Apr 2, 2026 · 1h 20m · neon-show
Why "Boring" Infrastructure is the Best Path to a $60B Company | Manish Jindal, Cloudflare & Arize
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The Neon Show, Arize President and former Cloudflare executive Manish Jindal shares essential operational frameworks for scaling enterprise infrastructure from early product-led growth to multi-billion-dollar public valuations. He breaks down why foundational backend plumbing outperforms volatile application layers, how dedicated observability bridges the production AI reliability gap, and the discipline needed to build enduring tech companies.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
Siddhartha as informed peer 3.8 Guest teaching 5.3 Guest disagreement 1.6 Siddhartha pushing back 1.8
speaking balance: gold is Siddhartha, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 16:15 Pushing back on winner-takes-all infra narrative
Manish directly disputes the host's premise that infrastructure markets are winner-takes-all, citing multi-billion dollar coexisting giants across observability, security, and databases.
Hardest push from Siddhartha ▶ 18:39 Host challenges enterprise software moats citing HubSpot dropSiddharth challenges Manish's claim about data-owning enterprise apps by citing HubSpot's dramatic stock crash as counterevidence.
Biggest teaching moment ▶ 58:15 Explaining why traditional APMs fail on AI agentsManish educates Siddharth on why green Datadog dashboards miss garbage hallucinated outputs in non-deterministic agent workflows, requiring dedicated evaluation layers.
Siddhartha holds their own ▶ 1:15:18 Host details fund M&A data across multiple infra companiesSiddharth demonstrates deep domain expertise by citing specific valuation multiples and buyer profiles for five portfolio acquisitions including Requestly, ZenDuty, and Logic.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Siddhartha as informed peer | Guest teaching | Guest disagreement | Siddhartha pushing back | Why |
|---|---|---|---|---|---|---|
| Manish Jindal's Career Journey and Startup Selection Framework | 3 | 4 | 1 | 1 | Siddharth opens by highlighting Manish's impressive track record across Splunk, Cloudflare, and Arize. Manish elaborates on his core framework for picking companies, emphasizing large TAM and founder conviction. | |
| Academic Background and Immigrating to the United States | 2 | 3 | 1 | 0 | Siddharth asks about Manish's academic and early career background. Manish shares his journey from engineering in India to Dell, consulting, and landing at Splunk via personal networks. | |
| Leaving Splunk Early and Core Career Lessons | 4 | 5 | 2 | 2 | Manish reflects on the mistake of leaving Splunk early due to a difficult boss and shares the lesson of sticking with high-conviction infrastructure winners. Siddharth reinforces this with the picks-and-shovels analogy and cites Databricks. | |
| Emerging Opportunities in AI Middleware, Observability, and Inference | 4 | 6 | 2 | 1 | Siddharth asks where real revenue will grow over the next decade. Manish outlines why middleware, AI observability, evals, and inference will be durable while applications face rapid developer churn. | |
| Infrastructure Moats and the Shift to Natural Language Interfaces | 6 | 6 | 3 | 5 | Siddharth pushes back against the plumbing thesis, arguing winner-takes-all dynamics in software and citing HubSpot's 80 percent stock drop. Manish counters by showing multiple billion-dollar infra players can coexist and explains natural language replacing traditional UI workflows. | |
| Cloudflare's GTM Playbook: From PLG to Enterprise | 3 | 6 | 1 | 1 | Manish breaks down Cloudflare's sequential GTM strategy, explaining how they resisted boiling the ocean and turned self-serve into an enterprise tier starting with a phone call from Bain Capital. | |
| Scaling Cloudflare: Milestones from $50 Million to IPO | 4 | 6 | 2 | 2 | Siddharth notes how modern founders prematurely rush upmarket. Manish details why Cloudflare deliberately turned down massive customers like Apple early on to protect product roadmap integrity. | |
| Defining and Identifying Genuine Product-Market Fit | 3 | 5 | 1 | 1 | Siddharth asks how to identify true product-market fit. Manish gives concrete historical examples from Splunk's log searchability, Cloudflare's web speed metrics, and Arize's agent observability. | |
| Joining Arize: Conviction in AI Observability and Operator Mindset | 3 | 5 | 1 | 1 | Siddharth asks why Manish joined Arize instead of starting a company. Manish explains his previous insight from Insight Partners, the open-source traction of Phoenix, and his preference for decade-long operator commitments. | |
| Structuring Cloudflare's High-Velocity Sales Organization | 3 | 5 | 1 | 1 | Manish describes Cloudflare's sales structure from 10 to 50 million, explaining why they hired hungry mid-market AEs rather than enterprise veterans with rigid playbooks to preserve velocity. | |
| The Evolution of Go-To-Market in the AI Era | 3 | 6 | 2 | 1 | Manish explains how selling AI differs fundamentally from traditional SaaS: instead of displacing existing vendor spend, AI GTM requires heavy education and DevRel to capture new AI project budgets. | |
| Arize's Open-Source Flywheel and Enterprise Adoption | 4 | 6 | 2 | 3 | Siddharth questions why Arize focused directly on large enterprise instead of mid-market. Manish explains that enterprise developers adopted open-source Phoenix due to data privacy constraints, pulling Arize into accounts like Wells Fargo. | |
| Global Go-To-Market Expansion Strategy at Arize | 4 | 5 | 2 | 3 | Siddharth pushes on whether expanding globally is premature. Manish defends the move by explaining that international expansion was strictly demand-driven rather than speculative cold entry. | |
| Talent Acquisition and Founder Evaluation Strategies | 3 | 4 | 1 | 1 | Siddharth asks how operators can evaluate early-stage startups amidst AI hype. Manish suggests stress-testing founders with provocative questions during interviews to gauge their response to criticism. | |
| AI Hype vs Reality: Bridging the Production Reliability Gap | 4 | 7 | 2 | 2 | Siddharth raises the narrative that 95 percent of AI projects fail in production. Manish gives a masterclass on non-deterministic systems, explaining why traditional APMs miss hallucinated agent outputs and why online evals are essential. | |
| Production Agent Case Studies: DoorDash and Air Canada | 3 | 6 | 1 | 1 | Manish walks through real enterprise agent case studies, including DoorDash's automated refund pipeline and Air Canada's costly hallucination failure that led to legal liability before adopting Arize. | |
| Distribution Channels and the Path to Fully Autonomous Agents | 3 | 5 | 1 | 1 | Siddharth asks when enterprises can fully hand over workflows from humans to AI. Manish draws a parallel to autonomous driving, arguing human-in-the-loop will remain necessary for the next few years. | |
| Agent Capabilities, SaaS Evolution, and Consumption Growth | 5 | 5 | 2 | 3 | Siddharth points out the contradiction between demo failures and SaaS market panic caused by Claude. Manish highlights Cursor and Claude Code as proofs of capability, projecting massive consumption spikes as seen with Sierra. | |
| Angel Investing Framework: Backing the Jockey and Middleware | 4 | 5 | 1 | 1 | Siddharth asks about Manish's angel portfolio choices like Portkey and Composio. Manish explains his philosophy of backing the jockey over the initial idea, noting early pivots in both Composio and Cloudflare. | |
| Infrastructure M&A Dynamics and the Value of Product Bundles | 7 | 6 | 2 | 4 | Siddharth cites specific M&A exits from his fund (Requestly, ZenDuty, Apica) and asks why infra tech gets acquired if software tech has no moat. Manish explains the multi-product bundle value proposition in M&A. |
Statements from this episode (35)
Jindal: Large TAMs give startups multiple chances to survive early mistakes
“I'm a big believer that if the market you are in, if the market has a big opportunity, you as a company will get a lot more chances to make it work, right? If you're in a market with a small dam you do, do one thing wrong and you're out of business more than l…”
Jindal: Startups usually avoid hiring MBAs and management consultants
“Startups usually do not hire people with like MBAs and who would come from management consulting for their, in their mind you guys, you know, all about process and data and analysis, but as a startup, you do none of that, right? A startup is all about taking a…”
Jindal: Never leave a winning startup over a bad manager
“Never leave a company you believe in for like, you know, these kinds of reasons, you know, you have to stick it out because things change really fast. And one thing doesn't change is like, you know, the trajectory of the company, like if you are an interest co…”
Jindal: Enterprise customers almost never replace foundational tech infrastructure
“Once you put the plumbing, you never change it. That's the reality of it, right? And unless and even if the plumbing breaks, you could just go whatever plumbing you have, you fix it. And I feel that a lot of the infra companies are bad, where if you can be par…”
Jindal: AI application startups hit $100M revenue far faster than infrastructure companies
“A lot of the AI application companies are growing way faster than the plumbing companies, like pick like Repplet or pick Lovable or pick, you know, any of these, Companies, they are able to go from, like, ten million to, like, hundred million within, like, a m…”
Jindal: Only 1 or 2 top AI application companies will survive
“They may be 20 companies would do well today, but off that maybe one or two will be successful in the long run, because it's very easy to challenge the mode of those companies, right? Because end of the day, their biggest mode is their consumer, right? Who's u…”
Jindal: Long-Term Tech Winners Are Easier to Pick in Backend Infrastructure
“In a long run, it's much easier, I believe, to pick a winner where you're sitting on in the back end versus on the front end, unless you are a data, you own the data, like Salesforce.”
Jindal: UI-Based Software Will Likely Be Replaced by Natural Language AI Systems
“In the future, I believe that you are not going to do clicks to get something done. You're just going to ask in a natural language, right? So, which means that like, I think this earlier software companies more than likely is going to get replaced by AI system…”
Jindal: Arize uses Cursor over BigQuery to eliminate internal data analysts
“Even in my current company all our systems data, like Salesforce, HubSpot, and everything even Gong data, we actually put into BigQuery, you know, every data, and then we have we have cursor sitting on top of BigQuery, and if anything I want, I just ask cursor…”
Jindal: Startups Cannot Excel Across Multiple GTM Motions, Geographies, or Products Simultaneously
“Actually, you know, I'm a big believer that you can't be good at everything. And when I, what I mean by that is if when you're building a company, right, you can't be good at serving all kinds of customers. You can't be good at serving a customer through a sel…”
Jindal: Cloudflare's First Enterprise Plan Was Just PLG With White-Glove Support
“So initial enterprise plan was nothing but The same product which you have on PLG with the white glove service, you know, because what happens with a lot of customers is they come through PLG because they love that it's very simple to start, very simple to use…”
Jindal: Cloudflare landed Bain Capital as first enterprise customer via website inbound
“So first customer, enterprise customer we have was Bain Capital. So Bain Capital was using ourselves a product and they were happy with it. And we just put a phone number on a website, you know, if somebody wants to call us and that phone number was pretty muc…”
Jindal: Signing giant enterprise customers early forces them to dictate product roadmaps
“Other challenge you could also happen is if you go bring a very large customer very early on, then that customer is going to force you product in a direction you may not want to go, right? I mean, if you bring in, I don't know, Apple as a customer, more than l…”
Jindal: Cloudflare avoided pursuing Akamai's customers during its first four years
“Like in the first three, four years, we did not go after any of the Akamai customers at all. But if they come to us, we will, but there are times where we did say no to the customers, like, no, we are not ready for you.”
Jindal: Only 2% of websites used CDNs around 2012–2013
“Back in 2012, 13, there was only two percent of the website who actually was using some kind of a CDN solution or somebody did also, but not.”
Jindal: Cloudflare delayed hiring enterprise AEs to protect 30-day sales velocity
“We hire, like, a lot of mid-market A's, not, we did not hire, like, your traditional enterprise A's early on... The reason for that was that we wanted to build a fast velocity motion which was, like, the deal size was about 50,000 dollar, and we wanted to do d…”
Jindal: Cloudflare rejected GTM leaders with rigid playbooks
“So, so in terms of even hiring the leadership team in the go to market, we were hiring people who were more open about like doing things in a different way versus like come with a playbook. So we never hired somebody who like, Hey, I have a playbook. This is h…”
Jindal: AI startups capture net-new budgets rather than displacing existing software vendors
“In the AI world is you are not displacing spend. You're not displacing vendor, actually. All you are doing is telling them that we have built something which will allow you to either make your AI work investment Or AI is gonna may help you in either reduce you…”
Jindal: DevRel and developer education are far more critical in AI GTM
“Your go to market, you have to lead with education more and education, which means that you spend a lot more time in building documentation and also your brand has to be more educational and then you can attract the developers to start using your product. And …”
Jindal: Enterprise developers prefer open-source tools over SaaS for data privacy
“What we realized is that it's the developer at the larger companies wants to use open source, and if you are a developer at a, let's say, born on the web company, you actually want more of a off the shelf self-sell product. It doesn't have to be open source, a…”
Jindal: Wells Fargo had 100 developers using open-source Arize before commercial deal
“Wells Fargo is a big customer of Arise, and they were, before Wells Fargo became a large customer of Arise, they were like hundred developers using the Phoenix product, and that's the reason why we were able to, you know, get into the Wells Fargo.”
Jindal: Large enterprises have more AI agents in production than tech startups
“Observability you need it when you have something in production, your agent is in production, and enterprises, like the large enterprises where the agents are in production more than, you know, like the, if you look at the Silicon Valley companies, they all ar…”
Jindal: Never expand into geographic markets without inbound customer pull
“Basically, you know, I'm a big believer in, in, you know, like, you know, when you think about the go-to-market is that you never want to go into a market cold, you know, because, you know, what happens when you go into the market cold is that you have to inve…”
Jindal: Prior domain sales experience matters less when hiring AEs
“What they have sold is less important because we can teach them how to sell our product. We can teach them about the industry. So we are not like only hiring A's who actually are coming from a similar space. We have A's who come from very different industries.”
Jindal: Execs should ask irritating questions to test startup founders
“If you want to make sure that you have the right founder, you ask tough questions in the interview process. Right? And you sometimes, ah, ask questions where it might be irritating to them, and then you see how they react to it. Right. And then you get a sense…”
Jindal: AI agents fail at scale without evals and observability
“How can you have a agent running at scale if you don't have the evals or AI observability in place, like otherwise, You are totally flying blind. And especially when the system is non-deterministic, which can do, you know, you have no idea whether it's perform…”
Jindal: Traditional observability tools like Datadog show green even when AI hallucinates
“In the world of agents, if you think about it you could be in a world where you actually, agent is perfectly, you would get, you know, you have a Datadog dashboard and everything is green on the dashboard that, you know, they, it's doing all the right thing, b…”
Jindal: AI agents should only be deployed in predefined workflows
“I mean, you want to put agents in any, anything where the steps are kind of defined to do the job.”
Jindal: AI underperformance stems from inadequate enterprise tooling, not model shortcomings
“So that's why, you know, like, I know the narrative in the market is that AI is not doing what it's supposed to do, but I think it's not that, it's more that the people just have not taken the time to put the right tools and tooling in place.”
Jindal: Claude Code and Cursor are the world's most powerful AI agents
“The two agents in the world right now which are, like, truly the most powerful is the cloud code and the cursor. Those agents are amazingly, insanely good.”
Jindal: Sierra.ai hit $200M ARR and added $50M in the subsequent quarter
“Sierra.ai got two hundred million ARR last quarter, and then they added fifty million ARR the next quarter, and it's not that they actually signed fifty million dollar of new deals. What happened was the agents they deployed in production for all these retaile…”
Jindal: Cloudflare was originally pitched as an ad-monetized network
“Business school idea was very much about like a more like ad revenue. So you basically build the network and a lot of traffic, and then you can make money through the eyeballs. That's kind of the idea versus like actually to be the security, the layer, the net…”
Jindal: Long-term independence likelihood for tech startups is very low
“Any company right now any company you pick, the likelihood for them to stay independent in the long run is very, very low.”
Jindal: ClickHouse acquired 13-person startup Langfuse for close to $1 billion
“You know, one of our comparator Lang Fuse this company out of Germany, I think they had, like, 13 employees, and they recently got bought by ClickHouse, and I mean, they did not disclose the number, but we believe, like, it's close to a billion dollar acquisit…”
Jindal: Software bundles create defensibility even when individual features commoditize
“And I think a lot of these companies who are acquiring, they feel like I want to build the best basket out there. Then I'm not displaceable, basically. But within the basket, few things may become obsolete or become like a commodity, and that's okay. But the b…”