1 00:00:00,356 --> 00:00:03,849 Hello and welcome back to Data Driven. This week, Frank is at KubeCon 2 00:00:03,977 --> 00:00:06,688 2026, and he is noticing a pattern. 3 00:00:07,233 --> 00:00:10,986 Kubernetes is about to make all of our artificial intelligence dreams 4 00:00:11,083 --> 00:00:13,039 come true. Now here's Frank. 5 00:00:14,451 --> 00:00:16,568 Hello, it's Frank here from 6 00:00:16,568 --> 00:00:20,353 Data Driven. And you 7 00:00:20,722 --> 00:00:23,577 don't know who I am. My name is Frank Lavinia, despite what the badge says. 8 00:00:23,689 --> 00:00:27,394 They got rid of the law. I'm here in the expo center 9 00:00:27,523 --> 00:00:31,310 before they start up. They're still setting up today. I don't think the 10 00:00:31,375 --> 00:00:35,164 expo center opens up for another hour or 2, but I figured I would 11 00:00:36,368 --> 00:00:39,723 do a quick livestream and I'm going to double down on my claim I made 12 00:00:39,723 --> 00:00:42,693 yesterday. Let me get my headphones off. I'm going to double down on my claim 13 00:00:42,693 --> 00:00:46,274 yesterday about how the AI world, the data world, saw how 14 00:00:46,402 --> 00:00:50,175 that has collided. I think the next logical step 15 00:00:50,239 --> 00:00:53,322 in the evolution of this is the collision of 16 00:00:54,269 --> 00:00:58,107 the AI world and the data 17 00:00:58,107 --> 00:01:01,753 world now colliding with DevOps. And from what I can tell, 18 00:01:02,861 --> 00:01:06,651 data operations, DataOps is going to collide slightly differently. I think you're going to 19 00:01:06,715 --> 00:01:10,264 have the monster that is data and AI or AI-related data 20 00:01:10,585 --> 00:01:13,926 implementations colliding with DevOps separately than from good 21 00:01:13,926 --> 00:01:17,523 old-fashioned DBA-type operations. The cool thing is 22 00:01:17,619 --> 00:01:21,377 that conferences tend to be a little bit ahead of the curve. So if you 23 00:01:21,554 --> 00:01:24,252 are wondering, like, where should I go next in my career or what I should 24 00:01:24,252 --> 00:01:28,085 do next? I would definitely investigate that space because regardless of 25 00:01:28,085 --> 00:01:31,837 what happens to one particular vendor because of a scandal, 26 00:01:32,238 --> 00:01:35,926 you're going to see AI is not walking back. AI is not going to slow 27 00:01:35,926 --> 00:01:39,437 down. AI, I think, is actually going to increase its footprint 28 00:01:39,437 --> 00:01:43,190 on-premise because the token 29 00:01:43,222 --> 00:01:46,316 prices, right? And that was the other theme that you see is control your token 30 00:01:46,348 --> 00:01:50,100 costs. A few years ago, you said token, people would have thought you meant 31 00:01:50,100 --> 00:01:53,596 Chuck E. Cheese or some kind of Bitcoin thing. But tokenomics now 32 00:01:54,526 --> 00:01:58,214 is a rising term and point of concern for many CFOs, 33 00:01:58,759 --> 00:02:02,431 definitely CFOs, probably also CEOs, as 34 00:02:02,527 --> 00:02:06,054 well as chief IT officers. So I think we're looking at 35 00:02:06,712 --> 00:02:10,079 a new world of AI where that is going to be a 36 00:02:10,159 --> 00:02:13,815 hybrid world first. And I say this out of 37 00:02:14,649 --> 00:02:18,400 personal experience, right? When I was at Microsoft, I was doing AI at Microsoft. 38 00:02:18,577 --> 00:02:21,592 I was all in on Microsoft, all in on the cloud. And then I realized, 39 00:02:21,624 --> 00:02:24,833 wait a minute, there's going to be some scenarios where I may not want certain 40 00:02:24,833 --> 00:02:28,091 types of data to leave my house, right? 41 00:02:28,524 --> 00:02:32,135 So then I set up my own home lab. I have the DGX Spark. 42 00:02:32,456 --> 00:02:36,259 And actually, fun fact, I actually now have a Qlik instance 43 00:02:36,548 --> 00:02:40,238 running in my home lab. And I tell you, it is, it has revolutionized 44 00:02:40,238 --> 00:02:43,047 how I use AI. And I'm not just saying that because I'm wearing a Qlik 45 00:02:43,047 --> 00:02:46,577 t-shirt. I'm telling you this as a home lab operator. And one of the reasons 46 00:02:46,577 --> 00:02:50,329 I like having a home lab It's not because it is fun, but it is 47 00:02:50,409 --> 00:02:53,934 also another home maintenance chore, right? But it lets me 48 00:02:54,191 --> 00:02:57,812 LARP, live action role play, as a CIO 49 00:02:57,972 --> 00:03:01,049 or a CTO, because that's de facto 50 00:03:01,706 --> 00:03:04,943 what I am at my house. Because I have a home lab, I have my 51 00:03:05,087 --> 00:03:08,676 family relying, the kids are relying on some of these services. My wife is the 52 00:03:08,676 --> 00:03:11,785 CISO because she works in information security. So she always 53 00:03:12,378 --> 00:03:15,936 lets me know when I do something wrong. But ultimately, I 54 00:03:15,984 --> 00:03:19,815 think one of the things that I've seen as a home 55 00:03:19,831 --> 00:03:23,278 labber is the impact of cost. If you're paying by 56 00:03:23,278 --> 00:03:26,981 tokens, the whole subscription thing where you pay $20 a 57 00:03:27,029 --> 00:03:30,380 month and you get unlimited or effectively unlimited, 58 00:03:30,893 --> 00:03:34,259 those days I think are numbered because when Anthropic IPOs, 59 00:03:34,660 --> 00:03:38,363 Wall Street is going to take a very different lens on a 60 00:03:38,427 --> 00:03:42,181 publicly traded company than a Privately funded startup. 61 00:03:42,534 --> 00:03:46,323 So I don't know that's going to happen days or weeks after the 62 00:03:46,531 --> 00:03:50,015 IPO, but it's going to happen within a set period of time 63 00:03:50,288 --> 00:03:53,964 after the IPO. So we're living in the golden age of 64 00:03:54,060 --> 00:03:57,817 cheap tokens because they're subsidized. That's coming to an end. 65 00:03:58,025 --> 00:04:01,653 And when that comes to an end, that $20 a month subscription could cost up 66 00:04:01,653 --> 00:04:05,474 to $200, right? And an economy like this where people are 67 00:04:05,474 --> 00:04:09,180 very anxious about AI taking over their jobs, Who is really going to pay 68 00:04:09,245 --> 00:04:12,975 that for their own personal use? I don't know. But I can 69 00:04:12,975 --> 00:04:15,765 tell you this, that the 70 00:04:17,505 --> 00:04:20,662 reason I joined Click was I saw that they had a vision and I saw 71 00:04:20,662 --> 00:04:23,931 that they had a vision to manage AI infrastructure, whether it's in the cloud, 72 00:04:24,350 --> 00:04:27,457 whether it's in on-prem, whether it's 73 00:04:27,811 --> 00:04:31,530 air-gapped, whatever. You as the administrator of the network, and I'm not, 74 00:04:31,546 --> 00:04:35,365 I'm not an IT admin, right? I haven't done anything remotely resembling IT 75 00:04:35,496 --> 00:04:38,831 admin work. Since the 1990s, and I've 76 00:04:38,831 --> 00:04:42,538 specialized into development, client development, mobile development, 77 00:04:43,021 --> 00:04:46,820 and then AI development. I know that, 78 00:04:46,965 --> 00:04:50,777 I know what it is like some, some, some principles never change, but the core 79 00:04:50,777 --> 00:04:54,201 thing here that I want to point out is these people are going to have 80 00:04:54,201 --> 00:04:58,000 to work together, right? Admins, AI builders, AI data scientists, data 81 00:04:58,049 --> 00:05:01,606 engineers, that sort of thing, as well as people who are consuming 82 00:05:01,606 --> 00:05:05,361 AI, whether they're end-level consumers or developer-level consumers. When I say 83 00:05:05,377 --> 00:05:08,517 end user consumers, people that just click on their phone or their app and then 84 00:05:08,517 --> 00:05:11,946 they use AI. When I say developer consumers, the people that get an 85 00:05:11,946 --> 00:05:15,167 endpoint and they can just consume, right? They hit the endpoint and they do what 86 00:05:15,167 --> 00:05:18,788 it is. What really matters and what Qlik makes really easy 87 00:05:20,134 --> 00:05:23,867 is the ability for administrators to take control 88 00:05:24,059 --> 00:05:27,680 over their IT infrastructure through AI. 89 00:05:28,369 --> 00:05:32,199 Because you can route based on content, based on query, based on 90 00:05:32,231 --> 00:05:36,080 certain metadata about a request or just based on the size of the request, 91 00:05:36,465 --> 00:05:39,593 all these sorts of things you have control. You can route it to a public 92 00:05:39,593 --> 00:05:43,299 API. You can route it to a server running locally. I think we're 93 00:05:43,299 --> 00:05:47,149 seeing the golden age of this is about to— one golden age is 94 00:05:47,149 --> 00:05:49,235 going to end and a new one is going to start. And I think we're 95 00:05:49,379 --> 00:05:53,197 entering the golden age of on-prem IT and self-hosted AI. 96 00:05:53,806 --> 00:05:57,576 Just regardless of what Sam Altman gets mad about, people using RAM 97 00:05:57,576 --> 00:06:01,268 for local AI use cases. But those days, I really think you're going to 98 00:06:01,284 --> 00:06:04,816 see kind of these orchestrated systems where low 99 00:06:05,057 --> 00:06:08,654 memory consumption AI models are going to make 100 00:06:08,654 --> 00:06:12,459 decisions automatically for agents as to where they 101 00:06:12,587 --> 00:06:16,409 get their intelligence from. So that's all I got. I'll walk you around real briefly 102 00:06:16,939 --> 00:06:20,519 around the floor as we have. I'm going to do my extra 103 00:06:20,519 --> 00:06:24,196 special tool here. We're at the Salt Palace, I think it's 104 00:06:24,196 --> 00:06:27,658 called. Here, and we have the IBM booth, Everpure. 105 00:06:28,400 --> 00:06:32,123 The Qlik booth is just over there. I'll be over there today, 106 00:06:32,429 --> 00:06:36,056 booth number 276, I believe. So stop by and say hello. 107 00:06:36,152 --> 00:06:38,086 In the meantime, I'm signing off.