Edge AI Isn’t the Future – It’s Here. These Companies Are Leading It
AI Semiconductor Race
The edge AI semiconductor market is being transformed in 2024–25 through its rapid growth from a disruptive niche to a mainstream technology. The growth as a result of the increasing deployment of AI inference directly on devices such as industrial sensors, smart cameras, autonomous vehicles, and smartphones has unlocked demands further downstream. The rationale of being able to process data at the edge is fairly simple: reduce latency, increase privacy, and normalise the strain on data centres. Development is happening at a furious pace; Intel detailed its new Core Ultra processors designed for edge AI in January at CES 2025 (Lindsay, 2025), IBM released Power11 chips for inference workloads ai in business usages, memory-chip companies like Micron are tackling the “memory wall” that limits the AI-architecture (AI on-device) capabilitie, and governments are not remaining passive. South Korea’s National Artificial Intelligence Committee pledged 9.4 trillion won by 2027 to support the domestic AI-semiconductor ecosystem; China’s US 8.2 billion National AI Industry Investment Fund seeks to accelerate chip innovation; and India’s 2024–25 budget doubled its semiconductor and display allocation to ₹6,903 crore, approved three new fabs in February, and extended R&D programs for chip startups. These breakthroughs in technology, government funding, and the larger move to decentralise AI workloads are moving the edge AI semiconductor market altogether from an interesting niche to a salient component of the AI computing practice.
Here are the top names powering the edge AI semiconductor boom:
1. Nvidia
NVIDIA disclosed quarterly earnings for Q1 FY2026 of $44.1 billion in revenue, growing 69 % year-over-year, from $26.1 billion in the same quarter last year, and dominating the AI semiconductor sector. Despite restrictions by the U.S. that prevented it from selling H20 chips to China, effectively costing it about $4 billion this quarter, it performed better than expected, plus provided a prediction of $45 billion in Q2 revenue.
On the edge engine front, NVIDIA is not letting up. It revealed at GTC 2025 its GB10 Grace‑Blackwell SiP coming to AI workstations, providing 1 PFLOPS FP4 performance in a small package, perfect for edge inferencing starting July 22, 2025. Also, while CES 2025 introduced developments that included the RTX 50-series chips and NIM microservices aimed at providing the capability for foundation models to run locally on RTX AI PCs.
2. Intel Corporation
Intel had revenue of $53.1 billion in revenue in 2024, with both the Network & Edge Group and the Client Computing Group moving toward edge AI semis. The company embraced edge AI, launching both its modular Open Edge Platform and Edge AI Suites bundles of software and hardware oriented toward retail, industrial, and smart cities, announced at MWC 2024. In Q1 2025, Intel launched the Tiber Edge Platform, with the Geti toolkit for computer‑vision model training at the edge. Also, from a report by Reuters, new CEO Lip‑Bu Tan is advancing a homegrown strategy to outpace Nvidia by focusing on edge AI devices and systems instead of purchasing many startups.
3. Google (Alphabet Inc.)
Alphabet is advancing the edge AI semiconductor space with its custom AI chips. In 2024, Google introduced Trillium, its sixth-generation TPU, which is optimised for on-device inference to improve energy efficiency, and with 4.7× more compute compared to the last TPU v5e and 67% improvement in efficiency. During Google I/O 2024, Google introduced Gemini Nano, tailored for mobile and edge devices, along with Trillium TPUs in preview via Google Cloud services. At the most recent Google Cloud Next 2025, Alphabet introduced Ironwood, its current seventh-generation TPU with an impressive delivery of up to 3,600× performance and 29× better energy efficiency versus its original TPU. When combined with intelligent models built for edge, Google is firmly established in the edge AI semiconductor space with plans to strengthen its position as a dominant provider of edge inference.
4. AMD (Advanced Micro Devices)
AMD is making a strong foray into the edge AI semiconductor segment with its flexible, power-efficient adaptive SoCs and embedded platforms. First quarter of 2025, revenue was $7.4 billion, gross margin was 50%, operating income was $806 million, net income was $709 million, with a growing embedded and edge AI revenue represented in its admission.
On February 6, 2024, AMD unveiled its Embedded+ architecture, which combines the first Ryzen Embedded CPUs, and Versal adaptive SoCs on the same board, bringing together the compute-rich power and capabilities of adaptive SoCs to develop new ways to simplify sensor fusion and establish a low-latency AI inference platform for industrial, medical, and automotive projects.
A few months later, on April 9, 2024, AMD introduced the Versal AI Edge Series Gen 2 second-generation adaptive SoCs with next-generation AI Engines sporting up to 3× TOPS-per-watt, and with integrated Arm CPUs for true end-to-end edge AI acceleration.
5. Qualcomm Technologies, Inc.
Qualcomm is strongly established in the edge AI semiconductor market, providing best-in-class AI acceleration across mobile, IoT, automotive, and enterprise devices. In 2025, Qualcomm made known their Edge AI Box, a plug-and-play combination of AI inference accelerators and 5G connectivity for smart cities, surveillance, and smart factory use cases. At Embedded World 2025, Qualcomm launched developer kits featuring Edge Impulse and RB3 Gen2, providing access to over 170,000 developers to proof and prototype AI models on microcontrollers and edge processors. In March 2025, Qualcomm announced a partnership with Palantir that combined their real-time data analytics with Qualcomm’s edge AI platforms for industrial and manufacturing use cases. These media releases, from Qualcomm’s newsroom, share a clear embrace of empowering AI at the edge of the network, making Qualcomm a leader in edge AI semiconductor derivatives.
6. Arm Holdings
Arm is still a foundational vendor in the edge AI semiconductor market, providing processor IP, AI accelerators, and development tools. Arm launched its first Armv9 edge AI platform in February 2025, the Cortex‑A320 CPU plus Ethos‑U85 NPU, which was capable of running on-device AI models with one billion parameters, targeting IoT and smart city use cases. In October 2024, Arm announced ExecuTorch, a PyTorch framework on its compute platform, which would allow efficient deployment of quantised Llama 3.2 AI models onto mobile and edge devices. Arm’s year-in-review report A-to-Z 2024, released in November 2024, further emphasised Arm’s advances in edge AI, including new Ethos accelerators, as well as the KleidiAI performance library for developers. All these announcements, from Arm’s newsrooms, reaffirmed a clear embrace of empowering AI at the edge of the network.
7. Graphcore
Graphcore is certainly disrupting the edge AI semiconductor market by simplifying the deployment of AI workloads in proximity to the AI data’s point of origin. The UK-based company was founded in 2016 and designs state-of-the-art Intelligence Processing Units (IPUs), along with a software stack called Poplar that provides APIs for AI workloads. In November 2024, Graphcore initiated its first recruitment drive since the acquisition by SoftBank in July 2024, with 75 new positions in silicon, systems and software, in increased capacity to develop next-generation AI compute platforms, which was reported on “Graphcore’s Blog”. SoftBank’s acquisition reflects confidence in Graphcore’s USA-developed IPU technology, which is seeing increased adoption in edge computing to help deploy large AI models out of traditional data centres. Although Graphcore does not manufacture its chips, and partners with foundries to design integrated chips for deployment in clients’ edge systems, which helps consolidate Graphcore’s points of engagement in the AI semiconductor market.
8. MediaTek
MediaTek is a powerhouse in the edge AI semiconductor market, providing AI-optimised SOCs for smartphones, IoT, automotive, and more. At MWC February in 2025, MediaTek launched Hybrid Computing device-cloud and RAN capabilities for low-latency Gen-AI at the edge. At Computex May in 2025, MediaTek’s CEO announced the company’s first 2 nm chip and collaborative efforts with NVIDIA to produce the GB10 Grace‑Blackwell Superchip, which includes merging MediaTek’s ASIC knowledge with the AI fabric of NVIDIA. And these are not just demos: MediaTek also announced in March 2025 the Genio 720 and 520 IoT platforms, which support generative AI workloads within smart environments. These are official releases demonstrating MediaTek’s vertically integrated approach.
9. Synopsys
Synopsys is an important behind-the-scenes player in the edge AI semiconductor market for its EDA tools and IP. Synopsys noted on June 19, 2025, in announcing a deep collaboration with Samsung Foundry, to successfully tape‑out HBM3-based customer designs with advanced sub‑2 nm technology nodes, incorporating its AI‑driven flows and 3DIC Compiler to accelerate development and improve power, performance, and area. Synopsys also achieved first-pass silicon success in developing its IP stack with TSMC’s 2 nm N2 process in late April 2025, establishing low‑power AI chips for high-efficiency edge mobile devices. Synopsys also collaborated with SiMa.ai in late 2024 to improve its SoCs for automotive edge AI and showcased its work at CES 2025.
These developments, advanced process support, high-efficiency IP, and ecosystem alignments—position Synopsys as an enabler in the edge AI semiconductor market, despite not building its silicon.
10. Huawei Technologies Co., Ltd.
Huawei is still a powerful player in the edge AI semiconductor market as it continues to deliver in-house AI rockstar silicon and state-of-the-art edge inference systems. In April 2025, Huawei started mass shipping its Ascend 910C, a dual-chiplet SoC (likely in response to Nvidia’s H100) with ~60% of inference performance, built on SMIC’s industry-leading 7 nm N+2 process. That same month, Huawei launched CloudMatrix 384, a supernode with 384 Ascend 910C NPUs connected through ultra‑high-bandwidth fabric, developed to provide high-powered edge and data centre AI. Beyond these flagship chips, Huawei’s Ascend 310—a 16 TOPS AI inference SoC has been deployed in real-world healthcare
Conclusion
The edge AI semiconductor market is no longer emerging; it is exploding. From global leaders in chip design like NVIDIA, Intel, and Qualcomm to smaller, specialised innovators like Graphcore and Arm, these companies are not only developing innovative chip designs, but they are also re-inventing where and how AI happens. As AI moves further beyond the cloud and into devices, factories, vehicles, and cities, the demand will (and should only) increase for silicon that is faster, smaller and smarter at the edge. We are already seeing not just real investment and government support, but also partnerships. It is clear that edge AI is the new normal, and these are organisations building the silicon that will make it a reality.
