Latest / The Series A Show with Fexingo: Early-Stage Funding, Pitch Decks, and Startup Milestones / How Series A Startups Survive the AWS Cloud Cost Shock
Transcript
- Lucas: If these conversations are useful for what you're building or running, remember that a small group of listeners keeps us ad-free every month through buy me a coffee dot com slash fexingo — that's F. E. X. I. N. G. O. Luna: Yeah, exactly. It really does help when you can focus on the work instead of chasing ads. Lucas: Let’s get into it. Andy Jassy made headlines this week saying Amazon Web Services is on track to become a one trillion dollar business, but he also dropped a warning that they’re going to spend big to get there. On paper, that sounds like good news for AWS shareholders, but for a Series A founder sitting in a room trying to justify their next eighteen months of runway, it’s actually a pretty scary signal. Luna: Right, because when the incumbent raises prices or changes the cost structure, the little guys usually eat the difference before they can pivot. Lucas: Exactly. The hidden angle here isn’t just that cloud computing is expensive; it’s that the entire model of early-stage software development has been built on the assumption that compute costs would either stay flat or drop as a percentage of revenue. That assumption is breaking. We are seeing a structural shift where the marginal cost of serving an additional user is no longer negligible, which fundamentally changes how founders should be designing their products right now. Luna: So you’re saying the old playbook of launch fast, iterate later, and worry about optimization at Series B is dead? Lucas: It’s not dead, but it’s dangerous if you don’t have guardrails. Look at the current market environment in late October 2026. Investors are getting pickier about gross margins. They used to care mostly about growth rates and customer acquisition costs. Now, they want to see that your unit economics are sound even if you scale slowly. If your cloud bill is growing faster than your recurring revenue, you aren’t building a business; you’re building a liability. Luna: That makes sense. But how do founders actually measure that risk before they blow their entire seed round? Lucas: They need to look at their infrastructure spend as a percentage of total operating expenses. Ten years ago, that number might have been five percent. Today, for many data-heavy or ai enabled startups, it’s climbing toward fifteen or twenty percent. The trick is to separate fixed costs from variable costs. Fixed costs are things like your engineering salaries and office rent. Variable costs are your cloud bills. You can control the former through hiring plans, but the latter is exposed to market forces like what Jassy just described. Luna: So the move is to treat cloud infrastructure more like raw materials than overhead. Lucas: Precisely. When you buy steel for a car company, you negotiate contracts, you hedge against price spikes, and you design the car to use less steel per unit. Cloud infrastructure needs that same discipline. Founders who wait until they hit a million dollars in annual recurring revenue to optimize their architecture are already too late. The compounding effect of inefficient code or unoptimized database queries hits harder when you scale. Luna: I’ve seen teams waste thousands on idle servers just because they didn’t shut them down after a demo. Lucas: That’s the low-hanging fruit, sure. But the real issue is architectural. Many Series A startups build their MVP on generic virtual machines because it’s easy to spin up. Those are expensive on a per compute hour basis. The smarter play is to evaluate serverless functions or containerized environments that scale down to zero when nobody is using them. It adds complexity to the codebase, but it protects your margin when usage dips. Luna: Does that mean we should stop recommending standard cloud providers to our listeners? Lucas: No, but we should recommend multi-cloud strategies or strict budget caps from day one. One concrete thing I want everyone to take away is the concept of the 'cost ceiling.' Every engineering team should have a hard limit on monthly spend that triggers an automatic review. If you hit ninety percent of your budget, development slows down until optimization happens. It feels counterintuitive to slow down dev velocity, but it’s the only way to survive a price hike without burning cash. Luna: That’s a great operational moat, actually. Most people think moats are about technology or network effects. Lucas: Operational efficiency is a moat because it’s boring and hard to copy. Competitors can steal your features, but they can’t easily steal your ability to serve ten thousand users for half the cost of their legacy stack. This ties back to Jassy’s comment about AWS spending big. When the giants raise prices, the lean operators win market share because they can afford to discount or reinvest in product while others bleed cash. Luna: So efficiency becomes a competitive advantage in a high-cost environment. Lucas: Absolutely. Let’s talk about AI specifically, since that’s driving a lot of the current cloud inflation. Generative AI models are incredibly compute-intensive. Startups building with LLMs often see their inference costs spike by thirty percent quarter-over-quarter if they aren’t careful. You have to choose between using a hosted API, which is convenient but pricey, or hosting open-source models on your own hardware, which is cheaper but requires significant engineering talent. Luna: And most early-stage teams don’t have the talent to manage their own GPU clusters effectively. Lucas: Correct. So the recommendation is to start with APIs for validation, but have a migration plan ready for Series A. That plan should include identifying which models can be quantized or distilled to run cheaper. It’s technical debt, but it’s the kind of debt that pays you back in margin. Ignoring it is like ignoring a leaking roof because you’re busy painting the walls. Luna: I like that analogy. Leaking roofs are expensive to fix later. Lucas: They are. And here is another subtle point: data storage costs. People forget about egress fees. When you move data out of the cloud to send to customers, that charges add up. If you are building a video or media platform, your egress bill could exceed your compute bill. Optimizing for compression and caching locally can save millions over two years. Luna: So the lesson is to audit everything, not just the CPU hours. Lucas: Every byte. Storage tiers, network transfer, database indexing, even the regions you deploy in. Deploying in multiple geographic regions increases redundancy but also doubles your baseline cost. Ask yourself if you truly need global availability at Series A. Often, a single region with a good CDN is enough to start. Luna: That seems like a no-brainer in hindsight, but founders are so eager to promise global scale. Lucas: They are. It’s a marketing play, not an economic reality. Investors know this now. They will ask you about your cloud strategy in due diligence. If you can show them a detailed breakdown of how you’re managing variable costs, it signals maturity. It shows you understand that growth without profitability is just a faster path to insolvency. Luna: It turns financial discipline into a trust signal with investors. Lucas: Exactly. And it buys you flexibility. If you have healthy margins, you can weather a price increase from AWS or Azure without panicking. You can absorb the shock and keep building. If you are already stretched thin, a five percent hike can wipe out your net income overnight. Luna: That’s a terrifying thought for anyone relying on venture capital to bridge the gap. Lucas: It is. Which is why we need to talk about negotiation. Yes, you can negotiate with cloud providers once you hit certain thresholds. Even at the Series A level, if you are committing to a year of spend, you can often get discounts. Don’t just accept the list price. Talk to your account manager. Show them your growth trajectory. They want to lock you in. Luna: So relationship management extends to your IT vendors too. Lucas: It absolutely does. Treat them like partners, not vending machines. If you commit to being a loyal customer, they will give you better rates, credits, and support. It’s a simple lever that many young founders ignore because they think it’s beneath them. It’s not. It’s part of the job. Luna: I agree. It’s smart business to protect your bottom line at every level. Lucas: One final piece of advice: monitor your costs daily, not monthly. Set up alerts for anomalies. If your bill jumps twenty percent in a week, something is wrong. It could be a bug, a leak, or a successful marketing campaign that outpaced your capacity planning. Either way, you need to know immediately so you can react. Luna: Real-time visibility is the only way to stay ahead of the curve. Lucas: Right. In a world where infrastructure costs are rising and investor scrutiny is tightening, the founders who survive will be the ones who treat every dollar of spend with extreme prejudice. It’s not about being cheap; it’s about being efficient. Efficiency is the ultimate defensibility. Luna: Efficiency as a feature. That’s a powerful way to frame it for builders. Lucas: It is. As we look at the rest of Q4 2026, watch how these cloud giants adjust their pricing. The winners will be the startups that adapted their architecture before the price hikes hit. Luna: And hopefully, those that did will have enough runway left to tell us about it. Lucas: We’ll be listening. Thanks for tuning in.