Introduction: Synchronization of non-invasive ventilation is challenging in extremely premature infants. We compared patient-ventilator synchrony between non-invasive neurally adjusted ventilatory assist (NIV-NAVA) using transdiaphragmatic (Edi) catheter and synchronized intermittent positive airway pressure (SiPAP) using an abdominal trigger. Methods: This study was a monocentric, randomized, crossover trial in premature infants born before 28 weeks of gestation, aged 3 days or more, and below 32 weeks postmenstrual age. NIV-NAVA and SiPAP were applied in a random order for 2 h with analysis of data from the second hour. The primary outcome was the asynchrony index. Results: Fourteen patients were included (median [IQR] gestational age at birth 25.6 (25.3–26.4) weeks, median [IQR] birth weight 755 [680–824] g, median [IQR] postnatal age 26.5 [19.8–33.8] days). The median (IQR) asynchrony index was significantly lower in NIV-NAVA versus SiPAP (49.9% [44.1–52.6] vs. 85.8% [74.2–90.9], p < 0.001). Ineffective efforts and auto-triggering were significantly less frequent in NIV-NAVA versus SiPAP (3.0% vs. 32.0% p < 0.001 and 10.0% vs. 26.6%, p = 0.004, respectively). Double triggering was significantly less frequent in SiPAP versus NIV-NAVA (0.0% vs. 9.0%, p < 0.001). No significant difference was observed for premature cycling and late cycling. Peak Edi and swing Edi were significantly lower in NIV-NAVA as compared to SiPAP (7.7 [6.1–9.9] vs. 11.0 [6.7–14.5] μV, p = 0.006; 5.4 [4.2–7.6] vs. 7.6 [4.3–10.8] μV, p = 0.007, respectively). No significant difference was observed between NIV-NAVA and SiPAP for heart rate, respiratory rate, COMFORTneo scores, apnoea, desaturations, or bradycardias. Discussion/Conclusion: NIV-NAVA markedly improves patient-ventilator synchrony as compared to SiPAP in extremely premature infants.