This study presents a machine-learning-accelerated design of high-performance platinum intermetallic nanoparticle fuel cell catalysts. Carbon-supported PtCo intermetallic alloys are promising low-platinum oxygen reduction reaction electrocatalysts for proton-exchange-membrane fuel cells (PEMFCs), but the trade-off between particle size and ordering degree makes it challenging to achieve high specific activity and large active surface area. By using machine learning to screen the configuration space, the impact of chemical ordering on thermodynamic stability was statistically quantified. Introducing Cu/Ni into PtCo provides additional stabilization energy by inducing Co-Cu/Ni disorder, facilitating the ordering process and improving the trade-off between specific activity and active surface area. The small-sized, highly ordered ternary Pt₂CoCu and Pt₂CoNi catalysts were experimentally prepared, showing a large electrochemically active surface area of -90 m² g⁻¹ Pt and a high specific activity of -3.5 mA cm⁻². PEMFCs with net-zero carbon emissions are promising energy conversion devices, but their large-scale commercialization is limited by the high cost of platinum-based electrocatalysts for the oxygen reduction reaction (ORR). The US Department of Energy has set performance targets based on platinum group metal (PGM) usage. Recent studies have shown that carbon-supported structurally ordered Pt-based intermetallic compound (IMC) nanoparticles are promising low-Pt catalysts for ORR in PEMFCs. However, achieving high or full ordering degree is desirable for alloy catalysts. High-temperature annealing is crucial to form Pt-M alloys with ideal stoichiometric ratios and overcome the energy barrier of disorder-to-order transition, but this can lead to sintering and reduced electrochemical surface area (ECSA). The seesaw relation between particle size and ordering degree was observed in a typical impregnation synthesis of PtCo alloy. To promote the ordering degree of IMC catalysts, machine learning methods have demonstrated significant potential in accelerating material discovery by efficiently navigating design spaces and predicting properties. Here, we perform machine-learning-accelerated computational screening to de novo design the element composition to increase the thermodynamic driving force for the disordered-to-order transition and promote the nucleation of IMC phase with high ordering degree. After systematic screening of the ternary Pt₂CoM alloys, two optimal solutions of Pt₂CoCu and Pt₂CoNi were obtained. The experimentally prepared Pt₂CoCu (Ni) IMC catalysts show both large ECSA of -90 m² g⁻¹ Pt and high SA of -3.5 mA cm⁻², leading to a high MA of -3 A mg⁻¹ Pt. The highly ordered Pt₂CoCu catalysts also exhibit enhanced MEA performance in practical H₂-air fuel cells. The study demonstrates that the introduction of Cu/Ni leverages the thermodynamic driving force for theThis study presents a machine-learning-accelerated design of high-performance platinum intermetallic nanoparticle fuel cell catalysts. Carbon-supported PtCo intermetallic alloys are promising low-platinum oxygen reduction reaction electrocatalysts for proton-exchange-membrane fuel cells (PEMFCs), but the trade-off between particle size and ordering degree makes it challenging to achieve high specific activity and large active surface area. By using machine learning to screen the configuration space, the impact of chemical ordering on thermodynamic stability was statistically quantified. Introducing Cu/Ni into PtCo provides additional stabilization energy by inducing Co-Cu/Ni disorder, facilitating the ordering process and improving the trade-off between specific activity and active surface area. The small-sized, highly ordered ternary Pt₂CoCu and Pt₂CoNi catalysts were experimentally prepared, showing a large electrochemically active surface area of -90 m² g⁻¹ Pt and a high specific activity of -3.5 mA cm⁻². PEMFCs with net-zero carbon emissions are promising energy conversion devices, but their large-scale commercialization is limited by the high cost of platinum-based electrocatalysts for the oxygen reduction reaction (ORR). The US Department of Energy has set performance targets based on platinum group metal (PGM) usage. Recent studies have shown that carbon-supported structurally ordered Pt-based intermetallic compound (IMC) nanoparticles are promising low-Pt catalysts for ORR in PEMFCs. However, achieving high or full ordering degree is desirable for alloy catalysts. High-temperature annealing is crucial to form Pt-M alloys with ideal stoichiometric ratios and overcome the energy barrier of disorder-to-order transition, but this can lead to sintering and reduced electrochemical surface area (ECSA). The seesaw relation between particle size and ordering degree was observed in a typical impregnation synthesis of PtCo alloy. To promote the ordering degree of IMC catalysts, machine learning methods have demonstrated significant potential in accelerating material discovery by efficiently navigating design spaces and predicting properties. Here, we perform machine-learning-accelerated computational screening to de novo design the element composition to increase the thermodynamic driving force for the disordered-to-order transition and promote the nucleation of IMC phase with high ordering degree. After systematic screening of the ternary Pt₂CoM alloys, two optimal solutions of Pt₂CoCu and Pt₂CoNi were obtained. The experimentally prepared Pt₂CoCu (Ni) IMC catalysts show both large ECSA of -90 m² g⁻¹ Pt and high SA of -3.5 mA cm⁻², leading to a high MA of -3 A mg⁻¹ Pt. The highly ordered Pt₂CoCu catalysts also exhibit enhanced MEA performance in practical H₂-air fuel cells. The study demonstrates that the introduction of Cu/Ni leverages the thermodynamic driving force for the