Welcome to you all. We are having this presentation in Trondheim for the first time. As you know, Trondheim is one of the larger silicon hubs in Norway. A lot of our technology stems from Trondheim and from NTNU, so we wanted to come meet our advisors, our Trondheim-based shareholders, and felt like this was a good opportunity to update everyone. Since we are not live streaming, it will be recorded, and we will publish a recording as soon as possible, practically after the session has ended. There is some light food. I suggest we will eat after the presentation has been held. We are thinking about 45 minutes worth of materials here, and we will also share some take-home gifts afterwards.
[inaudible]. You might already know me. I am the Group CEO, Ola Stene-Johansen. Together with me, breaking his virginity, is our CFO, Christoffer Løvdal . He has a background from real estate development and commercial real estate, so he is taking charge of that section of the company and talk more about Arctic Core, the umbrella that deals with data center developments. Today, I think we might explore some future possibilities and potential. For the first time, I think this disclaimer really comes in handy. Certain things that we say might not happen. We usually either skip this part, but today I think it's a bit more important to consider that future statements are not necessarily always precise.
I think we should state that our plan was to take questions from the audience during the presentation, and the questions from the stream would come later. If you have any questions, raise your hands and I will get you in. You can do that, yeah.
Many times I have been challenged to try and explain what do we do. Even my girlfriend has no clue what does Lokotech Group really do, and it is a bit of a difficult message to send sometimes. We try to explain it with a cartoon. It is extremely simplified, but it has five steps, right? You plug in your machine, the machine contributes to the network security, and you get rewarded. The different entities in our group, they deal with different parts of that value chain. The electricity in part is relatively complicated, but it is underneath Arctic Core. We focus on the cheapest power available in Europe, Northern Norway, and so forth, renewable, and has capacity. For the hardware side, we are, as you all know, been through a six-year R&D effort.
The silicon is now being sliced, diced, and packaged in Malaysia, and will be shipped to our both U.S. side and Norwegian side maps for testing and bringing up all the technology. The latter part, what turns the computational power into actual cryptocurrency, that's the software layer that PowerPool serves. It aggregates hash rate from all over the world, turns it into cryptocurrency by generating value to the different networks, and we charge a fee for that. If you follow us closely, you'll see that in Arctic Core now we have three data center projects. We've just handed off Halden, a fairly distressed asset, so we're now down to three. We are working on utilizing the [inaudible] data center as a project entity for a capital-light type of AI HPC play. It's in the making.
We haven't announced it yet, but when we have material developments, we will announce it. I was really hoping that we'd have everything signed before today, but as you know, legal and counterparty due dil takes a bit of time. That being said, I'm going to try and bring you through the whole history of the different group entities today. We start with Lokotech AS, a company I co-founded. We even have another co-founder here today. He's here, Christoffer. We spent the first years tinkering about, and we found, verified one of those secret sauce ingredients. This was back in 2019. We raised initial capital. We made the deal late in 2020. in May 2021, we successfully listed on Euronext Growth Oslo. The initial capital went into the first design phase. Trial and error goes to a completed design, commercially sized.
It was a very efficient design, but unfortunately, the footprint of the ASIC was too large, so commercially, it wasn't viable. We then initiated a redesign, trying to compress or make the design more dense so that the silicon will also be a commercial viable product. We were, and still are, quite risk-averse, so we validated this technology in MPW, in multi-project wafer. It was done in 2024, and the road towards a commercial scale design started. In 2025, we secured the financing. It's a really harsh equity market back then. I think we were the only successful equity deal on the exchange that quarter. But it secured the financial future of the technology. Iterative process continued, and we finally got a sign-off on the GDSII. That's the graphical design file of the actual chip Q1 2026. That led to a tape-out.
The most major milestone in this company so far, the road to mass production was initiated. We also signed an LOI for commercializing the AI part of this ASIC. We're still in negotiations with those counterparts. In Q3, the master was done, and the wafers were close to completion. Right now, today, the silicon has been shipped out of the foundry and is in Malaysia for processing and packaging, slicing and dicing, back grinding, and all that goes into taking the silicon wafer into actual chips, as you know them. Then what? So bringing up any chip, I'm sure there's engineers in the room that can verify it, but bringing up any chip is a challenging process. You're striving to get out of reset, as we call it. We obviously have a timeline of when the silicon will be completed and commercially produced at scale.
There's a little bit of an art how much time, how much tinkering do you really need to bring the technology up? It might also tell us that in order to be commercially viable in the future, you might want to respin a layer or two in order to improve yield or gain even better functionality. So those are questions that we work with every day in order to solve in the most efficient manner. We do think that we'll have the technology brought up before Christmas, but Christmas time is also a busy time, so we might slide into new year, early next year. This is a slide that I've been presenting over and over and over since at least since 2024.
As you know, for cryptocurrency mining, the electricity in part is very important. Power cost is your definitive largest operational cost. It's all about efficiency. How much compute, how much output, how much cryptocurrency can you generate for a limited amount of energy. We are still far ahead of the main competitor, the big elephant in Bitmain, Chinese-owned, privately owned. We think that even if we should end up with a parity product, if they can release something that is on par with us, we still have a Western Hemisphere mindset, quality of life features, gradual degradation, and so forth. I think clients are willing to pay for that. We verified the fact that these Bitmain machines tend to not. They're unreliable. Half of the machinery in PowerPool hosting or pilot site has been maintained, and that's since January this year.
It will take you 180 days of downtime to ship this machine back to China to get repaired and return again. That's something that will cost you revenue as a miner. So we believe that the technology is a lot better than what they produce today. We hope that the future generations will also give them a hard time competing. For mining, the market throughout the year has been stable. It's been rough, I'd say. Majority of the hash rates have gone a little bit down. Prices have been under pressure. But the markets are still strong. We're still looking at 2.4 petahashes, roughly speaking, same as we entered the year. On peak, I think Scrypt reached 2.9. If you're a new shareholder or you're considering to invest in Lokotech because of this journey that we've been on so far, these are the main points I'd highlight.
The design team is experienced. They've done 15 tape-outs with parts of this technology in it already. We have run a downscale prototype from the MPW since September 2024. We are offering quality of life features such as this gradual degradation. We have an on-chip microcontroller, and the fabric is highly programmable. There's also little technical risk connected to the process node that we use. 12 nm is a mature node, and the blocks from the standard library has been verified over and over and over again. But we do, under any circumstance, foresee some type of need to improve the yield by doing a single layer respin. It's not financially very burdensome, but the time it will take to reach optimal yield is most likely dependent on such an engineering group.
I'm going to try and take a helicopter perspective on this company and try and see that the market cap hasn't moved a lot since its inception. We've been investing in researching and developing an Edge AI inferencing ASIC for six years now. What if you could invest in one of the other AI companies six years ago and be a part of that journey? Our ASIC is made for what we call agentic AI. It's a little bit of a buzzword, but it all has to do with the fact that the chip can perceive and decide on what to do, and then act on that information. The market is huge and growing, and if you measure the conversions of three different verticals, we're looking at $150 billion a year. To our knowledge, there's no existing competition in this niche. The GPUs, they're too expensive.
They're too power-hungry. You can't really use it for industrial automation or IoT applications. The Edge NPU, they have no on-device learning. They can't really do agentic AI workloads, but they can inference on a known model. Most likely more efficient than we can, but not every AI application requires a large model to use a device. We can also do cloud-based AI. We use API calls. They're expensive, take time. There's at least a 100-millisecond ping. So it's not fitting for a lot of these applications. We continue to negotiate with our partners to secure a prosperous and long-term future of Arctic AI. We are securing IP that will be important in the future generations of an AI-only ASIC. Hopefully we'll get there soon enough. On the other hand, having prototype chips in hand really changes the narrative.
There's plenty of NVIDIA killers out there, but nobody has a chip in hand, and they're all thinking about going through the R&D phase that we've already done. The narrative really changes when you have that physical proof.
You said this summer that you expected the joint venture to be turned into an agreement, I think you said before August.
Yeah.
What kind of obstacles have you met?
Well, it's really about agreements, right? The complexity of the licensing agreements have increased a lot during this last month, and the more complexity you add, the more time it will take to conclude it. But the developments have been positive, and I am more sure now that it will turn into an actual JV than what I was. But I am optimistic sometimes, and parts of these counterparts, they are also very much busy in preparing this, receiving these physical chips.
We continue. We work, meet every week in order to formalize something that is as beneficial as possible to our shareholders. It's also a bit of a give and take.
What will be the timeline of the [inaudible] or anything like the collaborative partners and the physical chips?
I think if we had the physical chip in hand earlier, it would be a lot easier to close it. But then again, it's an opportunity that they don't want to miss, I think. So we are nearing a point that our negotiation standpoint only increases over time.
It will depend on the results? Physical chips results or?
No, I think it will depend on our ability to secure a prosperous future. If we are not able to get all the different IP holders in agreements, you can't really ensure a third, fourth, 10th generation. So that's the first step.
I can see that you got some results in the simulations and the MPW as well for AI.
Yeah, we have simulation data. We haven't announced it yet. But I'm staying firm in my statement that the first generation will be commercially viable as it is. With that being said, in order to commercialize it, you need first to benchmark the physical silicon with various AI workloads or platforms, if you like. We plan for this, so it does support PyTorch and Alexa and a range of others. We are, on the business side, starting to gain some traction on identifying which niches this chip can fit. It's a soft material to talk about an ongoing negotiation, but nothing that we claim has changed. Did that answer your question?
Like MPW results with silicon.
No, but the MPW didn't include the AI block, so it's not physically proven yet, and therefore we consider the maturity level of the technology to be lower. That being said, it piggybacks off the cryptocurrency economy. For AI purposes, you have a much more traditional sales cycle. You'll have to integrate this product into other OEM manufacturers, support within SDK, that's a software development kit, and compete for design wins. Maybe 12 months down the road, you'll see some actual revenue. Volume ramping will take time for AI purposes. But piggybacking on the cryptocurrency side, where enormous fully paid cash up front, 12 months in advance, you get economies of scale from the other vertical, which also makes me believe that the price is a leverage that will happen right out of the door.
Now we're at the electricity in part of that old cartoon I showed you, powering these type machines. You take it, Christoffer?
Yeah.
The timeline.
Yeah, the timeline. So 2019- 2021, we built our own modular data center unit from scratch, and we also acquired a location in Halden, which could prove to have some good potential for a small data center operation. We entered into a leasehold agreement in 2022, I believe, this was before my time, where we acquired two properties next to each other up in Kautokeino for data center purposes. In 2024, 2025, we acquired what we call the Mountain Hall in [Trosvikveien]. It's actually an old storage unit for Diplom-Is, if you know what that is, like ice cream. They used to store their products in there. We thought this was a good opportunity. We have good access to power there for data center purposes.
We have also entered into a leasehold agreement in Øksnes, a greenfield property, which, if the regulation proves right, is going to be very beneficial for us. We want to acquire properties where we have access to power, where we can test and show our machines. In 2026, Q2, the regulation was completed. The property is now available for our potential use. In 2026, this is last week, we disposed of the Halden property because we saw it was not beneficial for our uses at all. We also got a building permit for our test center up in Kautokeino. At the time, we have access to 2 MW to put up a test center. We want to test our equipment when the silicon comes in. We want to test how it performs under Arctic conditions. We want to test both air-cooled and hydro-cooled containers. We want that.
Then we got the permit two months ago, and we are well on our way to building that test site for our equipment, where we will test both our own equipment and other equipment, so we can give a, what do you call it, benchmark of our equipment compared to Bitmain's equipment and other-
Suppliers
suppliers in the market.
The most important thing we need to mention is that we need to explore opportunities of capturing the heat and reusing the heat.
It's politically a very sensible thing to do. We are in discussions with local business owners, proprietors on how they can be used on our site specifically. It's a demand from NVE for every future data center to explore those options. It was a part of the learning of doing it the first year in Kautokeino.
That's why we're-
Transformers have been shipped, I believe now.
The transformer was a little late. It was supposed to be shipped on 29th of the September . It got shipped yesterday, I believe, from Møre Trafo. It is going to be installed this week. So we will have our test site up and going in this
In this month.
month. Yeah.
This is it. The test machine has been implanted.
Yeah.
This is an illustration of what this modular data center unit looks like. We have the racks for the dry cooler and piping to cool all the equipment inside with hydro. It is also about putting up a heat exchanger there, so you can use that heat for something and get paid for the power twice if you would like. Under normal market conditions, heat is priced at around 2/3 of the power, so it's a significant upside to be able to utilize that heat. That' s Kattegat on the right.
This is a week ago, I think. I haven't gotten updated pictures, but they are very close to finishing the site. We also needed to dig a high voltage cable of 1.4 km. They said yesterday that they had 100 m left. There had been a lot of rain up there, so it filled up the trench where they built it, but it should be ready this week.
The first container is also being shipped. This week. In Mountain Hall, we have started liquidation of the interior. There's a lot of stuff to clean up before we can be turned into a data center. So liquidation has started. We're pulling out the copper and the aluminum to making it ready for a potential future data center.
2 MW is there. Drop A is untouched. We believe the site can handle up to 10 MW. It's a bit of a process to get energizing 10 MW worth of equipment, I'll tell you that much, but at least we have started the process.
Yeah. It's all about tearing out the old equipment from the Diplom-Is. They had a lot of coolers there, so of course, to freeze the ice. So we've worked well into the process of liquidating all that equipment and getting the money from the metal in those machines, so.
Now over to PowerPool. Ruben is not here today. He can't really watch the live stream either. But I'll happily take questions. So PowerPool is our cash cow. It generates a high 90% cash conversion from net revenue. It started in 2021. Ruben got his first client via Reddit on the tester software, and a year later they launched Bitcoin for Sharp in 2016. It's continued to launch new algorithms. That means we're diversifying the revenue stream, more blockchains, more opportunity to make money. Lokotech Group acquired a controlling stake in May 2024. We started to market the service via public panels and appearances and conferences, and the B2B sales started to shape up a little bit. We saw demand for so-called SOC, Security Operations Center compliance. That was done in 2025 after a full software reshape. We started to offer hosting services.
The pilot in Northern Norway was well-received, and we continue to develop that business. Important milestone lately is the launch of Equihash, and I'll let you know a little bit later why. We also launched custom mining firmware. It means that if you run a Bitcoin machine with stock firmware, you're losing out on 10% of the revenue, roughly speaking. You lose out on fully pool integrated fleet management. That means that you have a data center, and you have perhaps 10,000 machines running, and either way, you have to control them in a mass way. This is compatible with those type of solutions. I think it's also important to see the benefit of selling this firmware as a service to larger clients. They often don't like pool fees, and they want to negotiate on that.
If you bring something more to the table, they're more inclined to switch. We've so far completed all the documentation required also for SOC 1. There's been exactly zero clients demanding it. The documentation is now shelved for now. Why should we spend all the money for an audit? It's ready, it's done. We can do it if we want to, and we will do so if the clients ask for it. We're also launching solo mining. That means that you're not opting in to be a part of the greater pool. You're taking the chance of finding a block yourself. Still needs the pool software as a middle layer between the hardware and the blockchain. It's been something that both enthusiasts and also larger players would like. RandomX for Monero is also something we consider to launch.
Politically a little bit sensitive, but nonetheless, we are here to make money.
As I understand, you're aiming at substantial growth for PowerPool. What is your view on the capacity for doing that? The power, buildings, and also the political landscape. How do you foresee to do this growth?
PowerPool software runs on about 40 servers, about.
40?
40. 4-0. It's a software layer. We do not necessarily need to own the hardware that produces all this hash rate. We do not need to own the infrastructure for the power. That's the client's issue. It' s very scalable in that sense.
What we do see is it's a combined political thing, I believe. Chinese miners want to use Chinese pools. Russian miners want to use Russian pools. The U.S. miners want you to be a U.S. entity. Even though, we are serving 15,300 clients from 121 nations, and it continues to grow every day. We also see a tendency that new larger actors, they don't want to deal with switching pools through something that is already in operation. That is a tough sell. We have been trying. What they do is, "Okay. We want to bring you on for our next site." Then you are dependent on their development schedule, their acquisition of power assets, and so on. We are completely reliant on them executing on their strategy. The last half year has been rough for a lot of miners. What we do see, though, is that there is no churn.
If we go down in hash rate, it is because a client has ceased their operations, or there is some intermediary issue like a power outage or a trough hole that caught fire, and they bring it back up.
Back to Equihash, let us talk about hash rate. Hash rate is the computational power. How many hashes do you deliver per second? Therefore, hash rate. We launched Equihash because the underlying asset, Zcash, was looking more and more interesting to support. Since summer, in June, it has now become close to 20% of PowerPool's revenue. I think it underlines that being open to support new and different algorithms is important for a complete value driver also for PowerPool. Scrypt now is about 40% of our revenue. Network hash rate this year has gone down 2.5%, whereas our hash rate in the pool has increased. Same goes for SHA-256, the algorithm that supports Bitcoin amongst some thousand other coins. Network hash rate is down, our hash rate is up. We can't really do that comparison for Equihash, but it's becoming more and more significant.
We think that RandomX might have the same effect for us, so that's why we really want to dig into supporting new algorithms and other business development efforts.
The AI JV negotiations are ongoing, and it's developing in the right direction, in my view. We have also started, and we are pretty close, I believe, in signing a new agreement with some international actors that would want to enter Norway for HPC and AI compute. That's the big takeaway here because it's a capital-light model. They need boots on the ground that understands the local politics and the local energy markets. Then again, we don't have the capital to bring up an AI compute data center. We are talking NOK 50 million, NOK 70 million per megawatt. Those are the big boys that want to come and get access to cheap regional electricity and a very stable political environment. Little bit out on a limb here when I start to talk about something that hasn't happened, but I am very confident that we will get that business going shortly.
Very good talks going on with the different parties. As I mentioned, this in-house mining firmware is a catalyst for B2B sales. It looks better on their accounting if they do not pay at all. Then again, it looks a lot better on their accounting if they can earn 10% more and pay us 2% of that 10% increase. It is a bit of a mindset game for existing miners. If I get my wishes, we will staff up the sales department in that sense and start to knock on more doors because that mining firmware is superior to what the stock firmware is. It' s really a no-brainer for good clients.
Quick question on the SPDU. What is the status on that?
The SPDU is a hardware catalyst. It is a Smart Power Distribution Unit. Mining hardware today, they often do not even have an on button. The stock power supplies have a tendency to fail over time. The SPDU is a hardware unit that goes in between the grid and the computer. We can monitor, measure, and also predict when the machines are acting up. It is a low-cost item, and the only alternative is to have really expensive industrial power distribution units, then you have a separate software suite that you have to follow as well. The first sample batch is fully booked from clients around the globe, and is undergoing CE certification right now. When the certification is completed, we can offer it more broadly.
To summarize, we are listed, and I think the most important part of being listed for us is the AML/KYC landscape that is becoming more and more of an issue for both traditional finance but also for cryptocurrency businesses. What we can do is we can audit trail both from the hardware side, where was the chip produced, who put money, where is it operating, who is operating it, and we even can trace a dollar back to the hash. That is an advantage that nobody else in this market has, the way I see it. We are bringing the most efficient and the most robust chip to the market for hashing Scrypt. We have completed six years of R&D, which also includes agentic AI, and we talked about the synergies with the cryptocurrency business.
We have preorders for the mining hardware, and we hold to the sales because we really want to know the yield and the margin picture so we can price it correctly. As soon as we know, sales will open again. Our cash cow is performing well on all the fundamental KPIs. Clients are increasing, hash rate is increasing in a tough market, and we are looking at the same type of revenue figures as we were in the first half of this year, $2,500 roughly per day. Then Q&A.
Yeah. Opening up for more questions. Nobody? Yeah.
Regarding this timeline, you said that it would take some time to bring up because we have seen that we have millions of these hash rates of Q2 2026. It is like still three months. So how much you are expecting with the bring up and then we have currently we are doing the engineering wafers and now [inaudible] are they fulfilling this existing demand or the orders or how it is?
It's a very good question. I wish I had an answer. When are we shipping the first large batch? It really depends on how the bring-up efforts go. If we're lucky, it'll take a week. If we're unlucky, it will take six weeks. They are prepared. The knives have been sharpened. Then there's a question of, and a lot of these clients, they prefer it this way. They'd rather not receive the physical hardware. They want to receive the hash rate. So, "Please, can you host it for us? Can you run it for me?" That also ties into the CE certification. We don't know the electromagnetic compatibility efforts yet. That might mean that we'll have to change up how the systems operate. But in the meantime, we can operate them ourselves.
Yeah. That's what I wanted to ask. The other question is that we are producing the engineering wafers but we are doing that now. Other than the production, so engineering wafers and then we'll have a mass production. Then we have two batches or three batches we are doing that. Then have they earned this engineering wafers or are there programs to engineering wafers but engineering less than-
The production is staggered. That means that the first engineering wafers are done. The rest of the boat is close to being done, but we're not pushing them through production until we bring up a set of them. It has to do with a security feature. We have to wait for problem to provide the encryption keys needed to communicate with the chip. Every chip has their own key. Without doing so, you can't produce a product. So they're on hold. Sitting on the There's five out of 64 layers left.
That's a really big deal. Again, because the finance everything. The feedback is vital for this production [inaudible]
Yeah, but the clients have paid a security deposit back. We are also dependent on paying for because it's very capital intensive to produce this equipment. We don't want to use their money before we know the w orks. It's a matter of how
All this production. We secure all three batches with all of them [inaudible]
No, we are dependent on using the client funds, therefore we are waiting until we know.
We are not using clients money from that. We are using the Lokotech technology to produce these three batches, first two batches and then we get the results and then we can use the client money. We do not need to look other round of financing or anything. That is the-
No, we are not.
That is the-
We're healthy financially wise to do venture into the negative working capital.
[inaudible]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
More questions.
[Non-English content]
We have no guiding for that.
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
[Non-English content]
We asked on [5 Shack], we discussed a lot about this wafer probe test. Have you forgotten this one?
No. It's one of the results of the bring up on the first wafer. Without knowing how the wafer prober should probe, can't process the rest of the wafers. Well, we can, but it will be garbage, so we're not doing that. The first wafer goes to bring up, the second wafer goes to quality assurance of the prober, and from there on out, it's reasonably
Okay, one question on this Edge AI, because you said you have this unique proposition of having this on-device machine learning models. But how does the customer traction for it, or customers prefer to have something using periodically cloud and train the models or train models or are there more customers who prefer to have a on-device learning?
To take one very relatively simple example, industrial automation or predicting maintenance for industrial applications. Let's say you have 500 oil pumps spread around across an area. You want to know when should I plan for replacing one or making maintenance mode. Each and every oil pump will have its own differences, be it temperature, be it acoustics, be it age, composition of material. To have one standardized problem that can predict maintenance across all of them, it needs to be self-learning or self-calibrating is perhaps a better description of that. So that it sits and listens, and when it receives a signal that this I need to decide upon, it's sitting there perceiving, and then you analyze, okay, now I need to decide. Then it acts and sends back the report to whoever seems in charge of it. Today, you read manually.
The different niches have different requirements. Take another example, a smart bed. You want a bed that changes how it performs based on how you're moving around. We can make that today with a GPU. It'll cost you $35,000 perhaps. Run 240 W beneath the bed. It's going to produce some heat. It's a non-product. If you have a simple chip that can change how the smart bed operates over time according to how you change as a person, then you need that self-learning capability, or else it will eventually not function. Now, we have on-device security as well. If there's an application where data security is an issue, let's say you have a health wearable, we can be GDPR compliant out of the middle, out the door. Or if you're doing it in an autonomous underwater drone, you don't want it to be hacked.
It has some niche qualities that currently is not available.
More questions?
Yeah. I was surprised there are not more questions on the very exciting phase you are in now with the prototype testing. There was a question touching on how long time will it be and when will you know. You have answered that, and yes, of course, it is a difficult question because you will learn a lot when you bring it up, how good is it. My question is, what kind of testing have you. Will you be able to say with this prototype chip that it is TRL 6? Will you also test the chip in arrays and can we say a little bit more there and get you with a lot of people?
Sure. The bring-up boards are 3 by 3 array. It goes nine chips on top of on the bring-up board. We can also test and verify the in-between chip communication. That's an important feature of the hash plate originally, seeing that we like to have this gradient degradation. If a chip fails, it doesn't bring down the whole system. Do not have to pull it out of the rack and send it to China and hope we get something in return. We have a little bit lower output, but also a little bit lower power consumption. Yes, 3 by 3 array.
Any more questions?
What about this whole joint venture economics? Because you said this will be a share of the Arctic Core AI. How is the plan or future plans on the IP sharing and the-
What we're trying to achieve is an AI venture that is self-standing. It stands on its own legs. That means own staff, own plans for the future generation.
But in order to get there, the first hindrance is to make sure that the entity has all the IP that it needs in order to future-proof it.
I think we all acknowledge that a lot of the group and the recurrent size and scale is not equipped to challenge all the large players out there.
It's going to be that humble. Exactly how it will be commercialized after the first step, I am not sure. But I am sure that we will do so together with a partner that has established capacity. Be it sales and distribution, be it the software side of it. Not sure about that, who and when, but I can promise you that the initial talks is a lot easier when the physical chip is in hand. Take, for instance, Acuity, number one investment bank for semiconductor companies. They do not want to talk to you if you do not have a chip. The door is closed before you even attempt. There are so many brilliant ideas out there, right? And they are all claiming to be the next Amelia killer. But proof is in the pudding.
You are looking for a partner who can contribute more into the software SDKs and everything with some experience?
Either a software SDK partner or existing distribution and clients within the space. There are multiple companies that sell Edge NPUs with their own OEM systems out there. I am very surprised if they are not interested in also exploring competing technologies.
But then Lokotech will get finance or something or a personal opinion?
It has to do with timing, I suppose. If it happens before we turn the first orders around, that freestanding entity will require some capital. But by whom and how, it's [inaudible] premature to talk about that.
Okay. You said something about the group are looking for a December impact when Christoffer . Does it still need money or?
There's been no change in their mode of doing business and we are still executing M&A strategies elsewhere. They had a firm demand to show us their chip. I'll send them a chip next