Latest / Broadcast2Post by Key Code Media / RAM and SSD Prices Are Surging: What It Means for Broadcast, AV, and Creative Teams

Transcript

RAM and SSD Prices Are Surging: What It Means for Broadcast, AV, and Creative Teams · Broadcast2Post by Key Code Media

• 0:00
  Speaker A: Foreign. Welcome back to Broadcast to post. I'm still your host, Michael Kammes, and I've got a good one for you. Today we just had a great conversation with the CEO of editshare, Brad Turner. And we talked about a wide range of things, but we really focused on kind of the tech crisis in the media industry. That's the high price and shortage of the tech that we use in our media creation lives. Brad talked a lot about supply chain and all the other factors that go into the current crunch we're in. You're not gonna wanna miss this one. So let's get right to it. Brad, thank you so much for joining us. And we've got a lot of questions to get to, so why don't we just jump right in. What's currently driving the memory shortage that we're seeing across multiple industries? And why should our industry kind of take note?
• 0:54
  Speaker B: Yeah, so the big headline here is that the hyperscalers, Google Meta, Amazon, OpenAI, Anthropic are spending billions of dollars in capital expenditures to build out data centers. And those data centers obviously require space, they require racks, they require tons of servers in them that consume the base commodities that we in the broadcast and post production world need to put into our, those of us who operate on hardware we needed to put into, into our gear so that we can provide this sort of on premise performance that, that the broadcasters and producers expect.
• 1:35
  Speaker A: So because of that, you know, it's not just one component that our industry uses, it's multiple components across all the different technologies we use.
• 1:43
  Speaker B: Right.
• 1:44
  Speaker A: Whether it's for editing, whether it's for storage, storage, whether it's just for servers running processes in the background, it, it just runs the complete gamut. So let's get down to some of the, the technical aspects because most people think it's just video cards, right? Gpu because that's what AI uses, right. You're always told if you're going to use AI you need a lot of compute. But it's more than just GPUs, isn't it?
• 2:06
  Speaker B: Yeah, it really is. So GPU is like made up of a whole bunch of different things, things that are on the GPU card. But there's also just within a, in an AI server there's tons of memory. So it's called dram. Right. And they actually use a higher form of that memory that's not purely commodity. It's been tuned for AI and that's squeezing out normal DRAM production. Two is they need a lot of hard drives and they need both spinning hard Drives which is the traditional, you know, nearline drive that you would use in production. And they need nand. And NAND is what goes into NVME drives which are the very, very high speed solid state drives that power high end finishing applications. And I know we'll get to this later but we've seen just massive price increases and shortage of, of supply because of just the what seems like an unlimited appetite for all of those components.
• 3:08
  Speaker A: Speaking of pricing increases, you gave us a graph here which I think we'd like to share with everyone that kind of talks about that price increase you're seeing. Can you kind of talk us through this?
• 3:19
  Speaker B: What this shows is starting back In December of 2025 we started this is, this is indexed, this is December of 2025. And what you can see is over the period of 18 months hard drives went up by 89%. So something that cost you a hundred bucks now cost you $189 today. And you apply that same logic to DRAM and NVME. I mean NVME is up over 300% in a year. And so we really started and you can also see on this, this chart where you can start to see the issues come in and any publicly available data would show you this. It really started in earnest last fall and that, that was content that was coincident with an acceleration in data center build out by the, by the hyperscalers. So as they've, if you, if you matched up this increases along with their reported CapEx spending by quarter, they, they line up pretty well. So last fall, in fall of 2025 was, was really where we said holy cow, we've got a real problem here. This is becoming real.
• 4:30
  Speaker A: So Brad, because the media industry has a lot of niches in it, right? What kind of groups of users or folks are you seeing the most affected by this? In kind of the media industry.
• 4:42
  Speaker B: Yeah. So number one would be enterprise network attached storage or a san where we play right. We are right in the, right in the crosshairs of the hard disk drives and the NVME going up significantly I would contrast that and the price we'll go through the slide where the price increases have been but I would contrast that if you're you know, working on more prosumer or consumer based hard drives or equipment where they have, you've got more variability to source, more options to source components you're going to see, you know, and, or use older technology, you're going to see less of a price increase there. And so for example, if we were to build a desktop computer with enterprise drives, it's going to be significantly more expensive for us to build up with those drives versus you know, going on the Dell and buying a laptop. The other would be the difference between a high level, you know, hopper based GPU, that's the architecture in Blackwell, versus a run in the middle GPU, RTX or AMD type product. Those things have seen maybe 10 or 30% increases over the last year versus you know, 90 or 100% increases. So there's different tiers within the, within our space of equipment and depending on how big the availability is and how mission critical it is, you're seeing different levels of price increases and demand occur across that spectrum.
• 6:24
  Speaker A: And so when, when post facilities and broadcast facilities are even sometimes trickling down the consumer, what are some of the things that us as professionals need to expect in this climate? What does that mean for us because of these constraints?
• 6:40
  Speaker B: Yeah, so these shortages affect different manufacturers differently. What I'm seeing is number one, just longer lead times. We've been fortunate to not experience that, but we have heard tales of folks having, having to wait weeks and even months beyond what they historically had been able to get their time frames, had been able to get their gear and their hardware, which is causing, you know, planning problems. The second is the inability for certain vendors to honor prices over maybe a couple of weeks. Right. So that is introducing more volatility into budgets. And so customers are putting padding in in case the prices go up. And they're also doing longer term planning of hey, if, if these price increases and availability continue into 2027 and beyond, what does that mean for my mix between on premise infrastructure and software versus taking another look at the cloud and, and you know, for cloud, for, for both applications and for storage, is it worth another look?
• 7:59
  Speaker A: And I can add from a key code perspective because of the immense lines of products that we carry that many manufacturers are giving us price lists on a weekly or bi weekly basis and saying we can only honor quotes for several weeks at best. And so that means it falls on clients to understand that A, there's going to be lead time, but B, once you get a quote you kind of got to pull the trigger. And that's an unfortunate place to be in because you always want people to make informed decisions, but with the pricing influx they're there really isn't the time to wait because it's only going to get more expensive, unfortunately. And also it comes down to, well, if we can't get this specific part, if we're building A custom solution. What alternatives could we use that are still available, affordable, and will perform the task that's needed?
• 8:46
  Speaker B: We're in, we're in the place where we can actually, because we're dealing directly with commodities that we can. Things have to live up to our performance standards, but that we can swap things out and try different things. And we have to take testing infrastructure to know whether or not that's going to work. And we've made some of those changes, you know, without sacrificing quality. And not every player in this space has that flexibility.
• 9:11
  Speaker A: So Brad, at the beginning you kind of mentioned that because of product availability challenges as well as pricing challenges, many manufacturers have to try and perhaps re engineer their products to work around these, these shortfalls. How are some ways that the folks are trying to cope with that on the vendor side in the industry?
• 9:33
  Speaker B: Yeah, it depends. If you're a manufacturer, it depends on how bespoke your boards are and whether or not you've created some sort of specific motherboard that runs your equipment or you know, a backplane, A specific backplane or some other specific board that, you know, runs your inputs. We happen to be fortunate that we are, for the most part are using commodity architecture. Right. So we're working at the commodity level. And so because of that, in our case, we have been able to diversify our supply chain and test different brands of hard drives, memory, NVMe to make sure that, you know, we're, we're holding the performance, but we have potentially lower prices and you know, lower priced alternatives at the same performance because we're working at the commodity level. If you are again, somebody else who works in the broadcast space that has a very specialized board or chip, you're going to have less flexibility. Unless you've foreseen this issue and have already started developing multiple sources for your very specific board or chip that you're putting into your products. And so I, you know, while this has been a challenge for us, that, that's a particular challenge for, for those other manufacturers.
• 10:55
  Speaker A: So a lot of the technologies that we've talked about are integral to AI, and obviously AI is a hot topic, whether it be analytical AI or generative AI. Right. Those are really hot button topics. How would you say that this, the, the shortage and this cost explosion is affecting how AI is either being built for users or even the adoption of that AI.
• 11:22
  Speaker B: Yeah. So, you know, it's interesting, Michael, and it'd be interesting to get your perspective on this too. What you, I think you're just starting to see the Shortage of if you're working off of an LLM or you're by mcp, you're just starting to see shortage in compute at this point, right? The short built all this capacity and there was a talk earlier this year that we were in an AI bubble, okay. And that these prices were collapsing. You were actually seeing anthropic get the profitability and you're actually here seeing when, when you're accessing these models, these foundational models versus via APIs, you're seeing massive amounts of token usage and actually now complaints of hey, I've got a capacity shortage on my computer. That's crazy. That's completely different conversation than we had at the beginning of the year, right. So I actually, I think we're going to start seeing if you're an enterprise level operator that's producing and archiving a bunch of content and you've got an AI strategy, you're going to be watching your token consumption and being careful that you're not going to blow through all of your tokens for the year, right in a quarter. And so those enter. I think enterprise operators are paying attention. Anyone that is taking, you know, that sort of AI technology to the mid market, like where we play is definitely making an architecture decision between do I use an AWS or an anthropic or you know, an open AI to drive my AI strategy or do I go on premise and do I try to do it with an on premise model, open source LLM and or you know, more of a simplified like what we're doing, you have to take with analytical AI, you've got to take all these complicated models and compress them down into something that will run on premise, right? So there's a spectrum here in terms of the levels of technology and what are the tasks that you're trying to accomplish.
• 13:17
  Speaker A: Just as you mentioned that manufacturers are looking into alternative technologies or I don't want to call them replacement parts, but different ways to engineer something to still be viable, viable in the market. Right now I think the same thing has to happen with AIs. Instead of saying we're going to throw everything at a ChatGPT or a Claude and saying big LLM, do everything. I think we're going to start, we're going to continue to delve into the agentic workflows and the harnesses model, right, where you have some big brain to orchestrate things, but then sub agents to carry out those orders and once those smaller models are carrying out orders, you don't need a huge LLM to Carry out that small order, you can do it on prem with smaller GPUs. And so I think you know, looking at things like openclaw or, or my, my favorite of the day, Hermes, I think those are, are what organizations are going to have to look at as viable alternatives to like you said, blowing their entire spend in one quarter because someone used the highest reasoning. Model 24 7.
• 14:22
  Speaker B: Exactly. I think you're going to see these agentic models get more and more task focused, more and more specialized, more open source, where you're going to have these vendors who are provid focused models with controllable costs. Right? Because what they, what they, what enterprises are going to try to do is stuff users back into the equivalent of the $20 a month. All you can eat that those folks like ourselves, right versus hey, it's, it's a wide open budget with all these tokens. And so yeah, you're going to see that, that splitting of the market between the specialized and then the broad generalist high powered models that are pretty expensive to run.
• 15:00
  Speaker A: And while you know, open source, you know, don't currently rival, shall we say, you know, neck and neck with the frontier models, I will say that it is very promising to know that this is kind of the first technology revolution where open source is almost keeping pace with the big companies. And I think that's also going to help keep that pricing down a little bit because if there is too much of a gap, it's not worth it, right? The juice isn't worth the squeeze. So we'll use these open source models and someone will inevitably productize that to the point where it's a viable option.
• 15:38
  Speaker B: And in the gaps of like how do you make this actually work in your products and make it be a great user experience, that's where it's really going to shine, right? It's how do you actually bring it into a workflow that is compelling and adds value to the user, whether they're an editor or a technical director or a broadcast engineer. It's really got to be the polishing of the customer experience that really matters here. And so that's where folks like ourselves on the backs of these open source models will earn our keep.
• 16:09
  Speaker A: You mentioned earlier that one of the potential viable alternatives from these big capex expenditures is maybe going OPEX and using the cloud, right? So what market segments or what folks, what groups of users in the media industry are probably the best folks to start looking at cloud first? Is it hey, I need to do AI. Those of you who need to do AI should think about the cloud or is it those who just need storage, maybe they should use the cloud first. Who do you think's the, the, the most ripe to move to the cloud?
• 16:39
  Speaker B: Yeah, so I have a couple ideas from my perspective and it has to do with the trade off of price and performance, right? And what, what are your really your needs and you obviously in our space if you're on premise it's more of a deterministic performance, right. It will perform when you want it to perform and it, and it does it at a known cost. So we're very, very good editor has historically been very good and most on premise broadcast providers like for deterministic performance. But if you're willing to release those bounds a little bit, there are parts of the market that you know, a SaaS based workflow can actually make a lot of sense for your use case. So for example Media asset management, if you're a corporation that just wants to manage finished video and make it externally available and maybe put some watermarking on may or may not need to be a system of record like we have in media, Cloud's a very good option for that, right. The performance is good enough, it's got the right levels of security and anyone can access it through, you know, standard corporate security. Entrepreneurs are finding ways to find those crevices in the trade off between price and performance and security that I find is like actually very, very interesting and I think for things like AI now if you've got an immediate asset management in the cloud, let's take that example and you are comfortable allowing your assets to go to a public cloud. That's a great solution for you if you don't mind the latency and actually getting your analysis back. It's actually really good. And so I think if you think of that, that that triad of performance, security and price as your trade offs there's multiple different options in there that you can, you can move yourself toward depending on what your needs are.
• 18:31
  Speaker A: If you had a crystal ball, do you see this going through 2027, do you see this being as something like Christmas 2027, we should be happy for 2028 or kind of. What are you seeing?
• 18:45
  Speaker B: Yeah, this is like warm up for my board of directors meeting. Look, there's some news. If you squint hard and you look at the China capacity like so China has a pretty aggressive industrial policy, right. And they're able to move large amounts of capital pretty quickly. Now it takes a long time for a fab to get up and running. But they're not starting from, like, from a dead, dead start, right? Like, they are. They're revving up their industrial machine here. And so if these manufacturers can get this capacity online, I think first of all, you're going to see a flattening of the curve, like fewer price increases. It's going to flatten out. I think they're going to remain elevated, though, through 2027. I think you need to be in the 2028 before you truly see this business revert back to the price levels that you saw in the past. And the market says that because if you look at like Micron stock, they went from $500 billion valuation to a trillion over 50 days or something like that. Right. And so the market is pricing in a longer, longer time to return to normalcy. So my hope is again that China and the South Korean manufacturers see the profits to be made and they kick up the investment machine and by 2028, we see some slacking off on the prices.
• 20:12
  Speaker A: Brad, thank you so much for your insight today and all the facts and figures you've shared, and you're a fantastic illustration. Thank you so much for sharing it with us and we'll talk to you soon.
• 20:22
  Speaker B: All right, thanks, Michael.
• 20:24
  Speaker A: Thanks for watching. Broadcast to post. Don't forget to follow Keycode Media on social and contact us about your projects@kicodemedia.com.