The new TP65B110HRU is a 650V class bidirectional GaN switch that lets a solar microinverter collapse two converter stages into one, cutting switch count in half and dropping the electrolytic DC link capacitor that has long limited converter lifetime.
Every solar microinverter, AI server power supply, and onboard EV charger has, until now, shared a quietly embarrassing component: a fat electrolytic DC-link capacitor bridging the two converter stages inside. It is bulky, it runs hot, and it is one of the most common reasons a power converter eventually fails in the field. On March 23, 2026, Renesas announced a switch that, for the first time, lets designers route around it altogether.
The part is the TP65B110HRU, a 650V-class bidirectional GaN switch Renesas calls the industry's first to combine depletion-mode (d-mode) GaN with integrated DC blocking in a single device. In a conventional solar microinverter, current must be steered, blocked in one direction, then steered again — which is why designers string two unidirectional silicon or silicon-carbide FETs back-to-back and tack a large electrolytic capacitor between the DC and AC stages to stabilize the intermediate bus. A bidirectional switch does both jobs natively, so two of these GaN devices can replace a back-to-back Si or SiC pair plus the DC-link stage, roughly halving the active switch count and eliminating the intermediary DC-link capacitor.
The reliability pitch is what gives the topology claim teeth. Electrolytic capacitors are a well-documented weak point in power conversion: their liquid electrolyte dries out over time, their equivalent series resistance drifts upward with thermal stress, and they are physically the largest single component on many converter boards. Removing one does not make a converter immortal, but it removes a part whose end-of-life behavior designers have spent decades compensating for. That is the engineering problem Renesas is selling the TP65B110HRU against, not a generic "GaN is faster" speed story.
The numbers Renesas reports are concrete. In a real-world single-stage solar microinverter built around the new switch, the company measures greater than 97.5% power efficiency, weighted to California Energy Commission (CEC) test conditions. The part itself is rated for ±650V continuous AC/DC blocking, ±800V transient, 110 mΩ typical on-state resistance at 25°C, and dv/dt above 100 V/ns, in a top-side-cooled TOLT package. Because d-mode GaN is a normally-on device that normally demands a negative gate bias to switch off, Renesas co-packages low-voltage silicon MOSFETs so the switch can be driven from a standard gate driver, matching the drive simplicity of conventional silicon designs without the BOM penalty of enhancement-mode (e-mode) GaN.
The applications Renesas is targeting are exactly the ones where the DC-link capacitor hurts most. A single-stage bidirectional topology fits cleanly into a residential solar microinverter, a three-phase Vienna rectifier for AI server PSUs in data centers, and onboard electric vehicle chargers — three markets that have all been bumping up against converter efficiency, density, and lifetime ceilings for the same reason. EE Times framed the part in similar terms, noting that high-voltage GaN bidirectional switches simplify board layout and shrink the magnetics and passives that have to surround them.
The honest framing, though, is that this is still a vendor announcement. The 97.5% efficiency is Renesas' own CEC-weighted measurement on its reference design; independent benchmark numbers from competitors such as Navitas, Power Integrations, Innoscience, or EPC are not yet in hand. Rohan Samsi, vice president of Renesas' GaN Business Division, cast the part as a "major shift in power conversion design norms" — a marketing claim dressed as an engineering one. The TP65B110HRU is in volume production and Renesas is sampling the RTDACHB0000RS-MS-1 evaluation kit at APEC 2026, so the part is real and available. But the longer-term questions — d-mode GaN qualification maturity, unit cost against SiC, gate-driver ecosystem breadth, and reliability in the field over a decade rather than a datasheet — will only be answered once converters built around it actually ship and age.
That is the point worth sitting with: the DC-link capacitor was never the only thing that could fail in a converter, but it was the component whose failure was the most predictable. A bidirectional switch that genuinely removes it from the topology is, at minimum, a serious answer to a reliability problem the industry has been working around for decades. Whether it is also the "major shift" Renesas claims is a question the next few years of field data, not the press release, will settle.