CAN Answers to Cryptonomist – Interview

Canaan was one of the earliest ASIC pioneers. Today, as the hardware market becomes more commoditized and you face competitors such as Bitmain and MicroBT, how do you defend your competitive edge?

Everyone can see the competitive landscape of this industry. History is important, of course, but at the end of the day customers look at the product: whether the machine can make money steadily in real environments, whether it creates fewer problems, and whether it can keep running for a long time.

Canaan has gone through many cycles in this industry. When the market is good, the industry is very exciting. When the market is bad, it can be very harsh. The reason we have survived all these years and are still launching new products is actually simple: we have kept our attention on the product itself. A mining machine is not a PowerPoint product, and it is not a product that can be judged only by launch-event specifications. It has to run in mining sites, in homes, and under many different power and thermal conditions for a long period of time. In the end, customers judge you in a very practical way.

From this perspective, I have confidence in the Avalon product line. One important reason is that I personally keep a close eye on the products. I am deeply involved in product definition, and even in some specific design discussions. Whether it is our large machines for industrial mining farms or consumer products such as Avalon Home, there are many third-party reviews and real user comments available. For example, bitcoin mining influencer Red Panda Mining recently tested the A16. I had not heard of him, but his comments about the machine were very positive. Feedback like that is more valuable to me than any marketing statement, because it comes from real usage and from people who truly understand mining machines.

There are two major new directions Canaan has been focusing on since last halving in 2024. First, we’ve almost single-handedly nurtured a consumer-level market for home mining. Home mining is a direction I personally care about a lot. While it is relatively new, it carries deeper significance tied to Bitcoin’s foundational architecture. When I started my work on FPGA bitcoin mining solutions back in 2011, nearly all bitcoin was mined at home by individuals. Bitcoin mining should not exist only in large industrial mining sites. Quieter, safer, and easier-to-deploy home mining machines allow more individual users to participate in the network again. To me, this is directly connected to the spirit of Bitcoin’s decentralization. Internally, we often say that decentralization cannot remain just a slogan. In the end, real products have to lower the barrier so that everyone can participate.

Another direction is the integration of bitcoin mining with energy systems. Bitcoin mining has a distinctive characteristic: it operates as a highly flexible and interruptible computing load. This characteristic allows it to participate in grid balancing, absorb stranded energy when needed, give back as a curtailed energy source, reuse waste heat, and be very cost-efficient in some heating cases. Our work on greenhouse heat reuse, wellhead energy utilization, district heating, flexible load participation, and joint mining is essentially exploring the same idea: mining machines can go beyond being pure hashrate devices and become part of energy systems and compute infrastructure. 

In a market that is becoming more commoditized, our advantage must come from continuing to make better products and going deeper into real application scenarios. Scale matters in this industry, but product stability, real-world efficiency, service capability, scenario understanding, and long-term engineering accumulation matter as well. What cannot be ignored about Canaan is that we have engineering capability in industrial mining machines, while also having made significant investments in home mining products and energy-related scenarios. These directions may not always be the loudest in the short term, but I believe they will define the real differentiation in the next stage of the industry.

---

2. With Bitcoin mining becomes more institutionalized, do you see ASIC manufacturing evolving into a lower-margin, scale-driven business, or is there still room for meaningful innovation?

If we look at the traditional IT equipment industry, it is very common to see hardware margins gradually decline as the market becomes more standardized, customers become more concentrated, and supply chains mature. But the Bitcoin mining machine industry has its own special characteristics. The valuation logic behind a mining machine is straightforward and intuitive. Fundamentally, the value of a mining machine tracks the discounted value of its expected future mining profits. When customers buy a machine, they calculate how many coins it may mine, what the electricity cost will be, and how long the payback period will be. This economic model is simpler and more transparent than many IT products.

What really makes margins in this industry fluctuate sharply is often the movement of hashprice. Bitcoin price, total network hashrate, block rewards, transaction fees, electricity costs, and financing conditions can all change customer demand and machine pricing within a short period of time. So, I would not simply say that once the industry becomes more institutionalized, ASIC manufacturing will necessarily become a “low-margin, scale-driven business.” Scale is certainly important, and institutionalization will make customers more professional and more rational, but, ultimately, margins are still closely tied to the customer’s expectation of future returns.

I also want to correct one common misunderstanding: lower margins do not mean there is no room for innovation. In many cases, when margins are too high, the industry may actually have less motivation to innovate. In the past, we’ve seen that when ASIC manufactures, especially family-owned businesses, were busy booking chip inventory, shipping machines and making profits, they were more incentivized to over-extend themselves to catch the bull cycle instead of focusing on innovative products and services. But when margins become lower, the pressure becomes more real, customers demand cost-savings, and manufacturers have to seriously solve problems that were easy to ignore in the bull cycles. From this perspective, a lower-margin environment may force more innovation.

There are many places where this industry can innovate. Chip efficiency is still important, of course, but innovation at the system level is becoming more important: hash board design, thermal structure, power efficiency, firmware and tuning, reliability, serviceability, noise control, transportation, and ease of deployment. A mining machine runs in the real world, not only in a lab. What customers really care about is whether the machine can operate stably for a long time, whether problems can be repaired quickly, and whether the total cost of ownership of the whole system can be reduced.

The standardization of the industry itself will also create new areas for innovation. Large institutional customers will require clearer delivery processes, more stable quality systems, better after-sales service, more predictable power curves, and system solutions that fit different scenarios. Immersion cooling, high-density deployment, modular hash boards, and customized systems for different energy environments are not solved just by designing a chip. Our recent work with partners such as Tether and Bitfury on customized hash board modules is an example of exactly what happens when customers need more than a standard mining machine. They need computing modules that can be embedded into their own infrastructure.

Beyond bitcoin mining, the industry is moving in a new, significant direction: the integration of mining with energy systems. In the future, mining machines may not only be sold to traditional mining operators. Increasingly, miners may be utilized in more energy scenarios, such as waste heat reuse, wellhead energy, renewable curtailment, grid demand response, and even sharing part of the power and infrastructure with HPC and AI data centers. Many engineering problems need to be solved, and the business models are still evolving, but, in this sense, the room for innovation may be even larger than before.

Finally, we are enabling more community innovation by opening up our chip technology that’s been our key competitive advantage. With customers like Bitfury and more recently Tether, we have licensed our technology and helped these customers create innovative solutions to fit in their respective mining environment. This openness enables the customers to tailor-make their mining solutions with different form factors, swappable/recycle-able hash boards, higher power density, etc., which not only lowers their costs, but showcases our awareness of environmental sustainability. At the home level, we open-source software and give more access to our chips, so that individual developers and SMEs can either use our home miners to fit into their smart building solutions, or they can use our chips to create new products such as wall heaters or the heat source for hot tubs. The result is that people can offset the cost of heating their homes using a bitcoin mining heater that replaces traditional gas or electric heat sources like house boilers and heat pumps.

My view is that ASIC manufacturing will become more professional, and scale and operating efficiency will become more important. The industry will not lose innovation because of that. As the market moves from rough growth into a more refined stage, real innovation will expand from individual chip parameters to machines, systems, energy, services, and business models. For Canaan, this is exactly where we plan to keep investing.

---

3. Following the 2024 halving, many miners have faced tighter margins. How has this shifted demand for your machines, particularly in terms of efficiency versus upfront cost?

While seemingly true for the 2024 halving, a halving does not necessarily mean miners’ margins become tighter. If we look back at Bitcoin’s history, every halving has basically coincided with the industry entering its next stage of development. Over the long term, the scale of the industry has continued to expand, and miners’ overall returns have not simply declined. The real change is that the industry now demands more from its participants.

After the 2024 halving, our original view followed this logic. Of course, this cycle has also experienced some variables that were less prominent in the past, especially the introduction of bitcoin to a broader US monetary system, and geopolitics, tariffs, supply chains, and regulatory conditions. These variables created many unplanned fluctuations for the industry. But if we look at the result, in 2025, we saw bitcoin reach new all-time highs multiple times, and we also saw a large increase in network hashrate. The industry did not stop. It continued to develop at a higher speed.

I do not think the core change in this cycle is that miners are forced to make a simple choice between efficiency and upfront cost. The real change is that customers now have more comprehensive requirements for machines. In the past, many people assumed that a mining machine would run for a period of time, recover its cost quickly, and then be replaced in the next equipment cycle. In this cycle, more people realized that a good machine is not something that becomes obsolete after one year. If design, manufacturing, and maintenance are done well, it can continue to generate returns over a longer period.

This creates a very practical new requirement: customers want both excellent efficiency and a very durable service life. Efficiency determines daily competitiveness. Reliability and lifetime determine the real return over the full lifecycle. Miners now look more carefully at stability in real environments, thermal performance, power systems, ease of maintenance, and long-term uptime, rather than only looking at the purchase price.

For us, this is actually a change that fits Canaan well. Avalon has always placed strong emphasis on product stability and engineering reliability. As the industry matures and customers become more professional, this long-term accumulation becomes more valuable. After the halving, customers are not simply asking for cheaper machines, and they are not only looking for the best efficiency number on paper. They care more about whether a machine can make money steadily over its full lifecycle. This is now one of the most important directions in our product design and customer communication.

---

4. To what extent do you see mining centralization as a risk to the long-term health of the Bitcoin network, and what role can manufacturers like Canaan play in preserving decentralization?

Frankly speaking, I used to worry a lot about this question. The security of the Bitcoin network ultimately comes from a hashrate base that is sufficiently distributed and sufficiently open. If hashrate is controlled by a small number of people, institutions, or regions, of course that creates long-term risk.

But later I gradually felt that physical hashrate centralization is not as easy as people imagine. The reason is simple: electricity resources themselves are distributed. Mining needs a large amount of low-cost power, and low-cost power cannot be concentrated infinitely in one place. You can build large mining farms, but it is very hard to reduce all competitive power resources in the world into a single point. As the industry scale continues to grow, hashrate is actually forced to look for more regions and more types of energy.

Over the long term, the physical distribution of mining has a natural rebalancing force. If one region has mining-friendly policy, suitable electricity prices, and mature infrastructure, hashrate will go there. If tariffs, regulation, supply chains, or power conditions change, hashrate will move again. This movement has been very clear over the past two years. The United States once had very large-scale mining, then tariffs, machine flows, and supply-chain factors created changes, and part of the hashrate moved back toward East Asia. Now we are seeing some movement back to the United States again. Sometimes I feel this network has its own balancing mechanism.

What I worry about more now is concentration at the country level. If one country, or a small number of countries, gain too much influence over hashrate, mining pools, energy, and equipment supply, that is not good for Bitcoin’s long-term global nature and neutrality. Bitcoin should not become the network of one region. It should always remain a global network.

This is also why I care so much about home mining. In a sense, home mining is bitcoin mining returning to its roots. In the early days, many people participated in the network from their homes, offices, or small-scale environments. Later, industrialization was a natural outcome. But I do not want individual participation to disappear completely.

I have a relatively clear personal goal: I hope home mining can eventually account for around 20% of global hashrate. I do not think this is unrealistic. The heating demand of homes and small commercial spaces around the world is very large. If only a very small percentage of users participate, it could create a meaningful amount of hashrate. More importantly, this hashrate would be naturally distributed across different countries, cities, and households. That is very valuable for the decentralization of the Bitcoin network.

Of course, we cannot ask ordinary people to participate only because of ideals. The product must truly fit the home environment. That is why our Avalon Home product line is not just a smaller industrial mining machine. We have done a lot of work on appearance, noise, safety, user experience, thermal management, and household functions. It has to fit into ordinary life. Users should not have to tolerate a noisy, ugly, hard-to-maintain machine just because they want to mine.

I think there is also a bigger question here: can the heat generated by continuous computing serve ordinary people’s lives? Heat from electronic products has always been a challenge in the home environment. Computers, servers, mining machines, and future AI devices all generate heat. Our idea is to look at this problem from the other side. Since continuous computing will inevitably generate heat, can we design it as part of home heating, space heating, or other daily-life scenarios? Mining is a beginning. Future AI computing may face the same issue.

If this can be done, it is not just a new hardware market. It may bring both economic and social value. Users get a device that can generate income and serve household needs at the same time. The network gets more distributed hashrate. The energy system gains another way to connect computing with heat reuse. This is why I think home mining is very much worth long-term investment.

What Canaan can do is very concrete: continue to provide equipment suitable for participants of different scales, especially allowing home users and small-scale users to participate; lower the barriers in noise, deployment, safety, and maintenance through product design; and keep exploring the connection between mining machines, energy, heat reuse, and grid balancing. Decentralization cannot remain solely an idea. It has to be implemented through real products and real usage scenarios.

---

5. What are the next major breakthroughs you expect in ASIC design? Are we nearing physical limits in chip efficiency, or is there still room for disruptive gains?

If you had asked me this question a few years ago, I would have said very directly that mining chips were still far from physical limits. In fact, over the past few years, even after people started saying that chip efficiency was approaching its limit, the power efficiency, or PE, metric of mining machines, meaning energy consumption per terahash, still improved very significantly.

But if you ask me today, my answer would be more cautious. Progress will certainly continue, and there are still many things engineering can do, but the pace will be slower than before. One reason is that improvements in advanced processes themselves are slowing down. Another very real reason is that a large amount of global advanced semiconductor capacity is now being absorbed by AI demand. The rapid development of AI is a good thing, but it also raises the cost of wafers, packaging, and overall manufacturing resources.

This changes the economics for the mining machine industry. Reducing PE further is still technically possible, but it is no longer a pure engineering question. The economic model of mining machines is very simple, as we discussed earlier. In essence, manufacturers share a portion of customers expected future mining profits. Better product performance can certainly create value, but the manufacturer cannot take all of the incremental value. If wafer costs rise too quickly, part of the economic benefit from efficiency improvement will be offset.

Future breakthroughs may not only be reflected in how much the PE of the next chip improves. Chip efficiency remains important, but we will care more about system-level returns: machine reliability, long service life, power efficiency, cooling methods, deployment density, ease of maintenance, and the total return under different energy environments. Customers do not look at one parameter in isolation. They look at whether a machine can make money steadily over its full lifecycle.

I do not think progress will stop. This industry will definitely keep moving forward, and we can already see the results of continuous progress. For example, we have launched the A16XP model with 300T and 12.8 J/T. This shows that even when process improvement is slowing and manufacturing costs are rising, mining machine efficiency and product capability are still improving. It is just that the next stage of innovation will be more three-dimensional. In the past, people mainly looked at chips and efficiency. In the future, people will look more at the overall optimization of chips, machines, energy, operations, and application scenarios. For Canaan, this is actually a direction we are willing to invest in, because it is closer to real products and closer to customers’ true long-term returns.

---

6. Canaan has also explored AI chips in the past. How do you evaluate the trade-off between focusing on mining hardware and diversifying into AI or edge computing?

I think the question of whether we should move toward AI no longer needs much debate. AI will certainly change the whole computing industry. It will also change energy and compute infrastructure, and even human society. The real question is: what is the right way for us to get there?

It feels a little like you already know where the destination is, but the area between here and there is very wide, very complex, and quite chaotic. We cannot simply say that, because AI is the distant goal, we should immediately build an AI chip or some computing product today and call it completed. For Canaan, we need to find a path that connects with our existing capabilities, customer base, energy resources, supply chain, and engineering experience.

I personally have one view: bitcoin mining is actually a preview of the future AI token business. Bitcoin mining has a very special characteristic. Whether you mine a bitcoin in the United States, Asia, the Middle East, or somewhere else, its value is basically the same. Hashrate, energy, machine efficiency, and operational capability are eventually priced by one global market.

The AI industry today is not yet like this. Tokens from different models, different scenarios, and different service providers still have different value and quality. But if in the future many models become “good enough,” and the buyers of compute and tokens are no longer only human beings with limited time and attention, but agents with clear objectives that can automatically ask for prices and execute tasks, I think global AI token prices will tend to become more even and more unified. At that point, the AI token business and today’s mining business may be very similar at the underlying level. Both will create a competition for energy, chips, system efficiency, and global settlement capability.

I sometimes joke that when we get to that stage, these agents may very likely settle with cryptocurrency. For machine-to-machine global, small-value, high-frequency, automated settlement, the traditional financial system may not be the most natural tool. This may not happen immediately, but I think the direction is worth serious thought.

Our preparation today has two layers. The first layer is energy and infrastructure. Whether it is bitcoin mining or future AI computing, the essence is the same: both need low-cost, stable, schedulable energy and large-scale deployment capability. Our current investment in energy infrastructure, flexible load, grid coordination, and heat reuse is preparation for a broader computing business in the future.

The second layer is hardware itself. I personally keep thinking about what a future “AI token mining machine” should look like. It will not simply be the same as today’s data center GPU, and it will not simply be the same as today’s bitcoin mining machine. It may need to find a new balance among cost, efficiency, serviceability, deployment flexibility, model adaptation, and network settlement. Many of these questions have not yet been fully defined, and that is exactly where I see opportunity.

Therefore, I do not see focusing on mining hardware and transitioning toward AI as two mutually exclusive choices. Mining has given us very important training: how to design high-efficiency chips, how to deliver hardware at scale, how to build a business model around energy cost, and how to deploy and operate computing equipment globally. If the future AI token business really moves toward globalization, automation, and commoditization, these experiences will become very valuable.

What Canaan needs to do now is to maintain the competitiveness of our current mining machine business, while gradually connecting energy infrastructure, ASIC design, system engineering, and future AI computing. This process will not happen overnight, but I believe it is a very important direction for the company.

---

7. Given ongoing geopolitical tensions and export controls, how resilient is your supply chain today, particularly regarding advanced semiconductor manufacturing?

Over the past one to two years, we have made many adjustments, and these adjustments have already started to show results.

Let me start with manufacturing. Because a significant portion of our customer base is in the U.S. market, full-machine manufacturing in mainland China is gradually being phased down. This shift is not just a slogan. It is the real rebuilding of capacity, processes, quality systems, and delivery systems. We started doing this relatively early, so during this round of geopolitical and tariff volatility, although we also faced pressure, we were not completely caught off guard.

Our manufacturing layout in Malaysia has been a very successful step. The Malaysian factory is now supporting a large amount of manufacturing, and operating quality has been relatively stable. This shift is important to us because it allows us to serve global customers and also provides better flexibility when supply-chain and trade conditions change. For a hardware company, manufacturing migration is not finished by moving equipment to a new place. The real challenge is moving the people, processes, and know-how together, which includes the yield control, quality systems, delivery rhythm, on-site management, supplier coordination, and the front-line team’s understanding of product details. We invested early and deeply in this area. So the Malaysian factory is not a simple assembly replacement. It has already formed a relatively complete manufacturing capability.

U.S. domestic manufacturing has also been running for quite some time. Its scale is not as large as Malaysia for now, but it is no longer a trial facility. It is real production and real delivery. On a quarterly basis, our U.S. domestic manufacturing has already reached the level of several thousand machines. Having two production facilities certainly has cost challenges, but, from the perspective of customer service, delivery certainty, and supply-chain resilience, it has strategic value.

If we only look at full machine manufacturing, I think Canaan’s layout is relatively early and relatively solid. We did not wait until problems arose and then look for temporary substitutes. Today, we have manufacturing capability outside mainland China, and those capabilities are now supporting customer deliveries.

For wafers and packaging and testing, our principle is simpler: we fully follow compliance requirements. Where licenses are needed, we apply for them as required. Where we need to confirm processes with foundries, packaging and testing partners, and suppliers, we work closely with them. Advanced semiconductor manufacturing is not something one company can complete alone. It depends on long-term, stable partnerships, and it must respect the compliance requirements of every market and every partner.

Sometimes when we talk about this issue, it feels as if we have been fighting supply-chain battles for many years. In reality, many of these changes happened over just the past year or two. The external environment of this industry changes quickly. Tariffs, export controls, customer locations, manufacturing locations, and logistics routes can all affect delivery. What we can do is act early where possible, remain compliant, communicate clearly with key partners, and build a more distributed, resilient manufacturing system.

No hardware company can say it is completely unaffected by geopolitics. But from Canaan’s perspective, we now have large-scale manufacturing capability in Malaysia, continuously operating manufacturing capability in the United States, and stable relationships with foundry, packaging and testing partners. These are the real foundations of supply-chain resilience.

---

8. Do you foresee a regional fragmentation of mining hardware markets, where different jurisdictions rely on distinct supply ecosystems?

If you take this question to an extreme, it sounds a little like a movie scene: equipment moving through different regions, different markets with different rules, and different buyers using different paths. As a public company, Canaan is very strict on compliance. We do not pursue gray-market opportunities and have no interest in unethical profits.

I would approach this question in two parts. Regionalization is certainly happening Tariffs, regulation, customer location, logistics cost, after-sales service, and certification requirements will all cause different regions to form different delivery and service systems. U.S. customers, for example, will care more about domestic manufacturing, delivery certainty, and compliant routes. Other markets will also have their own power conditions, policy environments, and deployment methods. This kind of regionalization is real.

But it is difficult for the global mining hardware supply chain to become fully fragmented. Advanced chip design, wafer manufacturing, packaging and testing, key components, and quality systems cannot be independently replicated in every region. A legal, stable, and sustainable supply system still requires global cooperation and long-term accumulation. Especially for a company like ours that designs ASICs and builds full machines, the truly important thing is to build good products, manufacturing, quality, compliance, and customer service systems, rather than chasing short-term regional arbitrage.

Neither my personal focus nor the company’s long-term direction lies in the fragmentation narrative. Instead, we are focused on several strategic, longer-term priorities: improving the core product; enabling broader participation in the network through home mining; integrating mining with grid balancing, heat reuse, and ESG applications; and evolving Canaan from a mining hardware manufacturer into a broader computing infrastructure company, spanning both blockchain and future AI workloads.

Therefore, we will respect regional differences, strictly follow compliance requirements, and build manufacturing and service systems according to the needs of different markets. But strategically, we do not define the company as one that is trying to capture opportunities in one regional market. We want to build a long-term, transparent, and compliant computing infrastructure business. To me, that is more important.

---

9. Canaan has expanded into self-mining. How do you balance potential conflicts of interest between selling machines to customers and competing with them directly?

On the surface, we sell mining machines and also participate in mining ourselves, so it may look like we are competing with customers. But if you look deeper, what customers really need is hashrate, or exposure to future mining returns. They do not necessarily need to go through the whole process of site selection, construction, deployment, and operating of a mining farm.

Building and operating a mining farm is very heavy work. It involves power resources, land, grid connection, transformers, cooling, network, operations teams, regulatory compliance, machine repair, and cash-flow management. If any of the links have a problem, the final outcome could be affected. I have seen financial institutions fall into difficult situations after being forced to take over mining farm assets. The issue was not a lack of capital or a misunderstanding of Bitcoin’s long-term value. Rather, they discovered that what they had acquired was not a simple financial asset, but a highly complex energy and operating system. The United States has a deeply specialized business environment, where, in theory, such functions would be supported by mature service providers. However, at that stage, the industry had not yet fully developed that kind of infrastructure. As a result, what many customers really need is a lighter, more transparent, and more professional way to participate in hashrate, rather than taking on the full complexity of building and operating mining farms by themselves.

From this perspective, our development of self-owned hashrate and joint mining is not taking business away from customers. It is actually expanding the market that customers can participate in. We organize hardware, deployment, operations, and energy resources, so customers can participate in hashrate returns in a lighter way. This is a different business form from simply selling mining machines, and it can serve different types of customers.

I believe this direction will continue to evolve. Today, many customers buy mining machines or participate in hosting and joint mining. As a next step, customers may buy hashrate directly. Further down the road, hashrate may become a more standardized product, and financialized tools could be developed around hashrate returns within a compliant framework. This logic is not unfamiliar in traditional commodity and energy markets. In the bitcoin mining industry, it still needs time, scale, credit systems, and regulatory frameworks to mature.

So, when we look at self-mining, it is not just about mining more coins for ourselves. It is part of our preparation to move from a mining machine manufacturer toward a hashrate and computing infrastructure service provider. Only by truly participating in operations can we understand more deeply the pain points our customers face in power, operations, machine stability, cash flow, and risk management. And, in the end, these experiences will help us make better products.

Of course, there must be boundaries. Canaan’s foundation remains technology and products. We do not want to become a mining farm operator that only competes with customers for resources. We would rather play the role of hardware provider, hashrate organizer, and long-term partner. If customers want to buy machines, we will continue to sell good machines. If customers want hashrate exposure but do not want to carry the full pressure of construction and operations, we can also provide them with a more suitable way to participate.

I do not think self-owned hashrate and customer business necessarily conflict. The key is how the business model is designed. If designed well, it is not a zero-sum relationship. It connects mining machines, energy, operations, hashrate, and customer capital, and allows more customers to participate in this market in different ways.

---

10. Looking ahead three to five years, do you envision Canaan remaining primarily a hardware company, or evolving into a broader infrastructure player within the Bitcoin ecosystem?

Canaan’s goal is clear: we want to move from being a company that mainly sells mining equipment toward one that is a computing infrastructure service provider. If we break this direction down, it is an extension of what we are already doing today.

The first layer is still chips and hardware. ASIC design, full-machine engineering, supply chain, manufacturing, and delivery capability remain Canaan’s foundation. Whether the future is bitcoin mining, AI computing, or other blockchain-related computing, it will still require hardware systems that are highly efficient, scalable, and able to run stably for a long time. This is our core foundation.

The second layer is energy. In the past, when people looked at mining machines, they mostly looked at a machine’s hashrate and efficiency. But I increasingly feel that what really determines the economics of large-scale computing is the combination of chips, machines, and energy. Our current work in self-owned hashrate, joint mining, energy projects, grid flexible load, and heat reuse is essentially connecting computing with energy infrastructure. Future AI computing will face the same problem: as compute grows larger, energy will become one of the most important constraints.

The third layer is hashrate service. What many customers really need is exposure to hashrate returns, or in the future, broader computing capability. They do not necessarily need to buy machines themselves, build mining farms, hire operations teams, and handle power and compliance issues Because Canaan will gradually move from simply selling equipment to organizing hashrate, delivering hashrate, and operating hashrate. Mining machine sales remain important, but they will become part of a larger business system.

Home scenarios are also an important part of this direction. Today it is home mining. In the future it may be home AI or other forms of distributed computing. My view has been consistent: continuous computing inevitably generates heat, and the real question is whether we can design that heat to serve ordinary people’s lives. Avalon Home is an initial step. It connects mining, heating, home use, and decentralization. As AI computing moves into homes and smaller spaces, this model can naturally extend.

If you ask whether Canaan will remain a hardware company over the next three to five years, my answer is yes: hardware will continue to be our foundation. But we intend to go beyond the scope of a traditional hardware business. Our goal is to build a more integrated computing infrastructure platform, spanning chips, devices, energy systems, and hashrate services.

This is not a theoretical vision. We have established manufacturing layouts in Malaysia and the United States. We are investing in energy assets and self-owned hashrate. We are exploring new user scenarios through home products. At the same time, we are considering what kind of computing devices will be needed in the emerging AI-token era. Over the next few years, our focus is to connect these capabilities into a coherent system. Canaan will not remain solely a Mining hardware vendor. Our ambition is to become an infrastructure service provider that links energy, chips, and hashrate.