Hook: electromechanical breakers take milliseconds to clear a fault. At 800 VDC AI rack density, the energy you dump into busbars, PSUs, and copper in that window can already cook selectivity—and your neighboring racks. Solid-state circuit breakers (SSCBs) are designed to interrupt in a few microseconds. That gap is the real story behind this week’s Infineon × SolarEdge move.
What just happened
On 9 September 2026, Infineon and SolarEdge announced they are extending their partnership into solid-state circuit breakers for 800 VDC AI and hyperscale data centers. SolarEdge leads the SSCB design; Infineon supplies CoolSiC™ silicon-carbide JFETs as the switching core. The companies frame the SSCB as the missing layer between the solid-state transformer (SST) and the compute rack—not a replacement for conversion, but the protection that makes high-density DC distribution workable.
That extension builds on their November 2025 SST collaboration: medium-voltage AC (13.8–34.5 kV) straight to 800–1500 VDC, targeting over 99% conversion efficiency (vendor claim — to verify in independent rack tests).
One day earlier, on 8 September 2026, Enphase said 4 kW IQ SST power modules are now being built in Arlington, Texas, for full-scale IQ SST racks aimed at the same 800 VDC AI architecture—MVAC → 800 VDC, with racks assembling up to about 5 MW from hundreds of modules. Full-system demos are expected in November 2026; customer pilots are planned for 2027 (company timeline — to verify).
Same week, two layers: Enphase is scaling the conversion block (SST modules on a U.S. line). Infineon + SolarEdge are scaling the protection block (SSCB with SiC JFETs) that sits downstream of that conversion. Grid → SST → DC distribution → SSCB → GPU rack.
Why DC breakers are hard (and why millisec isn’t fine)
AC breakers get a free lunch: current naturally crosses zero every half-cycle, so the arc extinguishes. DC has no natural current zero. Open a mechanical contact under load and you fight a sustained arc. Clearing time stretches into the millisecond regime—plus contact wear, bounce, and limited ability to coordinate “selectivity” when dozens of high-power feeds share a dense 800 VDC bus.
At AI rack power density, fault energy scales with voltage × current × time. Stretch “time” by three orders of magnitude (ms vs µs) and you are no longer talking about a neat fuse event—you are talking about busbar damage, upstream trip cascades, and lost neighboring compute. Electromechanical breakers were never designed as the primary selectivity tool for this class of DC distribution.
What SSCB + SiC JFET actually changes
An SSCB replaces the mechanical contact with a semiconductor switch. No contacts to separate means no arc chamber dance—and Infineon/SolarEdge state the design goal as interrupting DC faults within a few microseconds. That is the selectivity window operators need if they want denser DC distribution without oversizing every upstream stage.
Why SiC JFETs specifically? For breaker duty you want very low on-resistance (conduction loss is always-on tax), extreme surge / avalanche ruggedness when you must turn off high inductive energy, and a device that can be cascode-configured for normally-off system behavior. Infineon’s CoolSiC JFET portfolio (including 750 V / 1200 V evaluation paths aimed at SSCB apps) is built around that profile—low RDS(on), high single-pulse current capability, and avalanche behavior useful when inductive energy must be absorbed or steered to a clamp (device ratings — to verify against final SSCB BOM).
Engineering map (simplified)
SST = collapse MVAC → HVDC conversion stages (efficiency + footprint).
SSCB = microsecond DC fault isolation between SST bus and rack (selectivity + density).
GPU rack = the load that cannot wait for a mechanical trip.
SST alone does not solve protection. A 99%-class converter that feeds an unprotected or slowly protected 800 VDC plant still leaves the distribution layer as the weak link. Conversely, a fast SSCB without an efficient SST still leaves you with multi-stage AC/DC conversion losses and a fat sidecar. The September announcements are interesting precisely because they land on different layers of the same stack in the same week.
Takeaway for readers
If you are tracking 800 VDC AI power, stop treating “SST” as the whole story. Conversion speed and protection speed are different engineering problems. Enphase’s U.S.-built IQ SST modules push the MVAC→800 VDC conversion narrative into manufacturing. Infineon + SolarEdge’s SSCB push says: without microsecond-class DC interruption, rack-dense 800 VDC distribution remains selectively fragile. The winners will be the stacks that co-design both—not the decks that only show converter efficiency slides.
We have covered SST / AI power architecture themes on the channel before—if you want the visual walkthrough of grid-to-rack conversion, see @TheEngineeringCore-v. This post is about the protection gap that sat quietly behind those conversion demos.
When your facility team specs the next 800 VDC hall: are they budgeting for SSCB selectivity—or still assuming the breaker catalog from the AC era will “just work”?
SOURCES
- Infineon press release — SolarEdge SSCB collaboration for 800 VDC AI DCs (9 Sep 2026)
- eeNews Europe — Infineon and SolarEdge bring solid-state protection to 800 VDC
- Enphase GlobeNewswire — IQ SST power modules built in Texas (8 Sep 2026)
- Transformer Magazine — Enphase starts US SST module production
Author: The Engineering Core · Blog: The Engineering Core · Figures marked “to verify” are vendor claims pending independent confirmation.
