The top applications of FPGAs are signal processing, wireless and telecom infrastructure, networking, data centre acceleration (including AI inference), cryptography, aerospace and defence, automotive, industrial control, medical imaging, and test and measurement. What these have in common is a need for high, predictable throughput and low latency, combined with volumes too low or requirements too changeable to justify a custom ASIC. Below are the ten most important areas, followed by eight more where FPGAs play a real but narrower role.
FPGA stands for Field Programmable Gate Array, a type of integrated circuit that can be configured by the user after manufacturing. FPGAs offer many advantages over traditional microprocessors for suitable tasks, such as high performance, low power consumption, flexibility and parallelism. In this post we explore some of the most common and interesting applications of FPGAs in various domains.
Key takeaways
- FPGAs win where data streams in continuously and must be processed in fixed time: radio, radar, video, packets, sensors.
- They are chosen over ASICs when volumes are modest, standards change, or time to market matters.
- They are chosen over CPUs and GPUs when latency must be deterministic or the data path is unusual.
- In consumer products FPGAs are mostly small helper chips; the main processors are ASICs or SoCs.
Applications at a glance
| # | Application | Why an FPGA | Example functions |
|---|---|---|---|
| 1 | Signal processing | Parallel DSP blocks, streaming | Filters, FFTs, beamforming |
| 2 | Wireless and telecom | Standards evolve, high data rates | 5G radio units, baseband, fronthaul |
| 3 | Networking | Line-rate packet handling | SmartNICs, filtering, low-latency trading |
| 4 | Data centres | Offload from CPUs | Compression, storage, security, search |
| 5 | Artificial intelligence | Custom precision, low latency | Inference at edge and in servers |
| 6 | Cryptography | Fast bit-level operations | AES, SHA, public-key engines |
| 7 | Aerospace and defence | Long life, low volume, radiation-tolerant parts | Radar, EW, satellite payloads |
| 8 | Automotive | Sensor fusion, fast iteration | ADAS sensor processing, camera bridging |
| 9 | Industrial and robotics | Deterministic real-time control | Motor control, machine vision, fieldbus |
| 10 | Medical, broadcast, test and measurement | High-quality streaming data | Ultrasound, video switching, oscilloscopes |
The top 10 applications
1. Signal processing
FPGAs are widely used for signal processing tasks, such as digital filtering, image processing, audio processing, video processing and radar processing. They can handle complex algorithms and large data sets with high speed and accuracy, making them ideal for real-time applications that require low latency and high throughput. As an illustrative example, a 32-tap FIR filter fully unrolled across 32 DSP slices at 300 MHz produces 300 million filtered samples per second, one per clock, with a fixed latency of a few tens of cycles.
2. Wireless, radio and telecommunications
FPGAs are used in wireless and radio applications for baseband processing, modulation and demodulation, and error correction. They are also used in high-speed telecom networking equipment, such as switches and routers. Cellular radio units and early deployments of new standards often use FPGAs, because the design can be updated as the standard and the operator's needs settle, before moving to an ASIC at high volume.
3. Networking
FPGAs are used for networking applications such as switching, routing, firewalling, load balancing and packet processing. They can process network traffic at high bandwidth and low latency, supporting protocols and standards such as Ethernet, TCP/IP, UDP/IP and the processing behind HTTP, SSL/TLS and VPN traffic. FPGAs can also adapt to changing network conditions and requirements by reconfiguring their logic. Low-latency financial trading is a well-known niche, where the fixed, sub-microsecond path through an FPGA matters.
4. Data centres
FPGAs are used for high-speed data processing and networking in data centre applications. They can accelerate workloads such as machine learning, networking and security, and improve the performance, efficiency and security of data centre servers where CPUs alone fall short. Public examples include Microsoft's Project Catapult in Bing and Azure, and Amazon EC2 F1 instances. More detail is in What is an FPGA accelerator.
5. Artificial intelligence
FPGAs are increasingly used for artificial intelligence applications, such as machine learning, deep learning, computer vision and natural language processing. They can accelerate neural networks such as CNNs, RNNs and GANs by exploiting their parallelism and reconfigurability, and they support many data types and precision levels, such as floating-point, fixed-point, low-bit integer and binary. In practice their main strength is inference; training of large models is dominated by GPUs and dedicated accelerators, and FPGA training is mostly a research and niche use. See FPGA based machine learning inference.
6. Cryptography
FPGAs are used for cryptography, the science of securing information and communication. They can implement encryption, hashing and signature algorithms such as AES, RSA, SHA and ECC efficiently, and can also perform key generation, authentication and random number generation. Because the algorithm sits in hardware, an FPGA can be updated when a cipher changes, which is relevant as post-quantum algorithms are adopted. See FPGA based cryptographic acceleration.
7. Aerospace and defence
FPGAs are used for aerospace applications such as satellite communication, navigation, control and telemetry, and are widely used in defence applications such as radar, electronic warfare, communications and cryptography. They can perform complex calculations with high accuracy and precision. For defence systems they offer high performance and low latency, flexibility and adaptability, resistance to radiation and harsh environments (in suitably qualified parts), and security and anti-tamper features. FPGAs in space can also be reconfigured after launch, which is valuable when a satellite cannot be reached for repair.
A correction on radiation: ordinary commercial FPGAs are not radiation hardened. Space and some defence systems use radiation-tolerant or radiation-hardened FPGA families, and designers add mitigation such as configuration scrubbing and triple modular redundancy, because radiation can flip bits in the configuration memory of SRAM-based FPGAs. Flash and antifuse based FPGAs are less sensitive to this effect.
8. Automotive
FPGAs are used for automotive applications such as advanced driver assistance systems (ADAS), autonomous driving systems (ADS), infotainment systems (IVI) and vehicle-to-everything communication (V2X). They can process sensor data from cameras, radars, lidars, ultrasonics and GPS with high speed and low latency, and can implement safety features and functions. Automotive-grade FPGAs are qualified to AEC-Q100, and high-volume production vehicles often move these functions into dedicated SoCs once the design is stable.
9. Industrial, robotics and automation
FPGAs are used for industrial applications such as automation systems (PLC, SCADA), robotics (robot arms, drones), measurement systems (oscilloscopes, multimeters) and control systems (motor, servo). They can control and monitor machines with high performance and reliability, and support interfaces such as SPI, I2C and industrial Ethernet.
In robotics and automation they provide real-time control and decision-making, with high performance, low power, flexibility and reconfigurability. FPGAs implement sensor processing, control algorithms, communication protocols, machine vision and AI, and are often combined with processors, memory, sensors, actuators and cameras; many devices integrate processor cores with FPGA fabric on one chip (SoC FPGAs). They are used in industrial, medical, military and educational robotics, where fast, reliable responses in complex, changing environments are needed.
10. Medical, broadcast and test and measurement
Medical: FPGAs are used in imaging systems (MRI, CT, ultrasound) and diagnostic systems (ECG, EEG) to process biomedical signals and images with high resolution and quality, and can help ensure patient safety and privacy. The original version of this post also listed therapeutic systems (pacemakers, defibrillators) and wearables (heart rate and blood pressure monitors); implanted devices such as pacemakers more commonly use ultra-low-power custom chips or microcontrollers, so FPGA use there is limited.
Broadcast and video: FPGAs are used in video processing and encoding for HD and UHD video. They offer high performance, low latency and flexibility for encoding, decoding, transcoding, switching, mixing and filtering, and support many video standards, formats, resolutions and frame rates, so products can follow changing market demands.
Test and measurement: FPGAs are used in test and measurement equipment, such as oscilloscopes and logic analysers, for real-time signal processing, triggering and data acquisition.
Eight more applications
Blockchain
FPGAs have been used for blockchain applications, which are distributed systems that store and verify transactions using cryptography. They can perform hashing and mining operations at high speed and with lower power than CPUs or GPUs. For Bitcoin, FPGA mining was a short phase around 2011 to 2013 before purpose-built mining ASICs took over; FPGAs have since been used mainly for newer or changing algorithms. The original claim that FPGAs support consensus protocols and smart contracts is kept here, but this is an uncommon use.
Gaming and retro computing
The original post listed console games (PlayStation, Xbox), arcade games (Pac-Man, Space Invaders), retro games (NES, SNES) and virtual reality games (Oculus Rift, HTC Vive). A correction: current consoles and VR headsets render graphics with custom ASIC SoCs, not FPGAs. Where FPGAs do shine in gaming is in recreating classic hardware, where the FPGA emulates the original console or arcade board at the circuit level, and in accessories, capture cards and development hardware.
Smart grid and energy
FPGAs are used in smart grid and energy systems, which need high reliability, security, efficiency and adaptability, for example in smart metering, power quality monitoring, renewable energy integration, microgrid control and demand response.
Power electronics
Fast control loops for inverters and motor drives, where switching decisions are needed in microseconds.
ASIC prototyping and emulation
Large FPGA systems run a chip design before silicon so that software can be developed early.
Scientific instruments
Particle physics trigger systems and radio telescopes use FPGAs to filter huge data rates in real time.
Storage
Compression, encryption and computational storage close to flash drives.
Education
Low-cost boards are the standard way to teach digital design and computer architecture.
FAQ
What is the most common use of FPGAs?
By value, communications (wired and wireless), industrial, aerospace and defence, and data centre markets account for most FPGA demand, as reported by vendors; exact shares vary by source and year.
Why not just use an ASIC?
An ASIC costs far more to develop and takes much longer, and cannot be changed after manufacture. FPGAs make sense at lower volumes or while requirements are still moving.
Are FPGAs used in phones?
Small, low-power FPGAs are used in some phones and wearables for sensor and display bridging, but the main processor is always an ASIC SoC.
For why FPGAs matter more generally, see Significance of FPGAs in the new world.
Your comments will be moderated before it appears here.