The power semiconductor market in 2026 assembles itself around three demand centers — electric vehicles, solar and storage, and AI data-center compute — and the interesting news is not which of the three grew, but that they now pull on the same component families at the same time. A SiC diode can serve a solar inverter in the morning, an EV charger in the afternoon, and a server power supply at night, which means the supply chain that serves one serves all three. That convergence changes how buyers should read this year’s market data, and it is the lens this article uses to connect the macro numbers to discrete devices you can actually order.
The Macro Drivers and the Devices They Move
The shortcut used by most trend coverage is to name a market size and stop. The more useful frame is driver-to-device: each macro trend maps to a specific discrete-device family, and that mapping is what turns a market report into a sourcing decision.
EV penetration drives the demand for high-voltage SiC and IGBT devices in the drivetrain and charger. Solar and storage growth drives bypass diodes, PV-inverter rectifiers, and grid-side SiC. AI data-center compute drives the 800 V server and rectification stacks that run on SiC MOSFETs and fast-recovery silicon. None of the three is a new category — all existed in 2020 — but the simultaneous growth is new, and it shows up in the supply chain as competition for the same wafer capacity.
That is why reading trend coverage in isolation underestimates the real constraint. The relevant question is not “how big is the EV market?” but “which family of diodes and MOSFETs does each EV sold in 2026 pull from the same wafer line?” The industry reports that answer that question well — the TrendForce and Yole analyses of SiC supply and demand — show the three drivers drawing from a shared base, and that shared base is the actual bottleneck the rest of this article traces.
EV, PV, and AI Compute: Three Demand Signals
Each of the three drivers has a measurable demand signal, and the signal is worth knowing because the sourcing consequence differs.
EV and charging infrastructure pull the highest-voltage devices: 800 V SiC MOSFETs in the charger, IGBTs in some traction drives, and SiC diodes in the onboard charge stages. The 2025–2026 reports from the major semiconductor trackers consistently place EV and charging as one of the two largest SiC demand segments, and the 800 V platform shift is what moved SiC from a specialty into a mainstream power-conversion material. For a buyer, the consequence is lead time and price pressure on exactly the 1200 V SiC diode and 650–800 V MOSFET families that were exotic five years ago.
Solar and storage are the second driver, and the most volatile. PV installation growth drives bypass diodes and inverter rectifiers, and storage adds the bidirectional PCS stacks that need both rectification and inversion. The diodic content per watt has actually risen as bypass counts and MPPT stages grew, so module demand tracks installation numbers but component demand tracks a richer bill of materials per module. The relevant trend for buyers is not just module volume but the rising count of bypass diodes and protection devices per module — a driver that compounds with installation growth.
AI data-center compute is the third and the most capex-heavy. Each GPU rack draws kilowatts and needs high-efficiency rectification, which pushes server power supplies toward 800 V DC inner rails and SiC-based PFC. The industry-report picture of 2026 shows data-center power electronics as a growth segment whose efficiency demand — the “1% pays the bill” economics — directly benefits low-loss device families. That demand overlaps EV charging on the same SiC MOSFETs, which is the supply-chain tension the next section makes concrete.
SiC and GaN Growth: Numbers and Reality
The market-size numbers from the trackers are directionally consistent and worth repeating with their source and date. TrendForce’s analyses of the SiC and GaN power device market, Yole’s SiC market reports, and Omdia’s power semiconductor trackers all put SiC revenue growth in the low double digits per year through 2025–2026, with the largest contribution from EV chargers and data-center power rather than from the older SiC industrial base. Those are estimates, not invoices: tracker methodologies differ in scope (substrate versus device, module versus system), which is why the reports disagree on the exact figure and why any single number should be read with the report’s definition of market in view.
The reality behind the growth is more interesting than the growth itself. SiC wafer yield and capacity, not device demand, are the binding constraint, and the reports treat that constraint explicitly. When OEMs announce multi-year SiC supply agreements, they are locking wafer supply, not just pricing — which is why 2026 sourcing strategy increasingly reads like supply-chain strategy rather than component selection. GaN is a separate story with a different curve: strong growth in chargers and power adapters, where the compactness argument wins, but a narrower role in the 100 kW-plus conversion space that SiC holds. The wide bandgap sourcing guide translates both curves into a practical buy-for-2026 checklist.
The honest caveat is that market reports describe a market, not your bill of materials. The revenue split between SiC, GaN, and silicon depends on whether the report counts substrates, devices, or systems, and the differences between trackers are large enough that citing a single “market size” without its definition is misleading. The skill of reading these reports is less about the headline than about the scope line that defines it.
Price, Lead Times, and the Supply Chain
The supply-chain reality in 2026 is a market where the three drivers above collide on the same bottleneck. SiC wafer capacity is the clearest constraint: the studies quantified in the reports above show demand growth running ahead of announced capacity through 2025–2026, which keeps lead times and price firm even as silicon itself stays abundant. For buyers the practical consequence is to treat SiC parts as planning items — qualified early, committed early — rather than spot purchases.
Silicon power semiconductors tell a different story. Commodity rectifier and diode capacity is broad, pricing is competitive, and lead times are short, which makes the silicon 50/60 Hz rectifier and general-purpose diode families the low-risk backbone of most designs. The interesting silicon trend is consolidation at the high end: fast-recovery and low-VF silicones are still produced, but the wafer capacity conversation belongs to SiC, so the premium silicon families hold their price while the commodity part stays cheap.
Regional supply adds a further layer. The reports show substrate and device production concentrated in Asia, with policy incentives in the US and EU aiming to add local capacity; the 2026 picture is one of diversification announcements that have not yet changed the near-term geography of the supply chain. For a buyer, the actionable question is not “where is it made?” but “can the supplier’s second source actually deliver at the same spec and lead time?” — which the second sourcing strategy guide turns into a qualification process rather than a slogan.
For a purchasing team the caveats translate into a working rule: use the market reports to set the negotiating context, not the component list. If the reports say SiC capacity is tight, the negotiation over a SiC MOSFET order is about allocation and lead time as much as price, and samples should be requested early in the design cycle rather than after the design freeze. If the reports say silicon capacity is loose, the same negotiation is conventional supplier management. The same report, read with the driver-to-device frame, tells a buyer which conversations will be hard in 2026 and which will be normal — which is the difference between a trend article and a sourcing input.
Reading This Year’s Market Data: Caveats and Sources
Market data deserves a reading procedure, because the same report can mislead and inform depending on the scope line. Four checks cover most errors.
First, identify the scope. Is the figure substrates, bare dies, devices, or systems? Substrate revenue and device revenue are different numbers with different growth rates, and an article that says “SiC market” without saying which is not yet useful. Second, check the date and vintage. 2026 planning material should cite reports published in late 2025 and 2026, not a 2023 projection dusted off as current. Third, compare two sources before repeating a number: TrendForce, Yole, and Omdia each measure a different slice, and their agreement on direction is more informative than their disagreement on exact values. Fourth, separate volume from value — a market growing in revenue because of price can look like growth in demand, and vice versa.
The named sources for this article are the tracker reports and market analyses published in the 2025–2026 window by TrendForce, Yole, and Omdia, plus the industry outlook pieces that cite SEMI data. They are directionally consistent on the three demand drivers and on SiC capacity as the binding constraint; the exact figures should be re-read from the latest edition of each report at writing time, because this article will age — which is the nature of a market snapshot and the reason the caveats section exists at all. For the device-level view that the market reports only gesture at, the SiC power device design guide and the data center power supply architectures report translate the macro numbers into component decisions, and the TVS category shows the protection side of the same 2026 bill of materials.