Build vs Buy Is Dead. Now It's Build vs Rent Your Extinction

Here is the question nobody in your boardroom is asking correctly: when you sign that three-year SaaS contract next month, are you buying software, or are you renting the right to be replaced by it later, on someone else's terms?

That's not a rhetorical flourish. That's the actual math now. GoodFirms just published its 2026 survey and the number is loud: 91% of software companies use AI to cut development costs. Read that again. Not "explore AI." Not "pilot AI." Use it. Daily. To ship code cheaper than a year ago. Meanwhile Google News is drowning in "Top Custom Software Companies 2026" listicles from London Business News, Netguru, LawFuel, all of them quietly repackaging the same outsourcing day rates as "AI development." Same vendor. Same invoice. New buzzword stapled on top.

The build side of the equation just got radically cheaper for anyone with real engineers. The buy side is trying to hide that fact from you with a rebrand.

The old build vs buy software calculus is obsolete

For twenty years the consultant pitch was simple: build costs you time and headcount, buy costs you a subscription and some integration pain. You ran the spreadsheet, TCO over three years, and usually buy won because building software was genuinely expensive and slow. That spreadsheet is garbage now. Not wrong, garbage. It measures the wrong variable.

The variable that matters in 2026 isn't cost-per-feature. It's ownership of the model. When you buy SaaS, you're not buying a tool, you're renting a black box that some vendor in California or Bangalore is currently rebuilding on top of AI agents, at a fraction of their old cost, while charging you the same price they charged in 2023. Their margin just went up. Yours didn't. And you have zero visibility into when they decide your entire product category is worth automating out from under you, because they own the roadmap, not you.

Meanwhile if you build, and you build smart with AI-assisted engineering, your cost-per-feature has also collapsed. But you own the thing. You own the data. You own the model behavior. You can point it wherever the market moves next. That optionality is the entire game now, and almost nobody is pricing it into the decision.

Why "buy" is the dangerous default in Sweden specifically

This is where Jönköping matters more than San Francisco.

Sweden has a high-trust, high-cost labor market. Our engineers are expensive, our social contract is strong, and our default cultural posture toward vendors is trust first, verify never. That's a beautiful thing for a functioning society. It is a terrible instinct for procurement in an AI transition. Swedish and Nordic founders are quietly outsourcing strategic software decisions to vendors in Eastern Europe and India right now, today, without ever asking the one question that actually matters: what happens to your product roadmap when that vendor's margins get eaten by the exact same AI tools that are cutting everyone's development costs? Because those tools cut both ways. The vendor's cost of delivering to you just dropped. Are they passing that savings on? Or are they pocketing it and selling you the same "custom software development" package at 2023 rates dressed up as "AI-powered delivery" like the listicles currently flooding search results?

I checked Swedish coverage on this topic today. There isn't any. Zero. Breakit, DI, none of them are running this story right now, which tells you something important: this conversation is happening in boardrooms, not in press. That's not Sweden being efficient. That's Sweden being quiet about a decision that deserves noise.

Build vs buy vs Sweden's cost structure

Compare the postures. In the US, venture-funded companies are building in-house AI-augmented engineering teams because equity absorbs the risk of getting it wrong and speed is the only currency that matters to their investors. In parts of Asia, particularly China and increasingly India's product companies, the state and the market are both pushing hard toward vertical ownership of AI stacks because nobody wants dependency on a foreign platform they can't control. In Sweden? We optimize for stability. We sign the multi-year contract because it feels safe, predictable, budgetable. It is the opposite of safe right now. A three-year SaaS contract signed in September 2026 is a bet that the vendor's product, pricing, and existence remain stable while the entire category gets rebuilt underneath it by AI. That is not a conservative bet. That is the riskiest bet on the table, dressed in a suit that looks conservative.

The EU, for its part, is not remotely ready for this. AI Act compliance conversations in Brussels are still mostly about risk classification and transparency obligations, not about the market structure question of who owns the models replacing entire SaaS categories. Nobody in Brussels is asking what happens to European digital sovereignty when every mid-size Swedish company's core operational software is a rented AI layer sitting on top of a vendor's proprietary model trained partly on that same customer's data. That's a policy gap wide enough to drive a truck through, and Sweden, sitting on some of the best engineering talent per capita in Europe, is sleepwalking into renting its own extinction instead of building sovereignty into its own stack.

Where this actually goes

Two to five years out, here's the trajectory I'd bet the company on. The SaaS category as we know it, the multi-tenant, seat-based, annual-contract model, starts fragmenting hard by 2028. Not because SaaS companies are bad businesses, but because the marginal cost of a company building its own AI-agent-driven internal tooling keeps dropping toward zero while the SaaS vendor's price stays sticky because that's what subscription pricing does. The gap between what you could build and what you're being charged to rent becomes impossible to ignore. At the same time, we get closer to systems that can genuinely reason across a company's own data, workflows, and context in ways that generic SaaS platforms structurally cannot match, because generic platforms are built for everyone, and your business is not everyone. This isn't AGI in the science fiction sense. It's something more boring and more immediate: agentic systems that can hold your specific business logic in working memory and act on it continuously. Once that threshold is crossed for a given function, whatever tool you're currently paying a subscription for in that function becomes a liability, not an asset, because it locks you into someone else's slower, genericized version of what you could run yourself. The companies that win this decade aren't the ones who bought the most SaaS or built the most software. They're the ones who correctly identified which functions were core enough to own outright, and which were commodity enough to keep renting, and who revisited that split every twelve months instead of signing a contract and forgetting about it.

The regulatory blind spot nobody's fixing

Sweden and the EU need a serious conversation about model ownership and data sovereignty that goes beyond GDPR-era thinking. Right now there is no meaningful policy pressure on SaaS vendors to disclose how much of their delivery is AI-automated, what that means for pricing, or what happens to customer data used to train the models replacing their own customer support and development teams. That's a gap regulators in Stockholm and Brussels should be closing loudly, not quietly. Until they do, the burden is entirely on you, the buyer, to ask harder questions before you sign.

What to actually look at this week

If you're serious about the build side of this decision, stop treating "build" as scary or slow. The tooling has changed completely in the last few months. OpenHands is worth a serious look if you want AI-driven development that your own team controls end to end, not a vendor's black box. For anyone drowning in legacy codebases they inherited from an outsourced build, Graphify turns your existing code, docs, and schemas into a queryable knowledge graph, which is the fastest way I know to actually understand what you're paying to maintain before deciding whether to keep renting it or bring it in-house. And if token costs are the thing making "build with AI" feel expensive to your CFO, rtk cuts LLM token consumption 60 to 90 percent on common dev commands, a single Rust binary, zero fuss. That number alone should change your build-vs-buy spreadsheet.

This is exactly the conversation we have with clients at HEIMLANDR before any contract gets signed. If you're weighing custom SaaS development against another vendor renewal, the first question isn't cost. It's ownership. We build fast, MVP development in weeks not quarters, and we build it so you own the model, the data, and the roadmap, not us. If you want to see what a real punk rock SaaS development company operating out of Sweden actually delivers when the incentive is your independence instead of our recurring invoice, that's the whole reason HEIMLANDR exists.

What to actually do about it

Before you renew or sign anything this quarter: list every SaaS tool your company depends on. For each one, ask three questions. Could an AI-augmented internal team build a leaner version of this in under three months. Is this vendor's core value the software, or is it the data network effect and integrations that software alone can't replace. And critically, what happens to your business if this vendor gets acquired, sunset, or repriced 40% next year because their own margins collapsed under AI competition. If the answer to question one is yes and question two is no, you're renting your extinction. Stop it. Build it. Own it. That's true whether you're comparing Claude Code tooling against a legacy dev shop, or evaluating AI agents against another SaaS seat license. Software development Sweden style should mean building sovereignty, not renting comfort.

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.