What is an ASIC chip? The dual engine of encryption Mining and the AI wave.

ASIC (Application-Specific Integrated Circuit) is a chip designed for specific tasks rather than general-purpose computing. By hardwiring hardware logic, it can achieve efficiency and performance far beyond that of general-purpose processors when executing specific tasks.

Its application in the field of cryptocurrency mining is well-known, such as Bitcoin mining machines. However, in recent years, with the wave of artificial intelligence sweeping the globe, ASIC has also shown great potential in scenarios such as AI inference and edge computing.

What is ASIC? Definition and Core Features

The core of ASIC lies in "specificity." Unlike general-purpose processors such as CPU and GPU, ASIC is hardware-optimized for specific algorithms, directly embedding the logic of specific algorithms into the hardware.

This design philosophy allows it to achieve extreme performance and energy efficiency in target tasks. For example, the energy efficiency of Google's TPU v5e can reach 3 times that of NVIDIA's H100, and AWS Trainium 2 achieves a 30%-40% improvement in cost-effectiveness for inference tasks.

However, the "specialized" nature of ASICs is also their biggest limitation. Once the algorithm changes or the application scenario shifts, ASICs may not be able to adapt, leading to poor flexibility. This also makes ASICs more suitable for scenarios where the algorithm is stable and large-scale deployment is required.

Application-Specific Integrated Circuit in Cryptocurrency Mining: Application and Profit Analysis

In the cryptocurrency field, ASIC miners are the absolute main force in the mining of digital currencies like Bitcoin. The ongoing BTC bull market in 2025 has allowed miners to return to a profitable mode.

Currently, the price of hash power at 1 petahash (PH/s) is about $60.49, an increase of 13.7% compared to a month ago. Miner earnings have also risen accordingly.

The following is an overview of the profitability of mainstream ASIC miners currently on the market (based on an electricity cost of $0.06 per kWh):

Miner Model | Hash Rate | Power Consumption | Daily Net Profit --- | --- | --- | --- Antminer S21e XP Hydro 3U | 860 TH/s | 11180 W | 35.66 USD Teraflux AH3880 | 600 TH/s | 8700 W | approximately 26 dollars Antminer S21 XP+ Hydro | 500 TH/s | 5500 W | 22.17 USD Avalon A1566HA 2U | 480 TH/s | 8064 W | 17.27 USD

Even small mining machines have found a space to survive in the current market. The Nerdminer NerdQaxe++ still manages to net $0.18 a day with only 76 watts of power consumption and a hash rate of 4.8 TH/s.

The once legendary Antminer S9 series is now in a state of loss under standard electricity prices, but at a low electricity rate of $0.02 per kilowatt-hour, these old devices can still remain profitable.

Market Trends and Policy Impact

The U.S. tariff policy on ASIC imports has triggered a shock in the industry. According to analysts, the tariffs could increase miners' costs of importing ASICs by 10% to 50%.

Some industry insiders are even concerned that the prices of new imported ASIC miners could surge by 5 to 10 times. This could slow the relative growth of Bitcoin hash rate in the United States, prompting miners to turn to the second-hand market or efficiency improvement strategies.

Mining companies are also actively adapting to this change. Bitdeer announced that its SEAL03 Bitcoin mining ASIC chip testing has reached an energy efficiency level of 9.7J/TH, meeting the 10J/TH target, with plans for mass production in the second quarter of 2025.

At the same time, Hut 8 has partnered with Eric Trump to launch a new Bitcoin mining project called "American Bitcoin", aimed at becoming one of the largest and most efficient pure Bitcoin mining companies in the world.

Growth Opportunities in the AI Inference Market

In addition to cryptocurrency mining, ASIC has also shown strong growth momentum in the AI inference market. According to VerifiedMarketResearch data, the global AI inference chip market size has reached 15.8 billion dollars in 2023.

The market is expected to grow to 90.6 billion dollars by 2030, with a compound annual growth rate of 22.6%. The penetration rate of Application-Specific Integrated Circuits in inference scenarios is continuously increasing.

Tech giants are actively developing their own ASIC chips: Google's TPU series has been iterated and upgraded, AWS has launched the Trainium series of chips, Microsoft is planning the Maia series of products, and OpenAI is also developing its first ASIC chip.

These developments indicate that ASICs are becoming key hardware for accelerating artificial intelligence applications, especially in large-scale inference tasks.

ASIC Industry Chain from an Investment Perspective

The ASIC industry chain covers multiple aspects including design, manufacturing, packaging testing, and application.

In the design and server side, companies such as Cambrian, Chipone, and Allwinner Technology are worth paying attention to. Chipone, as a global leading ASIC design service provider, ranks second in the number of IPs globally, and its customized mass production service shipment has exceeded 100 million units.

At the manufacturing and packaging end, companies such as Changjiang Electronics Technology, Tongfu Microelectronics, and Northern Huachuang provide key support for ASIC mass production. Tongfu Microelectronics specializes in Chiplet packaging technology and offers full-process services from wafer level to high-end packaging.

At the application and material end, Orbbec has self-developed efficient and low-power Application-Specific Integrated Circuit (ASIC) chips, focusing on edge computing and the Internet of Things; Ruiskanda has also self-developed ASIC chips for communication products, creating technical barriers in specific communication scenarios.

Risks and Challenges

There are also many risks in the ASIC field. The risk of technological iteration is paramount, as the customized nature of Application-Specific Integrated Circuits makes them highly dependent on the scenarios. If downstream demand changes too quickly, it may lead to a lag in chip iteration.

The cost and risks of mass production cannot be ignored. The initial design costs are high, and if the mass production scale does not meet expectations, it may affect profitability.

As major players accelerate their in-house development, third-party vendors face the risk of market share squeeze. Industry competition is intensifying, and technological barriers are continuously rising.

In addition, policy changes such as adjustments to import tariffs can significantly impact the cost and supply of Application-Specific Integrated Circuits (ASIC) miners.

Future Outlook

Bitcoin mining is just the tip of the iceberg for ASIC applications. As the AI inference market approaches a scale of $90.6 billion by 2030, the customization advantages of ASICs will also be demonstrated in more fields.

The success of chips like Google TPU and AWS Trainium has already proven the tremendous potential of ASICs in specific computational tasks. In the future, as technology continues to specialize, dedicated chips such as TPU, DPU, NPU, and even LPU will collectively drive the next generation of computing revolution.

ASIC may not be the most flexible solution, but it is becoming an indispensable cornerstone in specific scenarios that pursue efficiency and performance.