🇨🇳Panda Bonds: Renminbi Bond Issuance by Foreign Borrowers in China

Concept

A panda bond (熊猫债) is a bond denominated in onshore renminbi (CNY), issued by a non-resident of the PRC on the mainland market, governed by PRC law, with custody and settlement through CCDC or Shanghai Clearing House.

Four categories of non-resident may issue: sovereigns, multilateral development banks, financial institutions and non-financial enterprises. One route works in practice — the interbank market (CIBM) with registration at NAFMII, and PBoC approval for financial institutions. It carried 99.3% of volume in January–July 2026 (S&P China Ratings, 13 August 2026). A debut cycle runs 4–7 months; a follow-on tranche inside a live two-year registration takes 2–6 weeks. The debut budget is 0.5–0.8% of the amount raised, or 17–27 bp per annum on a three-year note. The three-year coupon for an investment-grade issuer in 2026 sits at 1.72–2.18%: Deutsche Bank priced at 1.72% on 29 May 2026 (Deutsche Bank), Kazakhstan at 1.90% the same day (AIFC), with the top of the range at Samruk-Kazyna's 2.18% in April 2026.

The market opened in October 2005 with two issues — IFC and the Asian Development Bank — and spent its first decade as a niche: Xinhua Finance puts cumulative issuance over 2005–2015 at CNY 11.3 billion, while the sum of ICMA's annual rows for the interbank and exchange perimeters gives CNY 19.0 billion for the same period (CNY 7.5 billion to 2015 plus CNY 11.5 billion in 2015 itself). By July 2025 cumulative issuance had passed CNY 1 trillion and by July 2026 CNY 1.3 trillion (Xinhua Finance via the Shanghai Financial Committee, 22 July 2025; BJD, 15 July 2026), with outstandings of CNY 516.24 billion at 24 July 2026 (Sina Finance). No material default has been recorded in the market's twenty-plus years — that is the assessment of an analyst at Founder Securities as reported by the trade press (January 2026); the regulator publishes no official default statistics for the segment.

What the news flow leaves out begins here. A 1.8% coupon is the price of renminbi, not the price of money. Swap the proceeds into dollars and the gap collapses: an offshore debt trader at a large Chinese bank puts the real post-hedge saving at roughly 10 basis points (The Paper, 6 April 2026), and for an emerging-market sovereign hedged renminbi comes out about 100 bp more expensive than the dollar (FDL Policy Note 33, March 2026).

Second: the 2026 record is measured off a falling base. 2024 closed at an all-time high of CNY 194.8 billion. 2025 delivered CNY 183.56 billion on Wind data (Chinanews, 14 August 2026) and only CNY 163.66 billion on Lianhe's (Lianhe Ratings) — minus 5.8% or minus 16% depending on the perimeter; over January–October 2025 the market ran 13.6% below the prior year (S&P China Ratings, 7 November 2025). That is where the "plus 90.3%" measured over January–May 2026 comes from (SCMP, 8 June 2026). Then there is scale: Hong Kong's dim sum market is materially larger than the panda market in 2026. Panda is the smaller channel of renminbi funding for non-residents, not the principal one.

This article holds the mainland debt perimeter in full. The price of renminbi money as such falls outside it — that is the subject of lending in China. Dim sum appears here only in comparison with panda: as an instrument in its own right it belongs to the Hong Kong domain. The mainland equity and exchange perimeter stays with the Shanghai article, and a foreign investor's access channels to Chinese paper with the country hub.

A panda bond is a way to raise renminbi from Chinese investors, not a way to borrow money cheaply. The headline saving of 230–260 bp against the dollar exists only up to the currency swap; after the swap around 10 bp remain for an investment-grade corporate borrower and a negative number for an emerging-market sovereign. The case is unarguable in exactly one configuration: when the renminbi raised is spent as renminbi.

Key parameters

Origins and chronology by stage

The market has passed through seven stages: an approval regime open only to development institutions (2005–2009), an easing on proceeds that changed nothing (2010–2013), a corporate workaround through NAFMII (2014), the sovereign wave under SDR and the developers' exchange bubble (2015–2017), a new constitution in Announcement No. 16 (2018–2021), liberalisation of the use of proceeds (2022–2023), and records on a changing cast of issuers (2024–2026).

2005–⁠2013: development institutions only

2005: the rulebook came before the deals. On 18 February 2005 the PBoC, the Ministry of Finance, the NDRC and the CSRC issued joint Announcement No. 5 (2005), adopting the Interim Measures for the Administration of the Issuance of RMB Bonds by International Development Institutions (《国际开发机构人民币债券发行管理暂行办法》); the citation is confirmed by the preamble to the 2010 revision on the regulator's site (PBoC). Eligibility was confined to multilateral, bilateral and regional development institutions. Both debut deals came in October 2005: IFC placed CNY 1.13 billion for ten years at 3.4%, and the Asian Development Bank CNY 1.0 billion for ten years at 3.34%, maturing 14 October 2015, with BOCI as sole lead manager and settlement through the predecessor of today's CCDC (ADB, 13 October 2005). ADB's stated purpose was to finance projects in the PRC and to reduce currency mismatches for borrowers without foreign currency revenue.

Where the name comes from. The instrument is named after the national symbol, on the same logic as samurai bonds in Japan, yankee bonds in the United States and bulldog bonds in the United Kingdom; that is the official formulation used by Bank of China (BOC, 6 January 2026) and NAFMII (Panda Bond Manual). The term is a market coinage: it does not appear in the text of the 2005 Interim Measures.

Why the market stood still. Four issues for ~$0.6 billion came over 2005–2009; Xinhua Finance puts cumulative volume over the decade 2005–2015 at CNY 11.3 billion, while ICMA's series gives CNY 7.5 billion to 2015 and CNY 11.5 billion in 2015 itself, or CNY 19.0 billion at the end of the period — different bases, and the annual series below follows ICMA.

The reasons read straight out of the Measures: the application went to the Ministry of Finance, was cleared with the PBoC, the NDRC, the CSRC and SAFE, and was put to the State Council for approval (art. 4); the NDRC and the Ministry of Finance tested size and purpose against industrial policy and the balance of payments (art. 5); the coupon was set by the central bank with reference to yields on comparable Chinese government bonds (arts. 6 and 16) — market pricing did not exist; two ratings were required, at least one Chinese, none below AA, along with at least $1 billion already invested in Chinese projects (art. 9); accounts had to follow Chinese standards (art. 11); and the working language was Chinese only (art. 21). Taking proceeds abroad was prohibited: Beijing feared pressure on the exchange rate.

2010: the wrong barrier came down. Announcement No. 10 (2010), published on 30 September 2010 (PBoC), used article 17 to permit renminbi to be transferred abroad under PBoC rules and, with SAFE's consent, to be converted and remitted in foreign currency. State Council approval, coupon approval, the two ratings and the $1 billion threshold all stayed. The result: two issues for ~$0.3 billion over 2010–2014 — fewer than in the frozen five years before. This is the key observation about the market: the binding constraints were admission and the absence of market pricing, not the destination of the money.

2014–⁠2017: corporates, sovereigns and the exchange bubble

2014: Daimler opens the market by going around the regime. The first foreign non-financial corporate issuer did not use the Interim Measures at all. It went through NAFMII as an issuer of debt financing instruments, where neither a State Council decision nor coupon approval was required. A CNY 5 billion quota was registered in December 2013 and the first tranche priced in March 2014, with Bank of China as lead (ICMA/NAFMII, 2021). The quota was raised to CNY 20 billion in February 2016; by May 2021 the programme had produced 30 tranches for CNY 71 billion — over 20% of the entire market at the time — and by 2026, 42 tranches and more than CNY 90 billion raised (BOC, 4 June 2026).

2015–2017: sovereigns under SDR, and an exchange bubble. The catalyst was political: preparation for the renminbi's inclusion in the SDR basket (the IMF decision came in November 2015, effective 1 October 2016). The PBoC states the link directly: "since 2015, in order to expand cross-border use of the renminbi and advance the renminbi's accession to the SDR basket, the PBoC together with the Ministry of Finance has intensified its promotion of the panda bond market" (Shanghai Financial Committee, 15 May 2025). But the 2016 peak was not inflated by sovereigns: of CNY 130.04 billion, CNY 83.84 billion came on the exchanges and, on CICC's estimate, more than 70% of exchange-traded panda bonds in 2016 were placed by property developers — offshore-incorporated Chinese housebuilders for whom a dearer dollar made borrowing renminbi at home cheaper (CICC via Sina Finance, 30 June 2025).

After external debt controls on developers tightened, exchange issuance collapsed from CNY 83.84 billion (2016) to CNY 11.6 billion (2017) and CNY 4.1 billion (2020). Real estate accounted for 29% of panda bond issuers over 2015–2019 and then all but vanished (S&P China Ratings, 7 November 2025).

2018–⁠2026: a new constitution and records

2018: the market's constitution changes. Announcement No. 16 (2018) of the PBoC and the Ministry of Finance was signed on 8 September 2018, published on 25 September 2018 and took effect on publication (PBoC); on the same date Announcement No. 15 (2018) repealed the 2010 Interim Measures and moved the subject into a single framework. The article-by-article comparison follows in the next section. In parallel, on 2 March 2018 the SSE and the SZSE formalised the Belt and Road bond pilot on the exchange perimeter through parallel notices on running it, expressly admitting foreign governments to the exchange market for the first time; by that same date seven issuers held a CSRC approval or an exchange no-objection letter for a declared CNY 50 billion, against CNY 3.5 billion actually placed (Zhong Lun).

Deals under the label ran before the rules were formalised: the Hongshi Holding bond placed on the SSE in January 2018 was billed as the first official Belt and Road bond (CNBC, 28 January 2018), but its issuer was Chinese and its project a cement plant in Laos — it was not a panda bond.

2019–2021: flat volume, changing cast. Volume stood at CNY 59.84 billion in 2019 and CNY 58.65 billion in 2020. Developers left; subsidiaries of foreign banks and multinationals arrived; then, against high global rates and renminbi volatility, offshore vehicles of Chinese groups came to dominate again. By August 2021 cumulative issuance on the interbank perimeter reached CNY 370 billion across 64 foreign issuers, and the share of foreign investors among panda bond holders was 21%, which ICMA and NAFMII call the highest of any segment of the Chinese market (ICMA/NAFMII, 2021; the source does not disclose the basis of calculation). On 11 August 2021 the PBoC suspended almost all rating requirements for interbank market issuers other than ABS (Caixin, 12 August 2021).

2022–2023: the money was released and the market doubled. Yinfa No. 272 (2022) of 23 November 2022 (effective 1 January 2023) allowed proceeds to be taken abroad without SAFE permission, and the NAFMII package of 2022–2023 simplified registration. The 2022 total was 52 issues for CNY 85.07 billion; 2023 delivered CNY 154.47 billion across 94–97 tranches from 37 issuers, up more than 80% on all three metrics and an all-time high (Lianhe Ratings). Lianhe identifies three drivers: the inversion of the China–US yield spread from mid-2022, the institutional easing, and the acceleration of renminbi internationalisation — in November 2023 the renminbi overtook the yen to become the fourth currency in global payments on SWIFT data. The 2023 profile was nevertheless domestic in substance: 79.6% of volume came from issuers with Chinese roots and 70.5% from issuers incorporated in Hong Kong, and the largest issuer of the year was Mengniu Dairy with 16 tranches for CNY 41.3 billion.

2024–2026: records, and a break in the composition of issuers. 2024 closed at CNY 194.8 billion on Wind data (CNY 197.8 billion on iFinD) across 109–110 issues, with the New Development Bank the largest single issuer; outstandings reached CNY 323 billion at year-end and, on IIF figures as reported by the trade press, overtook Japan's samurai market for the first time. 2025 brought a fall in issuance alongside record net issuance above CNY 100 billion and outstandings up more than 30%. PBoC Governor Pan Gongsheng: in 2025 foreign governments, development institutions, financial institutions and large companies placed more than CNY 170 billion of panda bonds, and outstandings rose 34% year on year (China Daily, 9 April 2026). The main structural shift: the share of genuinely foreign issuers — as opposed to offshore vehicles of Chinese groups — rose from 27% (2023) to 32% (2024), 39% (2025) and 50% over January–July 2026 (S&P China Ratings), and on first-half 2026 volume passed half for the first time.

Regulatory evolution

The regime has changed three times in substance: an approval regime for development institutions (2005, revised 2010), a registration regime for four categories of issuer (Announcement No. 16, 2018), and a monetary one that lifted restrictions on the use of proceeds (Yinfa No. 272 (2022), effective 2023) — with NAFMII calibrating the operating rules of self-regulation on top of that frame.

What Announcement No. 16 actually changed. A line-by-line comparison explains why the market only came alive after 2018.

NAFMII self-regulation is the second layer, and it is where the operating mechanics moved. For practice it matters more than the state level: this is where feedback deadlines, the composition of the filing package and the model covenants are set.

The exchange perimeter lives under a different act: companies incorporated abroad apply by analogy CSRC Order No. 180, the Measures for the Administration of the Issuance and Trading of Corporate Bonds (2015 revision), with registration at the CSRC and processing by the exchange. The only thing that has equalised the two perimeters is the money regulation: article I of Yinfa No. 272 (2022) extended a single proceeds regime across the interbank and exchange markets, on top of the regulatory split.

The tax regime has changed four times, always late. A three-year exemption of foreign institutional investors' interest income from corporate income tax and VAT was introduced by Caishui No. 108 (2018) for the period 7 November 2018 to 6 November 2021; Announcement No. 34 (2021) extended it to 31 December 2025 (PRC State Council); Announcement No. 4 (2025) reinstated VAT on interest from government bonds, local government bonds and financial bonds issued from 8 August 2025, leaving previously issued paper exempt to maturity (KPMG China Tax Alert, August 2025); and Announcement No. 5 (2026) extended the non-resident exemption over 1 January 2026 to 31 December 2027. Between Announcement No. 4 in August 2025 and the publication of No. 5 in January 2026 the market spent almost five months without an answer on whether the relief would survive.

Panda, dim sum and renminbi paper from domestic issuers

A panda bond is a non-resident's renminbi inside the PRC; a dim sum bond is a non-resident's renminbi in Hong Kong; a renminbi note placed by an issuer on its own domestic market is neither — panda status is set by the issuer's residence and the place of issue, not by the currency.

The distinction has practical bite. Renminbi bonds placed by an issuer on its own domestic market — renminbi government paper, renminbi bonds of a domestic municipality, renminbi issues by domestic corporates — are neither panda nor dim sum: they open neither the Chinese nor the international investor base, they simply convert domestic savings into renminbi form. Moscow negotiated access to the panda market for more than ten years, never agreed acceptable terms and placed at home instead (The Moscow Times, 22 May 2026). The mirror case: a renminbi bond issued by a Chinese resident is ordinary domestic paper, even when foreigners buy it through Bond Connect.

Tellingly, the channels add up rather than compete. Indonesia used both in 2026: offshore renminbi notes of CNY 6 billion (October 2025) and CNY 9.25 billion (February 2026), and then a CNY 7 billion panda in July 2026.

A comparison of foreign bond markets

Panda is one of seven national markets for non-resident bonds in local currency; what sets it apart is not the currency but the combination of approval-and-registration admission, PRC governing law and restrictions on the use of proceeds.

The volume figures in the table are market data rather than regulator statistics: they follow the Reuters/LSEG issuance tallies of 17 August 2026, the Goldman Sachs review reported by the South China Morning Post on 10 August 2026 and Bloomberg data of 1 July 2026.

When and why panda overtook samurai. Panda bonds outstanding first exceeded samurai bonds outstanding in 2024 (S&P China Ratings, 25 August 2025); the fact was recorded publicly on 28 February 2025, when the panda market was put at roughly $45 billion (SCMP). There are four reasons, in descending order of weight. First, the inversion of the rate cycle: ten-year Chinese government bond yields fell below US Treasuries in April 2022 and have not come back, while the yen stopped being uniquely cheap once the Bank of Japan exited yield curve control and negative rates. Second, the regulatory simplification of 2018–2024. Third, samurai's stagnation as an asset class: the 2026 increase rests largely on a single name, and without it the year is merely a seven-year high. Fourth, swap cost: the 2024 surge in samurai issuance itself pushed up the price of yen swaps and consumed the arbitrage (Japan Times, 5 October 2024).

One qualification, without which the comparison misleads: "overtook" refers to amounts outstanding, not annual issuance, and it does not extend to dim sum. The offshore renminbi market — roughly CNY 1.6 trillion outstanding and about CNY 1 trillion of annual issuance — is more than three times the size of the panda market in the same currency. Investors prefer the same renminbi with English-law documentation, a trustee structure and a London forum. Book depth confirms it: panda oversubscription runs 1.5–3.2x, while State Grid's dim sum deal drew a 13-times book — CNY 193.8 billion of orders against CNY 14.9 billion (SCMP).

Economics: why a 1.8% coupon is not cheap money

The economics of an issue is the difference between the renminbi coupon and the cost of alternative funding in the currency the issuer actually spends, net of swap cost and the issue budget.

The weighted average coupon on panda issues in the first half of 2026 was 1.85% (−8.75 bp y/y); three-year AAA paper priced at an average of 1.94%, 30.42 bp below the same period of 2025 (Far East Credit 远东资信, 7 August 2026). Dollar funding of comparable quality is quoted at SOFR plus 50–90 bp, or 4.15–4.45% (The Paper, 6 April 2026). The gross gap is 230–260 bp, and that is the number circulated as "the saving". After the hedge almost nothing is left of it, and three independent sources say so.

The spread between those readings comes from the issuer's credit curve and the point of comparison — and it sits inside the source itself. The ≈100 bp figure is FDL's average conclusion from market swaps for an emerging-market sovereign; the Kenya arithmetic in the same note is a single case measured against Kenya's expensive eurobond curve: a panda swapped into dollars costs 4.65% against 6.15% on the eurobond, renminbi 150 bp cheaper. Swapped into Kenyan shillings the same trade costs 11.958% against 11.38% on domestic debt, dearer than money at home. For an investment-grade issuer the advantage over the dollar compresses to a handful of basis points, and the issue budget eats them.

Hence three configurations in which the gain comes from something other than the coupon. The borrower who keeps the money in renminbi: a multinational with production in the PRC (Volkswagen, BMW, Mercedes, BASF, Bayer, Henkel) does no swap at all and takes the full 230–260 bp — BASF's 2025 panda coupon was 2.28% against more than 5.5% on its dollar debt, and the self-use panda bond structure works on top. Samuel Fischer, head of onshore DCM at Deutsche Bank in China, describes panda bonds moving "from a tactical option to a strategic component" of German issuers' funding (Deutsche Bank, 10 July 2026). The global bank: Deutsche Bank, UBS, UOB, BNP Paribas, Barclays, Crédit Agricole and Morgan Stanley put the proceeds to work without conversion, and diversification buys an issuance window outside the dollar cycle. The sovereign buying a relationship: Yu Yongqiang of JunHe names Belt and Road participation and relations with the Chinese authorities as the motives (law.asia, May 2026).

Pricing

A panda bond prices at a spread to the policy bank curve, principally China Development Bank, rather than to the Chinese government bond curve, because the spread between those two curves reflects a tax wedge rather than credit risk.

There are two reasons. Tax: interest on government and local government bonds is exempt from corporate income tax and VAT for domestic investors, while interest on policy bank paper and credit bonds is taxed at 25% corporate income tax plus 6% VAT; the CGB–CDB spread is in substance an implicit tax rate (ADB ABMF, "Fees and Taxation in the Inter-Bank Bond Market", 2020). Liquidity: the IMF records that policy bank paper is in fact more liquid than sovereign paper and responds more strongly to macro and monetary signals (Schipke, "The Future of China's Bond Market", IMF, 13 March 2019). The result: the CDB curve is a taxable risk-free base to which a credit spread attaches cleanly.

Yield levels have compressed faster than volumes have grown: the weighted average coupon ran 3.3% (2022) → 3.0% (2023) → 2.01% (2025, −22 bp on 2024) → 1.85% (H1 2026). Three-year AAA paper on the interbank market averaged 2.26% in the first half of 2025 (−24.2 bp y/y), and 2.17% at three years and 2.38% at five years across 2025 as a whole. The weighted average tenor of 2025 issues was 2.94 years.

Bookbuilding mechanics. Placement follows the NAFMII rules on DFI bookbuilding: centralised order collection by the bookrunner, announcement of a coupon range, order collection, cut-off. The pricing and allocation pilot launched in September 2023 gave overseas issuers three concessions: deal size may be adjusted after the book but only upwards or not at all; up to 20% of the deal is placed by flexible allocation in favour of market makers, long-term holders and ESG investors; and the issuer receives, during the placement, investor names, investor type, order size and the bid rate. The last point is what really separates the panda market from the euromarket, where the issuer sees the book through the syndicate: here the anchoring mechanism is institutionalised and disclosed in advance rather than left as a grey area.

Who the anchor investors are. The reference book breakdown is AIIB's July 2025 placement: bank treasuries 88.0%, central banks and official institutions 10.5%, asset managers and insurers 1.5%; by geography 57.5% offshore against 42.5% onshore, of which offshore commercial banks 47.0%, onshore foreign banks 27.0%, Chinese state and joint-stock banks 10.0%; more than 30 accounts, 12 new investors, syndicate Bank of China, CITIC Securities and Standard Chartered (China).

How a Chinese investor assesses a foreign issuer's credit. Since roughly 88% of a supranational's book and almost 40% of the market as a whole sit with bank treasuries, the real credit work is done not by a rating but by four mechanisms: the bank's internal limit on the name and the country; the paper's repo eligibility and its ChinaBond valuation; a zero or preferential risk weight for development institutions; and a development bank guarantee for sovereigns below investment grade. ICMA noted as early as 2017 that unfamiliar international names require additional credit work, and that remains a barrier (ICMA, 17 October 2017). The practical consequence: the market is more sensitive to Beijing's political relations with the issuer's jurisdiction than any rating can capture.

Tax as a pricing factor. The exemption of non-resident interest income from corporate income tax and VAT runs to 31 December 2027. Without it a non-resident pays 10% withholding tax (reducible under the more than one hundred double tax treaties the PRC has) plus 6% VAT, which treaty protection does not reach — some 15–16% on the coupon in total. On a 1.85% coupon that is about 28 bp of yield, a figure comparable to the entire credit spread of a panda bond over the CDB curve. There is no stamp duty or transaction tax on debt trades on the interbank market; withholding mechanics sit in the material on withholding tax, and headline rates in the material on taxes.

Two tax events change the pricing geometry of this market, and neither is priced in yet. First: Announcement No. 4 (2025) reinstated VAT on interest from government bonds and financial bonds issued from 8 August 2025, while preserving the relief for older paper to maturity. Because a panda prices to the CDB curve, and the tax wedge is precisely what separates that curve from the CGB curve, the base under the whole pricing system shifts as issues roll over: "27 bp over CDB" in 2028 will not mean what it meant in 2025, and spread series that cross 8 August 2025 are not comparable. Second: the non-resident exemption expires on 31 December 2027. Three consecutive extensions (2018, 2021, 2026) argue for a fourth, but each was signed at the very edge of the deadline — in 2025 the market spent almost five months in limbo. A bond maturing after 2027 has no guaranteed tax treatment for its full life.

Guarantees and credit enhancement

A sovereign below investment grade reaches the panda market only with a multilateral development bank guarantee, and what that guarantee buys is less credit quality than clearance through the internal limits of Chinese bank treasuries.

Development bank guarantees: Egypt and Pakistan

Egypt is the prototype. On 27 October 2023 it placed CNY 3.5 billion (about $500 million) for three years at a coupon of 3.51% under an unconditional and irrevocable guarantee from AIIB and the African Development Bank — the first African sovereign on the PRC interbank market, the first guaranteed sovereign panda and the first double development bank guarantee on the Chinese market (AIIB; AfDB). Bank of China was lead underwriter and bookrunner and HSBC joint lead; the proceeds went into a sustainable sovereign financing framework covering transport, renewable energy, water and sanitation, affordable housing and digital infrastructure.

Pakistan shows the evolution to partial cover. On 15 May 2026 it placed CNY 1.75 billion (about $258 million) of sustainable development bonds under a partial guarantee from ADB and AIIB. ADB records the first explicitly: this was the first bond with less than 100% cover to obtain a domestic AAA rating (ADB; AIIB). The book was five times covered; the coupon is reported by the industry press at around 2.5%, with no official confirmation of the rate.

What it costs. Neither AIIB nor ADB discloses the guarantee fee on individual deals. The only published estimate is FDL's, which names annual fees of 50 and 80 bp on the two guarantees in the Egyptian transaction; they add to the coupon and belong in any comparison with alternative funding (FDL Policy Note 33, March 2026). The saving reads off the coupons: Egypt, rated in the B category, paid 3.51% for three years at the end of 2023, when its dollar paper traded at double-digit yields. The constraint on replication is not price but the scarcity of guarantee capital at the development banks — which is exactly why the move to partial cover matters: an incomplete guarantee consumes less capital and the domestic AAA is achieved anyway.

Corporate guarantees and keepwell

Corporate enhancement works differently. The standard structure for a panda placed through a subsidiary or holding vehicle is a parent guarantee (跨境担保) with registration of the cross-border security at SAFE; that is how BMW Finance N.V. issues under a BMW AG guarantee and Air Liquide Finance under a L'Air Liquide guarantee. Under the holders' meeting rules the credit enhancement provider is expressly named among the reserve conveners of a meeting and, at the same time, stripped of voting rights as a material affiliate.

Keepwell deeds are not used on the panda market — and that is a structural argument in its favour. The keepwell deed arose as a way around the requirement to register cross-border guarantees after 2014: it is not a guarantee but a parent's promise to keep the subsidiary solvent, usually by way of a loan or equity contribution. On 19 March 2025 the Hong Kong Court of Final Appeal in Peking University Founder Group v Nuoxi Capital Limited confirmed that a keepwell is a legally binding contractual obligation and not "worthless window dressing", but rejected the issuer's liquidators' claim of more than $1 billion on causation and loss: performance would have come through a loan, so the SPV would merely have swapped one debt for another and suffered no net loss; the court said expressly that the outcome would have differed on a claim by the bondholders themselves (case analysis).

Panda bonds do not need the structure: issuer and obligation sit inside the PRC legal perimeter, so a direct guarantee is available. Keepwell is a phenomenon of the offshore market — dim sum and Chinese dollar debt.

Two routes: the interbank CIBM and the exchanges

A panda bond is placed either on the interbank market supervised by the PBoC and the Ministry of Finance with registration at NAFMII, or on an exchange under CSRC oversight; only the first route is economically alive.

The conclusion rarely stated out loud: the exchange route is dead for a genuine non-resident. The exchange share fell from roughly 5% in 2024 to 0.2–0.7% in 2026, and what remains is offshore-incorporated Chinese groups for whom the PRC is the home market in substance. A foreign sovereign, a development bank or a European industrial group has no alternative to the interbank market, and the mainland exchanges as a venue are covered in the material on Shanghai. Inside the interbank market there is a second fork — a public DFI against private placement notes: PPNs go to a narrow circle of qualified investors with lighter disclosure and cost less to prepare, but they do not create a pricing benchmark for the next tranche, and the benchmark is precisely what earned Deutsche Bank 21 bp between its March and May 2026 issues on the bank's own estimate (by coupon, 23 bp on the three-year and 19 bp on the five-year).

Instruments and tenors

A panda bond is issued not as a single instrument but as one of the debt financing instruments of the interbank market — from ultra-short commercial paper to perpetual notes — with a weighted average tenor of 2.94 years in 2025.

The interbank range: super and short-term commercial paper (超短期融资券, up to 270 days), commercial paper, medium-term notes (中期票据) — the workhorse of the segment, private placement notes for a narrow circle of qualified investors, perpetual and subordinated notes, asset-backed notes (ABN), and green and other GSSS instruments. On the exchange perimeter there is one instrument, the corporate bond (公司债券). The instrument also decides the depository: DFIs are held at Shanghai Clearing House, while financial bonds of overseas banks go to CCDC.

A chronology of firsts shows how the range widened. The market's first ultra-short paper was China Merchants Port on 16 November 2016: CNY 1.5 billion for 270 days at 3.19%. The first green panda from an overseas non-financial enterprise was China Power Clean Energy on 19 May 2017. The first subordinated perpetual was Beijing Enterprises Water on 11 May 2020: CNY 1 billion, 5+N format, coupon 4.00%, 2.16x covered; China Power International followed with a 3+N format at 4.35% on 3 November 2020. The first carbon-neutral panda was China Everbright Greentech on 26 July 2021. The first sustainability-linked bond (SLB) was China Shipbuilding Leasing in March 2023. The first science and technology panda (科创熊猫债) came in 2025, after the technology section of the debt market was launched. A case of its own is the IBRD issue of 31 August 2016 — Mulan bonds of SDR 500 million settled in renminbi, the world's first SDR-denominated bonds.

The GSSS segment is growing faster than the market: in 2023 green, social and sustainability pandas delivered CNY 21.7 billion, roughly double the prior year. Framework issuance has been available since 18 February 2025: a qualified issuer places a green panda under its existing international framework, and NAFMII recognises the PRC 2021 Catalogue, the EU Taxonomy and the ICMA principles — removing the need to build separate Chinese green documentation.

By tenor the market is short. The weighted average tenor of 2025 issues was 2.94 years, the standard is a three-year note, five years became regular only among banks (Morgan Stanley in 2025; Deutsche Bank, BNP Paribas and UBS in 2026), and ten-year paper appears only in the debut issues of 2005. The reason is not demand but the hedge: the onshore swap is expensive or unavailable beyond five years, which constrains issuer and foreign buyer alike.

Who may issue

Four categories of non-resident have the right to issue, each with its own registration body and its own set of requirements.

Tiered issuer classification (分层分类) was introduced by the NAFMII Guidelines, effective 19 October 2023 (NAFMII). The mature offshore tier (境外成熟层) — listed on a major foreign exchange, aggregate global bond issuance over the past 36 months of at least the equivalent of CNY 10 billion, no defaults in 36 months — gets a single DFI registration with no amount stated and comments within three working days. The basic offshore tier (境外基础层) registers product by product with an amount stated, and gets a reply within five working days.

Accounting and audit. CAS is accepted, as are equivalents recognised by the PRC Ministry of Finance — EU, Hong Kong, UK and Swiss standards. Article 16 of the 2018 Measures requires disclosure of material differences between the standard applied and CAS together with a reconciliation, and that is the single biggest driver of a long preparation phase. Where CAS is used, the audit goes to a PRC firm holding a securities licence. A non-resident auditor is admitted under article 19: a valid licence in its home jurisdiction, at least five years of experience auditing publicly traded securities, a consent letter and a filing with the PRC Ministry of Finance no later than 20 working days before the submission, and annually thereafter (audit in China).

A rating is not required on either the interbank market or the exchange: for panda bonds the requirement was removed by article 26 of the 2018 Measures, and the market-wide PBoC suspension of rating requirements followed on 11 August 2021. Estimates of the unrated share differ by sample perimeter. On Lianhe's data, 70 of 114 tranches priced unrated in 2025 — 61.4% by count and 65.3% by volume; of the 44 rated tranches, 43 carried AAA and one AA+ (Lianhe). A first-half 2026 review gives higher unrated shares — 65.6% (2024), 69.3% (2025), 60.85% (H1 2026). S&P China's sample covers a longer period: over 2023 to October 2025, 57% of panda bonds priced without an issue rating, and where an issue rating was present the issuer rating was AAA in 92% of cases (S&P China Ratings, 7 November 2025).

The national scale does not map to international scales: of the 77 issuers outstanding on 31 July 2026, 45 hold an international rating and 87% of those are investment grade — and domestically every one of them is AAA, while the six issuers whose international rating is speculative grade (BB+/Ba1 and below) hold domestic ratings in the AAA–AA+ range (S&P China Ratings, 13 August 2026). The regulator has now said so directly: a NAFMII notice of 21 July 2026, effective 1 August 2026, requires agencies to publish a mapping of the national scale against international scales and bars acceptance of reports from agencies that do not (Xinhua via QQ). The conclusion: a Chinese AAA is not an argument in a conversation with an international investor, and for a sub-investment-grade issuer the problem is solved by a development bank guarantee.

Procedure and calendar

An issue is a sequence of seven steps in which registration takes weeks while the accounts and the legal package take months.

The lead underwriter can only be a Chinese bank or securities house, and under article 8 of the Guidelines at least one syndicate member must have a subsidiary or branch in the issuer's country of incorporation. Syndicate size is tied to deal size: four lead underwriters from CNY 5 billion, three at CNY 3–5 billion, two below that. Bank of China has ranked first in panda underwriting for eleven consecutive years since 2014 as at January 2026 (Bank of China, 6 January 2026), and for twelve on the bank's own June 2026 figures; then ICBC, CCB and CICC, and among foreign houses Deutsche Bank (China), HSBC (China), BNP Paribas (China) and UBS Securities. Add two legal advisers (the PRC one mandatory under article 24 of the 2018 Measures), an auditor, a depository, a trustee (art. 25) and, optionally, a rating agency.

1. Preparation — 6 to 12 weeks: mandate, due diligence, accounts and the CAS reconciliation, translation into simplified Chinese, legal opinions; in parallel, the non-resident auditor's filing with the PRC Ministry of Finance.
2. NAFMII registration (or PBoC approval). The package is submitted only through the underwriter. Preliminary assessment and feedback take 10 working days, with the same period allowed for revisions; then a registration meeting of the expert panel (注册会议) with three possible outcomes: accepted, accepted with conditions, deferred. Comments on supplementary documents come within three working days for the mature tier and five for the basic tier; the practical cycle including responses runs 4–10 weeks.
3. Registration acceptance notice. Valid for two years (art. 9 of the Guidelines); inside that window tranches price without re-registration, self-directed for the first 12 months and on a notification basis thereafter. The use-of-proceeds plan is filed no later than three working days before the tranche documents are published.
4. Registration of the proceeds account, with the bank's certificate issued.
5. Roadshow and bookbuilding — 1 to 2 weeks. Oversubscription runs 1.5x to 5x: Deutsche Bank drew CNY 8.66 billion of orders against a CNY 5.5 billion deal in March 2026, Pakistan was five times covered in May 2026, UBS more than three times in August 2026.
6. Settlement, listing, secondary trading.
7. Post-filing: the registration form is updated within 20 working days after the tranche closes.

The calendar: 4 to 7 months for a debut issuer, 2 to 6 weeks for a follow-on tranche. Volkswagen priced its fifth issue under its DFI programme on 21 May 2026 (Clifford Chance), and Deutsche Bank came twice in 2026 less than three months apart. That is the argument for registering a programme rather than doing a single deal.

Documentation and investor protection

A package of eleven documents is filed in simplified Chinese through the lead underwriter, and what protects the holder is not an English-law covenant set but the combination of a mandatory trustee, NAFMII's model investor protection provisions and the self-regulatory rules on holders' meetings.

The source for the package composition is the NAFMII Panda Bond Product Manual (熊猫债产品手册), 5 July 2024. On a private placement (定向发行) the requirements are lighter: two years of accounts instead of three, a private placement agreement instead of a prospectus, and some documents may be filed in English by agreement between the parties. Foreign governments and development institutions file legislative and governmental authorising instruments instead of articles of association, and may submit an economic data report confirmed by a legislative or audit body instead of standard accounts.

Ongoing disclosure is conducted in Chinese as a matter of principle. The asymmetry runs in the issuer's favour: if it does not publish quarterly accounts at home, it need not publish them in the PRC either — a warning to investors suffices; but everything it discloses on any other market must be duplicated on the PRC interbank market, and for material events a Chinese version of the key content must appear within 7 working days. A change in the intended use of proceeds is disclosed no later than five working days in advance. The audit requirements — the foreign auditor's filing with the PRC Ministry of Finance 20 working days ahead and the CAS reconciliation certified by a PRC firm — make an opportunistic entry to the market practically impossible for a debut issuer (audit in China).

Covenants under PRC law are set not by the prospectus but by NAFMII model documents: the Examples of Investor Protection Provisions and the model text issued under Notice No. 219 (2023) of 25 December 2023. Three blocks.

Cross-protection (交叉保护条款). The trigger is non-payment by the issuer or its consolidated subsidiaries on other debt financing instruments, corporate bonds, offshore bonds or bank loans, once a cumulative threshold is met, defined as the lower of a fixed renminbi amount and a percentage of latest consolidated net assets. The mechanics: disclosure within two working days → the lead underwriter convenes a holders' meeting → a decision among three options: an unconditional waiver, an investor put, or a conditional waiver with compensation (additional security, a coupon step-up of an agreed number of basis points, a moratorium on new borrowing). The grace period runs up to 10 working days; if the breach is not cured, principal and interest fall due on the day after the meeting.

Pre-agreed restrictions (事先约束条款) are financial metrics and negative undertakings. The metrics are selected from a set: debt to assets below an agreed ceiling, return on equity above a floor, minimum net profit, current ratio, a cap on the growth of interest-bearing debt, a cap on the share of outstanding DFIs in interest-bearing debt; monitoring is quarterly or annual. Prior holder approval is required for: the sale or transfer of material assets above 10% of net assets; deconsolidation of key subsidiaries contributing 30% or more of revenue or profit; pledging or reducing a stake in listed subsidiaries; new debt-disguised-as-equity (名股实债) above 10% of net assets; pledging or transferring an asset pool; large guarantees to third parties; and debt restructuring affecting solvency. The approval threshold is a two-thirds quorum and three-quarters approval.

Change of control (控制权变更条款). The triggers are a change of controlling shareholder or ultimate controller, departure of key directors and managers, a change in the agreed board composition, and a downgrade or negative outlook resulting from the change of control. Disclosure within two working days, a put at 101% of par, a ten-day registration window, and payment within a month of registration.

The holders' meeting operates under the rules effective 1 April 2024 (NAFMII Announcement No. 27 (2023)).

Changing payment terms — price, rate, maturity, security — requires unanimity of holders under the Guidelines on default and risk resolution (2022, art. 19); debt-for-debt exchanges and other resolution methods go to a vote with the result disclosed within two working days. A consent solicitation mechanism was introduced in April 2022. The institutional infrastructure exists: 96 trustee institutions were registered on the interbank market at the end of 2024 (NAFMII, 23 April 2025).

A comparative treatment of forum and immunity is in Yingxin Angela Chen, Transnational Litigation Blog, 7 January 2026.

What is hard and what is left to negotiation. Drawing that line saves weeks of negotiating time.

Issue budget and the entry threshold

The budget is an underwriting fee tied to size plus a block of fixed costs for advisers, audit and infrastructure that does not scale with the deal — and therefore sets the entry threshold.

The tariff base is the schedule in the ADB/ASEAN+3 Bond Market Guide for the PRC: NAFMII registration and PBoC approval are free of charge; underwriting carries a regulatory floor of 0.4% (typically 0.4–0.6% for a corporate issuer, 0.15% for financial paper, 0.1% for government paper); CCDC registration is 0.6–1.15 bp by tenor (1.0–1.05 bp for one to five years) and SHCH 0.3–1.15 bp (around 1.0 bp for one to five years); payment servicing is 0.05 bp of the sum of coupon payments and redemption at CCDC and 0.3–0.5 bp of the issue amount charged once at SHCH (0.5 bp for paper longer than 180 days). Adviser fees are not regulated: a Chinese plus a foreign legal adviser for a debut runs CNY 2–5 million, and audit with reconciliation and translation another CNY 1–3 million.

Debut issuer, CNY 2 billion for three years, interbank route, CCDC as depository (midpoints of the ranges, estimate):

A follow-on tranche of CNY 2 billion inside a live registration: underwriting at the regulatory floor of 0.40% — 8.00 million; registration and payment servicing — 0.00 if the CCDC relief is adopted (0.22 otherwise); rating surveillance — 0.05; tranche-level legal work instead of a full package — 0.30; refreshed accounts and reconciliation — 0.20; trustee — 0.25. Total CNY 8.80 million, or 0.44% of size — 14.7 bp per annum, 40% below the cost of a debut. The saving comes not from a discount but from the two-year life of the registration.

What is proposed to be waived from September 2026, and what it is worth. CCDC has put out for public consultation a draft cancelling its issuance registration and payment servicing fees on panda bonds: the document was published on 27 July 2026 and the proposed relief window runs from 1 September 2026 to 31 December 2028 (Yicai, 27 July 2026; SCMP, 28 July 2026). CCDC had published no final notice as at 4 September 2026, so the budget above assumes the relief applies and carries the price of its absence alongside. On a three-year CNY 2 billion deal that is CNY 0.22 million — 1.1 bp of size, 0.37 bp per annum, roughly 1.5% of the budget: underwriting at the regulatory floor is thirty-six times the fee being waived. One consequence is concrete, though — the choice of depository becomes a priced decision: SHCH has announced no equivalent relief, and on its schedule the same deal would pay around CNY 0.30 million (1.5 bp of size) against zero at CCDC (on the ADB/ASEAN+3 schedule, estimate).

The entry threshold. There is no regulatory minimum size; the threshold falls out of the cost structure. A debut issuer's fixed block — legal advisers, audit and reconciliation, trustee, rating — is CNY 6.1 million and is independent of deal size; the variable part is 0.45%.

The curve approaches 0.45% asymptotically — 15 bp per annum, the price of underwriting, below which nothing goes. The fixed block adds 40 bp per annum over that asymptote at CNY 500 million, 20 bp at CNY 1 billion and 10.2 bp at CNY 2 billion; beyond that each additional CNY 500 million buys less than 4 bp. The practical debut threshold is CNY 1.5–2 billion, and below CNY 1 billion the load does not pay for itself. The market has arrived at the same point empirically: the median 2026 issue is around CNY 1.8 billion, a typical corporate tranche runs CNY 2–5.5 billion, and the smallest sovereign deal was CNY 1.75 billion (Pakistan, 15 May 2026, with a partial AIIB and ADB guarantee). For a follow-on tranche the threshold drops to roughly CNY 500 million: the fixed block shrinks to CNY 0.8 million, and all-in cost at that size is 0.56%, or 18.7 bp per annum.

The budget has to be compared with the net rate gap, not the gross one. A debut issuer pays 25 bp per annum of issuance cost against a post-swap saving of about 10 bp — meaning a swapping debut issuer is under water before the first coupon. Run the same arithmetic for a repeat issuer keeping renminbi as renminbi: 14.7 bp of cost against 230–260 bp of saving. The line between worthwhile and not lies in two variables — the currency the money is spent in, and whether a live registration exists — never in the coupon.

Use of proceeds and repatriation

Proceeds land in a dedicated account and may be transferred abroad, invested inside the PRC or on-lent to Chinese residents — provided the use matches the prospectus and the rules on the use of proceeds.

The governing act has to be named precisely, because the press names it wrongly. The rule in force is 《中国人民银行 国家外汇管理局关于境外机构境内发行债券资金管理有关事宜的通知》, Yinfa No. 272 (2022), issued 23 November 2022, effective 1 January 2023 (SAFE; PBoC); it repealed the 2014 and 2016 notices. The formulation "repatriation allowed since January 2022" with a reference to Announcement No. 2 (2022) is inaccurate: what appeared in December 2021 was only a consultation draft. More often still it is confused with Announcement No. 4 (2022) (27 May 2022, effective 30 June 2022) — but that one governs access for foreign institutional investors, not issuers' proceeds (CSRC).

The dedicated account (募集资金专户) is opened before pricing, in renminbi and, where needed, in foreign currency under the NRA regime, against the bank's registration certificate; every movement is tied to the purpose stated in the prospectus. SAFE registration is carried out by the Chinese lead underwriter on the issuer's instructions at the custodian bank, with updates within 20 working days after each tranche; previously issued paper still outstanding had to be registered retrospectively. Permitted outflows: transfer abroad, or purchase of foreign currency followed by transfer; payment of principal, interest and taxes; loans to Chinese residents; investment inside the PRC under FDI rules with repatriation of returns; transfers to the issuer's other accounts.

Article V allows both taking the money out and leaving it in the country (subject to the rules on direct investment and external debt); article VI permits currency derivatives only through a Chinese financial institution and only on the genuine-need principle; the notice itself contains no list of prohibited uses — it requires only that the use of proceeds match the prospectus.

Restrictions on the use of proceeds come from a separate self-regulatory act — the Rules for the Administration of the Use of Proceeds of Debt Financing Instruments of Non-Financial Enterprises in the Interbank Bond Market (NAFMII Announcement No. 25 (2022) of 22 November 2022, in force from 1 December 2022, NAFMII): proceeds go to the enterprise's production and operating activities in compliance with PRC laws, regulations and policy requirements (art. 5) and are spent as set out in the offering documents (art. 6); temporarily idle proceeds may be held only in safe principal-protected products liquid enough not to disturb the spending schedule (art. 7); and the proceeds account is not commingled with the issuer's other money (art. 10).

The rules are expressly stated to be a minimum requirement — other self-regulatory rules apply on top of them (art. 20). The Rules apply to interbank debt financing instruments; the exchange track runs under CSRC Order No. 180. The five-working-day advance disclosure of a change of purpose comes not from these Rules but from the NAFMII disclosure rules — art. 27 of the Rules on Information Disclosure for Debt Financing Instruments of Non-Financial Enterprises in the Interbank Bond Market (as revised 13 November 2023, NAFMII); the Rules themselves require only timely (及时) disclosure of the change (art. 6).

The self-use panda bond (自用熊猫债) structure is the main economic motive for German industrial groups: under article 4 of the PBoC's 2017 notice on macro-prudential management of cross-border financing, a loan extended by a foreign parent to its Chinese subsidiary out of panda proceeds is not counted in the subsidiary's external debt and does not consume its cross-border borrowing quota. Volkswagen, BMW, Mercedes, BASF, Bayer and Henkel use it; the operating-structure requirements sit in the material on a company in the PRC.

The honest number: roughly 30% of 2024 panda proceeds went offshore on Standard Chartered's estimate (Caixin, 18 October 2025) — the remaining 70% or so was never remitted abroad. The strategies are polar: Hungary (July 2025) converted its proceeds into euros, Indonesia (July 2026) kept 100% in renminbi on offshore accounts, Pakistan (May 2026) directed the money to domestic projects, and German groups fund their Chinese operations.

Market infrastructure

The market's infrastructure is two depositories split by instrument type, the CFETS trading system, and a syndicate perimeter in which Bank of China has held first place for more than a decade.

The legal anchor is article 11 of the 2018 Measures: the paper is held with a depository recognised by the PBoC; after closing, the issuer confirms the debt relationship to the depository, the depository registers the bond, and the issuer must ensure timely transfer of coupon and redemption to holders' accounts. Yinfa No. 272 (2022) is addressed to both depositories alongside CFETS, so both formally serve the panda segment; the division runs by instrument type, not by issuer type. The practical consequence for an issuer is that the fee relief CCDC has proposed from September 2026 does not extend to SHCH, where the bulk of corporate DFIs is placed; the general mechanics of holding securities are covered in the material on securities custody.

Infrastructure tariffs (ADB ABMF, 2020): NAFMII registration and PBoC approval are free; issue registration at a depository costs 0.6–1.15 bp depending on tenor and size; payment servicing is 0.05 bp at CCDC and 0.3–0.5 bp at SHCH; the CFETS trading fee is 0.5–2.5 per million capped at CNY 1,000 per trade; settlement is CNY 100–200 per operation. Against underwriting at 0.4–0.6% of size, the entire infrastructure block of the budget is measured in fractions of a basis point.

The underwriter ranking and its stability. Bank of China has held first place since 2014: eleven consecutive years per its release of 6 January 2026, and twelve on the bank's own data as at 4 June 2026, with cumulative underwriting above CNY 700 billion, more than 350 issues and more than 80 overseas clients (a year earlier: above CNY 570 billion, around 300 issues, more than 70 issuers).

The bank was involved in every issue by Mercedes-Benz (42 tranches), BMW (27), Shangri-La Asia (10), Veolia (8) and Bayer (4), and in every panda issue by National Bank of Canada, Crédit Agricole, Deutsche Bank, UOB and Barclays. Other regular leads are ICBC, CCB, ABC, Bank of Communications, China Development Bank, Eximbank, CICC, CITIC Securities and China Securities; the profile of these institutions is covered in the material on Chinese banks, and the market leader itself in the material on Bank of China. The first visible sign of diversification came in May 2026, when CICC took the lead on Kazakhstan's sovereign issue.

Foreign banks with PRC licences operate through locally incorporated entities and have sat in syndicates since 2015–2016: HSBC (China) and Standard Chartered (China) appear as leads on the debut sovereign deals of Korea (2015), British Columbia and Poland (2016). A separate lead underwriting licence for corporate panda bonds (Panda Corporate Bond Lead Underwriting Licence Type B) is issued by NAFMII: BNP Paribas (China) received it on 10 December 2018, becoming the third international bank with that right (BNP Paribas); Mizuho followed in January 2024; DBS (China) obtained a principal underwriting licence for non-financial corporate bonds and led Indonesia's 2026 debut. Syndicates also feature Deutsche Bank China, Crédit Agricole CIB China, Citibank China, Goldman Sachs Gao Hua, MUFG China, SMBC China and UBS Securities.

The role of foreign banks in the syndicate deserves a sober reading: they are almost always joint leads rather than sole bookrunners. Distribution runs through the Chinese state banks, which have the access to the domestic base — in Slovenia's placement 98.2% of the book went to banks and foreign investors took 0.6%. The other side of that is concentration risk that is abnormal by world standards: one institution covers 84–90% of deals across all four issuer categories, states itself that it "created the pricing logic of the panda bond market", and was first to implement every market mechanism — the single DFI registration (2021), flexible allocation (2024) and post-book size setting (2025). The choice of underwriter on this market is to a large extent an illusion.

Secondary market

Panda bonds trade over the counter on the interbank market through the CFETS system; first-half 2026 turnover was CNY 342.6 billion, and an annual turnover ratio of about 1.3–1.4x makes the paper a buy-and-hold instrument.

Trades are struck by request for quote, click-to-trade, with market makers or by direct negotiation; inter-dealer brokers, which NAFMII treats as a separate category, play a visible role. Direct retail participation is prohibited. A small number of issues goes through the Beijing Financial Assets Exchange (北金所) as a private placement venue — that is where Kazakhstan's sovereign issue was listed, alongside a simultaneous listing on AIX.

Turnover of CNY 342.6 billion in a half-year against outstandings of around CNY 500 billion gives an annual turnover ratio of 1.3–1.4x — orders of magnitude below the interbank market as a whole. The comparison with the external market shows the same scale of gap: FDL puts average monthly panda turnover at CNY 30 billion in 2023 and above CNY 50 billion in 2024, against roughly $110 billion a month in the secondary market for emerging-market sovereign eurobonds — some fifteen times smaller at around 7.1 CNY to the dollar. ICMA's 2017 diagnosis — "most panda bond investors intend to hold to maturity" — holds qualitatively, though the rise in trading institutions from 1,894 to 2,493 over a year points to gradual improvement.

Foreign investor access runs through three channels: CIBM Direct (since 2015–2016, registration with the PBoC, no quotas and no SAFE approval), QFII/RQFII (quotas abolished in September 2019) and Bond Connect Northbound (since 3 July 2017). At the end of 2024 foreign investors held CNY 4.16 trillion of interbank paper — around 2.7% of the market; 592 institutions worked through CIBM Direct, 830 through Bond Connect and 266 through both. Bond Connect monthly turnover in April 2026 was CNY 984.8 billion, CNY 46.9 billion a day on average, but the composition explains why panda paper is barely visible in that flow: 58% of turnover is policy bank paper and 22% government bonds, leaving the whole credit segment to share the remaining 20% (Bond Connect Flash Report, April 2026). The Northbound Bond Connect fee was cut by 60% in January 2025.

Panda bonds are eligible collateral in cross-border repo. Eligible collateral covers all cash bonds held by overseas institutional investors under Northbound Bond Connect, corporate paper included. The mechanics: request for quote through CFETS, outright repo with transfer of title, DVP settlement, a maximum tenor of 365 days and cycles from T+0 to T+3; the repo financing limit is up to 100% of renminbi sovereign and bank paper outstanding and 80% for other institutional investors (Bond Connect).

Indices. Global benchmarks do not include panda bonds: China's inclusion in the Bloomberg Barclays Global Aggregate (from April 2019) and in the FTSE WGBI covers only government bonds and policy bank paper. That structurally cuts off passive money and explains why the non-resident share of holders sits around 15% rather than rising. A local answer appeared on 4 January 2026: CCDC and ICBC launched the first panda market index series, 中债—工行熊猫债系列指数 — the ChinaBond–ICBC Panda Bond AAA Index covering 80% of issuers, and the ChinaBond–ICBC Panda Bond 30 Index tracking the thirty largest issues, selected for recognisability and liquidity (Sina Finance, 4 January 2026). An index published by the local depository together with the market's largest underwriter is unlikely to displace external benchmarks in global fund mandates any time soon, but it gives the market its first reference point.

The market in numbers and who buys

The market grew roughly 74% year on year in 2026 and passed CNY 200 billion, but 2025 was a falling year for it and demand is almost entirely Chinese.

Three qualifications change the reading. First, a large part of the 2026 increase is a low-base effect. Second, the dim sum market is 2.2–3.4 times the size of the panda market — CNY 358 billion against CNY 160 billion in H1 2026 (Goldman Sachs via SCMP, 10 August 2026) and CNY 683.04 billion against CNY 202 billion by mid-August. Third, and this one is structural: the share of genuinely foreign issuers, as opposed to offshore-incorporated Chinese groups, has risen from 27% (2023) to 32% (2024), 39% (2025) and 50% over January–July 2026; outstandings stood at CNY 506.4 billion on 31 July 2026, roughly 36% above a year earlier (S&P China Ratings, 13 August 2026); on Sina Finance's perimeter a week earlier, at 24 July 2026, outstandings were CNY 516.24 billion — the two bases differ by about 2%. On first-half volume Far East Credit puts the genuinely foreign share at about 54% — ahead of issuers with Chinese roots for the first time.

Demand, though, is domestic. Panda bonds trade on the interbank market, which a non-resident reaches through Bond Connect, CIBM Direct and QFII/RQFII, but the constraint is not access — it is mandate. Panda bonds are not in the global indices, so passive money goes past them into government bonds and policy bank paper, while an active manager runs into a rating scale that does not map. Hence 15% foreign participation in H1 2026, and 0.6% in Slovenia's deal; the historical high for the non-resident share was 2021 (21%), and it has fallen since.

The pricing anchor: FDL puts the premium of a sovereign panda over Chinese government bonds at around 100 bp, yet investment-grade issuers now price inside that — Kazakhstan took 1.90% with CGB yields near 1.4%.

One demand driver is fiscal: the exemption of non-resident interest income from CIT and VAT was extended on 15 January 2026 to 31 December 2027, so for a bond maturing after that date the regime is not guaranteed for the full life of the paper (withholding mechanics sit in the material on withholding tax).

Red chips: how foreign the market really is

For most of its history the panda market was not foreign funding but offshore Chinese funding: up to 75% of volume from 2014 to mid-2025 came from offshore-incorporated vehicles of Chinese groups, and the break came only in 2025–2026.

The legal basis of the channel is article 28 of the 2018 Measures: institutions of Hong Kong, Macau and Taiwan apply the Measures by analogy. That legalised the structure in which a Chinese group borrows renminbi in China through a Cayman or Hong Kong subsidiary. The scale: on CICC's estimate, from 2014 to June 2025 around 75% of market volume came from issuers with Chinese roots; in 2023 it was 79.6% of volume (Lianhe). The geography of incorporation in 2022 looked like this: Cayman 32.68%, Hong Kong 24.18%, international institutions 15.28%, the Netherlands 12.34%, Bermuda 10.81%, Macau and Hungary 2.35% each; in 2023 the share of issuers incorporated in Hong Kong jumped to 70.5%. On the exchange perimeter the skew persists to this day: by country of incorporation in 2025 it was Bermuda 20.7%, Hong Kong 19.5%, Cayman 9.5%.

The break can be dated. The share of genuinely foreign issuers ran 27% (2023) → 32% (2024) → 39% (January–October 2025) → 50% over January–July 2026 by count, and around 54% by first-half volume. On CITIC Securities' estimate, almost half of first-quarter 2026 issuers were purely overseas entities including supranationals; in May 2026, five of the eleven issuers of the month were "purely offshore borrowers" raising CNY 13.55 billion, more than half the month's volume (SCMP, 8 June 2026).

The change of cast changes what the product is for. A purely offshore borrower has no operations in China at all, so its motive is not industrial: Caixin records directly that the ability to move proceeds abroad became a more significant reason for coming to the market than the level of rates, and that by 2024 around 30% of the money raised was deployed outside the PRC. For that group a panda is not cheap funding but a channel for obtaining renminbi as such.

Panda market statistics cannot be read as statistics of foreign demand for renminbi without first checking the composition of issuers. On every database the "panda bond" perimeter includes offshore vehicles of Chinese groups, for which the PRC is the home market in substance, and before 2023 they supplied around 80% of volume. Two errors follow. The first is to treat market growth as an indicator of renminbi internationalisation — for most of its history it measured domestic money moving between pockets. The second is to compare the panda market with the samurai or yankee markets on gross figures without adjusting for that perimeter. The correct basis of comparison is genuinely foreign issuers, and that is only half of 2026 volume and less than a third of 2023 volume.

The issuer register

In more than twenty years the market has seen thirteen sovereign and sub-sovereign jurisdictions, six supranational institutions, around twenty foreign banks and more than fifty non-financial companies, with more than CNY 400 billion raised in the corporate segment alone.

Sovereigns and sub-sovereigns.

A coverage check: on Bank of China's June 2026 data the market has seen 19 sovereign panda issues, and the bank participated in 90% of them; the register above records 17 deals across 13 jurisdictions — the gap is probably repeat tranches from Hungary, the Philippines or the UAE.

Supranational institutions.

Bank of China has raised a cumulative CNY 116 billion-plus for AIIB, ADB and NDB; in 2025 AIIB, NDB and Afreximbank placed around CNY 35 billion for Belt and Road infrastructure and green projects.

Banks and financial institutions.

Foreign banks placed CNY 47.5 billion of panda bonds in 2024 — more than triple the 2023 figure (S&P Global, 25 August 2025).

Non-financial corporates.

The 2025–2026 deals that set the pricing reference, and two corrections to lists that circulate in the press. Portugal issued a dim sum bond in April 2026, not a panda — an eight-year note of roughly €250 million equivalent at a coupon of 1.77% (BOFIT, 26 June 2026); its ten-year euro paper was yielding around 3.4% at the time. That deal appears regularly, and wrongly, in lists of sovereign panda issues. The second correction is Brazil: the programme was announced in June 2026 and no bond had been placed as at September. The 2026 pricing corridor is set at the bottom by Deutsche Bank on 29 May 2026 (1.72% at three years) and UBS on 27 August 2026 (1.78% at five years), and at the top by Samruk-Kazyna (2.18%); the sovereign reference is 1.89–1.90%, from Slovenia, Kazakhstan and Indonesia.

Defaults and problem situations

No material default on a panda bond has been publicly recorded since the market opened in 2005; the regulator publishes no official default statistics for the segment.

The formulation belongs to Li Qinghe, an analyst at Founder Securities (January 2026): "no material default event has arisen in the panda bond market to date and the overall credit risk is controllable; because PRC monetary policy is relatively independent, the correlation of renminbi bond yields with the major global markets is low — so for overseas investors panda bonds can be a reasonable defensive asset" (Cailianshe, 6 January 2026).

The explanation for the clean record is structural rather than qualitative. First, the portfolio is skewed to the top of the scale: 88% of 2025 volume carried a domestic AAA, around 57% of issues had no issue rating at all, and around 11% had neither an issuer nor an issue rating — mostly development institutions. Second, and more important: the risky segment was cut out by regulation before the crisis arrived. Real estate made up 29% of panda bond issuers over 2015–2019 and around 15% of cumulative issuance across the market's history, and more than 70% of exchange issues at the 2016 peak; the tightening of external debt controls on developers over 2017–2019 collapsed the exchange perimeter from CNY 83.84 billion to CNY 4.1 billion before the wave of developer defaults. Third, the Chinese developer defaults themselves fell on bonds of their mainland operating subsidiaries, not on panda paper of the offshore parents, most of which had already matured by then.

A zero default rate over more than twenty years says nothing about loss given default. There is no precedent for enforcement on a panda bond against a foreign issuer under PRC law: article 22 of the 2010 Interim Measures expressly applied PRC law to default and disputes, but the provision has produced no actual cases. Two practical consequences follow. For the investor: a panda bond's yield contains no premium for uncertainty about the enforcement process, because the process is untested, and the absence of defaults over two decades itself creates moral hazard — issuers perceived as state-linked are priced more softly than they should be. For the issuer: a spotless record on this market will not teach Chinese bank treasuries how to work through a complex credit situation, and the first real event will be handled with no practice behind it.

Announced but never placed is a separate class of story that the statistics do not show.

The RUSAL issue also explains its own invisibility. The panda market is split into two perimeters with different regulators: NAFMII and the PBoC run the interbank one, the CSRC and the exchanges the exchange one. RUSAL went through the exchange segment and by non-public placement, so the deal does not and cannot appear in the interbank market's issuer registries, while commercial databases capture the programme only partly: the trade press, citing LSEG data, describes it as a single CNY 500 million 5.5% three-year deal in 2017 — that is, it sees the second tranche alone. Absence from the NAFMII registry is a property of how the market is mapped, not evidence that the placement never happened. The CSRC called the issue the first panda bond from a major Russian enterprise and the first from a company in a Belt and Road country; the legal anatomy of the deal was written up by the firm that advised on it (Zhong Lun).

The dysfunction of 2024 and early 2025 is the most serious market event of recent years (an assessment), and formally it was not a default. On CICC's estimate the PRC capital and financial account deficit reached $187.3 billion at the end of 2024 and the exchange rate broke through 7.3 to the dollar — and in those conditions foreign companies met de facto restrictions on remitting panda proceeds abroad, which depressed their issuance. The restrictions eased from the third quarter of 2025, and CICC names that, rather than the level of rates, as the main cause of the subsequent surge in issuance by foreign corporates (Cailianshe, 6 January 2026). Yinfa No. 272 (2022) itself did not change — what changed was how it was applied.

Place in renminbi internationalisation

Panda bonds cover the liability side of renminbi internationalisation: trade settlement makes the renminbi a settlement currency, Bond Connect an investment currency, and the panda market a borrowing currency — without which the move to reserve status does not happen.

The PBoC states the objective directly: the RMB Internationalisation Report 2023 and the Financial Stability Report 2024 set the goal of "enhancing the renminbi's function as a financing currency", with express support for panda bond issuance by overseas central banks, development institutions and multinational groups. The regulator's own metrics set the scale of the context.

The source is an interview with the head of the PBoC's Macro-Prudential Policy Department (first-half 2025 data, published October 2025); third place in payments is the PBoC's own assessment, and the interview does not disclose the basis of calculation. The key milestone for demand is measured differently: on the SWIFT tracker, which ranks currencies by share of global payments by value, the renminbi overtook the yen in November 2023 to reach fourth place. The two figures do not contradict each other — different sources, different dates, different perimeters.

The access infrastructure was built in parallel and determines who can buy panda paper at all.

The source is Bond Connect Milestones. By the channel's eighth anniversary (8 July 2025) it served 1,169 international investors from more than 70 jurisdictions holding CNY 4.35 trillion of onshore PRC bonds, after five years of 12% average annual growth. The mechanics of non-resident access to Chinese paper are set out in full in the country hub.

The link with the panda market cuts both ways. On one side, index inclusion secured passive demand for government bonds and policy bank paper — and that is precisely the curve a panda prices to. On the other, panda bonds themselves are in no global index, so passive money does not reach them: the non-resident share of holders fell from 18.6% in June 2025 to 14.6% at the end of 2025 and has held around 15% through 2026. Goldman Sachs puts the constraint plainly: the current issuance boom correlates not with expectations of renminbi appreciation but only with the relative cheapness of funding — and cheapness alone is not enough to internationalise a currency (SCMP, 10 August 2026).

What changed in 2024–⁠2026

The base act, Announcement No. 16 (2018), was not structurally revised in this period; what moved was the calibration — frameworks for sovereigns, the holders' meeting mechanism, the green perimeter, the tax exemption, rating discipline and fees.

Risks and constraints

The risks sit not in the issuer's credit quality but in the currency transformation, the administrative nature of the approvals and the narrowness of the investor base.

1. Currency risk and hedging cost — the primary one. The hedge consumes between 95% and more than 100% of the nominal saving; for tenors of five years and beyond the onshore swap is expensive or unavailable. Forward premium mechanics sit in the material on lending in China.
2. Risk that repatriation rules change. Yu Yongqiang of JunHe names a tightening of cross-border controls as one of two factors capable of collapsing demand, and the 2024 to early 2025 precedent bears that out: the rule did not change, the practice did. A deal whose economics rest on taking the whole amount out rests on administrative discretion, not on a rule.
3. The investor base is narrow and domestic. H1 2026: banks 39%, funds 37%, foreign investors 15%, down 3.8 percentage points from the mid-2025 high; Slovenia placed 0.6% abroad. A panda bond gives access to Chinese capital, not to international capital through a Chinese venue.
4. There is effectively no appetite for sub-investment-grade risk — FDL calls it "extremely limited". Hence the necessity of guarantees: ADB and AIIB for Pakistan (2026), AIIB and AfDB for Egypt (2023), at annual guarantee fees of 50 and 80 bp (FDL Policy Note 33, March 2026) that stack on top of the coupon. Guarantee capital at the development banks is scarce, so the channel does not scale at will.
5. Secondary market liquidity. The annual turnover ratio is around 1.3–1.4x; FDL puts average monthly turnover at around CNY 30 billion in 2023 and above CNY 50 billion in 2024, against roughly $110 billion a month in the secondary market for emerging-market sovereign eurobonds — some fifteen times smaller.
6. Rating inflation. On Lianhe's 2025 perimeter 65.3% of volume is unrated and 43 of the 44 rated tranches carry AAA; on S&P China's 2023 to October 2025 sample 57% of placements carry no issue rating and, where one is present, the issuer rating is AAA in 92% of cases. Six panda issuers whose international rating is speculative grade are rated AAA to AA+ domestically. The regulator acknowledged the problem in the NAFMII notice of 21 July 2026, and on 3 September 2026 S&P Global publicly described the practice as importing risk into the domestic market.
7. No enforcement precedent. Not one material default in more than twenty years — and therefore not one tested enforcement procedure on a panda bond of a foreign issuer under PRC law. A historical zero default rate says nothing about loss given default.
8. The sanctions filter. There is no formal filter and the practical one is absolute: the lead underwriter must be a first-tier Chinese bank or securities house, and every institution at that tier observes dollar compliance. Screening practice sits in the material on sanctions compliance.
9. PRC governing law, disputes heard in the PRC, no English-law protection for the holder: no negative pledge as a standard covenant, no self-contained list of events of default with an individual right of acceleration, and no trustee in the fiduciary sense.
10. The 2027 tax cliff. The exemption of non-resident interest income runs to 31 December 2027; without an extension the combined burden of some 15–16% on the coupon would consume around 28 bp of yield — the whole credit spread of the paper over the CDB curve.
11. Operating requirements. Disclosure in simplified Chinese only; a reconciliation of differences with CAS certified by a PRC firm; the non-resident auditor's filing with the PRC Ministry of Finance 20 working days ahead; disclosure of material events within seven working days; update of the registration form within 20 working days after a tranche closes.