When Kenya announced in October 2025 that it had converted its Standard Gauge Railway debts to China Eximbank from U.S. dollars (USD) into Chinese renminbi (RMB), the move was widely framed as a breakthrough for RMB internationalization and a clever way to reduce borrowing costs. Early reporting suggested the switch could save Kenya roughly $215 million a year due to interest rate reductions associated with RMB.
But this narrative does not tell the full story.
In a new policy note, we find that the largest source of debt relief did not come from the benchmark rate change alone. Instead, Kenya’s savings were driven primarily by traditional sovereign debt restructuring tools: removing interest rate margins, adding new grace periods, and extending maturities.
Read our new policy note, Kenya’s USD-to-RMB Debt Conversion Was Really a Restructuring
Kenya’s deal is done—but the terms may hold important insights for other borrowers such as Ethiopia, which is reportedly exploring a similar conversion of Chinese loans from USD to RMB.
In our analysis, we model the present value savings generated from Kenya’s switch from the dollar-denominated U.S. Federal Reserve’s Secured Overnight Financing Rate (SOFR) to China’s RMB-denominated Loan Prime Rate (LPR) benchmark, and compare changes in debt servicing costs related to each of the other restructuring terms: reduced margin on interest rates, extending new grace periods, and changing the loan maturity.
The USD-to-RMB narrative featured in the media isn’t entirely wrong. A benchmark switch could lower borrowing costs, if RMB-linked rates are lower than USD-linked rates. But it does not automatically create major cash-flow relief. In Kenya’s case, we estimate that switching from SOFR to LPR while retaining original margins would have generated only about $23.6 million in present-value savings.
What the narrative about Kenya’s switch from RMB to USD misses is the role of the other terms in the full restructuring package—and that’s most of the story. Under the full restructuring package, Kenya’s present-value savings rose to roughly $385.8 million, as illustrated by the chart below.
This means about 93.8% of the debt relief for Kenya came from other terms in the full restructuring deal—not the benchmark shift, as illustrated by the next chart.
In addition to analyzing Kenya’s restructuring deal, we modeled what a similar (hypothetical) restructuring package might look like for Ethiopia and the country’s Addis Ababa-Djibouti railway loans. Like Kenya, Ethiopia borrowed from China Eximbank for a major railway project. These loans were also made using floating-rate loans that became more expensive as global interest rates rose. And by 2025, SOFR-linked borrowing costs had pushed Ethiopia’s expected debt-service obligations higher, making an RMB conversion potentially attractive.
Our scenario analysis shows that a benchmark shift alone while retaining the margin would provide Ethiopia with meaningful but limited relief: roughly $169.1 million in present-value savings. But under a full Kenya-style package—margin removal, a four-year grace period, and a six-year maturity extension—estimated present-value savings would rise to roughly $778 million. Under that scenario, about $608.8 million or 78.2% of the relief would come from restructuring features, not the benchmark change.
With China making a concerted push to increase RMB internationalization, the lesson for borrowers is clear: RMB conversion should not be evaluated in isolation. Without margin reductions, grace periods, or maturity extensions, the cash-flow relief may be far smaller than advertised.
Kenya’s deal therefore offers a cautionary lesson for Ethiopia and other countries seeking relief from Chinese creditors. The most important question is not simply whether debt can be converted from dollars into RMB. It is whether Chinese creditors are willing to pair that conversion with the restructuring terms that made Kenya’s agreement financially meaningful.
Read the full policy note: Kenya’s USD-to-RMB Debt Conversion Was Really a Restructuring.
