In Conversation with AI Mentor: From a Neglected “Tech Outcast” to a “Key Player” in AI Data Centers

In December 2025, the IEC released IEC 60947-10, a standard dedicated to semiconductor circuit breakers. Almost at the same time, NVIDIA, in its whitepaper on an 800 V DC power distribution architecture, explicitly recommended replacing traditional mechanical circuit breakers with solid-state circuit breakers (SSCBs).

Meanwhile, an intriguing trend is emerging in device selection: just as SiC MOSFETs are enjoying their heyday, the JFET—once dismissed by the industry as “technically interesting but commercially unattractive”—is quietly making a comeback.

We sat down with our AI Mentor to explore this shift. Why can the JFET carve out its own niche in SSCB applications? What technical logic and industry shifts lie behind this comeback of a “normally-on” device?

Questions   &   Answers

1. The Misunderstood “Normally-On” Device: The Structural Advantages of the JFET

Interviewer: When people hear “JFET,” their first reaction is often: “It’s a normally-on device—troublesome to drive and prone to failure.” Yet in the SSCB arena, major international players are betting heavily on this very approach. What is the core logic behind this apparent contradiction?

AI Mentor: This is indeed a fascinating “perception reversal.” The JFET’s normally-on characteristic is a shortcoming in general power electronics applications—if the gate drive loses power, the device turns on, creating a serious safety risk. But in the specific context of a circuit breaker, “normally-on” actually becomes a design philosophy.

Think about it: what is the essence of a circuit breaker? It must “conduct reliably over the long term and interrupt instantaneously when needed.” Traditional mechanical breakers rely on the physical separation of contacts—they suffer from arcing, contact wear, and slow operation. SSCBs use semiconductor devices to achieve arc-free interruption, but semiconductors inherently have conduction losses—a decisive cost factor for a breaker that carries current continuously.

This is where the JFET’s structural advantage comes into play. As a unipolar device, it has no gate-oxide layer like a MOSFET, its conduction path is shorter, and its specific on-resistance is inherently lower. Onsemi’s EliteSiC JFET and Infineon’s CoolSiC JFET deliver 20–30% lower conduction losses than MOSFETs at the same rating. For data center power distribution systems running 24/7, this gap translates into substantial electricity savings every year.

Interviewer: But how do you solve the gate-drive reliability problem that comes with a “normally-on” device?

AI Mentor: That’s where the cascode configuration shines. By connecting a normally-on SiC JFET in series with a low-voltage silicon MOSFET, and driving the low-voltage MOSFET with a standard silicon gate driver, the overall device exhibits normally-off characteristics. Onsemi’s Combo JFET even integrates both devices into a single package, completely solving the gate-driver compatibility issue.

Even better, the cascode configuration inherently supports bidirectional blocking. The body diode of a conventional SiC MOSFET conducts in the reverse direction, so building a bidirectional breaker requires two devices in a back-to-back configuration. A single JFET-cascode device, by contrast, achieves bidirectional blocking on its own—simpler circuitry, lower losses, and better cost.

“The JFET’s normally-on nature is not a flaw, but a rediscovered degree of design freedom. In a circuit breaker, ‘normally-on’ means a shorter conduction path, lower losses, and a simpler bidirectional topology—exactly what an SSCB needs most.”

2. 10 µs vs. 0.8 µs: Where Is the Limit of Interruption Speed?

Interviewer: We’ve come across some academic work showing that JFET-based SSCBs can interrupt fault current within 10 µs, with some studies even achieving interruption of 400 V/180 A in 0.8 µs. That’s an order-of-magnitude improvement over mechanical breakers. But in real engineering applications, what limits further speed increases?

AI Mentor: Interruption speed is indeed the core selling point of an SSCB. Mechanical breakers operate in milliseconds; SSCBs work in microseconds—a difference that is critical for protecting semiconductor loads and limiting fault energy.

But faster is not always better. There is a “detect–decide–act” chain that must be balanced.

First comes fault detection. You need to accurately determine, within a few microseconds, whether you are dealing with a genuine fault or a normal inrush surge. This calls for high-precision current sensing and intelligent algorithms. Detect too quickly and you risk nuisance tripping; too slowly and you lose the protective benefit.

Second is the device’s own turn-off capability. The JFET’s turn-off behavior is actually more controllable than a MOSFET’s—as a current-controlled device, it turns off more gently and offers better dv/dt controllability, which is highly favorable for suppressing turn-off overvoltage. However, turn-off speed is limited by the response time of the gate-driver circuit and the speed at which gate charge can be extracted.

Finally, there are system-level constraints. The faster the interruption, the higher the di/dt, the greater the overvoltage induced by parasitic line inductance, and the harder the snubber circuit design becomes. In practical engineering, therefore, 10 µs is already an excellent level of performance; 0.8 µs is more of a validation under extreme laboratory conditions.

Interviewer: Self-powering is also an important research topic for SSCBs. Could you elaborate?

AI Mentor: Self-powering is a very promising direction. Conventional SSCBs require an external auxiliary power supply for the gate-driver circuit, which adds system complexity and cost. The self-powered approach exploits the energy of the fault current itself to drive the turn-off: when a fault occurs, the massive fault current flows through an energy-harvesting circuit that converts it into gate-drive energy, triggering the JFET to turn off.

The benefit is that the breaker still operates reliably even if the external power supply fails completely. For a device like a circuit breaker—“rarely used, but must work when called upon”—self-powering greatly enhances system robustness.

3. An Industry Shift: From “Tech Outcast” to “Key Player”

Interviewer: You mentioned that the JFET was once seen as “technically interesting but commercially unattractive.” Now both Infineon and onsemi are pushing JFET product lines, and Infineon plans to ramp up CoolSiC JFET production in 2026. What’s driving this shift?

AI Mentor: I believe it’s a precise match between the device and its application scenario.

Over the past fifteen years, the main battleground for SiC devices has been high-frequency switching applications such as EV main traction inverters and charging piles. These applications require high-frequency PWM modulation, and the MOSFET’s voltage-driven nature, mature ecosystem, and wide selection of gate-driver ICs have made it the absolute mainstream. Despite its low conduction loss, the JFET has struggled for competitiveness in the general market because of the driving complexity imposed by its normally-on characteristic.

But the SSCB is a completely different scenario: low-frequency operation, long-term conduction, extreme sensitivity to conduction loss, and undemanding gate-drive frequency. This combination amplifies the JFET’s strengths while minimizing its weaknesses.

More importantly, the rise of AI data centers has created entirely new demand. NVIDIA’s push for an 800 V DC distribution architecture calls for large numbers of solid-state circuit breakers in the distribution circuits. These breakers carry current continuously, so conduction loss translates directly into cooling costs and electricity bills. At the same time, data centers demand extremely high reliability, and the JFET’s freedom from gate-oxide degradation—a known pain point of MOSFETs—becomes a key selling point here.

So it’s not that the JFET suddenly got better; it’s that it finally found a battlefield truly suited to it.

Interviewer: China seems to be relatively lagging on the SiC JFET front. Some commentators argue that this is not a question of technical capability but of commercial judgment. What’s your view?

AI Mentor: That assessment holds some truth. The JFET manufacturing process differs considerably from MOSFETs—it requires dedicated ion implantation and annealing steps to form the channel. Chinese manufacturers have invested heavily in SiC MOSFETs, with production lines, processes, and talent all built around them. Shifting to JFETs means additional R&D investment and production-line adjustments.

But I’ve also noticed some positive signs. Sicred, for instance, has moved its 650 V/1200 V SiC JFETs into mass production, and PN Junction Semiconductor has showcased a JFET-based SSCB solution at trade shows. China is by no means behind in SSCB system-level research—Hunan University, Xi’an Jiaotong University, and Huazhong University of Science and Technology all have active teams in this area.

The key issue is supply-chain collaboration. JFETs require supporting gate-driver chips, packaging solutions, and system-level validation—this cannot be solved by a single point of breakthrough. My suggestion is that Chinese vendors learn from onsemi’s Combo JFET approach to lower the system adoption barrier through cascode integration, while working closely with universities and downstream users to achieve breakthroughs in specific applications such as data centers and energy storage.

4. The Future: The Co-Evolution of Smart Circuit Breakers and Power Devices

Interviewer: From a longer-term perspective, what are the technology evolution directions for JFETs in SSCBs?

AI Mentor: I see three trends worth watching.

First, voltage-class upgrades. JFETs today are mainly concentrated in the medium-to-low voltage range (650 V–1700 V). As 3.3 kV and 6.5 kV SiC devices mature, medium-voltage DC applications—such as rail transit and marine power—will open up new space. This places higher demands on the JFET in terms of series voltage sharing and long-term reliability.

Second, intelligent integration. The future SSCB will be more than a “switch”—it will be an “intelligent protection unit” integrating current/voltage/temperature sensing, fault diagnosis, communication interfaces, and even predictive maintenance. This raises new requirements for the device: lower parasitic parameters, better thermal performance, and deep integration with the gate-driver circuit. Here, the JFET’s structural simplicity turns out to be an advantage—it lends itself more readily to monolithic integration or modular packaging.

Third, competition and complementarity with GaN. GaN devices excel in high-frequency applications, but their current voltage and current ratings are not yet sufficient for high-power SSCBs. In the future, we may see hybrid architectures in which “GaN handles high-frequency sensing and auxiliary circuits, while the SiC JFET handles the main power path.”

“The ultimate form of a solid-state circuit breaker is not simply replacing mechanical contacts with semiconductors, but evolving ‘protection’ from passive response to active prediction. The JFET’s low loss, high reliability, and structural simplicity align perfectly with this evolutionary direction.”

Conclusion: A Technological Value Rediscovered

As our conversation drew to a close, our AI Mentor shared a telling detail: at the launch event for Infineon’s CoolSiC JFET, engineers described the device as “back to the future.” The JFET was the first field-effect transistor structure ever invented, yet in the era of power semiconductors it was overshadowed by the MOSFET. Today, on the specific battlefield of solid-state circuit breakers, it is returning in a more mature and precisely targeted form.

Perhaps the value of technology has never been linear. Some innovations must wait for the right scenario; some advantages must be re-understood. As AI data centers surge on 800 V DC power, and as the “dual carbon” goals drive deep transformation of power systems, the encounter between the JFET and the SSCB is precisely such a fit—“late, but right on time.”

This shift is both a challenge—our accumulated experience with JFETs is still limited—and an opportunity: in the emerging SSCB arena, the rules of the game are not yet written, and anything is possible.

References

[1] IEC 60947-10: Semiconductor circuit-breakers for overcurrent protection

[2] Infineon CoolSiC JFET product manual

[3] onsemi EliteSiC Combo JFET application note