Why GaAs HBT won the “Handset War” but could rapidly decline with new market requirements
Before the rise of gallium arsenide (GaAs) HBTs in the late 1990s, mobile power amplifiers (PAs) relied heavily on silicon bipolar junction transistors (Si BJTs) and early GaAs devices. First-generation phones used Si BJTs operating from 450 to 900 MHz. While effective at higher supply voltages (from 4.8 to 6V), these devices struggled to deliver power and efficiency as handsets transitioned to lower voltages (from 3 to 3.6V). As an alternative, GaAs MESFETs offered improved efficiency at 900 MHz but required a negative gate voltage – necessitating charge pumps that increased cost, complexity and board space. While now famous for base stations, low voltage silicon LDMOS (laterally diffused MOS) was also used in some early mobile handsets. It was cheaper than GaAs and more “rugged” (less likely to combust if the antenna was mismatched). But, it was physically bulky. The “die size” required for a certain power level was much larger than GaAs, making it difficult to fit into the shrinking “flip phone” designs of the mid-90s.
Thus, the GaAs HBT became the “disruptor” of the late 90s because it combined the best of all worlds:
- Single Supply: Unlike the MESFET, it used a positive voltage (E-mode), so no negative rail was needed.
- Size: It had much higher power density than LDMOS, allowing for smaller PA modules.
- Efficiency: It worked beautifully at 3.2V, perfectly matching the shift to Lithium-ion batteries.
While multiple incremental improvements have further improved GaAs HBT performance and pushed up its maximum frequency, it seems to be struggling to extend much beyond 6 to 8 GHz, depending on the manufacturing process. New frequency bands considered for 6G (also known as FR3 bands, which include multiple bands between 6.425 and 15.35 GHz) seem out of reach for GaAs HBTs.
However, this is merely the tree hiding the forest. For many applications requiring a few watts of power above 6 to 7 GHz, D-mode gallium nitride on silicon carbide (GaN-on-SiC) is the only viable option. While GaN-on-SiC is an excellent technology for delivering tens to hundreds of watts from UHF to Ka-band, its high cost and supply chain constraints are prohibitive for low-power applications (a few watts), such as Satcom, CPE, and fixed wireless access (FWA).
This challenge has been partially circumvented by the rise of active electronically scanned arrays (AESA) and massive MIMO in the telecommunications sector. These active antennas use tens to hundreds of antenna patches to achieve the required equivalent isotropic radiated power (EIRP). Since each antenna is fed by a lower-power PA than omnidirectional antennas, technologies such as silicon germanium (SiGe), GaAs pHEMT, CMOS, and silicon-on-insulator (SOI) can effectively be utilized. While this approach is delivering reasonable performance, it comes with an incredible system complexity (size, number of components, system calibration and usage). A more reasonable and optimal approach would be to use fewer antenna patches fed by higher power, more cost-effective solutions than GaN-SiC PAs. This is where E-mode gallium nitride on silicon (GaN-on-Si) RFGaN-LV1 can have a play as well.
The Disruption: GaN Goes Mainstream
We are at a pivotal turning point in the semiconductor industry. Gallium nitride (GaN) is shedding its reputation as a niche, expensive material reserved for high-power base stations. It is now evolving into a high-volume, cost-effective integrated solution destined for the next era of smartphones. This transition marks the end of an era, effectively challenging the decades-long dominance of traditional GaAs HBT by offering a “final peace deal”: a hybrid high-performance III-V on silicon.
We are now seeing Finwave and GF’s co-developed GaN-on-Si technology on GF’s RFGaN-LV1 process technology doing to GaAs HBT what GaAs HBT once did to silicon bipolar: offer a more integrated, efficient, and CMOS-compatible way to power the next generation of devices to enable the next frontier with 6G and Satcom low earth orbit (LEO) constellations.
The New Frontier: Partnering to bring E-mode GaN technology at commercial scale with GaN-on-Si technology
In the semiconductor world, it is incredibly rare to witness the birth of a truly disruptive technology that shifts the fundamental “substrate” of the industry. We are currently seeing exactly that with the emergence of Finwave and GlobalFoundries’ (GF) co-developed 5V E-Mode GaN-on-Si technology, built on GF’s RFGaN platform.
This represents a radical departure from the status quo in two ways:
- E-Mode 5V Operation: Unlike traditional GaN-on-SiC which is “normally-on” (depletion-mode), Finwave and GF’s E-mode technology with a MISHEMT topology is “normally-off” and functions at a sub-5V supply. This allows it to be powered directly by the battery in mobile devices—a feat previously reserved for GaAs HBT.
- The Scale of silicon: By utilizing silicon rather than expensive SiC or GaAs substrates, this technology can be manufactured on standard 200mm silicon wafers, with far more advanced process technology and tooling than traditional III-V fabs. This brings the cost-efficiency and massive scale of CMOS foundries to high-performance RF.
Built on over 10 years of R&D led by prominent MIT researchers, Finwave’s proprietary GaN-on-Si technology makes the impossible, possible:
- Etch-stop technology: Enables enhancement-mode operation with excellent RF performance.
- Low contact resistance: Facilitates envelope-tracking operation below 5V for handset applications.
- MISHEMT technology: Provides superior channel scalability for mmWave frequencies and beyond.
Initiated in the summer of 2024, the technology transfer from Finwave to GF achieved a major milestone at the end of May by passing the manufacturing milestone 4 (M4) gate. This phase successfully demonstrated device performance and established the process controls required to accept partner prototypes. The next major milestones will be the M5 gate for technology qualification, followed by the M6 gate for mass production readiness. With the successful passage of the M4 gate, Finwave has begun working on its first power amplifier and MMIC designs and intends to collaborate with selected RFFE partners on advanced power amplifiers for next-generation handsets.
Redefining RF Performance
The capabilities of the RFGaN-LV1 E-mode process shown below are undeniable. While GaAs HBTs struggle to maintain performance as they push beyond 7 GHz, the RFGaN-LV1 process excels.
Key performance benchmarks include:
- Exceptional power density: Consistently delivering in excess of 0.8 W/mm.
- Superior efficiency: Achieving a power added efficiency (PAE) of over 65%.
The first generation (Gen1) of the technology serves as the initial stake in the ground, with a path to subsequent generations—promising a steady roadmap of incremental improvements in performance, cost reduction, and deeper system integration.
The Road Ahead
The shift to this novel GaN-on-Si technology does more than just boost technical specifications; it fundamentally rewrites the architecture of mobile connectivity. This is only the beginning. With Fmax well above 140 GHz, this process can deliver performance up to 40 GHz, supporting countless existing and new applications requiring a few watts of power.
The smartphone market is no longer the only high-volume driver in the industry. The rapid emergence of high-speed, low-latency LEO satellite constellations alongside FWA, is creating a massive new market for customer premise equipment (CPE).
While a projected 10 million in CPE shipments expected by May 2026, which may seem modest compared to the ~1.1 billion smartphones sold annually, the hardware complexity tells a different story. A smartphone typically contains a handful of antennas. In contrast, a single LEO CPE is a sophisticated active electronically scanned array (AESA). A single can feature up to 1,280 antenna patches, each requiring its own dedicated RF front-end (RFFE).
Similar math applies to FR2 FWA CPEs and 5G base stations, which utilize anywhere from 256 to over 1,000 antenna patches to maximize the equivalent isotropic radiated power (EIRP).
The RF semiconductor industry is at an inflection point. Solutions that grant the ability to innovate cost-effectively at scale are imperative to enable this. GaN-on-Si is paving the way for the next wave of RF innovation, transitioning GaN from niche applications into high-volume commercialization for high-power and consumer RF applications.
About Finwave Semiconductor
Finwave Semiconductor is a fabless semiconductor company, headquartered in Waltham (MA, USA) that uses innovative transistor designs and breakthrough process technologies to unlock the full potential of Gallium Nitride (GaN). Founded by prominent MIT innovators and driven by industry leaders, Finwave is shaping the future of RF Communications through revolutionary advancements in energy efficiency and performance for applications in markets such as Aerospace and Defense, mobile infrastructure, smartphones, medical devices, and cloud computing.
About GF
GlobalFoundries (GF) is a leading manufacturer of essential semiconductors, enabling AI at scale from the cloud to the physical world. Through deep partnerships with customers, GF delivers differentiated, power‑efficient and high‑performance solutions for automotive, aerospace and defense, data center, smart mobile devices, internet of things and other high‑growth markets. With global manufacturing operations across the U.S., Europe and Asia, GF is a trusted and holistic technology partner for customers around the world. GF’s talented, global team remains focused every day on security, longevity and sustainability. For more information, visit www.gf.com.
