Conversation
Highlights

How Mike Weider Is Embedding Cost Into the Software Development Lifecycle — Before the Cloud Bill Arrives

Somewhere between his third exit and his fourth company, Mike Weider tried the VC path. Investing, advising, board seats. “I found that maybe there’s some people that are great at being helpers to others,” he said in a recent episode of BUILDERS. “I was constantly feeling like I wanted to get in there and actually do the work.”

Two years later, he’s back. The company is Frugal. The problem: AI is quietly destroying the gross margin structure that made SaaS businesses work — and the engineering layer has no tooling to do anything about it.

The Margin Math Nobody Is Running

Most founders understand that cloud costs are painful. The sharper insight Mike is building on is structural: the SaaS profitability model depends on near-zero marginal cost per customer. “You spend this money initially on the platform, and then each incremental customer is typically free or not meaningful,” he explained. “And so once you hit certain levels of scale, you get to profitability.” That’s the model. 70–90% gross margins, 80% being average. It works because the cost base is mostly fixed.

AI breaks that assumption entirely. “The average gross margins of an AI native company are more like 50, 60, 70%,” Mike said. “And so if software companies had a hard time getting to profitability at 80% gross margins, if you take away 20%, making life a lot harder.”

But the margin compression alone isn’t the whole problem. What changes the severity calculus is that token costs are variable per customer. Every inference call is a cost event. “The token cost of serving that next customer is material,” Mike said. “And so what we saw was that cloud costs in the past have been painful, but now could be maybe an existential problem.”

That shift — from line item to existential risk — is the founding hypothesis.

The Pattern Mike Recognized from His First Company

Mike’s first company was in application security. They built one of the first vulnerability testing solutions — a world where the standard workflow was: engineers build software, nearly ship it, hand it to pen testers, get a list of holes, cycle back through development. Slow, expensive, and disruptive to teams that had already mentally moved on.

The industry eventually solved this by shifting security left — embedding it into the development workflow rather than bolting it on at the end. That’s the paradigm Mike is applying to cost.

“Most developers, they write their code, they ship it to production — 30 days later or whatever, you get your cloud bill, and the feedback loop is broken,” he said. “Developers don’t know what things cost when they make these code changes. And by the time these bills show up, well, everybody’s moved on to something else.”

The parallel is precise. In both cases, the problem isn’t that no one cares — it’s that the feedback arrives too late to change behavior. Frugal’s bet is that cost visibility embedded upstream, in the workflow itself, changes the decisions engineers make before the bill arrives.

Why the Product’s Data Requirements Killed PLG Before It Started

One of the more useful things Mike does in this conversation is trace his GTM constraints directly back to the product’s technical requirements. Frugal needs access to source code, AWS billing data, and observability data. “Those things are extremely sensitive,” Mike said. “An individual developer is not gonna have the authority to grant access to some other third party to do that. They’re going to need approval.”

The authority problem compounds with the resonance problem. Cost visibility lands differently depending on where you sit in the org. “An individual developer may not be, you know, hugely concerned with the costs of their company, but the CFO cares, the head of engineering probably cares, the CTO cares.”

Both problems point to the same conclusion: PLG is structurally off the table, at least for now. There’s no wedge an IC developer can pull without executive sign-off, and even if they could, cost optimization doesn’t resonate as a bottom-up pain point. Mike accepted this early and built a top-down motion from day one rather than trying to engineer around the constraint.

The Sequencing That Keeps AE Compensation from Burning

Frugal launched in May 2025. For the first six months, the founders ran GTM themselves — content, network outreach, building the initial pipeline of design customers. No GTM hires. In November 2025, they brought on their first non-engineer: a head of growth.

The mandate was deliberate. “That person’s job was to sort of coordinate and warm up the inbound and to have them ready — three months of building the inbound machine in advance of hiring the first salesperson,” Mike said.

The sequencing logic is straightforward but rarely executed this cleanly. An AE dropped into a cold pipeline defaults to cold outbound — the highest-cost, lowest-leverage way to use that seat. The growth hire’s sole function was to prevent that scenario. “Initially a lot of handheld work from the founding team, and then hired a growth person to be the advance party, warming up the engine and the brand and the inbound machine in advance of the first salesperson.”

Three months of runway between the growth hire and the first AE. That gap is the product.

What Mike Learned Cutting Google Ads — and Keeping Cold Calling

Early on, Frugal ran Google Ads. They cut it. The channel failure is instructive: when buyers don’t have the vocabulary to search for your solution, search-intent channels can’t find them. A category that doesn’t exist yet can’t be discovered through keyword queries. “We did a little bit of advertising at the beginning and sort of realized that it’s probably not a channel that’s going to work for us,” Mike said.

Content is the long play — videos, written content, organic search — and organic traffic is starting to move. But the channel that surprised Mike most: cold calling.

“That is actually surprisingly working, and still kind of baffles my mind that cold calling still works in 2026,” he said. He’d been skeptical. “My personal experience is that I would never answer the phone for a number I didn’t recognize.” He had to be convinced.

The more important point is how to measure it in a new-category context. When buyers don’t know your category exists, a call that doesn’t book a meeting still plants a flag — you’ve shared a link, you can pixel the contact, you can retarget with content. The old metric for SDRs is meetings booked. That made sense when buyers already knew the category. When they don’t, measuring pipeline influence across the full funnel is the right frame. Mike’s broader point: “The go to market practices need to evolve and obviously adapt, but that doesn’t necessarily mean we abandon everything that worked in the past.”

Frugal is also going to its first trade show in 2026. The logic is the same: category education requires presence in the places buyers already congregate.

Positioning When You’re a Category of One

Mike was on the board of Cloud Checker, one of the first FinOps tooling companies, later acquired by NetApp. He knows that landscape from the inside, which is why Frugal isn’t fighting it.

“There are a bunch of tools out there that measure stuff, but not a lot of products that help action cost optimization in your engineering workflows,” he said. FinOps dashboards tell you what you’re spending. They don’t tell engineers why, and they have no mechanism to intervene upstream.

Rather than positioning against those tools, Frugal occupies the white space they leave behind. “It’s not so much competitive analysis, it’s more like competitive positioning,” Mike said. “It’s most important maybe why we’re different than those things.”

That framing — different, not better — is the right move when adjacent categories already have established vendors and mindshare. Fighting for “better” invites direct comparison on a market incumbent’s terms. “Different” reframes what the buyer should be evaluating entirely.

The Vision

The endgame Mike is building toward: cost becomes a standard input to every engineering decision, the same way security and code quality are today.

“Right now, when software is being built today, cost is not really in the discussion,” he said. “We see an absolute truth and vision that we know is going to be true — that’s going to be flipped. Every single engineering decision will have cost feedback, just like we do for security, quality, other issues.”

The forcing function is already visible. AI-native companies are watching their AI costs exceed their cloud costs. “I think that’s going to be true for everybody pretty soon,” Mike said. “And that is going to drive everybody to be really focused a lot more on their unit economics and on making engineering decisions with cost context. And that’s the future that we want to help bring about.”

The window to define that category is now.