A power supply built on the wrong material chemistry pays for capability it never uses. Silicon is the cost baseline; silicon carbide buys high-voltage and high-temperature headroom; gallium nitride buys switching speed and compactness. The three are not rivals for the same job — they map onto different stages of the same supply. A charger uses GaN in the high-frequency stage and silicon in the diode output; a PFC uses SiC where the voltage and efficiency demand it, and silicon where it does not. This article compares the three chemistries by stage, walks the charger, PFC, and server cases, and gives a decision map readers can run against their own system.
Three Chemistries at a Glance: The Property Table
The three materials sort by three properties: voltage class, switching speed, and temperature headroom. Silicon supports a modest voltage and speed at the lowest cost; silicon carbide supports high voltage and high temperature at a premium; gallium nitride supports the fastest switching at a moderate voltage. The property table is the whole comparison:
| Property | Silicon | SiC | GaN |
|---|---|---|---|
| Voltage class | Up to ~900 V | 1200-1700 V+ | Up to ~650 V |
| Switching speed | Slowest | Fast | Fastest |
| Temperature headroom | Moderate | Highest | Moderate |
| Cost | Baseline | Premium | Premium |
| Best stage | Anywhere cheap | High-V / high-freq power | Compact high-freq |
The table is the material decision in one view: the requirement picks the column, and the duty picks the row. A designer who reads the table before the product has already framed the chemistry question correctly.
Charger Case: Why GaN Shrinks the Brick but Not the Diode
The GaN charger story is the clearest case of chemistry-by-stage. The gallium nitride part sits in the high-frequency power stage, where its fast switching shrinks the magnetic components and therefore the brick. The same charger’s output rectifier is still a diode — and often a silicon Schottky — because the output stage does not demand the frequency that GaN provides. The GaN charger article makes this split explicit: GaN shrinks the brick in the AC-DC front stage, and the diode at the output stays silicon because it is a rectification role, not a switching one.
The lesson of the charger case is that the chemistry follows the function within the same product. A buyer who assumes a “GaN charger” is all GaN misses that the output diode and other stage components are selected by their own duties. The fast recovery diode article covers the output-stage rectifier choice, and the GaN gate-drive article shows the drive the GaN stage demands.
A worked stage-loss example makes the chemistry decision numeric. A PFC stage switches 10 A at 120 kHz against a 400 V bus. The silicon 650 V MOSFET dissipates a switching energy per cycle that, multiplied by 120 kHz, produces a switching loss of several watts; the SiC part at the same duty cuts that term sharply, and the difference is the watts the heatsink no longer has to shed. Over the product life at the operating power, that watts-per-device difference is the value that pays the SiC premium. The same stage at 40 kHz, by contrast, sees a much smaller switching term, and the silicon part holds the loss close enough that the cost baseline wins. The crossover is the frequency and the current, and the loss worksheet is the calculator that finds it.
A second worked case applies the map to a charger. A 65 W fast charger runs a GaN power stage at several hundred kilohertz, shrinking the flyback transformer and the brick; its output rectifier is a silicon Schottky because the output duty is low-frequency rectification. The same charger, redesigned at a lower frequency, could use silicon throughout and simply be larger; the GaN premium is paid for the size reduction, not for electrical necessity. The two cases together show the decision is always stage-specific: the chemistry that shrinks the brick is not the chemistry that rectifies the output, and a designer who applies one chemistry to the whole box is overpaying for capability in the stages that do not need it.
PFC Case: Where SiC Earns Its Price
The power-factor-correction front-end is where SiC earns its premium. A PFC stage switches at high frequency against line voltage, and the conduction and switching loss at that duty is where a silicon part heats up and the SiC part stays cool. The 650 V SiC PFC article and the SiC Schottky PFC article work the two SiC device types in the same stage: the MOSFET for the switching, the SBD for the diode roles.
The PFC case shows why SiC is the natural fit when the frequency and the loss terms are high enough: the premium is justified by the watts saved, the cooler junction, and the smaller heatsink. The decision is the loss-math comparison run at the real frequency, and the loss worksheet is the tool that computes it. Where the PFC duty is modest, silicon remains the cheaper answer; where the stage is pushed, SiC takes over.
Server and Industrial Cases: The Silicon Fallback
The server and industrial supplies are where silicon still wins by default. A server power stage at a moderate voltage and a conventional frequency has no pressing need for GaN’s speed or SiC’s voltage class, and the cost baseline makes silicon the right call. The same is true of many industrial supplies where the duty is steady and the frequency is modest — the wide-bandgap premium buys nothing the requirement did not ask for.
The fallback is not a failure of the new materials; it is the correct engineering result of matching the requirement. The server power article and the PV inverter rectifier article show the systems where the stage mix keeps silicon in the picture alongside any wide-bandgap parts. The design that names the stage, computes the loss, and then chooses the chemistry is the one where silicon falls out of the decision honestly.
The stage map condenses the whole article into a runnable table:
| Stage | Typical duty | Best chemistry | Why |
|---|---|---|---|
| AC-DC front, high freq | Fast switching, compact | GaN | Speed shrinks magnetics |
| PFC, high V + freq | High voltage, switching loss | SiC | Voltage + cool junction |
| Output rectifier | Low-freq one-way | Silicon Schottky | Rectification, cheap |
| Server / industrial | Modest freq, steady | Silicon | Cost baseline |
The table is the article in one view: each stage names its typical duty and the chemistry that fits it, with the reason in the last column. A designer who reads a system stage by stage, filling in the duty of each block, produces the chemistry choice for the whole product without a single new measurement.
The decision map for a reader’s own product runs the same way: list the stages, note the duty of each, and let the table assign the chemistry. The demanding stage — the one that runs fast, hot, or high-voltage — is where the wide-bandgap premium goes; the rest of the stages stay silicon. The result is a supply that is neither over- nor under-specified on chemistry, which is the entire point of the three-material, three-jobs framing this article has built.
Reading the Chemistry-Versus-Stage Decision Map for Your Product
The decision map for a reader’s own system runs in four questions. First, what is the highest voltage in the power stage — does it demand a wide-bandgap class? Second, what is the switching frequency — does it demand GaN’s speed or reward SiC’s lower switching loss? Third, what is the thermal budget — does the design need SiC’s temperature headroom to fit the box? Fourth, what is the cost target — does the premium clear the requirement-based test?
The map resolves most systems in two steps: name the demanding stage, then pick the chemistry that stage needs. A charger’s AC-DC stage that runs fast and hot points at GaN or SiC; its output diode stays silicon; a PFC at high voltage and frequency points at SiC; a conventional server or industrial supply at modest duty stays silicon end to end. The wide bandgap primer and the SiC power device guide are the two anchors for the wider material, and the SiC MOSFET families plus the SiC SBD families supply the parts a SiC stage lands on. The chemistry-versus-stage map, applied stage by stage, is what keeps the material choice honest and the supply efficient.
The closing rule of the chemistry-versus-stage view is that the material market converges on the same discipline the device market has always used: match the stage duty to the component, and let the cost sheet confirm the match. GaN, SiC, and silicon are not competing families for one job; they are three tools for three stage duties, and the best supply is the one that uses each where it belongs. A reader who takes the property table and the stage map away from this article has the whole decision method in two small reference cards. And those cards work for the next product, the next frequency, and the next voltage class without any new physics. That is the practical value of reading them.