In this episode, NG Zhang, founder and CEO of Canaan Inc., shares his journey in Bitcoin mining technology, from early ASIC development to modern innovations like heat reuse and energy integration. We talk about the evolution of mining hardware, industry trends, and future prospects.
Timestamps:
(00:00) – Intro and Early Days of Bitcoin Mining
(03:40) – From FPGA to ASIC Technology
(06:22) – The Avalon A16 Series
(09:07) – Reliability and Durability of Mining Machines
(12:15) – Failure Rates and Useful Life
(14:40) – Primary Market and Secondary Market for Mining
(16:40) – Small Scale vs. Large Scale Mining
(22:00) – What % is home mining today?
(23:00) – Heat Recovery
(29:00) – Different methods of cooling in Bitcoin Mining
(31:40) – Mining machine form factor
(32:40) – Stratum V2 Support?
(37:35) – Market Dynamics: Public vs. Private Miners
(39:50) – Energy Grid Integration and Bitcoin Mining
(44:25) – Efficiency gains are slowing down in bitcoin mining
Links:
Stephan Livera links:
- Follow me on X: @stephanlivera
- Subscribe to the podcast
- Subscribe to Substack
Transcript:
Stephan (00:00.631)
Hi everyone, welcome back to Stephan Livera podcast. Today we’ve got a special one,
My guest today rarely does podcast interviews. NG Zhang is the founder, chairman, and CEO of Kanan Inc., the company behind the world’s first Bitcoin ASIC mining machines, the Avalon series, shipped in twenty thirteen. NG Zhang has a background in electronic engineering and software from Behang University. NG, thank you for joining me today. And let’s you know, I think it would be great to start with a bit of a background on yourself and tell us what what was it like in the early days of Bitcoin ASIC machines?
NG Zhang (00:37.134)
very happy to join the the interview. Yeah. I think I can I come into Bitcoin business. I think at at the very beginning it’s a hobby, it’s not it’s a bit not a business. it’s start like twenty twenty eleven. I read some a very short news about some someone invited a cryptocurrency and very little
technical informations but it really attracted me from that time and maybe I spent about one week to set up my own family mining farm at the at the time. Yeah. And then I I think the the community at that time is is it’s really very very very good it’s fantastic. and
Back to that time I sell some small electronic products on the internet. on the internet. So you know back to like ten ten ten or something years it’s really hard to sell sell goods over the over the seas. So we use we use bitcoins to do training at that time.
So at first time not a minor, I’m a I’m a user of the crop currency, very very, very early stage. and then I think I can do something for the community at the time. So I think at late twenty twelve I decided to start the the Avalon project. it’s project to
try to design manufacture ASIC based Bitcoin miner. So I think in twenty late September we started project and after four and a half months our first generation of Bitcoin mining machine because ever one ship to our first customer. I I can’t remember the
NG Zhang (02:59.467)
Our first customer is Jeff Gardik, it’s a co developer at that time. I think it’s it’s really very very important because it’s you know, when I get the first box of the ASIC chips, the computing power is is equal to about half of the t entire Bincoin network at the time. So it’s really unbelievable.
Yeah. And you can you can do solo mining use only one machine and tour three blocks a day. Yeah. Then that’s that’s a early early time, yeah. So
Stephan (03:38.644)
Yeah, and as I understand, at that time it was transitioning from, you know, there was this transition in the early days it was, you know, CPU mining, then it was GPU mining, and then it was FPGA, and then after FPGA was the ASIC machine. So do you want to just talk to us a little bit about the insight there of transitioning from FPGA to an ASIC?
NG Zhang (03:57.47)
sure. I think early days we use I think very in the very early days we we we use CPU mining. and then we we have the CPU, GPU miner, the CG miner. Then you can use GPU instead of CPU, it’s tens of the acceleration. and because I study in col college so I I have the
scale to build FPGA boards for for mining. I think the FPGA mining is lower the power consumption, maybe 5x or 10x at that time, but the the performance is the same and also have a higher cost. So but you know if if algorithm is very easy to be implemented in FPGA that then it means it will be perfect
a real ASIC. So I think the F FPG period is a really short. It’s about maybe one year or two years at most. then we start to convert to to the ASIC technology. I think the very at the very beginning is the it’s all about the time to market. You know, from an idea then you need to assemble your team to build an ASIC.
and it’s only weeks left for for you to do that. at the same time you need to find every electronic components, the cases, heat seen, everything, and find a factory to to help you to manufacture them or you need to handle it by yourself. Yeah yeah. I I think very significant percentage of the machines is assembly by by ourselves. So yeah. And
ship them to your your customer. you have no channels, you you have nothing knowledge about how to trade overseas. You know, y I’m I’m only a student at that time. So every time I remember that that maybe three three months it’s it’s really you know it’s extremely busy. only three or two four hours for for a day to sleep, but it’s really the most interesting time in my entire life.
Stephan (06:27.191)
So let’s bring it forward now to the modern era of A6. I know you have the Kanan Avalon A sixteen series. So can you talk us through a little bit on this A sixteen series, how it compares to prior generations, and maybe just comment a little bit on the joules per tera hash efficiency.
NG Zhang (06:49.985)
yeah for for A16 is our latest model. I think a simple number is we have a 300 terahash. first model breaks through the 300 milestone at twelve point eight jars per terahash. And we also have the cost effective version of H16, it’s 282 and about 13.8 jars.
I think this is a very major improvement for us, especially compared with the previous generation. but I I think I but I always tell tell customers not to look only at a lab number. miners do not just buy a number for for on a slide and i i you buy it fleet ec economics so that means stable heart rate uptime thermal
this kind of things is is really important. if like ninety-nine percent uptime vers versus like seventy or eighty percent uptime will change the the game fundamentally. So we I think from the eight thirteen, fourteen and fifteen we continue to upgrade the the machine to have a best quality.
because this in these days the machines will last for a much much longer time than the previous stage. So yeah. So this is w what we designed the the machines. and if you you know it we have we we already have some test in in the world. So you can check the the videos from the YouTube. I think the very
if if if if only can think about one, most attracting me is the the noise, the sound is much, much lower than the previous model. Yeah. We we can do interview. even even even it’s next to a machine is running at like a economic model. Yeah. It’s it’s under new.
Stephan (09:02.285)
I see
Stephan (09:11.213)
So while you’re there, I mean I guess we’re talking about the reliability and the durability aspects. can you speak to how Kanan’s Avalon machines stack up against competitors such as you know, bitmains and miners or microBTs what’s miners? Can you speak to how Kanan’s Avalon machines are from a reliability perspective?
NG Zhang (09:35.182)
yeah. you know I would be very careful about comparing like failures with competitors, unless we use the same methodology, otherwise it becomes the the marketing is not engineering, you know you know I’m an engineer guy. So to be honest, real ability is no i it’s there’s no magic. It comes from from the design, the materials.
how many testing you you you have done, manufacturing, even the software, firmware, and you you need to do after sales feedback to to deal with the after sales. So in the in the has to be honest I think in the past the industry just moved too fast. So sometimes time to market was measured in hours or days, not weeks or months.
So a new generation came out, you need to ship it in like twenty-four hours, without any testing because your customer asks you to do that. You know. It’s every day is it’s so much money. So and all but luckily the machines were replaced quickly. So many companies include us r relied heavily on experience and stronger
components. there are ma many overredients and lack of design in the in the real machine. So that works in a fast cycle industry but sure it’s not enough for the current stage or next stage. So now I think the product cycle is becoming much longer so efficiency improvement is sl is slowing. This is is a real situation. at first
I have been in this industry for over 10 years, but first first time I feel the efficiency improvement is slowing. So the machines need to run for more years. that means the industry must improve design, validation, mental testing, burning. also we have water cool models, so you must test the water block reliability, also better firmware testing and codic.
NG Zhang (11:59.974)
control standards. So internally we we are pushing toward a very low annual repair rate this is our our goal so my engineering target is to around like low single digits or even less than 5%, 4%, 2%, this is our goal to an annual repair.
Stephan (12:20.321)
And that’s for the failure rates or that that under five percent number that you’re talking about. So just to summarize, I guess as you’re saying, in the early days of Bitcoin, there was such rapid efficiency improvements that the customers wanted your machines as quickly as possible. And I guess what the point you’re making is that there was maybe less focus on testing, burn-in, making sure that the like the uptimes and low failure rates, this kind of thing. Whereas now, as I understand, obviously the the efficiency gains are slowing down.
NG Zhang (12:23.487)
Right. yes.
Stephan (12:50.435)
Compared to the early years of Bitcoin and so the focus is on useful life for these machines. So NG, can you tell us what do you see as the useful life for an Avalon A sixteen nowadays?
NG Zhang (13:03.781)
it’s I think the the answer is really really really interesting. user life depends on I think heavily on the environment. if you put the machines in a good IDC or data center like environment, our modern miners is able to run like five years or more. It’s very, very easy. Yeah. But the reality is many real mining sites are dusty.
or or just in in the wild, you know, you can you can see through the you know the the the fans, you can see the outside. so this is a real environment. For those environments customer want you still need to have three years for the a stable operation. So this is really what is really challenging for for for us. Yeah.
Stephan (13:52.943)
I see. So as I read you then, if it is a challenging environment, then it might be as low as three years, but otherwise it might be more like five years is a typical useful life nowadays. and then I guess obviously that also will vary based on the the cost of electricity for that miner, because as I understand, the normal dynamic is that, you know, if you have
you know if your cost of electricity is higher, then you need a more you need a more efficient machine to make it work. Whereas like people who have like a very low cost of energy, maybe they can get away with using a less efficient machine because of the extreme low energy cost. Is that how you’ve seen it? And then that plays out into the let’s say the primary market for Bitcoin mining and then also the secondary market when it’s resold.
NG Zhang (14:29.133)
Mm-hmm.
NG Zhang (14:45.077)
Yeah, I think this there’s yeah, compare the electricity rates is is one side. the other side is the scale and the stability for your electricity. usually like we have flat electricity rate like the no common
actually cost in like Texas, like there’s maybe around like three, four, five cents. yeah, this will give you a very stable operation. if we move to like Africa, it may be like one or two cents. but you must deal with like 20 or 30 percent of like the power down for the from the grid because the infrastructure is not that good.
or even like other countries that that have unstable politic environments there’s they can like ban you from from anywhere at any time. So this is different. So if you have like we assume that you have very low electricity cost and it’s very stable then I think you should buy the you should buy new machines. This is my opinion.
and you should buy the cost effective models because your electricity is rate is low. but second hand machines or very legacy machines is not encouraged because you know the cheap electricity always have a smaller scale. So you must you want you surely want a total
you know, the the the T C O for the or the total revenue for from the the side. So very very no, you know the the the the bad machines will not give you a a very very good return. So by by new machines, but the cost effective model is is is is better, yeah.
Stephan (16:51.278)
See. now we were talking about small scale miners. I think it’s probably a good spot also just to ask about that. So there’s this you know, there are home miners and small scale miners, and maybe there’s discussion around things like heat recovery or heat reuse. can you tell us about any products you’ve done for this category for the home or small scale miner?
NG Zhang (17:12.873)
I think we before we talk about the home manner, it’s a very big topic. we start from the the big public manners versus small private manners. I think maybe it’s it’s not very you know not as public thought. large manners and small private manners have different advantages. This is my
large miners care about like the the fleet economics, financing, they use financing right. like big scale delivery, schedule, warranty, service, everything. and smaller miners usually care about flexibility, unit price, easy of deployment because they have maybe no IT team for the operation, only the
owner itself will help deal everything so that they need easier for deployment, also fast support and it’s and local power opportunities so both models will continue to to next Bitcoin money is very geo far geographic so very a very large size does not always have the or always have no
not always have the the lowest cost. sometimes the infrastructure grid connection will make the and also financing will make the cost higher than private small miner. a smaller one usually have access to local like surplus power or regional energy opportunities that a large public miner cannot use. So
I think this is a very health very healthy model, economic model for Bitcoin. that means it’s have natural decentralization for the deployment of the mining hardware. So yeah, we want industrial scale, but we also want decentralization. So we we we are very friendly for both large and small miners. Also the price for the large scale
NG Zhang (19:36.722)
or like two hundred machines, five hundred machines, the gap is really narrow in these days. So yeah, this is a s a real situation.
Stephan (19:46.422)
Yeah. And then as I understand you have product lines that are targeted for smaller scale miners. can you talk a little bit about some of those the different lines that you have? ‘Cause you have the Avalon A sixteen and then you have the some of these other product lines.
NG Zhang (20:04.845)
Yeah, you know we we built built my home miner and small scale miner from the very beginning of Evalon brand from maybe maybe Evalon 3 I think. Yeah. yeah so today we have reopened the the the home miner production line. we have several products in the category. Avalon Mini 3,
I I think I introduced from the the the bigger one from to the smaller one. we have our Avon Q is the biggest for home mining. it’s 90 terahash today and I think it’s going to upgrade in the next few months. And then it’s mini three and we have nano3s. and we also have a few new products, it’s coming out this year.
so the lineup is from higher performance to lower performance. but the lower the lower the hash rate it will be more friendly for the for for a home. like our ministry is it’s electronic heater style, it has very very good out outlooking. So why why we do the home
serious I I think the first answer is decentralization. bitcoin is not com complete of ordinary people can only hold coins but cannot participate in the mining layer. So my my long term I have a long term dream it’s it’s very clear. we I would like to see home money become meaningful part of the global hash rate, maybe like 20%.
for the total network heart rate. that that means we we need to sell two two million avalon q machines. So it’s a very tough work.
Stephan (22:07.118)
So actually, out of curiosity, what would you estimate home mining is today as a percentage of Bitcoin’s overall hash rate?
NG Zhang (22:16.269)
I think it’s still low. we are I I think the the the milestone is like we maybe like start from like two percent, no twenty percent. Yeah. I mean tax is not a very very very big scale in in this industry, right? So yeah, we I think the product itself is it’s it’s barely good enough for
Stephan (22:26.476)
Yeah. So you’re trying to ten X that that number. Yeah.
NG Zhang (22:45.261)
for for that target. but the marketing channels it’s different it’s different for from the like the our traditional to be business. So it’s really we need to learn learn something new. Yeah. So yeah.
Stephan (23:03.128)
I see, yeah. And I guess you’re you’re selling different features in a way because maybe i it with the whole heat recovery, heat reuse idea, it makes more sense in cold countries, to to to sell these home miners because obviously they can get a benefit from the heat reuse aspect, which maybe lowers the cost of mining for them. is that something you’re looking at? Is that how it works?
NG Zhang (23:13.261)
Yeah.
NG Zhang (23:30.188)
Yeah, both heat recovery, we we started this topic from the home miner. you know every watch used by a miner eventually becomes heat. so normal electron heater converts electricity into heat and a Bitcoin miner a bitcoin heater converts electricity into network security, Bitcoin reward and heat.
So so it’s a much more interesting use of an and energy. you know, to achieve that goal, a home miner is not just a smaller industry miner. it has to look good, has to be easy to use and has to be safe. yeah, it needs a very good materials and reliability. it c must last three or five years in your home.
at least so for me I think the noise requirement is very it’s very restrict. I test our home minor in my bedroom. I need to have entire sleep overnight when the minor manner is running next to my bed this is the the target.
also is this is a standard, it’s not targeted. We all we already achieved that. So this is the standard for home home product. and and also I want to see say say something further than that. I think we are entering an age of continuous personal computing. Bitcoin mining is the first obvious example.
You know, I think you are using every every every day, every everyone is using AI today, right? And also I think in the next stage most people will have their private cloud. it I think the safest place is to install the the workstation or the server in your your room, but sure it it must be very quiet, just like home minor today. So
NG Zhang (25:51.886)
I think we this this production line in in individuals home open new concept for the the computing device in for for for personal use. Yeah. And then I think we can talk about the bigger one. Yeah, the the bigger one. you know we
today we are changing the we are we are mo we are working together with the the ESG staff. yeah, so firstly it’s very the mining the minor running is very clean. it’s it will not product any waste for from its running. So
and also it’s very c low cost compared with many industry laws. So it can turn the electric waste electricity into a valuable dig digital asset and it’s very useful for for renewable energy. and we the heat recovery in the large scale is is very different to to home miner.
we need to pro produce hot water. the standard for the hot water is is really high. I think we we need to start like eighty I think number is eighty one degrees Celsius. so it’s it’s really high. I think most of the the home water heater can generate like eighty degrees Celsius. It cannot higher than that. But we need to produce eighty one.
and also and maybe eighty five at the most. So it requires many special design for the system and it it must last for three five years. So we I think we already spent like over one year working on on this. It’s a different industry field I think. So we already deployed many
NG Zhang (28:19.249)
I I think it it’s about like three three or five sites using the miners to generate hot water to service the the town, many yeah, the the in the in the real world. Yeah. It’s it and it’s all
Stephan (28:38.35)
I see. I see. So I guess what you’re getting at is it’s it’s not just a home miner thing. Actually even at larger minor scales, they are looking at heat reuse and in this case water heating as a way of maybe lowering the cost, or getting the benefit in another way so that you can make it work as an overall business model.
NG Zhang (29:03.085)
I think the ma the major target is to is to generate the heat. And then it’s Bitcoin mining. This is a different.
Stephan (29:10.454)
Yeah. So I guess different business models are out there. I guess the other aspect is i when it comes to bitcoin mining, people can talk about the different methods of cooling, right? Like air cooling, hydro, which seems to be also gaining in popularity, and immersion cooling. So can you just talk to the different methods of cooling and how you see that as, you know, from a canine perspective?
NG Zhang (29:32.698)
yeah, because there’s the the one cooling method that cannot win everywhere. So still today I think the air cooling is the the most universal because it’s simple, lower cost and easy to deploy and easy to maintain. it works for many types of customers, and our machines is optimized to running in the web.
So it it’s good. but for larger in industrial sites, especially in in like the in US or Europe, hydrocooling is becoming much more important. the direction is high density machines, because the hydrocyte is more expensive for every rank.
So often like our two U class machine can it’s already like ten kilowatts. So I think it’s very e it’s equal equal to the modern GPU servers. So you maybe you can just use use them before the the
expensive GPUs come comes comes in your IDC. just use the the the the miners to to adjust consumption the the electricity and generate heat. It’s it’s it’s really good. Yeah. yeah so also I I think it’s it’s good for like if you have minus it next to some living area
the hydro system will generate lower noise and lower maintenance. It’s also a benefit for for that. also we have the immersion version for the machines. yeah. So we we have all the types. we we don’t make decisions for our customers. our customers can can decide.
Stephan (31:51.181)
Yeah. and all on the form factor, as I understand, it’s still most of the mining is done in this so-called shoebox form factor. there’s been some talk, now maybe it’s just talk, I’m curious to hear your react reaction. Are there people interested in this idea of like a server rack style form factor for Bitcoin mining?
NG Zhang (32:07.814)
yeah.
NG Zhang (32:14.815)
I think except some old sites, the new sites is all convert to two use style, yeah. Because it’s standard factor for the water cooled servers. They can reuse the infrastructure
For something else. Yeah. So today I think that maybe like ninety nine percent of the machines is already the two to is two U factor. Yeah. This year.
Stephan (32:48.396)
Yes, there’s a it’s a shift there. now another area or topic I was interested to ask you about is Stratum V two, which is you know a new protocol for Bitcoin mining. is Kenan looking at this, would this be would it make sense to have native firmware support for it on Avalon machines?
NG Zhang (33:09.517)
yeah, I think I I knew talk more about this topic. yes. I think the the short answer is it’s it’s yes. but there are two layers. you know, one layer is the first is secure and efficient mining communication. there’s no
encryption in the street and v1 and in v2 we have it’s we have encryption so it’s it’s safer. so the second layer I think is the is the is the the JD the drop declaration this is that really matters but this is a lockdown in the
Stephan (33:48.152)
Job declaration, yep.
NG Zhang (33:56.182)
Usually it’s not done in the minor from where it’s done in the like the bridge or or the the mining router or controller in the in the sides. for KN yeah, we you know we we are looking at this from an engine we are looking at this from an engineering perspective. because many existing avalon machines is using very concentrated
MCU controllers. It’s running RTOS. It’s only have like four four megabytes, no no gigabytes, megabytes of memory. so it’s for that generation the the even the the it c you can only run the modern in incorporated communication because of because there are no enough computer power. So it it’s the limitations from the hardware. this is why we
Reflecting the our first firmware architecture and also we are moving at the same time we are moving to a more capable associate platform, our K2 230 associate chip. we are improving the SDK and modernizing the control software, improving the everything, and we can have the capability to implement more algorithms like the community compute communication
encrypted for for the i in the in the in the firmware. So the street w2 support is it’s now on our to do internal to do list. but I want want to be clear the manual selected transaction templates should not mean very small minor controllers run and a full node. So a red architecture is a local gateway or something that as I mentioned. Yeah.
And yeah and you know we we do some we already published the the code for our new platform late last year. but I don’t think the it’s it’s it it’s I have very very good code quality so we are re re re rewrite everything, nearly rewrite everything this year. I think the next
NG Zhang (36:22.253)
three to six months, you can yeah keep an eye on our GitHub and we we are doing we will have a new firmware platform.
Stephan (36:36.192)
Interesting. And so are you saying then that the y you might look at Stroud and V two support in that three to six months time on the new firmware?
NG Zhang (36:43.541)
Yeah. in fact we we are moving to one hundred percent we internally I said a number of hundred about hundred percent open sourced firmware and it and also it’s a model design so we could easily install new feature support on our new firmware platform.
now it’s actually really hard because you know it’s not a modular design if you you modify some some some something and maybe the the can adjust cannot run it. So it’s a very legacy structure for the all our old platform. So we are trying to build a new one in the next three or six months. Yeah. Also the Swiss M2 support is on the total list, yeah.
Stephan (37:40.067)
Yeah, gotcha. okay, so that’s on the the firmware side and then the the the to do list. Okay, so I guess zooming out a little bit about the broader miner mining industry. I mean, we were touching on this a little bit around how there are these, you know, big public miners. Some of those have, let’s say, partially or fully pivoted to AI and HPC, and then you have these smaller and private miners. Can you talk to us a little bit about
Just the difference in how you’re going to the market there because I guess it must be different when you’re selling to these big hyperscalers versus selling to retail and smaller miners. How does this shift affect you from a sales marketing perspective?
NG Zhang (38:19.213)
yeah I I think for as I mentioned if we we assume we are talking about the 2B models. yeah this the large miners care about you know that mentioned like the financing delivery schedule everything warranty outside support yeah
the smaller miners is more close to like the home miners you need to do marketing or then we just don’t know what is able minor. Yeah yeah we we we all buy we the only thing I know is it’s speed mat. So this is this is the really challenging stuff. so we we need to do more marketing for this this group of customers.
And also they are care about the the cost and the flexibility, easy f deploym deployment, y if you have some software to manage of our my my twenty machines, this kind of stuff. Or and also have the local RMA side is is better.
for for these kind of customers because the larger one will have their own you will you always they will have over repair station outside but you cannot have that for for smaller one. Yeah. so this is different for for larger one the smaller one.
Stephan (40:05.014)
Now, when it comes to energy and grid integration, there’s been a lot of discussion about this also, especially in recent years, things like grid balancing, demand response, heat reuse at scale. I know you have been commenting or the company has been commenting a little bit on this. Can you expand on that, on how you are viewing this Bitcoin mining and energy grid integration story?
NG Zhang (40:31.777)
Yeah, I I think these years people discussing money is not only
as an pure energy consumer. money is also a very flexible energy commission tool today. So first it’s very clean to top relation. second it’s very useful for the new renewable energy. we have some POC sites installed
using battery and the wind power. mostly it can run like twenty two hundred twenty days per year. but because the wind and solar is very voluntary, so the power is not stable enough, we need to do special designs and write software codes for to combine the the the the the the wind
the power the battery and the machines so also we have some cooperation with with the college we are doing research pro products together to have the the best algorithm for for that so this is there’s no like
NG Zhang (41:58.294)
Yeah, i we need to do some some more research on this this field. but sure it can can mining can absorb surplus and improve the economics for renewable products. Yeah, so our work with simplist an example for direction, generation storage and the mining
Together with power scheduling, so a hat rate follows the energy conditions, run at different running levels. so this idea is not to compete with the the grid. The idea is to follow the the power and the the the use energy that otherwise may have very very low value.
This is whatever I’m doing. and we we just mentioned heat recovery is another important part. if mining produces hot water that can heat homes, greenhouses or industry processes, this is a real ESC use case. yeah, it’s it’s much better for then like you can
you can you you you the the to to let the the cap and the bottle together you know so so this is I I and I would add add one more thing that the home money is already the best heat reuse is also ESG so yeah ESG is you can you can do a very large scale project also you can do it in very tiny project. Yeah. So
Mm. Also looking forward, I think the AI AI data center is will create huge electricity demand. they also need redundancy, backup power you you pay for the further grade for the line for backup. so and also they need load management. I think mining can work together with infrastructure as a flexible computer load.
NG Zhang (44:13.869)
can use the backup power for for mining and just cut off the the miners when you you need the backup power. So the physical infrastructure is becoming more simpler high density power, liquid cooling ranks and
energy and management software. Yeah, this is the the future for the the the money industry.
Stephan (44:43.734)
Excellent. Well, I think those are the key areas that I was interested to talk about. Do you have any closing thoughts before we finish up?
NG Zhang (44:52.435)
NG Zhang (44:56.301)
I’m gonna see. Yeah. I think for the
Yeah, I I think for the you know about the yeah, we’ll be talking about some something about the the next generation or next one to years about the the mining the industry industry. Yeah. I I think twenty twenty six is the first time I really really feel the the efficiency curves is slowing down in in Bitcoin mining.
it’s not that only a canon issue. it’s it is I I I’m talking about the reasons. i it is semiconductor economics. you know the advanced nodes are becoming more and more expensive today. AI benefits a lot from high density because it needs larger, more complex chips and packages.
Bitcoin money is different. The SHC two five six is is very simple and parallel so the die size is never a big problem. so for every transistor’s price, I think the low lowest point is like the twelve nanometers, many years ago. and then the price for a each transistors is going up, it’s not going down.
when you upgrade your your nodes. but the efficiency and the performance still improving significantly from the more advanced nodes. So the whole industry is continue to upgrade rapidly. but today you know as as I mentioned like from four nanometers to three then two nanometers the performance
NG Zhang (46:59.945)
improvements is is low, much lower than the yet that yesterday and the cost is skylocked high today. the only benefits is you have you can achieve more density which is useless for for money. So this is the reason why the the first time I really feel the EVC curve is slowing down. but so
Sure, we we I think we we already down the R and D for the next generation today. we also have like t fifteen, twenty percent of imp improvements for the next generation, with a reasonable cost. but I think that the you know, last day last generation there’s like thirty percent of improvements.
and also much higher cost for next generation. So in some ways I think this is good because the industry is finally mature enough to build infrastructure not the not just replace the machines all day. So yeah. So if you I think if you anyone want to join the this industry still a good timing to to buy machines.
but yeah you know it’s it’s a bear bear market now. So so but I think the for have a better profit, usually we people invest in the bear market and harvest in the bull market. So this is many many times. Yeah. This is maybe the the last thing I I want to talk today.
Stephan (48:49.677)
Yeah, well thank you for sharing your insights with us. and yeah, listeners check it out. It’s a th my guest is NG Zhang of Kanan. NG, thank you for joining me today.
NG Zhang (49:01.582)
Welcome. See you next time. Bye bye.