Project TBD Memory Vault

2026-04-29-vivian-bliss-dustin

49 min read

Vivian + Bliss + Dustin

Vivian’s Background & Expertise

  • Private equity and late-stage VC background covering semiconductor and deep tech in US

  • Based in Taiwan, collaborated with Taiwanese semiconductor companies and US startups needing Taiwanese components/TSMC capacity

  • Expertise areas:

    • Complete semiconductor value chain knowledge (beyond just TSMC manufacturing)
    • Current hottest applications: AI data centers and agentic AI
    • Pain points across Taiwan companies and US startups

Semiconductor Value Chain Structure

  • Most companies do only one specific thing (except IDM model like Intel/Samsung)

  • Key layers beyond chip design and manufacturing:

    • Substrate manufacturing
    • Cooling systems
    • Power connections
    • Passive components
    • Testing

  • Chip design = only layer with tons of startups (low capex, IP-based business)

  • Other components require 3-5 years factory buildout with heavy capex investment

Current Supply Chain Coordination Challenges

  • No centralized platform - entirely network-based sourcing via phone calls and relationships

  • NVIDIA’s approach to finding suppliers:

    1. Internal teams identify suppliers in Taiwan/other countries
    2. Invest in startups/VCs to access intelligence networks
    3. Identify adjacent manufacturers who could pivot capabilities

  • Companies want 2nd/3rd supplier redundancy for:

    • Cost reduction and margin improvement
    • Supply security (high demand causing shortages)
    • Competitive blocking (securing capacity to prevent competitors)

TSMC Capacity as Gatekeeper

  • Advanced node capacity (2nm, 3nm) already booked by Apple, Qualcomm

  • Startup validation process:

    • Need insider connections for initial conversations
    • Technical discussions with TSMC senior people
    • Formal supply chain division evaluation

  • TSMC capacity allocation = signal of startup legitimacy to investors

  • US startups struggle: excellent chip design capability but no manufacturing connections

  • VCs like MVP Ventures bridge Taiwan-Korea foundries to US startups

Market Intelligence Pain Points

  • Components companies lack comprehensive industry visibility

  • Need to understand:

    • Competitor activities and client relationships
    • Cross-sector demand patterns (substrate manufacturers want to know power generation trends)
    • Whether high demand is cyclical or structural

  • Current information exchange: person-to-person, no centralized database

  • Companies engage VC/stock market analysts for broader intelligence

  • Memory space example: manufacturers unsure if 3-year capacity expansion will have buyers

Compliance & Legal Considerations

  • Big companies have comprehensive in-house legal teams + external counsel

  • US government sanctions require proof components don’t source from China

  • Tier 1 suppliers audit all sub-components for compliance

  • International law firms (Baker McKenzie) handle complex cases

  • Potential platform opportunity: automate compliance workflow for faster auditing

  • Target stakeholders would be in-house legal teams

Next Steps

  • Dustin to schedule follow-up with Vivian in 2 weeks

  • Key connections needed:

    • In-house compliance teams at semiconductor companies
    • Procurement professionals at semiconductor companies
    • External law firms handling semiconductor compliance

  • Resources to obtain:

    • Map of semiconductor supply chain with key players
    • Complete list of AI data center components

  • Consider Taiwan/Korea trip for July research

Chat with meeting transcript: https://notes.granola.ai/t/5393b51e-2dc4-4175-be89-ebdddaa7e5c0

Transcript

Them: Semiconductor. I want to know more about robotics because it’s more like the word of the physical AI, especially for the robotics right now. I think it’s a good time for now to do this kind of a thing. So I want to join this team and to see what they are doing, what they are thinking. Because the TRI is more like the lab. It’s just like they did my under the Google. So they are like developing very advanced technology. So it’s more like the technical side. So I’m more like interested in technical side. Yeah. Hi, Dustin. Hi, I’m sorry I was late. Now I’m good. How are you doing? I’m good. I’m very very good where are you right now back at school? Yeah, I just come back to Philly today and I’m gonna have like another. Like good project or like this cipher letter. So. Yeah. I was going to save that background doesn’t look like filly. No slide to Google background. Anyway, thanks for carving out a little time to chit chat with us and help us. Yeah sure I can be helpful. Of course. I don’t know if I missed anything. Now we’re just talking about robotics and deep tech for that and summer plans. She’s gonna be in mountain view so we’ll all be around which is nice. Amazing. Yeah. Super amazing and you have to feed us information on what kind of problems you guys are facing so we can build around it. Sure of course. I will be like the internal spy to let you know. You know. Exactly no no not spy you’re sourcing custom built solutions for your company. The attacker. It’s called design partnership not espionage but anyway we just first of all really great to meet you second of all really wanted to connect with you because we are. Of course. Some excited guys. Who have a lot of passion and interest in critical supply chains and in particular semiconductors. That’s where we’re focusing our analysis right now and so we really wanted to just understand like what are the. What are the problems being faced right what we’ve done just to be like I’m gonna just give you all the information and then we can work together on this what we’ve done is we’ve done an incredible amount of like analysis desktop research and push the AI models all the way to the brink to try to uncover like pain points and then design what we think are interesting business solutions around those pain points. What we have not done. He’s gone out and talked to people and like actually had conversations and understood how procurement in the semiconductor supply chain works and how. Fabs and. Foundries and all these different players actually interact with each other. What are the pain points that they’re experiencing? What are the problems that they’d be willing to pay for? What are the problems where they’re like it’s kind of annoying but like I don’t really care that much. And so. I don’t quite know the right place to start and drill down in the conversation just to kind of kick off the conversation. But would love to learn more about first of all like what was your specific background prior to coming to business school? Second what kinds of things stood out to you as noteworthy? In your prior experience? And then third what are the specific pain points, the specific inefficiencies and specific things that you observed where forgetting feasibility. But like if this was solved it would be awesome and someone would be willing to pay for it. Sure so my background I’m coming from the private architect and the late stage visit background and I covered the semicolon doctor and the deep tech area in the area in the US. So my friends is investing in all like the seven conductor space because we are basing the Taiwan so we collaborate with a lot of like Taiwanese semicolor companies and also some US dollar locks because they need some like Taiwanese components and they need the TSMC capacity. So this is what I did and for my expertise I think there are a couple of things. The first one is. I know the whole value chain of the semiconductor because of the semiconductor is a very huge topic. So find everyone is talking about the semiconductor. I think most of the people will only think about like a TSMC but actually the TSMC is only doing the parts of the semiconductor value chain. They are just like manufacturer factory. They are trying to. Turn like the idea into the physical product like to the chip and the chipset. But for like whole value chain there are like different kinds of companies and layers. We can talk about this one later. So my expertise like I know what kind of layers in a semiconductor now and the second one is what is the hottest topic right now in a semiconductor. I think maybe you know like the most like. The hottest like application now is more like the AI data center and like the agentic AI. So for example for the agentic AI, it will relate it to the CPU buffer CPU. The background of let’s say here is not only like the MD or Intel but also like the whole value chain like who is like cooperating with the intel and empty to try to produce the chip because the cherry base not only like the foundry you need to like have a lot of substrate or different kinds of components or passive components to produce the final part out of the chape. And the second expert at the what is the pain point or like regarding the Taiwan companies or the US startups. So I think for this semiconductor industry for now there are a couple of like parties. The first one is more like the, I would say it’s more like the consumers who is buying like the semiconductor products and it’s more like the CSP. So we’re only talking about like the area. We don’t talk about like the smartphones because. And what is CSP? Of CSP it’s just like the Google Amazons made. Got it all service riders gotta got it got it. So yeah, still the terminology of this kind of company is called CSP. So it’s just like they provide a lot of like call service to the different kinds of customers. So for now it’s the biggest CSP including like Google, meta, amazon. So they are buying a lot of like centrally are buying a lot of like GPU. They are buying a lot of stacks. So they are the actual buyers of this like demand right now. And then the second part is it’s more like who is producing like the AI data center. So we know it’s like one of the biggest players in this space is on video. But nvidia is more like producing like the GPU. Into the AI data center. So actually the value chain is more like the Nvidia, they have the GPU and they cooperate with the TSMC to try to esthetic to produce the GPU. For them and then the Nvidia will also like try to source different kinds of components. For example like the cooling system, the power connections are like the substrate and then the Nvidia will X the ODM. The ODM is another terminology. It’s more like someone just help you to combine with all the components. ODM stands for what original device manufacturer original design manufacturer design. Yeah. Got it. So the audio and the biggest player I can key the names for you later after the meeting. So for the ODM is more like the biggest players in Taiwan and they are the very largest player for the OTM. So like the Hong Kai, they are the biggest like ODM. So they will try to have the recipe from the Nvidia and them to come by with all the components into AR data center. And then beta will sell this added to the CSP. So it’s more like one of the value chain right now. It’s more like vertical value. And how do these companies all coordinate with each other? Like is there any sort of centralized like platform which they can like see like what offers are being made for these subcomponents from other companies. They can discover alternate suppliers. Or is this all just like people working the phones, people drawing on their like pre existing relationships and just like dealing with each other manually. Yeah. As an interesting question. So actually they’re just like leverage layer Network and just call on the phone and decide what kind of suppliers they are going to source. So for example for the Nvidia list line a video they were like identify which kind of companies they want to source. For example in the cooling system they will have the company a for example. But in the meantime they will also try to identify the company B and C and D and E. So it’s doing right now is there are a couple of the approach. The first one is like their internal team to identify the different kinds of the suppliers in Taiwan or in other countries. And the second approach they are doing right now is they invest in a lot of like busy phones and also invest in some startups and trying to use this kind of network to tell them who is like the invisible. Player in these markets. So then the Nvidia will try to source like the second supplier or the third supplier for them too. So wait are they are they investing in these startups so that they can alternate suppliers or that they can provide the intelligence on who else. So they will invest in these kind of startups. And then for example so the Nvidia will like invest in the for example like Nvidia is like developing the photonics right now. So they will invest like a couple of like the US startups in the photonics areas. And then there’s kind of the companies will have the intelligence and then maybe they will like tell to the reader and tell like, oh, I know what kind of companies maybe in Taiwan they’re actually doing one of the components right now and maybe a team they will try to combine with the kind of intelligence and see oh maybe this kind of a traditional like manufacturing capability can transfer into the semi capability. So for example in Taiwan actually there are a lot of like companies they are not doing AI data center supply chain before they just have the capability. So for example capabilities in the power generations or in the substrate. But then the Nvidia will approach them and to tell them like oh you can convert this capability into the AI data center supply chain and then we can try. So from this out this kind of like Taiwanese companies they will like try to use. The R&D to develop different kinds of products and prototype and then fit to Nvidia and then Nvidia will decide whether to integrate them officially in their supply chain. So I think in the semiconductor supply chain the sourcing right now is not in the platform. It’s more like leverage the network. So people working in the semiconductor just like in the investment space or like in the supply chain space they know like tons of companies who have the capability to do the different kinds of components and then they will try to bridge them. To the largest like buyers like Nvidia, MD, Intel or the CSP. And so if you’re like one of these startups and you’re trying to attract an investment from Nvidia or whomever like is your whole value basically saying like I’m on the ground, I know all the players involved here. I know like all these factories in Taiwan that could be retooled from power generation into something related to data centers. And like that is like the value you’re offering or is it like the product vision? Like what exactly are these people pitching when they’re trying to get these investments? Yeah, so I think for this kind of a startups I think most of like year startups right now there are struggling to find the real partners outside of the US. So most of like the US startup in the semiconductor they have a very excellent capability like how to design the shapes. But they have the problem is like they don’t have the capacity to produce the chip so they just have ideas. So I know for now there are like a couple of the usbc especially in the deep tech area. They will approach this kind of a startups and tell them, oh, I know someone in the Samsung or someone in the TSMC and then I can help you to get the capacity. But in change of this one you need to give me like the series a investment ticket and at the lowest like valuations or what else. So this is the first approach. This your startups, they want to like know more about this space because this face is super complicated. There are millions of companies in this space and if we are like only doing the chips, you don’t know who is doing like the chipset or like doing equal and sustain this space. So the deep tech basically they will like try to connect them with some components and multiply this kind of a deep tech basic. Their internal team. I know there are like a lot of like partners. They are actually taiwanese or like korean. So they have like a lot of like local Network to help this kind of startups to connect with them. So that is one of the approach they are doing right now. And I think for like this kind of your startups, they are trying to do is like the div tag is very hot right now. So if you have the right thing. So for example like a couple months ago there are a lot of like startups that are doing more like the ASIC. ASIC shape. So they have a lot of interesting story and they also have the investor like from the hedge funds because the hedge fund also want to buy this kind of inference chips from these startups. But the startups don’t have the capacity to produce the very advanced shape. So dance, I know there’s a VC approach them and then should tell them, oh I have the connection in the TSMC. I can help you to bridge them and to talk to TSMC. And if TSMC commit to give you the capacity then that means your ideas will be successful. So the hints here is more like right now a lot of like basic investors. They will rely on like Samsung’s or TSMC their intelligence. To verify whether this semiconductor startups idea will be successful or not. So if NC agree to give them the capacity, that means TSMC or grade dislike technical approach. It sounds like reasonable. If this and the reject is kind of a technology doesn’t sounds like reasonable in like maybe three years or five years. So I think that struggling point for the US dollar right now. I know it’s that trying to approach like the foundry is the most important part. And once they like have the capacity from the foundry, they also need to like the different kinds of components and they need to source. But I think most of like the founders in this kind of semiconductors startups. They only have the technical background. They don’t have a lot of like supply chain background. So don’t know how to approach these companies and they don’t know how to like deal with like Asian companies. So what I can do is only to rely on the visa investors to do like visa investor will tell them what you should do a, b, c, d, e and then you I can bridge with you a, b, c, d, e. Yeah. It is a pain point right now for a lot of like deep tech startups that are trying to do. Sorry one thing is like I know some of the not trying to produce the real products. They’re only trying to sell the whole startups to like the video or internal to cash out. So like maybe they don’t have the pressure. To produce the real products. They just need someone to like say yes to their ideas. So I’m curious more about like the capacity allocation. Like it seems like what you said like the capacity is like sort of a signal. It’s kind of like a stamp of approval for a startup that like, okay, we have capacity from tsmc. That means like you think you should think that we’re legit because they’ve approved us and they’ve given us this like rare resource. Like how are those decisions made? Is it like again like you have someone on the inside you call someone you use some relationship you already have with them or is there a more formal like business channel where this is being done? To double click on that is it more, as Bliss said a signal, which is like we have capacity like therefore we’re legit. Or is like tsmc actually king making companies because like. Whoever has capacity. Will succeed whoever doesn’t have capacity won’t succeed like they’re kind of just deciding. For the first cautions are things yeah still the first touch point will be like you will need to have some insider person. And so for example there is startups. Noir is a basic company called mvp is basic Net and they’re very famous right now because they are trying to bridge the Taiwan and Koreans like the founder and semiconductor companies to the US startup. So like invest a lot of like in a deep tech in semi and robotics. So while they are doing right now is the founding partner of the MVP they won’t share. He has a very decent connection with the TSMC. Very senior. People. So what I is doing right now is like for example he will call someone in a TSMC and tell him like oh here is a startup there doing a and I want to bring the team to Taiwan to make with you and to see whether you are like comfortable of this idea and what do you think. And then so they will have like a couple of like discussions on the technical side. And then it will switch to the more like formal process. So let’s see what engage like their supply chain divisions to evaluate whether to grant the advanced node capacity to them. Because right now over the capacity especially in the vest no like the two nanometer and the three nanometer. It has already been booked by apple or like Qualcomm very company. So if the TSMC wants like squeeze the capacity that means like they feel like oh this technique technology is very important like in the next five years or 10 years. So the reason why the TSMC is willing to grant the capacity to them is the need to like have the pipeline in the semiconductor space. So you only like someone in this house. To make the phone call and to have the connection. It’s very hard for you to just just try to like find someone is doing like the supply chain manager intensity and tell him like oh I have a product and do you want a test? Can you give me a look capacity? I think it works but from the top down it’s more like defensible and quicker in this space right now because object semiconductor startups they want to scratch and have the capacity. And as for the second questions like just five points if TSMC say yes that means this company will be successful. But I think in other ways it’s more like so this deep tech company like the semiconductor companies in their page book in their pitch deck they will like say oh we have already have the TSMC capacity. For how many units? So because of the deep tech investor will ask them that oh you have awesome ideas but how did you produce your chips? Maybe you can find the intel but it depends on like. Whether the Intel rate is good or not right now. So most of the busy investors they were as it’s more like the do you have any connection or can you like have the capacity from the TSMC? Yeah. So it’s a kick. This is crazy helpful and like fascinating and it’s just showing how much we have to learn. When I think about the market overall. And I need to get to know this better I think about like chip designers like Nvidia and then you have the fabs like TSMC it’s really there’s really TSMC and Samsung. Yes. Like there’s not that many. Right. But then what you said is like everyone just thinks about the fabs but the supply chain is like way more complicated there’s this substrate manufacturing and all this stuff. Is it true that the market is like really just an oligopoly and the supply chain is really just driven by like TSMC, Samsung, Nvidia, Apple whatever. Or is there like more of a dispersed supply chain when you push. You know, I don’t know which direction but when you push like down below the TSMC level is the supply chain and the number of participants. Is it more dispersed or is it really just like concentrated where there’s like one group that’s doing this group that’s doing that like you mentioned. Quantum like quantum is doing everything. Like is it like one company’s doing everything or is there more dispersion when you push deeper up the chain? First adding one company Only doing one thing. So, for example, it’s like test and the only do the manufacturing. As if you are like talking about a testing, there will be another company. So in a semiconductor, it’s like one commonly doing one thing. Except like the idea model. Idea model is more like the Intel and Samsung because they have like their chip design and they also have the fab. So only Intel and Samsung, they’re more like IDM model. They are doing everything. But like the tsfc or most of like the semiconductor value chain companies, they are only doing one thing. And so they’re only doing one thing? So. They’re only doing one thing and the companies that are buying from them are they only buying that one thing from one person from one party or do they usually have like redundancy where there’s like a handful of different suppliers? Yeah, so I think. They are, they want to like buy from a couple of suppliers, but in reality they only can buy from one or two suppliers. So for example, I believe like on video or over the CSP, they don’t want like want to double down on the tsfc. They want to have a different sources. But right now they cannot have other solstice. They just have the tendency. But they are also trying to secure the intercapacity right now because the Intel, your radar is very awesome. They are improving a lot. So for like Nvidia and also like for other like apple Qualcomm, they are also trying to secure like Intel capacity. And I think for over the companies in a semiconductor value chance, they want to have like a second supplier, the third supplier because it’s well like impact their price. So if there is only one company doing these components, then the company can ask the higher prices for now. So for example, like right now the substrate or the PCB in this semiconductor is very hot because the demand and supply, the demand is very higher, but the supply is limited. And also like in memory space. So the demand is higher because the inference AI and agentic AI, but the supplier is limited. So I think for this supply chain in the salmon conductor, there are a lot of companies, but there are only like a few biggest players in different domains. Like in the manufacturing domains, they’re only like TSMC, Intel and Samsung. And like maybe in the cooling system there are only for two or three biggest players. And I think only one space there has a lot of like startup is in a chip design. I think it’s very interesting. So in a different kind of layers only in the chip design, they have a tons of life startups. And it also like a lot of deep tech investors. They are also interested in the chip design. Layers. The calcified chip design layers, you don’t need to have a lot of like cat packs. You don’t need to like have a lot of upfront cash outflow to build up your factory. But in other components like inequalizing power generations, PCB passive components, you all need to invest a lot of like capex to build out the factory. And to bid on the factory, you need like at least three to five years to build on the factory. But for the chip design, it’s more like the IP business. You just have like a lot of engineers sit there and to design your IP, you don’t need the factory. And so it sounds like it sounds like just playing back what I’ve heard from you there’s a huge growth in the chip design area because that’s a lower capex business. Yeah. I assume the dependency for those chip designers is then upstream on these higher capex manufacturing chains. It sounds like. All these companies want to have more options for suppliers, not because they’re worried about like redundancy and a breakdown in the supply chain, but because it eats away at their own bargaining power where they have to basically pay a higher margin. Yes. So it sounds like they all want different. Companies that they can go to. It sounds like what Nvidia is doing is actually going out and trying to identify adjacent manufacturers that could easily pivot their focus. And it sounds like. Yes. Yeah, that’s all super interesting. So then when you step a layer deeper than basically if all the growth is in the chip designers and you try to just think through the logical layers now you push beyond the chip designers. Everyone knows about the fabs, right? But then like let’s not think about the fabs. Let’s think about the cooling, the substrate, like all these other things. What does their supply chain look like and what are the pain points that they’re experiencing? I think for this chip design companies, it depends on their business model. So there are two kinds of a business model in the chip design in a like more general speaking. And the first one is they only. Like provide their IP. To their buyers. So for example, I just tell you the recipe. How to cook. And the second business model is like I provide. The end product to you. So I not only provide a recipe to you, but I also provide a dish for you. Like I provide like risotto for you. Not only for the recipe over the risotto. So I will give you risotto. Nvidia is the second right? Like. Nvidia as a second. Yeah. And but for the chip design startups, they also have like this kind of two different business models. And for the Nvidia, they are not only want to like invest in the first business model companies, but also for the second business model company. Because the chip, they have a lot of like different kinds of functions. Like for example the low power IC, the low power chip or like the ASIC is more like the inference chip. Right now it’s more like the everyone talking about TP, the CPU is for like a genetic AI. So there will have a lot of different kinds of the shape, the specialized chick. So what Nvidia is doing right now, there are not only developing in-house, they are trying to buy a lot of startups. And then they don’t need to like build from scratch. They just buy the startups. So first start out these two business models were like affect like how they cooperation model with the different. Like supply chance. Suppliers. So for example, if you only provide the IP or provide a recipe, you don’t need to like cooperate with the TSMC. You just like trying to find buyers who is company inspire recipe and sell your IP to them. But right now I know most of like the chapter size they are doing the second business model. Because like for the business model will affect their revenue stream. So the first one is more like the loyalty fee. So maybe they will charge like 2% or 1.5% of the end chip value. But for the second best model they are selling the units. So maybe the per chip is like $1. And for the ads and they’re one rack, you will have like a lot of like chips. So the VC or the investors, they are trying to look is what kind of business model startups can generate more cash flow once they have like guarantee. That can be successful. Are people focusing. It sounds like both of those business models then rely on actual manufacturing capacity. Right? Like at some point you can’t at some. Yes. It still sounds like, and this is I don’t totally understand if you are a chip designer, whether you’re strategy one or strategy two, you need someone to make your chip. And the most valuable thing is tsmc’s got to have got to put you in its capacity and if they don’t put you in your capacity then like it doesn’t matter how good your design is, you’re not going to, it’s not going to get made. What about. Huh, how do I ask this? Like are people paying attention to the companies that are actually making things? Like the manufacturing companies? Like how are they operating? And are people investing in that? It sounds like Nvidia is investing in that because they have a vested interest in keeping that ecosystem going. But like. I don’t know bliss, do you see what I’m trying to pull on? I can’t find the words. But. I think because like most of all the components manufacturers in the semiconductor space they are mature companies. So most of them are listed companies. So in the stock markets it’s super crazy. For this company’s share price. They grow a lot. For example in back then NASDAQ or NYSE the company like XTI or AOI they are doing more like the photonic things or the material things their share price are super crazy or like in the memory space like the sendix or the micron, their share price are also super crazy. So I think in terms of like talking about whether someone is investing in this space, it can separate into like more like in the VC investment part or it’s more like the stock markets more like the secondary market. So I think because most of like the components companies they are super mature. So their investors are more like the hedge fund or like. Stock market investors. But for the startups like most of the startups are in the chip design layers. So their investors will be more like the deep tech VC. Or maybe. What are the problems? Do you know what the problems that those legacy companies are facing is? Oh yeah I think one of the pamphlets that are facing right now is there are some trying to fight their customers or potential clients. By themselves. So for now if you’re all like the top players in the different segments then you just sit here. So like a CSP or a video will approach you because everyone knows everyone in this space. But it’s only limited to like a few players in this space. But there are also some different barriers they also want to like have the largest customers or the clients like maybe a median manufacturer that have a decent capabilities. They also want to have like the CSP clients, but I don’t know how to approach them. So is there any interest from any of these parties in like trying to move away from this very just sort of like you have to know people and like kind of do your calls to like figure out where the capacity is and like which factories can be retooled. Like are there any interest in like people changing this system or are the big fabs foundries just happy to like sit there and just get these calls and like again as Dustin said like play king maker here like do people complain about that or is it a way for people to kind of gatekeep themselves and like protect themselves from new firms? I think for the big like fab when we’re talking about the fab, I don’t think they have the urgency. To find a second players because that is not their responsibility. All they need to do is try to keep their pipeline. They just need to keep updated about latest technology. Like for example TSMC only need to know like all the photonics will be the next generation technology in the semiconductor. So like just like make sure they invest in the photonics and they have like photonics clients in their pipeline. But I think the players like the CSP or the largest like chip designer like AMD or Nvidia, they have the pressure and urgency to find a second players or the third suppliers a lot of like substitutes because there are two reasons. The first one is they want to decrease their cost. They want to like squeeze their. Cost materials. They want to increase their margin. So always like if they can find somewhere, it’s also can do this kind of product and then they can negotiate the price with the original players. And the second reason is the demand is very high now. So there is a lot of like supply issues in a different sub sectors like in the PCB or in a memory. So for less clients, CSP or the largest like designer, they will need to secure the different kind of suppliers. And the third reason is like for example for the Amazon versus like Google. So if Google now’s like by over the capacity from the potential players, then their competitors will buy over the capacity. So what they are doing right now is try to like buy over the capacity they can buy. So that’s the reason why they’re capex is increasing over the quarters because they are trying to secure it over the potential capacity and then they can block over the competitors. So that’s the reason why Google and Amazon are super active right now. They are trying to approach different kinds of like components manufacturers and try to talk to them and try to help them. Interesting. My initial reaction, I’m going to keep saying it fucking crazy helpful. Thank you for your time and your perspective. Where my brain keeps going. It’s just that it sounds like no one’s paying attention, but it sounds like there’s. If you kind of like pick your way up the supply chain enough. I don’t know. I think for the platform you are doing right now, I think it’s super interesting. But I think. The target audience will maybe will be like the chip designer in the startup space. Because the first reason is. They’re a lot of like chip designer companies in the startup space, but there are not too many component startups. In a semiconductor space. Question to you. Does it need to be a startup like I actually think that. I think the problem that we want to address and we don’t want to be stuck to a solution, right? We want to really understand the problems. But I think. These chip designers. Are not dealing with the physical movement of goods. And they don’t have negotiating power accordingly. And so my question is. Like. Are. The manufacturing companies actually more. Demanding of something that helps them better. Produce their product cheaper, more whatever. Yeah my brain is just going to like how do you how do you service those guys because they’re not getting the attention right now. And like what are the problems that they’re facing? It sounds like what they’re facing is just. Like a competition for margin. I think. I think. The competition for margin is not their first priority. I think the first party is to secure the capacity right now and to try to block their competitors. But are these companies securing capacity or are they. Like if we talk about again, I don’t even know enough to make things up but like a substrate manufacturer. Is that company competing for capacity or are they just trying to like. Make sure that they continue to get the tsmc contract because tsmc is yeah. For the soft track, they’re clients will be like CSP or Nvidia. Not the TSNC. Because only the manufacturing factory. So the customers of the TSMC will be like the CSP or the Nvidia. Like who will need to like have the chip as a product? Interesting. So Nvidia is taking the chip or like the semiconductor from tsmc. Yeah. And then they’re doing their own manufacturing operations to package it into a chip. Yes. So there was, yes. So for example like the chef. So he will tell like the TSMC like oh you need to help me to produce the tube and the chip will be like one of like the main dishes. Of the whole meal. And then the video need to serve their CSP clients. He can now just provide for example the noodles. He need to also provide like some meal or vegetable or some salsas. So the immediate need to source the cells for like maybe from the substrate components companies. And then to help them. To produce. And so those companies are all basically everyone that sells to Nvidia in that layer. Is competing. For capacity. They just want to keep the pipeline. Yes. So yes. Yeah, but a little bit complicated because in a semiconductor there are also like have the tier one or tier two suppliers. So not all of the components. Suppliers. They can direct sell to the Nvidia. It’s just like in the auto industry. So there will be the tier one tier two tier three suppliers. So for example in the substrack in this layer there will be like a tier one supplier ABC and the ABC will have like the direct relationship with their clients like with their CSP. Or with Nvidia or the AMD and they’re also like tier two companies. But the tier two companies are more like the median size or small size company. They maybe need to like. Need to have leverage the tier one relationship with the CSAP and then to sell their product indirectly. To their clients. But I think so it is the different kinds of process in a different soft sectors. But for all the components, I think if we want to solve the problem that is the first thing is there are a lot of like components in the semiconductor. So for example for the AI data center only an ad it has entered there are like different kinds of components inside. And the second part is the different kinds of components. They have different kinds of players who are producing these components. So it’s just like maybe the 10 multiply 10 metrics. There are a lot of players in this space. This is super helpful. I’m just taking those. So I have a couple questions that I want to do you still have a little bit of time. Oh yeah sure. Perfect I have a couple questions that I want to circle back to. But before we talk about that. Do you have any idea what kind of compliance. Regimes are faced? By these companies meaning you think about like in the united states there’s all sorts of things about checking what part of the supply chain comes from China and whatnot. Is this a thing that these semiconductor companies are dealing with? And again thinking less about the chip designers because like. And maybe we should be thinking more about the chip designers honestly. I’ve kind of written those off but maybe that’s a mistake. But I’m really thinking about like the manufacturing firms. Like do they face compliance burdens? Yes. Is that a pain point? To be honest I don’t think it is the time point. Okay let me take a step back. I think it depends on whether your target customers is the big company or the median company. If your target customers are like the big companies I don’t think it is their pinpoint because they have a very comprehensive compliance team. In house. And also like hire a lava like legal counsels to help them to deal with this kind of compliance issues. I think it’s a pinpoint for them but they already have the solution. So if you are targeting more like compliance issues in a big company then your stakeholders will be like the in-house legal. Of a semiconductor company and it will build a different perspective. Because I know like the most of all of like the CSP video it’s the largest companies and also like the components company. They have the in-house legal team and they also have external legal counsel to help them to deal with over the compliance issues. So for example a Taiwanese companies they are doing like a power controller. So if they want to sell their power control to the Nvidia then nvidia will have their in-house legal team to audit. Of the compliance. And for this Taiwanese companies they also have an in-house legal team to orders of the issues. And it is very important right now because like for example for the power generation is the product. But behind the power generation. There are a couple of like the components inside. So the Nvidia or the US governments is telling them you cannot buy any products that their components are from china. So then they have a very like comprehensive audit over the components. So while they are doing it right now is their in-house legal team will ask like their managers or different directions to provide the proof that they don’t source the components from the china. Is that something they ever hire external firms to do consulting firms? Yeah yeah they were hired legal councils. To help them to rebuild the compliance issues for now. Yeah because like the US governments for example US government has a sanction list like what kind of like the Chinese companies are no no. So it’s definitely not. And then there are some like gray area companies. But I know there are a lot of international legal house. They have like the team to help their clients to do this kind of a service. Want to do the platform to solve the compliance issues. I will imagine your stakeholders will be the legal counsel in house. Yeah. And it will be different from like working in the supply chain. And are there any like particular law firms that are like known for this? Like are they based in taiwan, maybe in the US? Are they just multinationals? Like I think just multinational companies for example in taiwan like the bakery mackenzie. They are the international law firms. So they have like the corporate team will help their clients to do this kind of things. But I think. They’re in house legal like in TSMC and like the audience companies I know they have like thousands of people in house. To deal with this kind of different things. But I think if you can provide a platform to them then why not? Then they can just quickly rebuild of the compliance issues. And I think it’s very important because. There are tons of components of every product and it takes a lot of time for the in-house people to review and audit of this kind of a things. But they’ve been a big company. It will be the top down approach. So. Plus you have any more questions on that front before I hijack? So I mean is the idea that you could sell to these like law firms to help them. Do the auditing themselves or just cut the firms out entirely. Or both that’s a good question. I’m not sure. Because I’m not sure. Yeah. For us to decide I guess. Yeah. I’m not sure because if the clients know if a legal house they are using a platform to help their clients to solve the problems then. Why not just like use your platform? Yeah this is super interesting. Vivian are there any questions we didn’t ask you that we should have. Like what is your platform want to solve the pain point? It’s more like in the compliance side. I think for us we’re really focused on understanding a pain point. We don’t have a platform yet, you know, and there’s there’s a lot of different things that we could build. The beautiful thing about bliss being so remarkably talented and us living in the age of AI is it’s really not that hard to build things. So we just we just want to understand like what is the real pain point where if we were able to build a solution here it would unlock. And so there’s kind of two things that we’ve been thinking about generally and we just don’t know. The biggest thing we need to do right now is like we need to be talking to people because we’ve done everything we can on online. And now it just comes down to talking to people. But one is. Effectively a B2B marketplace for the supply chain, B2B supply chain marketplace in the semiconductor space, right? Like you need to get component X from person y like who are these people can you identify more redundancy? Can you manage those transactions? Like that’s a whole business. The other one is. Can you automate the compliance workflow? For these companies? Right and there’s a lot of overlap right because like you need to in order for compliance purposes you need to share your supply chain. You need to like. Put in like your tier one, tier two work up the chain and understand where are things coming from. But it’s such a it’s such a, we don’t know, but a hypothesis is like that’s a pain in the ass. There’s so many components of the semiconductor supply chain. Can we help with that? Can we use large language models to automate that workflow to ease that burden to allow companies to more efficiently demonstrate fidelity in their supply chain? So to kind of like synthesize my response, what does the platform are building? We don’t know. We’re not building a platform yet. We’re trying to understand problems. These are the two problems that we’re kind of picking at. I think there is a pen pause to semiconductor battle train is facing right now. It’s like there are like different kinds of components companies. But like don’t know what other sub sectors companies they are doing right now and what is the capacity issue and what is a supply demand issue right now it is more like the secret information in this industry. So. Can you say one more time? I didn’t quite understand that. So the paint is. Price yeah so let’s make another company is a very like secret industry. So right now it’s more like there is no. Yeah. Like database. For all the companies to know for example I am the like cooling system manufacturer company. And my biggest clients is a video. So I only need to deal with the. Maybe in like two years ago. But right now I only need to know what kind of like the issues other components they are facing. Right now. Because then I can know whether I have the bargaining power to adjust my price or whether I have the bargain power to negotiate with the Nvidia to say. And so to go really slow there, he said what other components? So basically you’re saying like. You’ve got your substrate manufacturer and you’ve got your plastic casing manufacturer. Yeah. Yes. And they’re both selling to Nvidia and you’re saying the substrate manufacturer wants to understand. What I’m sorry to be very like I really want to understand. It’s more like for the first time for the substrate companies. What they are doing right what I need to know is there are two issues. The first one is they need to know their subtract competitors. While they are doing right now whether anyone is talking to Nvidia or anyone’s talking to an MD but for this one I think in this industry everyone will exchange the information personally. So we don’t have the database. So we just like have the chit chat or have like a phone call or the sales. They will have some reports to say oh what is going on in this abstract company. But the second point is like the substrate companies they may want to know. What is going on on other subsectors. For the substrate company maybe they want to know what is going on in the power generations because it’s very important right now what they want to like predict their like demands in this industry is like depends on how many orders the CSP or the Nvidia the order. But for this company they have the factories they need to like care about their capex that need to care about their budget planning. So they need to know what is really going in this industry. So for example. A lot of life subtract from the substrate companies. But maybe most of the orders are cancellable. So in the last minute the CSP will tell this company oh we don’t this one. So for this company they will need to have the intelligence to know what is going on in a comprehensive whole value chain. They cannot just hear their customers saying. So what they are doing right now is this components companies are also trying to engage into like the VC investors ecosystem or like the stock markets investors system or the sell side report analysis because they try to gather all the information to judge whether their business judgment is right or not. There is like the examples I think is very helpful is in the memory space because in the memory space. Over the companies or of the clients they are asking the memory manufacturer to increase their capacity. But they don’t know whether after like three years once their factories has been down where their clients deal this capacity. So they will need to know what is going on in this whole valuation. Whether like all other components they are also see or list demand is very. Like huge right now and we are not just like the cyclical industry and we are not just like our clients they are formal. So they just like ask a lot of others from us. So I think there is pamphlets can try to think about how to solve is that because this industry is super complicated and all the information are exchanged by person by person there is no database and over the database are audited. Like the memory price like the supply units is super audited. So the companies cannot rely on this metrics to decide their business judgment or their capex investment. I think it’s the one if you want to target the traditional component manufacturing space because not only the Taiwanese company there are a lot of us companies they are doing the traditional components manufacturing and are trying to do the transformation now to enter the AI space. So I think they are open to use this kind of platform for this intelligence or to tell them who is your potential clients in this space. Yeah, this is crazy helpful and where your brain is going is like very interesting to us. And it’s saying we’ve thought a lot about. I unfortunately need to run for another call. Yeah sure. But what So I would love to keep the conversation going. Like, I’d love to just, like, put another time on the calendar in a couple weeks and, like, just keep talking to us about. About you. I have a handful of questions that I. Definitely want to ask. And bliss, I think. Do you have another minute to. Hang on? I can have a few more minutes. Yeah. So we can, like, run those down. But I think the real thing that. But also probably soon. That I’d love to know is. Like, it sounds like the right next step is for us to talk to people on the one hand in compliance in the semiconductor space. And on the other hand, in procurement. And I would just love your thoughts on, like, who we should be trying to connect with. If there’s anyone in your Network that we could have a conversation with. But, yeah, it’s exciting to, like, start to get into the weeds and start. You know, fashion tree. About it. Yes. Yeah, sure. Of course. I need to jump, but. Vivian, I will. We’ll put time on your calendar. For, like, two weeks from now. And bless, I will see you in a debrief this because there’s a lot of, a lot of things going through my mind. Yeah. And then we can say if you have any questions, you can just, like, text me. Yeah, because the 700 is super complicated, so I don’t know whether I’m helpful or not because it’s super complex. No, you’re, it’s super helpful. I’m like, and again, I said, I know. It’s so hard to, like, give them one more. Totally. I know Bliss also has his own questions, but, like, my questions are like, you know, first of all, like, who can we speak to? Because that’s really important. But second of all, it’s like, is there some sort of map that we can, like, read to understand this? Oh, yeah, sure. Yeah. And then second of all, like, is there some sort of list of, like, all the different ingredients and components that go into it? Because I think if we just had those, like, documents to look at, that’d be really helpful. Yeah. Anyway, I really need to jump. But, Vivian, thank you so much, bliss. You’re the best. I’ll see you in a little bit. Yeah. See you guys. Bye.

Graph View

Backlinks