 I am Jonas Vermorella, I'm the CEO of FlowCAD. We are software companies that specialize in quantitative supply chain optimization. So it's, we only recently become involved with Bitcoin Cash, but I personally have been a long time Bitcoin enthusiast discovering Bitcoin in 2011. And as the CEO of a big data company, it was always obvious to me that Bitcoin should and would ultimately scale on chain. So I was actually very surprised to see that many people were actually that worried that about the feasibility of on chain scaling. For me, it was always very, very evident from the start. So when, in December, when I go really tired about people being obviously misinformed about the nature of the challenge, I decided to put forward an idea of having blocks of one terabytes. So why do we want blocks of one terabytes? Well, because if we have blocks of one terabytes, then we have 50 transactions a day per human for 10 billion humans. So this is, I believe, the end game for Bitcoin Cash. It's basically, we can onboard the entire world, all of mankind, and if we can do that cheaply, then basically, this is it. We have the, Bitcoin Cash can become the world currency, and there is no strength attached. There is no layer to involve or any kind of scaling mechanism that is not Bitcoin Cash. Bitcoin Cash should and will scale on its own. And, okay, so I came up with a fairly long blog post that contains way too many numbers for this talk. So if you're really interested, you can look my blog, thermal.com, terabytes, and you will have a very lengthy post at a terabyte mining rig. So that's basically, I put together numbers of how to build a machine that would be capable of sustaining four terabyte blocks for 20 years. So that means that's hardware cost, energy cost, bandwidth cost, and I came up with a grand total of $25.6 million. It may sound a lot, but again, I'm coming from a supply chain background and just to give you an idea, a commercial aircraft is $100 million, and people don't think, oh, aircraft are so expensive that nobody can afford to fly. I mean, who came with an aircraft today? Okay, so you see, lots of people can afford an aircraft, and the key idea is actually an aircraft can serve the needs of many, like a bridge. It's expensive to build a bridge, but you can use, many people can use a bridge for a very, very long time. So being expensive is not a problem as long as it's, everybody doesn't require its own aircraft, you know, and that's the same idea. But so this $25 million was my own estimate, but I realized, looking back at this estimate, that actually it was a very, very hasty estimate. It's not optimized at all. And actually we can do a lot better, and that's exactly what I'm going to do today until somebody smarter than me figure out how to make it even better. Yes? Okay, and then, so we have, the first things that you need to pay for if you want to have your Bitcoin mining rig is the bandwidth, and actually 20 years worth of bandwidth. So you need 50 gigabytes, by the way, not just that for terabyte blocks. It's not that much, actually. And back in December, I was thinking, how much is it going to cost? And I figured out the retail price for backbone internet bandwidth, and I came up with a bill of $60 million. It's quite pricey, but actually, if Bitcoin Cash becomes the world currency, this is not the retail price we are talking about. This is the big guy's price. The price for bandwidths that the Amazon and Google are actually paying, because by that time, there will be plenty of Bitcoin Cash companies that will be themselves super large. So, and just to give you a price point number, Facebook has invested $400 million to build a transatlantic cable that will bring them 120 terabits of bandwidth. So if we just linearly adjust the price, and it's going to operate for 25 years, if we just linearly scale the price, Facebook for the same bandwidth would only pay $200,000. And that's, so that's the real price point that really matters, because large, their cost system of Bitcoin Cash will have bargaining power to negotiate better bandwidth prices if it becomes the world-wide currency. So actually, out of my initial estimate of $25 million, we are down to actually something short of $20 million. So it's, we have already reduced our cost quite a lot. And by the way, transatlantic cable, it's the worst case. I mean, anywhere on earth, but when you want to cross the Pacific Ocean, bandwidth is going to be cheaper. So that's, I took the worst case, not, not a, and thinking further, you know, the future of bandwidth, it's looking very, very good. I mean, I believe that bandwidth is the most solved problem for Bitcoin, literally. Two years ago, a lab in Netherlands managed to have over one kilometer, 255 terabits per second, over a single strand of optical glass fiber. So that's two orders of magnitude better than what Facebook is getting right now. So really, I mean, and that's again, so that mean that by the time, you know, that Bitcoin Cash has become the world currency, this bandwidth problem is going to be, I mean, like nothing, literally nothing. That's the easiest problem to solve for, to scale Bitcoin Cash on scene. So what about the CPU cost, you know, to validate transactions? And back in December, I estimated that I found some notes on the internet about the fact that validating a transaction costs actually two milliseconds of CPU. And so when, and because terabyte blocks mean that you have seven million transactions per second to validate, I ended up with 512 Intel Xeon fee boards to cope with the throughputs. For those who do not know Intel Xeon fee, it's basically many core boards. So you have 64 CPUs on a single board. And, but actually my estimation was actually incorrect because nowadays the software back in Bitcoin has improved and actually quite a lot. Now to the pure CPU cost, it's more, it's closer to actually 100 microseconds. So it's 20 times better. So it turns out that you actually need a lot less of those boards. 25 would do just fine. So again, if we look at the grand total that was about 20 million dollar, we are now down to 17 million because well, we don't need to buy that many Intel fee board too bad for Intel. And, but what about the future of CPU? Actually, if you want to really be efficient and serious about computing, you need to go azix. I mean, it should not surprise anybody in this room. Minus, Bitcoin minus, for years have used azix and it has been a tremendous amount of hashing power to this network. And actually, to validate transactions you need what is called ECDSA, Cryptographic Signature Validation. And you can do it with azix as well and I was obviously not the first person to think of it. There is a lab that has working prototype to do just that and the performance are just astounding. I mean, we are talking of here a quarter square millimeter with 40 nanometer process, a vanishingly small consumption which is just 46 milliwatts and you can validate transaction in 1.5 microsecond. It's so efficient that actually one card that would look like GPU card that you would plug in a PCI slot, probably a 5.0 PCI slot so that you have the bandwidth would be sufficient to validate on the single machine, one card, those seven million transactions per second. And this is not the future, it's already been done, it's just not mass produced, that's it. Then what about the data storage cost? Okay, back in December, I thought, oh, one terabyte every 10 minutes, okay? So I need 20 years worth of one terabyte blocks and that in total, it's about one exabyte. And oh, that's a lot of data to store. And I say, okay, I'm just going to count how much storage I need for that but the reality is Bitcoin doesn't need the world blockchain. It only needs what is known as the UTXO dataset. And actually, if you do that on the Bitcoin network right now, the UTXO dataset is about 50 times smaller than the world blockchain. And my understanding is that this ratio will only grow larger as time passes for many obvious reasons. People don't want to fragment what they own over millions of unspent outputs. It's very complicated to scale for all the parts that are attached to the blockchain. So bottom line, we only need much, much less storage. And again, we can save actually over 12 million dollars worth of hardware equipment by just actually doing only the storing only the UTXO. So we are now back to a grand total of 4.4 million dollars for this terabyte mining rig. It's starting to look much better. But again, if we look at the future of that storage, it's looking very, very good. I mean, it's a bit maddening. We have three different hardware technologies that fiercely compete and that's relentlessly improved. We have SSDs which are integrated secrets and they are improving like crazy. The only things that prevents price decline is that pretty much all the cloud computing providers are migrating everything to SSD. So there is like a production shortage that keeps the price actually slightly increasing but it's just because demand is skyrocketing. But otherwise, it's progressing like crazy. And there is now, Samsung is actually shipping 30 terabytes. Drives recently there was another player that announced 100 terabytes SSD drive and then we have a magnetic recording. And again, we have 100 terabytes drives that are coming and we have optical. By the way, my favorite is optical and there is a company that has actually, that has a working prototype for something that would give you disk, the same size of a DVD that would be 360 terabyte per disk. And it's right once and it's actually incredible. It's a five dimensional data storage. So it's three special dimension and two optical dimension which are polarization and light intensity. So that gets you something that is super, super dense and it's made of fused quarts which is extremely cheap. And the latest tests indicate that this storage medium would be actually stable for 10th of thousands of years which is exactly what we need for long-term blockchain storage. It's pretty cool. And again, this is working. This is not like theory they have actually. It's just pre-industrialization. Okay, now what about the UTXO management cost? In my big terabyte mining rig, I needed back in December I was 266 nodes, not big coin nodes servers if you wish. And why did I need 256 servers? Well, I had 512 integer on board to plug. So that is it. Plus I had such a gigantic storage layer that I needed many servers to attach this storage layer to that many servers. Because I've reduced the number of integer on board to plug, because I've reduced the storage layer and I've actually reduced the complexity of many things, I actually need much fewer nodes to actually keep this mining rig going. And actually with 32 nodes, we still have four terabytes of RAMs. That's plenty to manage those terabyte blocks. And then if you do some clever tricks for IOPS, then you're good. And I can get back to that, but it's maybe beyond the scope of this talk. And so if we can also reduce the cost. And then the grand total cost for this mining rig is going down from $4.4 million to actually $1 million. Okay, so I believe back in December I made a full of myself by being incorrect of a factor of 20, which is pretty bad engineering. When you're just almost correct by factor of 20. But this is Bitcoin, so let's say I was almost right by a factor of 20. So what do we have now? We have the amortized cost because it's everything's paid for 20 years, energy, bandwidth, everything. So we have something that costs $50,000 per year over 20 years. So actually it starts to be very cheap. For example, I'd look at I'm running a company. I don't have an employee now that costs less than $50,000 per year. Maybe it's because I'm overpaying them. I don't know. I think I am probably, I don't know. But the point is having one person in Paris where we are located costs more than that. And so it's definitely business wise, it's cheap. I mean, if you think that a company cannot afford one more employee ever, no, yes it can, it can indeed will. So that's all the price that I've described. They are also, all those hardware prices, they are also decreasing by about 30% per year. If you compound 30% price decrease over 20 years, you end up with 1,300 times of price decrease over 20 years, that's insane. But guess what? This is exactly what did happen compared to the last 20 years. So it's not that insane because it has been happening for the last 50 years. But even if we are very skeptical and pessimistic, let's assume that the price will only decrease by factor 10, which is like a complete massive slowdown of all this progress. I don't think it will happen, but it might, who knows. If companies run out of ideas and ingenuity, who knows. But let's say, I mean factor 10 is very safe because basically it's what is already announced on the roadmap of pretty much everybody. And we'll say everybody, I mean Samsung, Intel, I mean big names who are not just hand waving like me about numbers. They have shareholders, they are public companies. So when they announce that something is coming, usually, I mean this thing is coming for real. This is not a marketing gimmick, unlike ICOs, let's say. So in March, and the good news about having 20 years worth of terabyte blocks that it takes 20 years to get there. You cannot have 20 years worth of terabyte block to manage before 20 years. So it will take time for us to take it there, like 20 years. Oh, and bottom line, I mean sorry, we don't have the slides anymore. But bottom line is if we assume that we have this 10X price decrease, then we are down to $5,000 a year, which is very, very cheap. Agreed, it's not, it's more expensive than a Raspberry Pi. Yes, yes, yes. But it's still fairly accessible. It will not endanger the decentralization capabilities of Bitcoin, it will be still that cheap for anybody who is truly into this thing to buy it. Yes, it will be hobbyist and it won't run on your iPhone 17 or whatnot. But still, decentralization is not at risk at all. By the way, we are shameless plug, we are hiring, we are seeking people to join the Terab, which is an initiative funded by Kongik to basically make this thing real with a software component, a big data software component to scale ZOTX or set. So if you happen to be a super talented software engineer, I mean just drop me a mail. Thank you very much. All right, okay, we got a few minutes for questions. Do we have quality questions to reflect the quality of the presentation? Any questions? Okay, here we go. Hi, you said miners only need the UTXO set. That goes and flies in the face of what people think today. Do you see that causing a problem in the future? I mean, the world is bad. There is plenty of people who think things that are vastly incorrect for many reasons. I mean there is, even in France, there is still many people who believe that the earth is flat. I came, you know, and according to the last survey, it's something like 15% of the population. So you see, whatever, you know, at the end of the day, it's better engineering that wins the game, you know. Google won the search engine, not because people believed in the algorithm of Google, it's just because it was better, obviously. So even if you believe that the earth is flat, you know, it doesn't make it flat and whatever. So yes, yes, many people will complain and that say that it's unfair or whatnot, but it is not. I mean, Bitcoin can run and will run just running on the UTXO set. And by the way, I suspect that if it becomes a world currency, there will be tons of companies that can afford to have a copy of the full blockchain, you know, because this data is highly valuable. There will be tons of companies that makes tons of money on top of this data. I'm just thinking of, you know, what do we need to preserve the properties of being a currency? I'm pretty sure there will be, you know, tons of companies that specialize in blockchain analysis and whatnot, and they will have the full copy, 100 of them. But here I'm talking of how do we keep it decentralized? Any other questions? Hi, so our company actually runs a wallet. We're edge secure, so you've probably seen our wallet somewhere. We run a bunch of nodes, full nodes, for our customers to use to query the balances of their transactions and so forth. That ends up taking a lot of bandwidth as well. In addition to just having the blockchain, we also have to index that and serve that. So does this calculation account for any of that sort of expense as well? No, not at all. Again, if you're serving one million users, you will have to pay for that on top. You see, again, but it's okay. You know, if Gmail, you know, Google is offering Gmail, it costs, you know, truckload of money to store all those emails and serve them, you know, it's not free. But guess what, you can make a living out of it with advertising and everything. And if you have one million users, you will be making, you know, millions per year to cover your expenses. I mean, just to give you an idea, Facebook is making about something like $25 per user per year and with you running a wallet, you will be, I guess, making even more because, you know, wallet is very close to financial stuff and so basically for you, if you have many users doing many queries, you know, hardware costs will not be an issue. I mean, you will be, you know, crazy rich and you will be able to afford something 10 times better than that, so don't worry about it. I'm a believer. Any more questions? One more question for Johannes in the back. Thank you for the presentation. It was surely very motivating and I'm sure everyone in the room here is very excited about one terabyte blocks. It would be also good for everyone to learn more about what Lokad does. I was wondering if you could just briefly tell us what Lokad does. Lokad does quantitative supply chain optimization. I'm not sure if all people are interested, but the bottom line is a couple of decades ago, the world has gone global, which is a tautology, but so supply chains have become super complicated. You know, you buy an iPhone, this iPhone has been manufactured in 20 countries. Apple has thousands of suppliers. You need to move all those parts around, assemble them, and it's the beauty of global business. It makes things very efficient thanks to specialization and the world decided that half of the things that were physical would be manufactured in China but sold overseas. Just imagine it's super, super complicated and you need to take decision on when to buy, when to order, how to distribute your inventory and basically you have Amazon that is putting tremendous pressure on everybody because they are delivering anything overnight. When you think of it, it's very, very hard to achieve. And so what my company does is to kind of put supply chain on, I would say, let's use the buzzword on AI autopilot. It's not exactly AI, it's more like machine learning, very smart machine learning, but it still takes a lot of human insight to make it work for real. So this is it. If you have more question about supply chain, just come to me. And by the way, supply chain are powered by commerce. It's only free commerce, it's only because we have free commerce that supply chain actually works. And so that my interest is that if I can make Bitcoin cashierality, there will be even more commerce and even more complicated supply chain for my core business to optimize. It's good.