Returning to its acquisition-driven growth playbook, Microchip has agreed to acquire Israeli edge AI chip startup Hailo. This is the first acquisition to be announced since longtime Microchip CEO Steve Sanghi began his second stint at the helm last year. The terms of the transaction were not disclosed.
Tel Aviv-based Hailo was founded in 2017, making it one of the most mature edge AI chip startups with around 100 current customers, primarily in the industrial and automotive markets. Hailo’s first publicly disclosed customer is HP, which supplies Hailo-based accelerator cards for its point-of-sale systems. The company had raised $344 million in funding.
The company’s first-generation product, the Hailo-8, launched in 2019 and offered 26 TOPS with notable power efficiency. The company’s NPU mixes compute and SRAM on chip, with a clever compiler that analyses AI models to assign compute and SRAM in a way that minimizes data movement. The Hailo-15, launched in 2023, is an SoC with an NPU, CPU, and DSP designed for IP camera systems, and the Hailo-10, launched in 2025, is designed for generative AI in the device.
Notably, Hailo partnered with Raspberry Pi in recent years, making its M.2 accelerator cards available on the hobbyist market. The startup’s developer community has lately reached 10,000; 80% of these are working with Raspberry Pi hardware, Hailo CEO Orr Danon told EE Times in a previous interview.
View All Microchip has acquired a dozen or so companies in the last 20 years, most notably analog chip company Micrel in 2015, microcontroller maker Atmel in 2016, and Microsemi in 2018. Two AI software companies were also acquired recently: VectorBlox with its AI software stack for FPGAs in 2019, and AI model optimization startup Neuronix in April 2024.
The portfolio fit for some of these companies seemed questionable at the time, and Hailo is no different. Microchip’s strategy is characterized by acquisition-driven growth rather than obvious technology synergy, and it will acquire companies offering practically any embedded technology to enable cross-selling opportunities. Cross-selling appears to be the limit of its ambitions; Microchip does not seem to want to verticalize or move up the value chain by offering systems or solutions, a common strategy among some of its competitors.
Microchip’s key selling point is that it keeps its product lines running for decades to meet the needs of its customers in high-rel applications like military, aerospace, automotive, and industrial, where design-ins are considerably stickier than in consumer markets. An embedded system in a factory or an aircraft might be installed for 10 or 20 years, or longer in some cases.
Microchip mentioned Neuronix in its announcement, the AI model optimization company it acquired in 2024. At the time of acquisition, Neuronix’ team was put to work shrinking models primarily for the former Microsemi PolarFire FPGA line, which is aimed at high-rel markets. Neuronix is at the upper layers of the AI stack and is likely relatively hardware agnostic (all hardware likes smaller models), so synergies could apply with Hailo’s software and hardware stacks. There does not appear to be any immediate synergy with VectorBlox, which is an AI software stack for FPGAs, or with MPLAB, Microchip’s integrated development environment for its microcontrollers.
What Microchip saw in Hailo was likely some combination of a mature software stack with a big user base, albeit mainly hobbyists, and multiple generations of proven hardware with 100 customers in industrial and automotive sectors, which are likely already on Microchips’s books for other parts of their embedded systems.
Acquiring Hailo makes sense if considering an AI accelerator/NPU was a missing piece of Microchip’s portfolio rather than assuming any synergistic motive. It seems likely, given its track record, that Microchip will maintain Hailo’s hardware product lines as separate offerings rather than try to develop new SoCs using Hailo’s IP in combination with other IP from its portfolio, but that option nevertheless remains on the table.
Edge AI is very much a growth market, unlike some of the long lifecycle markets Microchip currently targets, and AI technology moves very fast, much faster than other areas of embedded tech. Hailo will need continued investment in hardware and software, especially software, to succeed. Whether Microchip can or will provide the conditions needed for Hailo to flourish, particularly the investment in engineering and R&D, remains to be seen.
Another open question is: following NXP’s acquisition of Hailo competitor Kinara 18 months ago, can the embedded market support independent edge AI chip companies or are we faced with a musical chairs situation where the likes of Axelera, Sima, and MemryX are matched with companies like Infineon, Renesas, STMicro, and TI? Who will be left without a seat when the music stops?
The Microchip-Hailo transaction is still subject to regulatory approval but is expected to close by the end of Q3.
Read also:
[Warning Shots Fired as AMD Announces New Data Center GPUs](https://www.eetimes.com/warning-shots-fired-as-amd-announces-new-data-center-gpus-at-advancing-ai-event/)
[Etched Raises $300M with $1B in Pre-Orders](https://www.eetimes.com/etched-raises-300m-with-1b-in-pre-orders/)
[From Rhetoric to Metrics: Raghib Hussain’s First Year as Altera CEO](https://www.eetimes.com/from-rhetoric-to-metrics-raghib-hussain-first-year-as-altera-ceo/)