# AI Gateway Token Traffic Growing 140X Year Over Year | Aghi Marietti

Markdown copy of https://realalpha.app/v/lEcT1VvKIis/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

Aghi Marietti argues that deterministic AI gateways are becoming essential checkpoints for securing agents and controlling enterprise token costs.

- Show: MTS
- Hosts: [Sophia Dew](https://realalpha.app/people/sophia-dew/)
- Guest: [Aghi Marietti](https://realalpha.app/people/aghi-marietti/) is CEO and co-founder of Kong.
- Published: Oct 23, 2026
- Length: 27 min

## The best minutes

- [0:55](https://realalpha.app/v/lEcT1VvKIis/?t=55) [AI gateways give enterprises a checkpoint to stop agent attacks](https://realalpha.app/s/ai-gateways-give-enterprises-a-checkpoint-to-stop-agent/) (7 min)
- [8:06](https://realalpha.app/v/lEcT1VvKIis/?t=486) [Enterprise AI is spreading across models as token use surges](https://realalpha.app/s/enterprise-ai-is-spreading-across-models-as-token-use-surges/) (8 min)
- [15:18](https://realalpha.app/v/lEcT1VvKIis/?t=918) [AI cost controls can tie token budgets to business workflows](https://realalpha.app/s/ai-cost-controls-can-tie-token-budgets-to-business-workflows/) (3 min)

## Also worth watching

- [20:17](https://realalpha.app/v/lEcT1VvKIis/?t=1217) [Enterprise AI adoption will arrive in step changes](https://realalpha.app/s/enterprise-ai-adoption-will-arrive-in-step-changes/) (4 min)

Skip the sponsor reads, the token-budgeting exchange, and the closing product discussion of model routing and Volcano.

## Summary

Marietti is CEO and co-founder of Kong, which he says operates one of the largest enterprise AI gateway fleets. He describes Kong’s gateways as a way to connect APIs, models and agents, and to observe, rate-limit or block AI traffic.

Marietti argues that AI agents create new security risks and that gateways can give enterprises a deterministic point to observe and block traffic. He says enterprise AI use is spreading across more models and providers, while token volumes are rising even as costs per token fall. He describes Kong’s cost-management tools as a way to map spend to teams and workflows, set granular budgets and connect usage to outcomes. He sees wider agent deployment, better AI identity and registries, and clearer evidence of ROI as priorities for enterprises.

### Minute by minute

- [0:44](https://realalpha.app/v/lEcT1VvKIis/?t=44) **Agents expand the attack surface.** Marietti says attacks are becoming more frequent and sophisticated as attackers use models. He argues that agentic coding and enterprise agents create a largely non-deterministic attack surface, leaving activity defenders may not see. He says the risks are reaching Fortune 500 boardrooms.
- [2:57](https://realalpha.app/v/lEcT1VvKIis/?t=177) **Gateways give defenders a control point.** Marietti says defenders need security products, but AI security is also an R&D and infrastructure problem, not just an IT issue. He describes gateways as deterministic checkpoints that can provide observability and boundaries for AI traffic. He says a kill switch that can stop traffic is a major missing capability.
- [5:34](https://realalpha.app/v/lEcT1VvKIis/?t=334) **Gateways connect token and execution paths.** Marietti distinguishes the token path, where enterprises consume models, from the execution path, where agents use APIs and MCP servers. He says Kong brings these paths together through gateways that can inspect traffic and rate-limit, throttle or block it. That can stop unwanted spending or prevent an agent from making an unauthorized API call or payment.
- [8:18](https://realalpha.app/v/lEcT1VvKIis/?t=498) **Enterprises are adopting more models.** Marietti says 1,200 people attended Kong’s summit, including infrastructure, AI and platform leaders, and that platform engineering is taking a larger role in governing AI traffic. He says open-weight models now account for about one-third of enterprise model usage, while sophisticated organizations use dozens of models and some Kong customers govern hundreds.
- [11:17](https://realalpha.app/v/lEcT1VvKIis/?t=677) **Most enterprise work remains unagentified.** Marietti says the summit’s Moving AI theme came from customers describing their work. He says Kong’s previous-year bet was agentic enterprise, but most companies still mainly use chatbots and coding harnesses. He estimates that 95% of enterprise work has not been agentified.
- [13:48](https://realalpha.app/v/lEcT1VvKIis/?t=828) **Token use could reach quadrillion scale.** Marietti expects more enterprise workflows to become agentic and model counts to keep growing. For the enterprise traffic Kong can see, he estimates usage at about 500 trillion tokens, moving toward a quadrillion, with growth of 140x year over year. He says falling token prices will not remove the need to secure, govern and control costs as inference use expands.
- [15:40](https://realalpha.app/v/lEcT1VvKIis/?t=940) **Cost tools map spend to outcomes.** Marietti says Kong’s AI cost-management product was among the most requested offerings after companies faced budget pressure. He says Kong can use traffic metadata to map costs across agents, APIs, teams and repositories, then show workflow ROI. Customers can set budgets and block usage when a limit is reached, down to an individual or repository.
- [18:47](https://realalpha.app/v/lEcT1VvKIis/?t=1127) **ROI matters more than forecasts.** Marietti says predicting token use is difficult, but the bigger question is whether that spending creates a positive economic outcome. He points to productivity gains in R&D and engineering, while noting that enterprises have not automated all their general and administrative work. He says the next step is connecting token consumption to ROI.
- [20:17](https://realalpha.app/v/lEcT1VvKIis/?t=1217) **AI deployment grows in step changes.** Marietti says enterprise AI adoption can take months to reach production, then token use can spike when a workflow goes live. He expects continued multi-model use and growing demand for registries and agent identities. He says token vaults can let agents use credentials without seeing API keys or billing cards.
- [23:43](https://realalpha.app/v/lEcT1VvKIis/?t=1423) **Model routing lowers inference costs.** Marietti says intelligent routing can move requests from higher-cost to lower-cost models while preserving the same output, which he says can save large enterprises hundreds of millions of dollars annually. He describes Volcano as Kong’s early bet on an agentic cloud for building agents, complementing its work on APIs, agents and MCP.

## Transcript

MTS, hosted by Sophia Dew, with Aghi Marietti.

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

[0:00](https://realalpha.app/v/lEcT1VvKIis/?t=0) The numbers of tokens, by the way, is growing exponentially. For what we can see, we are running about 500 trillion tokens going to quadrillion soon because it's growing 140x year over year. And that is like 100 times bigger than a year ago. So the amount of tokens enterprises are using is insane. Welcome back to MTS. Today, we are joined live with Augie Marietti, the CEO and co-founder of Kong, who also goes by the name API Godfather. I like this. I like how you actually ended up getting that title. But, Auggie, welcome to the show. Yeah, thanks for having me, Sophia. It's great to be here. Thanks for joining.

[0:41](https://realalpha.app/v/lEcT1VvKIis/?t=41) Great to have you on again. I know last time you were here, we talked a lot about what you guys are doing, what you guys are doing on Kong, all the infrastructure you're building. But why I'm excited to have you on today, I feel like everything that happened within the last 24 hours is, I don't know, quite scary or novel. And I just wanted to get your thoughts on all of the security incidents that are happening with PHP, with Railway, all of this stuff. What's unique about what's happening now? Like, what's something that the mainstream public should be aware of? So, yeah, as you say, in the last 24 hours, you know, we had P2P, two shells, a bunch of gateways got compromised.

[1:20](https://realalpha.app/v/lEcT1VvKIis/?t=80) And it's increasing. The frequency is increasing. i think the attackers are just becoming more sophisticated by leveraging models and the defense is not is like you have now nationwide type of attacker powers by leveraging models and the the companies are not uh they haven't like defended their self through a model security models i think that's that's what's happening and you know all the agentic enterprise and the gentix coding is It's creating a whole new real estate, a whole new surface that is at times non-deterministic. Most of the time it's non-deterministic. And you don't know what actually is in between those.

[2:06](https://realalpha.app/v/lEcT1VvKIis/?t=126) And that creates a lot of empty spot to be attacked and to be more vulnerable than before. I think at Kong, we run one of the largest enterprise AI gateway fleet. And not only we show identity to agents, but we also we do re-limiting, blocking and, you know, kind of having a quill switch. And this is this is becoming a large risk into boardrooms of, you know, the Fortune 500s. Yeah, no, this is really fascinating. I mean, you have on one end, you have the attacker versus the defender. And for a lot of people, especially if you're using open models, if everyone has the same model and access, then really it's who can pull together resources to pay for compute.

[2:50](https://realalpha.app/v/lEcT1VvKIis/?t=170) And as you alluded to, a lot of these hackers aren't necessarily an indie hacker in their basement just trying to attack a company. Oftentimes you have entire countries that are pulling resources together to be able to attack American companies. And it's getting into some of these national security questions as well. And so let's talk a little bit about what the defender needs. You alluded to a few things that around kill switches, around permissions, around tracking. What does the defender need to be successful here? Yeah, at the core, you know, we're not a security company, but definitely they need to.

[3:26](https://realalpha.app/v/lEcT1VvKIis/?t=206) So you can divide it in three ways, at least how we see it. Obviously, I need to have a very strong security providers, like the usual suspect for security specific, call it endpoint security. That's evolving. So there is one piece of like, let's, the season is to buy their own security products. Then, but now it's becoming, it's not just an IT problem where you buy identity companies and IDP and all these things were used to secure and authenticate a workforce, but now it's becoming an R&D problem. What is changing is also becoming an infrastructure. problem. It's not just an IT or a security type of problem because the agents are more of a code and an infrastructural problem. So that's the thing where you have to have a very strong advanced security platform. You'll have to think differently from your IT posture where it's not just IDP providers or all this type of, you know, Okta, all this kind of thing. They're still there for the workforce, but the new workforce is going to be agents.

[4:30](https://realalpha.app/v/lEcT1VvKIis/?t=270) And three, infrastructure is where it's the most vulnerable. That's why we're seeing a lot of attacks on infrastructure, like software infrastructure, because that was not ready for an agentic world. And so what we are seeing, and we didn't like, naturally, like we got pulled into,

[4:43](https://realalpha.app/v/lEcT1VvKIis/?t=283) **Aghi Marietti:** is that gateways are becoming very strategic because in the code, gateways are like deterministic AI choke points.

[4:48](https://realalpha.app/v/lEcT1VvKIis/?t=288) And by being a deterministic AI checkpoint, no matter what happens, you have boundaries and you can block if you can capture. So what do they need? the AI observability to see what's going on, how it all connected. But what is really hard is to have the kill switch. There is no, most of the enterprise or nation, they don't really have a kill switch today. That's, I think, is the next frontier. Yeah, that's what the next frontier will be. This is really interesting when you say around the gateway. I know that's what you guys are quite famous for as well and what a lot of enterprises use, Kong for. And so when it comes to the actual gateway, how you say that's a point, a very specific point where you can actually determine what's going on, where you start to assess, like, where, like, especially for a very long horizon tasks, you can actually assess it during that switch, what's happening within the gateway. Where does the kill switch come in? Why is that a really hard problem?

[5:44](https://realalpha.app/v/lEcT1VvKIis/?t=344) So there are different types of gateways, and I touch, there are two paths forming in an enterprise, right? There is the intelligent path, which is the token path, and people buying inference, consuming models. And there is the execution path, which is agent consuming APIs, MCP server to get stuff done. So actually, there is two paths in enterprise, token path and execution path. Pong brings them together. In the execution path, you have APIs, And so you have API gateway. That's what we started with. In the token path, you have AI gateways, MCP gateways, and agent gateways. Assuming you have a fleet of all gateways inside enterprise for in and out AI traffic, that obviously proxy API requests, MCP2 calls, tokens, and you have all sorts of observabilities.

[6:33](https://realalpha.app/v/lEcT1VvKIis/?t=393) Where the kill switch comes is where it can be a cost problem, it can be a security problem, where you can intercept that traffic, whether it's human-generated traffic or a gently generated traffic and you can just do rate limiting or throttling or just block and within a second the gate close and that that transaction doesn't happen anymore what is an agent mcp that's the uniqueness right they they are i go back to ai chuck points and you have the checkpoint you have the gateway so there's no other technology in the world other than gateways and so that's why they're becoming strategic in the api world you have gateways for powering mobile applications between the backend microservices and your mobile apps that's fine but in the i world token paths its cost and an execution path it could be risk if it's not secure or cannot be blocked because agent can go and ping api and do some damage or make some payment they should not have i think that these two things are getting merged um and uh yeah it's a new paradigm is a new architectural shift where you just gotta deploy all sort of gateway as quickly as possible to make sure you can block when you have to block.

[7:41](https://realalpha.app/v/lEcT1VvKIis/?t=461) I want to shout out our sponsor, Lovable. You know the app you've been meaning to build? The internal tool, side project, or product you'd otherwise lose a weekend to? With Lovable, you get software you can ship today. That includes an editable code base you can inspect and change, plus two-way GitHub sync. Lovable also handles the backend and infrastructure. including managed Postgres, auth, storage, hosting, payments, and more, so you don't have to wire it all up yourself. And through Lovable's MCP server, you can create and deploy your project directly from the agents you already use. Turn ideas into software people love with Lovable.

[8:15](https://realalpha.app/v/lEcT1VvKIis/?t=495) Lovable.dev. Now, on to the episode. You're starting to see this new paradigm. You're having, I mean, traditionally, it's humans who are calling the different APIs, and now you have agents who are calling the different APIs, which opens up a whole new slew of potential problems or architectures that we haven't really thought about. I want to get your thoughts because I know what recently just wrapped was the annual API AI Summit. This wrapped a few weeks ago that Kong does. This is your guys' biggest in-person event ever. I want to talk to you specifically about your keynote, what you shared on the stage, because I think it talks a little bit about where we are right now and where the world is heading to.

[8:55](https://realalpha.app/v/lEcT1VvKIis/?t=535) So can you just summarize some of the high-level learnings there? Yeah, we did our annual user conference. 1,200 people was a mix of infrastructure leader, AI leader, and platform leaders. We have this concept of be AI movers. It's like Salesforce are trailblazers, and you create a movement, right? And we have this concept of AI movers. You're not really moving tokens or moving API requests. You're moving AI. But obviously, everybody can move AI, But then you have token maxing, security risk. You have to move AI safely, secure, and economically efficient. That's the only way to move AI, to make it long-term positive outcomes.

[9:37](https://realalpha.app/v/lEcT1VvKIis/?t=577) So that's how we help our customer. That's how we thought there was a general thinking where platform engineering is going back in charge into the global 2000s to govern AI traffic. Because when the traffic went kind of out of control, then governance comes back in. And so these teams, this type of budget gets back in charge. So that's another change we've seen in the last year. Another one is models. So there's all open-weight model. Probably you've seen a lot talking about it. There was no open-weight model a year ago at our last summit. Like just talking or what, today enterprises are using like one third are now open-weight model usage.

[10:25](https://realalpha.app/v/lEcT1VvKIis/?t=625) But the crazy stuff is, on average, like a sophisticated, advanced organization has at least dozens of models inside. And some of them, we govern hundreds of models, all type of it. It's not like, oh, one big one used OpenAI and Entropic. No, they have dozens and they use also different inference providers. So it's going to this like what happened in the cloud was one model late and then it gets microservices distributed and then it got multi-cloud. In the AI era, it's the same. We started with one model to consume more model to get a bit of closer than open wake model and even going to tune and train our own models with our own proprietary data.

[11:04](https://realalpha.app/v/lEcT1VvKIis/?t=664) And so it's a perforation, a large, like an organism of this nervous system of dozens and a hundred of models. That's another change that happened, that it wasn't even discussed or a problem to govern about a year ago. Wow. Yeah, I never even thought about that. Just the fact that all the things that you have to think about and connect and govern today in 2026 just wasn't totally true back in 2025. So just the speed that things are changing, especially in enterprises, is something we haven't totally seen before. I do want to ask specifically about your keynote. You had the title, Moving AI. I want to understand, where did that title come from?

[11:42](https://realalpha.app/v/lEcT1VvKIis/?t=702) It was like, you know, we're on a mission to build the connectivity layer of AI. And it's very, obviously, it's engineering, it's a product. But to connect humanity around it, like what actually you do, you're actually helping the world move AI. And so it comes actually from our customers to come together. to the idea, like, you know, we are moving AI. And I was like, wait a sec, what did you just say? I said, that's amazing, like, that's it. Like, if you go in a bar and meet somebody and say, oh, I'm, you know, platform engineering, you know, I'm really limiting APIs and token control cost. But if you say I move AI, like, it's very, it's like, okay, tell me more about it.

[12:23](https://realalpha.app/v/lEcT1VvKIis/?t=743) It's much higher, I think, higher humanitarian. so that that resonated very well from with the folks in the audience and so we decided to kind of make the theme of this year i like that this is the main theme of this year how did this differ from last year's theme what was last year's theme what was this year's theme what specific trends do you notice that inform what the theme of the conference will be so last year we all the bet we made actually happened a year later but it happened faster than what we thought last year thing was agentic enterprise agentic enterprise nobody people weren't quite talking yet but like now it's all over now enterprise are like one percent to agentic enterprise what they really use is the chatbot and coding harness that's it a little bit of agentification in support business unit but 95 95 of the real enterprise has not been agentified which is the whole opportunity but we thought we say like we were going to the age of ai connectivity that was the big thing Because AI had to, in the digital world, you know, was networking.

[13:23](https://realalpha.app/v/lEcT1VvKIis/?t=803) In the AI world is connectivity. And we're going through different models, different protocols. And we needed help to connect models, agents, and APIs together. So that was the big bet. Enterprises were starting to become agentic in some work streams. And companies were going multimodals. We did not envision open-way, but we envisioned a multimodal world. So that's kind of the bet we made a year ago. And then, like, here we are. Wow, that's interesting. So you guys had the certain bets a year ago that you were too early on a little bit or you just were able to predict before. What's that right now?

[13:57](https://realalpha.app/v/lEcT1VvKIis/?t=837) So if you were to have to predict what maybe next year, 2027's big API AI summit would be about, what trend do you see happening that isn't totally true right now? Yeah, it's the same like direction of a genetic enterprise, but more pieces of organization will be actually identified. So that's one, meaning not just support chats, but it will be other work stream in our workflow. The numbers, I think, of models will keep growing exponentially. The numbers of tokens, by the way, is growing exponentially. Like we, for what we can see, because some enterprise are here gapped, we don't see everything.

[14:34](https://realalpha.app/v/lEcT1VvKIis/?t=874) **Aghi Marietti:** But for a subset that we see, we are running about 500 trillion tokens, going to quadrillion soon, because it's growing 140x year over year.

[14:44](https://realalpha.app/v/lEcT1VvKIis/?t=884) And that is like 100 times bigger than a year ago. So the amount of tokens enterprise are using, it's insane that the cost per token is going down. And I think next year is going to go even more pragmatic that the cost per millions of tokens going down. Inference will still be exploding. So we're going to live in this enterprise war that is going to be more multimodal, cheaper tokens, cheaper level of intelligence, and more about how do we secure it? How do we govern it? How do we cost control it? Those are going to be, I think, the next thesis to making a genetic enterprise an economically efficient outcome.

[15:23](https://realalpha.app/v/lEcT1VvKIis/?t=923) Yeah, how do you cost control all of these tokens? I think that's one of the big areas of concern for a lot of enterprises right now. I mean, as you mentioned, even though the costs of tokens are going down, which can be a good thing, the overall costs that enterprises are still spending and their token spend is just going up. So despite, it's like a Jevons paradox that you're starting to see with token spend. And so what are you hearing about some of the concerns around cost management? So we preview our AI costs or token cost management products at Summit are going GA in a few weeks. That was the most, I think a top two ask or excitement was around AI cost management was that one.

[16:08](https://realalpha.app/v/lEcT1VvKIis/?t=968) And the reason is, you remember like in April, May, we went through this token maxing and everybody ran out of budget by the time it was May for the fiscal year. That was a big thing about Uber running out of budgets. And we are never, you know, we weren't that different on our own. So the beauty of how we approach the problem is not just a dashboard because everybody can show you AI cost dashboards. But because we intercept traffic around all these entities between MCP, agents, APIs, we are able to have metadata that nobody has on how actually costs get mapped, what are the margins, but how they get used all the way down to a repo.

[16:44](https://realalpha.app/v/lEcT1VvKIis/?t=1004) And then you can apply ROI cards. You can see the return on investment of this type of workflow, what outcome it created. So there is this level of dashboard, which is one piece, which is all the AI cost management. but also the granularity of connecting and mapping the different costs across the organization between the token part and the execution part. And then you can block. And that's unique to call. Because if you add a budget to, I don't know, R&D in France to use this amount of tokens,

[17:17](https://realalpha.app/v/lEcT1VvKIis/?t=1037) **Aghi Marietti:** you can then block when the budget gets hit all the way down to individual that maybe was consuming the most token

[17:23](https://realalpha.app/v/lEcT1VvKIis/?t=1043) and they weren't very productive, but all the way down to the repo or to the pull request. It can go super granular. And so that level of stack was very excited. And the beauty of this, it was not just a platform. Even CEOs of public companies were in need of showing this to their board because they were like, okay, how is all this cost coming together? And so that's something our enterprise customers are very excited about. We'll continue monitoring right after this message from our sponsors. 11labs, AI that communicates at human level across every channel and modality. 11labs.io slash mts Scale your startup on Neon.

[18:04](https://realalpha.app/v/lEcT1VvKIis/?t=1084) Millions of developers and startups have already chosen Neon for their backend. Start on the free plan or get up to $100,000 in credits for your startup at neon.com slash mts. Special thanks to our sponsor Kong, the AI connectivity platform. Connect APIs, LLMs, agents, and systems with serious security and governance. konghq.com. Finally prove what your AI investment is worth with Lairdyn. Head to Lairdyn.com today to start measuring your AI's bottom line. MongoDB, trusted to run critical applications by 75% of the Fortune 100, the intelligent data platform for the era. Apply today for access to up to $15,000 in Atlas credits, dedicated technical support, and exclusive events and go-to-market opportunities.

[18:48](https://realalpha.app/v/lEcT1VvKIis/?t=1128) mongodb.com charge b billing and monetization infrastructure for the ai era charge b.com i am curious though what's the hardest part about this it seems like when you're actually tracking ai spend it's incredibly difficult to predict how much it's going to cost it's it's it's now budgeting is getting easier to budget is your eyes getting more difficult to predict. However, I don't know if it's a predicting problem or it's an outcome problem in the sense. If I know that we're going to get a positive outcome by spending all these tokens at some point soon, then I don't need too much about the forecast of token spend as long as below the outcome value because I know eventually that will happen.

[19:39](https://realalpha.app/v/lEcT1VvKIis/?t=1179) I think now the issue is that I can still spend tokens. But I still don't have identified a positive economical outcome of what I'm getting out of that. Yes, there is R&D productivity, there is engineering productivity, but an enterprise, have they automated the whole, you know, G&A functions? No. So that, I think that's the next step. It's more like, how do we create ROI on these token consumptions? and since we can forecast in the meantime we just budget the hell out of it yeah how do we actually budget and predict what the roi is going to be i did want to ask i think something that we talked about throughout this interview is just the scale that this is happening what's unique about kong is because you guys are talking to enterprises all of the time you almost are at the forefront of being able to predict pretty well about where all of this is going to go and so just wanted to at your thoughts and just like big picture, where are we when it comes to AI usage in enterprise?

[20:34](https://realalpha.app/v/lEcT1VvKIis/?t=1234) How do you see this evolving over the next few years? Yeah, it's funny, like enterprise were not cool. And now with AI, enterprise went back cool. Enterprise sales, enterprise go to market. It's, yeah, four years ago, it was all like PLG. And now it's like with AI, everybody's attacking on enterprise. I think because the real, and I touched your question, the real, I think the real that question like all these all companies growing very fast kind of buying and reselling inference to each other whatever would they actually be in business 10 years from now it's not clear but i do know that you know united will stay in business 10 years from now and and so and coca-cola will be in business 10 years from now right so that's why enterprise i think are becoming first of all because they're the largest are in this span in the world but but but also you have a durable grow of helping them move to a genetic enterprise and stay in business.

[21:29](https://realalpha.app/v/lEcT1VvKIis/?t=1289) In this genetic enterprise, as I mentioned before, the scale of token usage is not linear because they're now going multi-models and open-width models to lower their token span with modern routing, intelligent routing, all of those things. It's not linear, though, like you have startups where everything grow and the token spot growth.

[21:52](https://realalpha.app/v/lEcT1VvKIis/?t=1312) **Aghi Marietti:** They're more like step functions. Like it might take six months to put an AI on a gentle workflow in productions.

[21:59](https://realalpha.app/v/lEcT1VvKIis/?t=1319) When it does, boom, it's just you have 50 million, I don't know, mobile app user or whatever, and it just spikes. So that's the interesting thing. As our AI gateway goes in production into we have a thousand plus enterprises, you can see very large, large step functions. That's one. Like when they do figure it out, workflow, it really spikes. But it just takes time to move technology, new technology into, you know, large, long, multi-decades organizations. So that's one. Big token spend, what happens? Open weight, multi-model, still multi-cloud, that doesn't change. A huge need for registries and identities.

[22:39](https://realalpha.app/v/lEcT1VvKIis/?t=1359) Okay, I have all this agent, I have all this MCP, I have all this API, I have all these models. I need registry to see as a system or record how it all fits. That's a new growing needs. And identities. The way you identify agents, authenticate an authorized agent is different than the workforce. That's why I think IDPs are very important, all these IDP companies, but they will be less important in a genetic enterprise as enterprises are now requiring things like tokens vaults. So you can extend that token to an agent and the agent will never see your API keys, will never see your credential. You'll never see your billing cards because nobody want to give those to them because they could do some damage.

[23:23](https://realalpha.app/v/lEcT1VvKIis/?t=1403) So the way you do registry, cataloging and identity of all these AI assets, it's changing a lot. And IDPs is one of the vulnerable spots that I don't know how valuable it would be in the very long run. Security is changing with these tasks before. Yeah. And those are the things we're seeing. Yeah, no, it's really interesting to get your perspective on that, on what's being seen, because, as I mentioned, you are at the forefront. You get to see what's actually happening. One last thing I wanted to ask you, Augie, before we actually head out. We just had Marco on a few days ago talking about Volcano.

[23:57](https://realalpha.app/v/lEcT1VvKIis/?t=1437) I know you guys launched a ton of new products this month. Pretty exciting. What are you most excited about that you want people to know? I think model routing and intelligent routing is a massive technology that we have at Kong that we see how enterprise can move token expenses from high cost token per million to lower cost token per million and get the same intelligent output. And that at scale, that saved hundreds of millions of dollars on an annual base into a large enterprise. So seeing that happening into that scale of cost saving is very gratifying it for the business that we support, all these customers.

[24:42](https://realalpha.app/v/lEcT1VvKIis/?t=1482) I think that it's like the underrated routing in the enterprise. because of the scale and the amount of money you save is very gratifying when you see it. Volcano, it's our kind of bet in the office that the CTO led by Marco to be on the other side of the house. We're always on the producer side of the house, APIs, agent MCP. This is to be on the other side to actually agent concentra and agent building. So we build Volcano's agentic cloud to build all of that. That's like a bet. And now you can all come together, right, between supply and demand. And so I'm very excited about that. It's very early.

[25:19](https://realalpha.app/v/lEcT1VvKIis/?t=1519) We'll have to do, you know, a big lunch and all of that. So those, I think, are, I think, things are top of mind, right? Very exciting. Well, Augie, really wanted to thank you for joining us today on MTS, actually getting the API godfather to tell us about where we are and where all of this is going. Such a pleasure. So thank you so much for joining. Thanks, Sophia. Thanks. Thank you. A huge thanks to MTS sponsors. Blitzy, autonomous software development for enterprise code bases. Ship 5X faster. Blitzy.com. AdQuick, make your brand a billboard. Out-of-home advertising as easy to scale as digital.

[25:55](https://realalpha.app/v/lEcT1VvKIis/?t=1555) AdQuick.com. Arena, measuring AI performance in the real world. Arena.ai. Support for the show comes from VCX, the public ticker for private tech. The U.S. stock market started history's greatest wave of wealth creation. From factory workers in Detroit to farmers in Omaha, anyone could own a piece of the great American companies. But today, our most innovative companies are staying private longer, which means everyday Americans are missing out. Until now. Introducing VCX, a public ticker for private tech. Visit GetVCX.com for more info. That's GetVCX.com. Carefully consider the investment material before investing, including objectives, risks, charges, and expenses. This and other information can be found in the innovations fund prospectus at getvcx.com. This is a paid sponsorship.