China didn’t solve the branch campus problem. It refused to have one.

The Branch Campus Trilogy, Part 2 of 3 - What the SFCU model actually built — and what compounded over twenty years

Part 1 of this series ended with a question. One country looked at every branch campus failure across Malaysia, Qatar, and Singapore and built something structurally different. Which country? I asked readers to guess in the comments.

My visiting professor who taught us branding, Sandeep Budhiraja,  got there within an hour. China, with State-directed co-creation.

“State-directed co-creation” is a better frame than most people would land on — it’s closer to the actual mechanism than the usual “China copies everything” or “authoritarian advantage” narratives. But it still undersells the specificity of what China decided in 2003. The decision wasn’t ideological. It was architectural. And the architecture is what compounded over twenty-one years into something no branch campus model anywhere else has produced.

Here is what China built, and why it matters for what India is doing right now.

The 2003 decision

In September 2003, China’s State Council brought into force the Regulations on Chinese-Foreign Co-operation in Running Schools. The timing is worth sitting with. Malaysia’s branch campus experiment was five years old and already splitting into survivors and a long tail of single-faculty outposts. Qatar’s Education City had just opened. The model the rest of the world was running: host country invites foreign university in, offers tax incentives and access to local students, lets them operate however they see fit.

China looked at that model and said no.

Not publicly, not ideologically. It said no through regulatory design. The 2003 regulations did not prohibit foreign universities from operating in China. They prohibited branch campuses. What they permitted — and required — was a joint venture between a foreign university and a Chinese institution: joint governance, dual degrees, a research mandate, and a purpose-built campus.

Four requirements. No other host country imposed all four simultaneously.

First: the Chinese university had to be a co-equal governance partner, not a facilitating host. The foreign university did not have unilateral exit rights the way Yale did in Singapore. Both partners owned the institution jointly, which meant both had skin in the game and neither could walk away cleanly.

Second: every undergraduate student had to receive two degrees — one from the foreign partner, accredited in their home country; one from the Chinese Ministry of Education. This made graduates portable globally while keeping them Chinese-credentialled. It also meant Beijing had a direct stake in the quality of what was being taught on its soil.

Third: research was not optional. The regulations required Sino-Foreign Cooperative Universities — SFCUs — to operate as research institutions, not teaching outposts. This single requirement is what separates the SFCU model from Reading Malaysia’s single-faculty real-estate programme, from Michigan State Dubai’s 85-student experiment, from the entire long tail I documented in Part 1.

Fourth: purpose-built campuses. Not shared floors in corporate buildings. Not converted office parks. Labs, libraries, residential colleges — the physical infrastructure that serious research institutions require.

Any institution that cleared these four requirements was, by construction, a survivor. The three traits that determine branch campus survival — physical infrastructure, research orientation, host-country co-investment — were not optional outcomes of the SFCU model. They were entry requirements.

Figure 1. Branch campus model vs China’s SFCU requirement across the four structural dimensions.

What the model produced

The first SFCU was the University of Nottingham Ningbo China — UNNC — which opened in autumn 2004. Nottingham signed the joint venture agreement with Zhejiang Wanli Education Group in March 2004, cleared Ministry of Education approval, and enrolled its first cohort before the year was out. It was the first foreign university to receive legal status as an independent campus in China.

Two years later, Xi’an Jiaotong-Liverpool University — XJTLU — opened in Suzhou in 2006 with 164 students. The partners were Xi’an Jiaotong University — one of China’s original nine national key universities, founded in 1896 — and the University of Liverpool. Where XJTLU chose to open matters, and I will come back to it.

China then did something Malaysia and Qatar never did. It paused. No further SFCU approvals were issued until the government had evaluated how UNNC and XJTLU were actually performing. Seven more followed over the next thirteen years: NYU Shanghai (2011), Wenzhou-Kean (2011), Duke Kunshan (2013), CUHK Shenzhen (2014), Guangdong Technion-Israel Institute of Technology (2015), Shenzhen MSU-BIT (2017). Nine total. Thirteen years. Each one assessed before the next was approved.

Malaysia licensed thirty-plus branch campuses in ten years. China licensed nine SFCUs in thirteen. Now look at what the nine built.

Figure 2. Three pacing strategies, three outcomes. China approved two institutions, evaluated them for five years, then continued.

XJTLU today has approximately 9,000 students on campus in Suzhou and a further 3,000 completing degrees at the University of Liverpool. It has published 11,851 scientific papers with over 159,000 citations. Every undergraduate earns two degrees. Its scores on the Gaokao — China’s national university entrance examination, the single all-or-nothing exam that determines university placement for every school-leaver in the country — run 40 to 60 points above the cutoff for key national universities. China’s best students are choosing XJTLU over conventional national universities. That is not a participation trophy. That is a revealed preference from students with options.

Duke Kunshan in its latest admissions cycle drew 11,884 applications for 550 places, including 5,882 international applications from 150 countries. UNNC’s science admission scores in 2016 ran more than 60 points above the key university cutoff. These are not institutions absorbing students who couldn’t get in elsewhere. They are institutions students are actively choosing over their alternatives.

The detail most analyses miss

Here is what I believe most coverage of the SFCU model gets wrong by omission.

XJTLU is not in a university town. It is in Suzhou Industrial Park — a purpose-built economic zone roughly an hour from Shanghai, housing more than 5,000 international enterprises and Fortune 500 R&D centres. The park was designed from inception as a Singaporean-Chinese government collaboration to create a high-technology manufacturing and innovation cluster. Think of it as the Indian equivalent of HITEC City in Hyderabad — except built with the explicit intent of mixing industrial R&D with academic infrastructure. XJTLU didn’t locate adjacent to industry. It located inside it.

XJTLU’s research strategy explicitly orients toward Suzhou Industrial Park. Engineering, computer science, and business students do internships at firms inside the park. The applied research agenda is shaped by the industrial problems those firms are trying to solve. The university is embedded in its ecosystem the way an IIT would be if it were built inside HITEC City, not in Powai.

The other SFCUs follow the same logic. Duke Kunshan is twenty minutes by high-speed rail from Shanghai. NYU Shanghai is in Pudong, the financial and technology district. CUHK Shenzhen is in the city that houses Huawei, DJI, Tencent, BYD, and the largest concentration of electronics manufacturing in the world. None of these locations are accidents. They are deliberate placements in China’s most innovation-dense industrial clusters.

The contrast with GIFT City is stark. India’s seventeen are located near financial services infrastructure, not technology or manufacturing R&D. Even in the best case — even for the three I believe will scale — their research agendas will not plug into an industrial ecosystem the way XJTLU plugs into Suzhou Industrial Park. The pipeline does not exist.

Twenty-one years of compounding

The SFCU model has been running since 2004. The way I find it most useful to understand what happened is in three waves, because the compounding mechanism is not obvious until you see how each wave fed the next.

Wave 1, 2004 to 2014: building the institutional base. UNNC and XJTLU grew from pilot operations into genuine universities. Gaokao scores for SFCU applicants rose every year — which meant the institutions were attracting the research-capable students who would eventually become the next generation of researchers. In 2014 the nine SFCUs formed the Sino-Foreign Cooperative University Union. The critical output of this wave was not the student numbers or the paper count. It was proof that the model held. China had not needed to unilaterally close a single institution. Compare that to Yale-NUS, the Singapore experiment that “worked” until the host country decided one day that it didn’t.

Wave 2, 2014 to 2020: the graduate bridge. The first SFCU cohorts had graduated and were moving through their careers. The class of 2017 data — the first full XJTLU cohort tracked through to graduation — showed approximately 85% going overseas for master’s degrees. That looked like brain drain. It was not. It was a knowledge bridge, and the structural distinction matters.

This is where the haigui — literally “sea turtle,” a Mandarin pun on the word for returnee, used to describe Chinese graduates who study abroad and come home — phenomenon becomes structurally interesting. The SFCU model produced a new kind of haigui: one who had received undergraduate education inside China, not abroad, and had therefore maintained deeper domestic ties throughout. They came back with a Western graduate credential and an intact Chinese professional network. That profile is different from the student who leaves at 18 and returns at 26, somewhat disconnected from the domestic landscape they left behind.

Wave 3, 2020 to 2026: the output. This is where I want to be careful, because I believe the popular narrative overstates the direct connection.

The South China Morning Post recently grouped what it called China’s Fantastic Four — the founders of DeepSeek, ByteDance, DJI, and Unitree — as a cohort partly defined by being educated in China. Liang Wenfeng (DeepSeek) studied at Zhejiang University. Zhang Yiming (ByteDance) at Nankai. Frank Wang of DJI at HKUST, then Shenzhen. None of them are SFCU alumni.

Here is what I believe is actually going on, and why the mechanism matters more than the lineage.

The SFCU research mandate forced institutions that had previously been teaching-focused to hire serious research faculty, build lab infrastructure, and publish. Those faculty became the professors and PhD supervisors of the next student cohort. DeepSeek’s R1 core team was trained almost entirely at Peking University and Tsinghua — both rank among China’s top two universities, broadly equivalent in national prestige to the IITs within India’s system — not SFCU institutions. But the supervisors of that team were themselves products of China’s broader internationalisation of higher education, of which the SFCUs were a leading edge.

One important clarification: the 985 programme — China’s national university excellence initiative, launched in 1998, six years before the first SFCU — was simultaneously strengthening Peking University and Tsinghua through a parallel track. The SFCU model and the 985 programme were not the same thing. What the SFCU model specifically contributed was a culture of international research quality and industry-connected applied research that propagated across the entire Chinese higher education system — including the 985 institutions — rather than producing a direct alumni line to any specific company.

The chain runs: SFCU model raised the research floor across Chinese institutions → stronger research faculty → stronger PhD cohorts → the Fantastic Four’s companies. Not a straight line. A rising floor, compounding over two decades.

China did not plan DeepSeek. It planned the conditions that made something like DeepSeek possible. There is a difference, and the difference is instructive for what India is deciding right now.

Figure 3. The compounding mechanism across three waves — each fed the next. India is starting with none of these structural inputs.

What didn’t go to plan

I want to be fair to the complexity here, because the SFCU story is not a clean success narrative.

The Thousand Talents Plan — China’s programme to recruit overseas-trained Chinese researchers back to domestic institutions — generated a geopolitical backlash in the United States that significantly damaged international research collaboration. Several SFCUs have faced scrutiny over academic freedom and national security concerns. NYU Shanghai was named in a September 2024 report by the US House Select Committee on Strategic Competition Between the United States and the Chinese Communist Party, alleging that joint US-China academic institutes create security vulnerabilities. These are real constraints on the model, and they are tightening.

There are also legitimate questions about academic freedom inside SFCUs — whether joint governance gives Chinese partners more curricular control in practice than the formal agreements suggest. These questions do not have clean answers, and the piece would be dishonest if it pretended otherwise.

For India, this geopolitical dimension is not hypothetical. If India were to build an SFCU-equivalent model today — inviting Western universities into joint ventures with Indian institutions under demanding structural requirements — it would do so in a world where US academic institutions are under congressional pressure not to deepen partnerships that resemble China’s model, and where the political optics of “India copies China on universities” create their own narrative risk. The structural logic of the SFCU model is sound and transferable. The geopolitical environment in which it was designed is not the environment of 2026. India needs a version that achieves the same structural outcomes through different political framing.

The compounding was real. The planning was imperfect. Both things are true.

What India should be asking — and isn’t

The comparison India’s policymakers should be making is with the SFCU model. From the public record — the UGC’s regulations on foreign university entry, the IFSCA’s GIFT City framework, the statements from education ministry officials — I see no evidence that comparison is being made. The conversation in New Delhi is about attracting foreign brand names and the revenue they generate for institutions that need international students. The conversation that produced the SFCU model was a different one: what structural requirements would force those institutions to produce research, develop local faculty, and connect to industry?

The lesson from China is not “be authoritarian and mandate things.” That framing lets India off the hook and is also wrong. The SFCU requirements were regulatory design, not executive decree. India’s regulatory architecture — the UGC, the IFSCA, state-level education regulators — is fully capable of the same. The question is whether there is will to write requirements that are actually demanding.

Four requirements forced quality at the point of entry. Joint governance eliminated unilateral exits. Dual degrees gave the host government accountability for standards. Research mandates meant institutions had to produce knowledge, not just credentials. Location intelligence meant that knowledge connected to industry rather than sitting in an academic silo.

India’s seventeen foreign universities at GIFT City and elsewhere meet none of these requirements. Branch campuses, not joint ventures. Single foreign degrees, not dual degrees with Indian accreditation. Undefined research mandates. Financial zones, not industrial clusters.

The gap between what China built in 2004 and what India has permitted in 2024 is not a gap of twenty years. It is a gap of structural design.

The demographic dividend argument makes this urgent in a way that often gets lost in the policy conversation. India’s working-age population share peaks somewhere between 2040 and 2045 depending on the projection — UN population data places the window there — after which the dependency ratio rises and the compounding advantage narrows. Compounding in higher education, as the SFCU model demonstrates, takes at least fifteen years to produce the tech output that matters. The arithmetic on starting in 2030 instead of 2025 is not forgiving.

Part 3 is the blueprint. What would a structurally analogous Indian model actually look like? Which state has the governance capacity to pilot it first? What are the realistic failure modes, and what is the minimum viable version India could operationalise in the next two years?

If you are advising a foreign university on India entry, or building a GCC with a research mandate, I am developing this blueprint and would value the conversation.

That is Thursday’s piece.

Part 2 of 3 — The Branch Campus Trilogy

B S Ashwin · freshwin.in