Build vs Buy Is Dead. In 18 Months You Build Everything.

Your SaaS Stack Is Becoming a Liability

Right now, your company is paying for somewhere between 15 and 80 SaaS subscriptions. Every single one of them got a price hike this year. Most of them got a price hike last year too. Meanwhile, the cost to build the exact functionality you're licensing has dropped 30-40% year-over-year, and the trend is accelerating. The crossover point where building custom software is cheaper than buying off-the-shelf SaaS is not a 2030 thing. It is happening right now, in 2026, and if you're a Swedish company paying US vendors in a weakening krona, it is already past you.

The entire build vs buy software framework was built on two assumptions: building is expensive, and building is slow. Both are collapsing. A senior developer with proper AI tooling today outproduces a five-person team from 2024. Not by a little. By multiples. And that means every procurement decision you made two years ago needs to be re-examined with completely different math.

The Numbers Don't Lie, But Your CFO Might Not Know Them Yet

Clockwise published their 2026 ERP development cost data and the numbers are stark. Custom ERP builds that would have cost $300K-$500K in 2024 are coming in at $150K-$280K today. That is not a marginal shift. That is a structural change in the economics of software ownership. At the same time, Gartner's latest enterprise software spend projections show SaaS licensing costs continuing to climb 8-12% annually, compounded. Do the math over three years.

Year one: SaaS is still cheaper. Year two: it is a coin flip. Year three: you are paying double what a custom build would have cost you, and you still don't own anything.

The U.S. mid-market has already figured this out. The data coming out of American companies in 2026 shows a hard swing toward custom software development. They are pulling budget away from SaaS renewals and putting it into internal builds, custom SaaS development, and purpose-built tools that do exactly what the business needs and nothing it doesn't.

Swedish companies? Still locked into multi-year contracts. Still paying in SEK that buys less dollar-denominated software every quarter. Still telling themselves that "we're not a software company" as if that's an excuse rather than a confession.

The "Building Takes Too Long" Argument Is Dead

This was always the real objection. Not cost. Time. "We can't wait 18 months for a custom build when we can have Salesforce running in 6 weeks." Fair point. In 2022.

In 2026, MVP development timelines have compressed to 4-8 weeks. Not for toy prototypes. For production-grade minimum viable products that you can ship to real users and iterate on. The combination of AI-assisted coding, mature open-source frameworks, and infrastructure-as-code has turned what used to be a 6-month project into a 6-week sprint.

At HEIMLANDR, we see this every week. Founders come to us expecting a 4-month timeline and we deliver working MVPs in a fraction of that. The tools have changed. The speed has changed. The old mental models have not caught up.

Look at what is trending on GitHub right now. OpenHands (74,000+ stars) is a full AI-driven development platform. ECC is pushing 186,000 stars as an agent harness that optimizes AI coding performance across Claude Code, Codex, Cursor, and more. RTK is a single Rust binary that cuts LLM token consumption by 60-90% on dev commands. These are not research projects. These are production tools that developers are using right now to ship software at speeds that make the old timelines irrelevant.

When a single senior dev with these tools can do the work that required five people two years ago, the "building takes too long" argument does not just weaken. It inverts. Now buying takes too long, because you are waiting for a vendor's roadmap to deliver features you could have built yourself last month.

The Sweden Problem: Paying Premium for Someone Else's Priorities

Let me be specific about what I see from Jönköping. Sweden has a particular version of this problem that is worse than the US or even most of Europe.

First, the currency exposure. Swedish companies are paying for Salesforce, ServiceNow, Workday, HubSpot, and dozens of other tools in USD. The krona has been weak and volatile. Every renewal is a surprise to the budget. This is not a technology risk. It is a financial risk that most CFOs are tracking poorly because it is spread across 50 different line items.

Second, the consulting dependency. Sweden has a massive IT consulting sector that has historically benefited from the complexity of integrating and customizing third-party SaaS. Companies paying $150/hr or more for "custom development" that is really just configuration and glue code between vendor products. That $150/hr bought you roughly 1x productivity in 2024. Today, with AI tooling, the same skilled developer produces 3-5x. But the hourly rate hasn't dropped. The consulting firms are pocketing the productivity gains.

Third, the cultural factor. Swedish enterprise culture is consensus-driven and risk-averse. "Nobody gets fired for buying Salesforce" works the same in Stockholm as it does in Chicago. But the risk calculus has flipped. The risky move now is continuing to rent software you don't own, can't modify, and that gets more expensive every year. The safe move, increasingly, is to build.

Compare this to what is happening in Estonia, where government digital infrastructure is largely custom-built. Or to Finland, where companies like Supercell have always built their own tools. Sweden talks a big game about being a tech nation, and we do produce incredible startups. But our enterprise adoption of custom development lags behind where the economics say it should be.

Breakit has been covering the growth of Swedish AI startups, but the bigger story is the one nobody is writing: the slow-motion value destruction happening inside Swedish enterprises that are over-indexed on SaaS licensing.

What the AGI Trajectory Means for This

Zoom out for a moment. The cost collapse I am describing is not a one-time event. It is a curve, and the curve is steepening.

Today, an AI-augmented developer is 3-5x more productive than a 2024 developer. In 18 months, with the trajectory of models and tooling, that number is likely 8-15x. In three years, we may be looking at 20-50x. I am not speculating about AGI here. I am extrapolating from the capability gains we have seen in the last 12 months and the tooling that is already in open-source pipelines.

What this means for the build vs buy equation is simple: it is not an equation anymore. When building something custom takes less time and costs less money than evaluating, procuring, and implementing a vendor product, the rational choice is always to build. And we are arriving at that point for a rapidly expanding set of use cases.

CRM? Buildable. Project management? Buildable. Internal dashboards and reporting? Trivially buildable. Customer portals? Buildable in weeks. The tools that justified SaaS subscriptions because they were too complex or time-consuming to build in-house are becoming straightforward custom development projects.

At HEIMLANDR, this is what we are building toward with our AI Agents practice. Not just building software faster, but building the agents that build software faster. The meta-layer. Because the companies that own their own development capability, augmented by AI, will have a structural advantage over companies that are renting functionality from vendors who are themselves scrambling to figure out the same transition.

The Regulatory Gap

Here is where EU policy is actually relevant, and not in the way Brussels thinks. The EU AI Act is focused on regulating AI use cases, classification, and compliance. Fine. But nobody in Brussels is asking the more important question: what happens to European competitiveness when the cost of software development drops by an order of magnitude and European companies are still locked into American SaaS contracts?

There is a digital sovereignty angle here that policymakers are completely missing. Every dollar a Swedish company sends to Salesforce is a dollar that could be funding domestic software development capability. Every custom-built system is an asset on the balance sheet. Every SaaS subscription is an expense that evaporates the moment you stop paying.

Sweden's Digitaliseringsrådet should be sounding alarms about enterprise SaaS dependency, not just talking about AI ethics and broadband coverage. The strategic question is not whether AI is dangerous. The strategic question is whether Swedish companies will own their digital infrastructure or rent it from San Francisco indefinitely.

What to Look At

If you are a CTO or founder processing this, here is where I would spend my attention this week:

OpenHands: AI-driven development that is actually usable in production workflows. This is where the "developer productivity multiplier" claim becomes tangible. Get your senior devs to spend a day with it. Measure what happens.

RTK: A CLI proxy that reduces LLM token consumption by 60-90%. If you are already using AI coding tools, this is a pure efficiency play. Single Rust binary, zero dependencies. The kind of tool that makes the economics work at scale.

Daytona: Secure, elastic infrastructure designed for running AI-generated code. If you are going to build more and buy less, you need infrastructure that is designed for the volume and velocity of AI-assisted development. Daytona is purpose-built for exactly that.

Hoppscotch: Open-source API development ecosystem. If part of your build-over-buy strategy involves replacing Postman, this is the answer. Web, desktop, CLI. On-prem or cloud. One less subscription.

What You Should Actually Do

I will make this concrete.

This week: Pull your full SaaS spend report. Every subscription, every seat, every annual contract. Calculate the total in SEK and then calculate what that number was two years ago. The growth rate will surprise you.

This month: Pick the three SaaS tools with the worst ratio of cost to actual value delivered. For each one, get a realistic estimate of what it would cost to build a custom replacement that does only what you actually use. Not what the vendor sells. What you use. The gap will surprise you even more.

This quarter: Run a pilot. Take one of those three tools and build the replacement. Not as a side project. As a real initiative with a real team, proper AI tooling, and a deadline. Four to eight weeks. Ship it. Learn what your organization is actually capable of building when you stop assuming building is hard.

The companies that make this shift in 2026 will own their software infrastructure by 2028. The companies that wait will spend 2028 trying to renegotiate SaaS contracts from a position of weakness while their competitors operate with lower costs and higher agility.

The Uncomfortable Truth

I run a software development company in Jönköping. I have skin in this game. When I say the economics of custom SaaS development have fundamentally shifted, you should weigh that against my obvious interest in selling development services. Fair.

But here is the thing. I am not telling you to hire us. I am telling you to build. Whether you build with your own team, with us, with another firm, or with some combination. The message is the same. Stop renting software. Start owning it. The cost structure has changed. The speed has changed. The tools have changed. The only thing that has not changed is the institutional inertia that keeps signing renewal contracts because that is what we did last year.

Build vs buy is not a framework anymore. It is a reflex from an era when building was genuinely harder and more expensive than buying. That era is ending. The question now is not "should we build or buy?" The question is "what should we build first?"

I know what we are building. The question is whether you do.

Fredrik Brunnberg is the CEO of HEIMLANDR.IO, building AI and software solutions from Jönköping, Sweden. This is the daily HEIMLANDR briefing. If you found this valuable, share it with someone who builds things.