substance + de-controlled governance + local data boundary + international capital = investable compromise.

But this formula only holds when it is real. If it is purely formal — registration in Singapore, real team in China, IP in China, data flows back to China, Chinese parent controls the board, overseas entity has no customers and no P&L — that is Singapore washing, and it has been proven not to work. If it is real — management and engineering teams in Singapore, independent P&L in the Singapore entity, data does not return to China, customer contracts governed under Singapore or local law, board with international governance, Chinese parent is non-controlling, IP licensing is transparent and auditable, third-party audit can be executed — that is Singapore architecture.

An under-appreciated parallel: Arm China. The story of Arm China (安谋科技) is the clearest instructive example of a China-origin joint venture structure losing control — and the inverse reference for DayOne. In 2018, Arm carved out its China business into a joint venture, Arm China, with Arm holding approximately 49% and Chinese investors approximately 51%. Subsequently, the two sides entered into serious conflict over the appointment and removal of the CEO, profit distribution, and technology licensing. Arm China CEO Allen Wu refused to step down for two years, leaving Arm unable to control the day-to-day operations of its own Chinese subsidiary — an extreme case in which "legally non-controlling" became "substantively without control." Arm China's lesson: when control boundaries are unclear, the "joint venture plus non-controlling" structure itself becomes the source of risk. For DayOne to succeed, it must clearly define whether GDS's 35.6% is accompanied by any form of hidden control arrangement.

7. Singapore's Historical Role Shift: From Port to Holding Company to Trust Layer

Singapore's historical role has always been as a middle layer.

The first phase: trade port — goods, finance, shipping, insurance, and contracts converging here. The second phase: regional headquarters — multinationals placing their Southeast Asia and APAC headquarters in Singapore. The third phase: capital and tax structure node — funds, family offices, holding companies, and IP management taking form here. The fourth phase, now beginning: trusted control layer for digital infrastructure and sovereign technology.

Singapore's government is reopening data centre capacity. IMDA's Green Data Centre Roadmap targets at least 300MW of additional capacity in the near term; EDB and IMDA launched a second round of Data Centre Call for Application in December 2025, releasing at least 200MW of capacity.^[20]^[21] In parallel, MAS, SGX, and Nasdaq are advancing a dual listing framework.^[28]^[29]^[33]^[34]^[35]

Singtel's RE:AI partnership with Mistral AI demonstrates another sovereign infrastructure pathway: Singapore's local telecoms operator, data centre capacity, and sovereign AI cloud combined with European open-model capability, serving finance, manufacturing, public sector, and healthcare sectors.^[30] This tripartite structure — European technology, Singapore governance, Asia-Pacific market — provides a replicable template for non-US-non-China technology collaboration.

But Singapore is not the only jurisdiction attempting to become the credible deployment location for China-origin technology:

Dubai/Abu Dhabi (UAE) — through sovereign funds like MGX and the ADGM free zone structure, the UAE is actively competing for China-origin AI and fintech companies. MGX is reported to be among TikTok USDS's investors.^[65] UAE advantages: zero personal and corporate income tax, relaxed immigration policies, and strong bilateral relations with China. Disadvantages: legal system maturity below Singapore's, FATF compliance history concerns, and a position far outside the Five Eyes trust circle.

Hong Kong still plays a central role in Chinese company overseas listings, but its geopolitical neutrality has been affected by international perceptions since the National Security Law, and its independence as a data governance framework is constrained.

Australia, within the Five Eyes framework, offers "highest trust but narrowest channel" — serving Western government or defence customers almost requires this path, but the market is small and costs are high.

By comparison, Singapore's differentiated advantage lies in being the only jurisdiction that simultaneously satisfies common law, deep financial markets, geographical and cultural proximity to China, a neutral geopolitical positioning, credible regulatory institutions, and ASEAN-hub regional market access. This combination is difficult to replicate by any competitor in the near term. But "difficult to replicate" does not equal "requires no maintenance" — Singapore must demonstrate the value of its trust layer in every concrete case.

8. Industry Sensitivity Classification: Not All China-Origin Internationalisation Is the Same

Different industries have different requirements for de-control and Singapore architecture. A useful classification framework:

8.1 Low to Medium Sensitivity: Consumer Brands, General E-Commerce, Cross-Border Supply Chain

Representative cases: Shein, PDD/Temu, Miniso, Pop Mart.

Primary risks: labour and supply chain transparency, consumer data, tariffs and small-parcel policy, platform governance, product safety, China supply chain dependency.

Singapore's value: headquarters relocation can improve the narrative, but supply chain, supplier origins, and profit attribution will still be examined through. Consumer brands need global supply chain compliance certification and localised narrative capability, not a Singapore control layer.

Conclusion: Internationalisation is feasible, but "headquarters relocation" is insufficient to eliminate political risk. Singapore is a valuable base, but not a silver bullet for supply chain controversies.

8.2 Medium Sensitivity: Data Centres, Cloud Infrastructure, Industrial AI, Enterprise Software

Representative cases: DayOne, Singtel-Mistral.

Primary risks: data sovereignty, customer sensitive data, hyperscaler dependency, cybersecurity, capital source scrutiny.

Singapore's value: real headquarters plus international investors plus non-controlling Chinese parent plus customer data not returning to China plus local contracts and SLAs plus third-party security audit plus regional infrastructure positioning. This is the layer where Singapore trust layer value is most direct and most fully realised.

Conclusion: The DayOne template is most applicable at this layer. Singapore can provide substantive value, provided there is real substance.

8.3 High Sensitivity: AI Agents, Foundation Models, Automated Decision-Making, Sovereign AI

Representative cases: Manus, TikTok.

Primary risks: algorithmic control, training data, content governance, model weights, talent and source code, national security, foreign acquisitions, export controls.

Singapore's value: local deployment plus customer-controlled keys plus model or software audit plus non-controlling Chinese parent or further isolation. But a Singapore registration address alone is entirely insufficient — Manus has already demonstrated this.

Conclusion: Registration address is essentially irrelevant; control and data boundaries are the core. AI company internationalisation must be planned from the product architecture design phase for auditability, not restructured on the eve of an IPO.

8.4 Extreme Sensitivity: LEO Communications, Orbital Edge Compute, Remote Sensing, Launch, Satellite Control

This is the section most directly relevant to this report's readership — and the one with the least developed playbook.

LEO Communications Networks

LEO communications networks serve national broadband, maritime, aviation, emergency response, defence, and remote area coverage. They are, in substance, non-terrestrial communications infrastructure, not ordinary network services. Customers will ask: where are the gateways? Through which jurisdiction does user data flow? Who holds the regulatory licences? Can the terminal be shut down? Can the parent country's government influence the network?

The Qianfan (G60) constellation presents a case fundamentally different from data centres. Its international expansion follows a dual-track "government-to-government agreement plus commercial services" path, not capital structure restructuring. Target customers are primarily state-owned telecoms in target countries, with contracts negotiated at the G2G level:^[81]^[82]^[83]^[84]

Qianfan's model is "sovereign to sovereign" — it does not depend on Singapore architecture, but on bilateral agreement frameworks between the Chinese government and target governments. Advantage: no complex equity restructuring required. Disadvantage: virtually zero possibility of entering Five Eyes markets or serving high-sensitivity Western customers, and launch costs (currently approximately $7,000–8,000 per kg, potentially down 30% with liquid rocket deployment) remain above SpaceX Falcon 9 levels (approximately $3,000–5,000 per kg).^[85]

For China-origin LEO companies seeking to enter international markets, Singapore's possible roles include: APAC compliance and market entry, neutral business headquarters, maritime/aviation managed services, multi-vendor architecture layer, customer data and key-control layer. What Singapore cannot do is make a China-origin LEO core network acceptable to Five Eyes or high-sensitivity customers — this is not a registration address problem, it is a question of technical control and network sovereignty.

Orbital Edge Compute / Orbital Compute

Orbital compute is more sensitive than data centres because it layers together AI, space, on-board processing, remote sensing, communications, data sovereignty, inter-satellite links, dual-use attributes, and export controls. This is the intersection of multiple "extreme sensitivity" dimensions.

The most dangerous narrative is: a Chinese company building a global space data centre to serve overseas sovereign customers. The more realistic pathway is: China-origin engineering capability plus Singapore-auditable mission software, ground segment, data governance, and edge workflow, deployed in APAC for non-sensitive use cases.

Suitable for early internationalisation: on-board data compression, remote sensing pre-processing, disaster response, maritime monitoring, agriculture and energy use cases, mission control software, open architecture devkits, testing and validation, optical ground station coordination, non-sensitive orbital edge SDKs.

Not suitable for early engagement: military high-resolution remote sensing, RF reconnaissance, sovereign communications control layers, encrypted communications cores, high-sensitivity AI models, restricted-export chips, propulsion and launch control, sanctioned-country customers.

For orbital compute companies following the DayOne template, the requirements must be stricter, more transparent, and more modular than DayOne's — because the technology combination itself carries higher dual-use attributes. Investors and partners will ask not only "where is the data" but also: "do the on-board processors use export-controlled chips?", "is the inter-satellite link protocol independently controlled by the overseas entity?", "does remote sensing data resolution exceed the commercial service threshold?" These questions must have clear answers at the architecture design stage — not at the pre-IPO restructuring stage.

The DJI warning: DJI is a case that no registration address and no equity restructuring can resolve. DJI's core drone technology, market share, and government customer controversies have led to successive listings on the US Entity List and the Department of Defense special entity list. This outcome was not caused by poor corporate architecture design — it was determined by the dual-use attributes of the technology category itself. For orbital compute companies, an honest early self-assessment is warranted: if we are essentially DJI, would a Singapore architecture change any outcome? If the answer is no, then the most important work is not structural reorganisation, but proactive self-limitation of use-case boundaries.

Launch Market

Launch is a highly sensitive industry. Rockets, propulsion, guidance, launch sites, and orbital insertion capability all carry deep national security relevance. The likelihood of China-origin launch companies using Singapore to access US capital or sensitive markets is extremely low.

Singapore's viable role: insurance, finance, legal contracts, commercial launch brokerage, non-sensitive rideshare coordination, APAC customer development, payload integration, post-launch data services.

Conclusion: Launch hard capability is not suitable for "Singapore architecture" packaging, but launch-adjacent commercial services are.

Remote Sensing and Earth Observation

Remote sensing sensitivity depends on resolution, use case, customer type, and data processing location.

Low-sensitivity scenarios: agriculture, flooding, forest fires, port congestion, marine environment, insurance claims, ESG monitoring. Singapore can serve as a platform for data governance, commercial analytics, maritime insurance, disaster resilience, ASEAN customer layer, and audit and end-use screening.

High-sensitivity scenarios: military targeting, high-resolution real-time surveillance, RF/SAR dual-use military applications, sanctions evasion, critical infrastructure monitoring. Singapore architecture cannot provide adequate protection for these scenarios.

Conclusion: Viable, but use-case segmentation is essential, and high-sensitivity scenarios should not be approached.

9. Fragile Balance, or New Normal?

The answer is: fragile new normal.

It is becoming a new normal because none of the parties have better alternatives: US capital still wants exposure to high-growth infrastructure; Chinese companies still want overseas valuations and international customers; Chinese regulators do not want to lose control over strategic technologies; US regulators do not want sensitive technologies and data under Chinese control; Singapore wants to remain a regional headquarters, capital market node, and trusted infrastructure platform; Southeast Asian and Middle Eastern customers want alternatives to both US and Chinese-only options.

It is fragile because the equilibrium point is unstable. Trigger factors, ranked by impact weight:

Most critical triggers (structural):

• Chinese regulators determine that core technology is flowing out (the Manus pathway)
• US regulators determine that the Chinese parent retains de facto control (CFIUS review)
• Customer data flows back to China or becomes accessible to Chinese regulators

Important triggers (manageable):

• Board composition or veto rights expose hidden control
• Company serves defence or high-sensitivity customers
• Supply chain touches sanctions or export controls
• Singapore entity has no real substance

Background triggers (geopolitical layer):

• Political events alter regulatory tolerance
• Founder identity, nationality, residency, and historical funding structure attract scrutiny

This is therefore not a stable free-market equilibrium, but a dynamic structure continuously repriced by regulation, capital, and geopolitics.

Before discussing the "fragile new normal," it is necessary to distinguish two frequently conflated concepts: 35.6% as a number is transitional; "Chinese parent as a non-controlling major shareholder" as a structural category is the new normal.

On why 35.6% is transitional: TikTok's trajectory provides the clearest reference. ByteDance went from majority control to 19.9%, driven not by commercial judgement but by political pressure — when scale is large enough and customers sensitive enough, US regulatory and political dynamics will systematically compress the Chinese parent's shareholding ratio. DayOne, if successful, will face the same pressure as its hyperscaler relationships scale. 35.6% will continue to fall — not a pessimistic prediction, but a structural trend.

But this does not mean the "new normal" is an illusion. What is stabilising is the structural logic: the Chinese parent retains economic upside (at whatever shareholding ratio); the overseas entity earns governance credibility (making it acceptable to international capital and sovereign customers). This logic is stable because neither side has a better alternative — complete decoupling represents asset loss for the Chinese parent, and complete exclusion means forgoing an entire class of high-growth assets for US capital. "Fragile new normal" does not mean the structure disappears; it means that its specific numerical parameters (shareholding ratio, control boundary, data segregation degree) will continue to adjust under regulatory and geopolitical pressure.

The most accurate summary judgement:

The DayOne model will spread, but only companies with real assets, real business, real governance, and real data boundaries will pass through.
The Manus model will be treated with caution, because it proved that relocating addresses cannot automatically sever regulatory attributes.
The TikTok model will become the extreme reference point for high-sensitivity platforms: when political pressure is sufficient, de-control can be imposed by regulation.

The "pincer movement" of bilateral US-China investment restrictions is the institutional foundation of this fragility: the US restricts capital flows into Chinese sensitive technology via outbound investment rules (effective 2 January 2025); China restricts sensitive technology outflows via foreign investment security review and technology export controls (the Foreign Investment Security Review Measures in force since 2021, with the Catalogue of Technologies Prohibited or Restricted from Export continuously updated).^[26]^[56]^[87]

There is one critical question this report has not yet addressed directly: do segregation costs (friction costs) destroy the commercial logic of the non-controlling structure?

The answer depends on the industry — and this distinction is itself central to understanding why DayOne is the best starting point for this structure and orbital compute is the hardest endpoint.

Data centres (DayOne) have a physically inherent segregation advantage. The essence of a data centre business is: data is where it is stored. In data centres in Singapore and Johor, the physical boundary of customer data is determined by the server rack — no additional technical architecture is required to achieve "data does not return to China." This segregation is intrinsic and costless, enforced by physical law rather than compliance effort. DayOne's cost of implementing Singapore architecture consists primarily of governance structure and personnel localisation, not technical restructuring. This is the fundamental reason why data centre businesses have a more favourable commercial logic under this structure than AI model companies or LEO operators.

AI model companies (Manus-type) and LEO operators face a fundamentally different situation. Model weights, training data, inference infrastructure, on-board algorithms — these things cannot be isolated by physical boundaries. TikTok reportedly spent billions of dollars rebuilding Project Texas (the predecessor to USDS), retraining algorithms, rebuilding data pipelines, and reconstructing the entire infrastructure on Oracle servers. This is not a "compliance cost" — it is a partial reconstruction of the original business model. For orbital compute companies, the software update pathway for on-board processors, access permissions for ground control systems, and the downlink encryption chain for remote sensing data all require explicit segregation design, each with real engineering costs.

This distinction points to a conclusion: the DayOne model can be implemented at the data centre layer at relatively manageable cost; at the AI layer the cost is higher and the architecture more complex; at the orbital compute layer, segregation costs may be the most underestimated variable in the entire framework. An orbital compute company pursuing Singapore architecture must design segregation boundaries at the first day of product architecture — not at the pre-IPO restructuring stage, when the cost may already exceed what the business model can absorb.

10. Does Singapore Still Have a Position? Yes, But It Must Be Earned Through Substance, Not Narrative

The old Singapore value proposition included: low taxes, easy incorporation, common law, financial centre, regional headquarters, Mandarin-English bilingualism, proximity to China, proximity to Southeast Asia. These retain value, but are no longer sufficient. Singapore's new value must become operative.

Before discussing Singapore's new value, an honest statement is necessary: Singapore does not have the depth of top-tier technology talent and startup ecosystems that China and the US possess, and it has not produced world-class technology originator companies in large language models, semiconductors, or LEO constellations. This is not a deficiency — it is a structural reality.

Singapore is a city-state of 6.8 million people without China's multi-tier engineering talent pool (complete density from top researchers to operations engineers) or Silicon Valley's capital, immigration, and failure-culture ecosystem. Attempting to transform Singapore into a "miniature technology innovation hub" is neither realistic nor necessary.

This is actually Singapore's advantage for the trust layer role, not its disadvantage. The capabilities Singapore has repeatedly demonstrated are precisely what trust layers require: deep financial markets, common law enforceability, internationally credible regulators, regional market access (ASEAN 10 plus broader Asia-Pacific), legal and compliance service ecosystems, and cross-border contract execution capability. These are not capabilities led by high-tech companies — they are capabilities led by financial, legal, logistics, and government institutions.

Singapore should therefore become the credible control layer and regional market layer for DayOne-class architectures, rather than attempting to be the originator of technological innovation. Technology comes from China; engineering comes from China; but governance, capital, customer access, and data boundaries are established in Singapore. This division of labour is not Singapore's second-best option — it is its most authentic and most defensible competitive position.

If Singapore tries to exceed this role — for example by subsidising large language model development or attempting to independently develop commercial rockets — it will face competition with larger-scale capital and talent ecosystems where it has no structural advantage. The logic of Singtel's partnership with Mistral is correct: technology provided by a French company, Singapore providing the deployment environment, regulatory compliance, and regional customers. This division is worth replicating across more sectors, not questioning.

Real substance: Singapore entities must have real employees, real products, real management, real customers, real books, and real decision-making authority. Without substance, Singapore is only a shell.

Data boundary: Sensitive customers will ask where the data is, who controls the keys, who sees the logs, who pushes updates, and whether data returns to China. Singapore must provide a data isolation and customer control framework — answering "the Singapore entity holds the service contract" is not sufficient.

Governance credibility: Board composition, shareholder rights, veto rights, management appointments, audit, risk committee, and compliance officers all need to be explainable. Particularly when a Chinese parent still holds 30–40%, "not consolidated" does not equal "no control" — more granular governance proof is required.

Capital market bridge: If the SGX-Nasdaq Global Listing Board materialises, it will make Singapore a new "co-listing bridge" where companies can simultaneously access Asian and US capital rather than choosing between SGX and Nasdaq.^[28]^[29]^[33]^[34]^[35]^[78]^[79]^[80] Whether DayOne becomes one of the GLB's first companies will be a critical signal for whether Singapore's capital market value proposition can be delivered.

Regional market substance: Singapore must help companies actually enter the APAC market, not merely hang a headquarters sign. DayOne's SIJORI logic — Singapore plus Johor plus Batam forming a triangular digital infrastructure zone — is the clearest expression of this "from headquarters to market" transformation.^[7] Singapore holds the control and capital layers; Johor and Batam hold the capacity layer; the regional market covers Southeast Asia and South Asia. This has a high degree of structural similarity for orbital compute, LEO communications, AI data centres, and maritime space services — much sensitive infrastructure in the future cannot all be located on Singapore island, but it can have its architecture, contracts, governance, data boundaries, and regional market entry defined by Singapore.

Sovereign customer vocabulary: Singapore is best positioned to translate China-origin technology into the language of international procurement: mission-sensitive, auditable, sovereign data workflow, customer-controlled, multi-jurisdiction, end-use screened, open architecture, independently governed. This is especially important for orbital compute, LEO communications, AI, data centres, and maritime space services.

Singapore also needs to face its genuine limitations honestly: it cannot help any company eliminate Chinese regulatory long-arm jurisdiction if core technology and data are still deemed Chinese sensitive assets; it cannot substitute for US CFIUS review if the Chinese parent's control is in question; and it cannot provide "unlimited exemption" on technology categories themselves (such as high-resolution remote sensing, LEO control layers) — DJI-type cases demonstrate that the dual-use attributes of certain technology categories represent a fundamental problem that registration architecture cannot resolve.

11. Direct Implications for Orbital Compute Founders and Investors

If a China-origin orbital compute company wants to internationalise, the wrong question to ask is:

Can we set up a company in Singapore?

The right questions to ask are:

1. Which product lines are suitable for international markets? Which technologies must remain in China?
2. Does the overseas entity require the Chinese parent to de-control? To what degree?
3. Must data be processed locally for the customer or in Singapore? How should the data return path for on-board processing be designed?
4. Can we do an open architecture? Can customers independently audit code and hardware BOM?
5. Is third-party code, model, security, or supply chain audit required?
6. Which customers should not be pursued? Which capital sources should not be accepted?
7. Is a Singapore-based engineering team required — not just legal and finance personnel?
8. Can the overseas entity operate independently without remote control from the Chinese parent?
9. Do the on-board processors use export-controlled chips? If so, to which countries?
10. Is the company's technology category inherently "DJI-type" (no solution from registration address) or "DayOne-type" (architecture can provide a solution)?

If the first few questions do not have clear answers, spending three months thinking through these questions will deliver more value than spending three months on Cayman incorporation plus Singapore entity establishment.

DayOne's value is that it demonstrates China-origin infrastructure capability can be recapitalised. Manus's value is that it demonstrates sensitive technology cannot be laundered through a change of address. TikTok's value is that it demonstrates that when control becomes a political question, shareholding ratios will be rewritten by regulators. Arm China's value is that it demonstrates that a joint venture plus non-controlling structure, when control boundaries are unclear, can result in neither party having effective control. DJI's value is that it demonstrates that the dual-use attributes of certain technology categories are fundamental problems that architecture design cannot resolve.

Orbital compute is more complex than all of these, because it simultaneously involves every sensitive dimension that appears in the cases above.

12. Singapore Trust Layer: A More Actionable Framework

For Singapore, the genuinely valuable role is not helping Chinese companies "appear international" — it is helping them "become auditable."

A Singapore Trust Layer oriented toward space-AI, orbital edge, and sovereign infrastructure requires at minimum seven modules:

Module 1: Corporate Control Layer
Is the Chinese parent non-controlling? Are there hidden veto rights or golden shares? Does management have local decision-making authority? Is the international shareholder structure transparent? Do independent directors have genuine blocking capability?

Module 2: Data Boundary Layer
Does customer data remain with the customer locally or in Singapore? Is transmission back to China prohibited? Who controls the encryption keys? Who holds logs and audit trails? Is the data downlink path for on-board processing authorised by the customer? Has cross-border data policy been third-party audited?

Module 3: Technical Audit Layer
Code audit and SBOM (software bill of materials); hardware BOM and chip provenance; security update processes; backdoor review; model and algorithm documentation; third-party penetration testing; export control compliance review for satellite hardware.

Module 4: IP and Licensing Layer
Where is the IP? Does the overseas entity own or license it? Is the licence revocable? Can the Chinese parent unilaterally cut off technical support? Does the overseas entity have self-developed modules? Risk assessment of IP being designated a "sensitive technology asset" by Chinese regulators.

Module 5: End-Use Screening Layer
No service to sanctioned customers; no service to high-sensitivity military applications; no service to restricted remote sensing missions; no supply of encrypted communications cores; clear commercial/civil/research use cases defined; customer qualification certification process established.

Module 6: Capital Eligibility Layer
Which investors can invest? Does US capital trigger outbound rules? Is it suitable for Middle Eastern sovereign capital? Is it suitable for strategic industrial capital? Is it suitable for customer-funded pilots? Will the nationality and ultimate beneficial ownership of investors trigger CFIUS?

Module 7: APAC Market Layer
Singapore as regional headquarters; Johor, Batam, Australia, Middle East as infrastructure or market extensions; Southeast Asian customers; maritime, energy, disaster response, insurance, agriculture, and port commercial scenarios; local partners and local contracts.

This is not a legal services menu — it is a strategic productisation framework. It helps companies answer business structure questions before the legal opinion stage, and transforms "what Singapore can provide" from a vague narrative into a testable specific checklist.

13. Core Judgement: Singapore Washing Is Dead; Singapore Architecture Is Just Beginning

In the past, many Chinese companies used Singapore for three things: as a tax and registration address; as packaging for overseas fundraising; and as a narrative tool for "we are not a Chinese company." That era is ending: Manus demonstrates that regulators will look through registration addresses; Shein demonstrates that supply chains will look through headquarters; Didi and Ant demonstrate that data and financial infrastructure will look through listing structures; TikTok demonstrates that content and algorithms will be forced into control restructuring; Arm China demonstrates that a joint venture structure with unclear control boundaries can result in both parties losing effective control.

But Singapore has not lost value. On the contrary, Singapore's value has increased — it has simply become harder to earn.

Singapore's future role is not to help Chinese companies appear less Chinese — it is to help China-origin technology and infrastructure capability become something that international capital and sovereign customers can understand, audit, and accept. This is a higher-order, heavier, more demanding, but also more defensible role.

Singapore's greatest risk is not being distrusted by both China and the US simultaneously (this risk is real but manageable) — it is that at the critical moment of differentiating "Singapore washing" from "Singapore architecture," it accepts too many unqualified cases and dilutes its own credibility. Every Singapore-registered company that is seen through by Chinese or US regulators will damage the "Singapore trust layer" narrative as a whole. This is also a real pressure on DayOne's IPO: if it succeeds, it will provide a template for those who follow; if it runs into regulatory difficulty, it will cast a shadow over the entire framework.

14. Final Conclusion: The New Compromise Is Not the Destination — It Is a New Battleground

DayOne is giving the market a lesson: overseas capital has not completely closed; China-origin does not mean uninvestable; a registration address alone is insufficient — you must be an architecture designer; the more sensitive the technology, the earlier you must design control, data flow, customer use cases, and audit architecture.

But it is also telling everyone: this balance is not stable.

DayOne has reached its current position because it sits at the AI infrastructure and data centre layer, with real assets, real overseas business, a non-controlling Chinese parent, and an internationally explainable capital structure. TikTok can continue to operate because it was forced, under political pressure, to restructure US operational control — at the cost of billions of dollars in valuation discount and two years of political uncertainty. Manus was blocked because the Singapore relocation failed to convince Chinese regulators that core AI technology had ceased to be a Chinese sensitive asset.

For Singapore, the true strategic opportunity is:

Move from holding jurisdiction to trust architecture.

For China-origin companies, the real question is:

You can retain the economic interest of the Chinese parent, or you can retain Chinese parent control — but in sensitive industries, having both is increasingly difficult.

For orbital compute, LEO communications, space-AI, and the next generation of sovereign digital infrastructure, this question will arrive earlier and more sharply than it has for data centre companies — because the technology attributes of these industries are themselves closer to the core of national security. DayOne provides a template, but the template orbital compute requires must be stricter, more modular, and more transparent than DayOne.

This report's final judgement:

Over the next five years, the internationalisation of China-origin sensitive technology companies will be neither simple overseas expansion nor complete decoupling — it will be a series of real, difficult, auditable control restructurings.
Singapore still has a place, but it must upgrade from "middle ground" to "trusted control layer."

These two sentences may be proven overly optimistic in five years — if US-China relations deteriorate further, the middle ground will continue to narrow. They may also be proven overly pessimistic — if more DayOne-type successes appear, Singapore architecture will become a robust enough paradigm. Whatever the outcome, one point is certain: in orbit, there is no neutral ground. Between the US and China, there is no permanent grey space. The companies that survive will be those that understood this earliest and began, from the first day, to design their control boundaries seriously.

Key Nodes to Watch

1. Whether DayOne completes its Singapore and US dual listing. If successful, it will become the landmark case for China-origin infrastructure de-control. If it runs into difficulty, the Manus lesson will be reinforced again.

2. Whether GDS continues to reduce its DayOne shareholding. If reduced below 20%, DayOne's "China association" will diminish substantially and the capital market narrative will become clearer. If maintained at 35.6% with hidden rights, it will become a sustained focus of market scrutiny.

3. How the CSRC handles DayOne's overseas listing registration. DayOne, as a company incorporated outside China with overseas operations and a non-controlling Chinese parent, will have its registration outcome directly reflect Chinese regulators' attitude toward the "proactive de-control" model.

4. Whether TikTok USDS can maintain US regulatory trust on data, algorithm, content governance, and software assurance. This is the extreme sample of de-control for high-sensitivity platforms, and its operational status will serve as the reference point for subsequent similar architectures.

5. The final disposition of the Manus transaction, and whether China establishes a clearer approval mechanism for overseas acquisitions of AI companies. This will determine the international boundary conditions for China-origin AI companies.

6. Shein's final IPO destination. Regulatory look-through of a consumer platform will continue to test the effective limits of registration address strategy.

7. The profile of the first companies on the SGX-Nasdaq Global Listing Board. If DayOne-class companies appear, Singapore's capital market bridge value will increase substantially; if only small companies list, the GLB's strategic value will be limited.

8. Whether the US outbound investment rules expand to cover broader AI infrastructure, space, satellite communications, or data centre-related sectors. This will directly affect US capital's assessment of China-origin infrastructure company investability.

9. Whether orbital compute companies begin proactively adopting Singapore trust layer structures. This is the critical verification point for whether this report's judgement extends from data centres to space-AI.

10. Whether China establishes clearer approval mechanisms for overseas fundraising involving AI, spatial computing, commercial space, and data infrastructure. After the Manus case, the industry is waiting for clearer regulatory boundaries — this will have a greater effect on the overall pace of internationalisation than the current "case law" approach of ad hoc decisions.

This report is compiled by the Singapore Space Agency independent research team. Views and analysis are based on publicly available sources and do not constitute legal, tax, securities issuance, investment, or regulatory compliance advice. Matters involving overseas listings, foreign investment access, cross-border data flows, export controls, national security reviews, and equity restructuring require independent assessment by qualified lawyers, investment banks, auditors, and regulatory advisors.

Singapore-registered entity · UEN 53448796C · spacesgp.com