build vs buySaaSinternal toolsvibe codingindie hackersstartup costsAI coding agents

The SaaS Cancellation Wave: When Building Your Own Tool Beats the Subscription, and When It Doesn't

Muhammad Tayyab

October 14, 2026·13 min read

Person holding a credit card while typing on a laptop — photo by rupixen via Unsplash

Pieter Levels cancelled almost all his business SaaS, Marc Lou replaced Beehiiv with a vibe-coded newsletter tool, and McKinsey says 32% of organizations skipped at least one software purchase to build in-house. Here's a total-cost framework for deciding when building beats buying.

In the space of a few days in October 2026, "cancel the subscription and build it yourself" stopped sounding like a hot take.

Marc Lou left Beehiiv (October 9). He posted that he'd vibe-coded his own newsletter platform "with a few prompts" because Beehiiv had raised prices "for the 3rd time in 2 years". In his words: "My first yearly bill was $1,000. My future bill would have been $4,000." He now uses Amazon SES and MongoDB and expects to pay "less than $300/year."
He open-sourced it as MailCheap (October 10). He released MailCheap, "self-hosted newsletters on Amazon SES", at about $0.10 per 1,000 emails, and called it "92% cheaper than Beehiiv (for me)."
Pieter Levels said he'd gone further (October 11). He wrote: "I've cancelled almost all of the SaaS subscriptions for my business." What's left, by his list, is AI inference, Hetzner hosting, Cloudflare storage, Backblaze backups and taxes. He added: "I think we'll be facing a tidal wave of cancellations once people find out how much money they can save." In September he had written that he'd reduced his SaaS dependencies "to just a handful of companies by vibecoding my own replacements".
Remote's CEO replied the same day. Job van der Voort, co-founder and CEO of Remote, replied to Levels: "We are doing this at @remote; almost no new software subscriptions. Cancelling everything we can. Still a lot to go, but the end is in sight for about 80% of all SaaS."

The enterprise data points in the same direction, but more modestly than the headlines suggest. McKinsey's State of AI in 2026 survey (1,719 participants in 97 countries, fielded May 4 to June 8, 2026, published August 25) found that "32 percent" of respondents "report that their organizations have decided against purchasing at least one software product or feature because they were able to build the functionality in-house using agentic coding tools." That's at least one purchase skipped, not 32% of companies replacing their SaaS stack. Coverage of the report also noted that the share rises to nearly half among McKinsey's "AI high performers".

So the trend is real. But "I cancelled Beehiiv" is a very different decision for Marc Lou than it is for a two-person team with a sales pipeline to run. The rest of this post is a framework for making that call on total cost, not license fees.

The license fee is the smallest number in the build decision

When you replace a subscription, you're not just replacing the invoice. You're taking over everything the vendor was quietly doing for you:

Build time. Even with AI coding agents, someone has to specify, review, test and deploy it.
Maintenance. Dependencies, API changes, browser changes, framework upgrades and the bug you'll find at 11pm before a launch.
Security. Authentication, secrets, rate limiting, input sanitization, access control, and patching when something is disclosed.
Uptime and on-call. If it goes down, you're the support team.
Domain-specific operations. For email, that's deliverability. For payments, it's reconciliation. For HR or payroll, it's regulation.
Compliance. Data retention, privacy requests, audit trails, and whatever your customers' security questionnaires ask for.
Opportunity cost. Every hour spent maintaining an internal tool is an hour not spent on the product your customers pay for.

MailCheap is a useful example because its README is unusually honest about this:

No support. "This is a one-time open-source release. Issues are off, and setup questions go unanswered." There are "no updates planned".
Setup takes real work. You need MongoDB, Amazon SES, Upstash QStash and a host, plus Redis and Blob storage if you want shared rate limits and image uploads. Marc estimates "about an hour, most of it waiting for DNS and for AWS to approve production access", and new SES accounts start in a sandbox (200 emails a day, verified recipients only).
Deliverability becomes your job. The README walks you through DKIM, a custom MAIL FROM domain for SPF alignment, DMARC ("Gmail and Yahoo require DMARC from bulk senders"), domain warm-up, and keeping complaints under 0.1% and bounces under 2%.
Check the real sending price. AWS's SES pricing page lists outbound email at $0.10 per 1,000 under à la carte pricing. It also says new SES accounts with no metered activity since June 1, 2025 start on the Essentials plan at $0.16 per 1,000 from July 21, 2026, and you can switch to à la carte. Still cheap, but check which plan you're on.

None of that makes MailCheap a bad idea. It's a great fit for someone who already understands email infrastructure. It just shows what "92% cheaper" includes and excludes. Others have already noticed the gap: within days, Migma AI launched a hosted version of MailCheap aimed at people who don't want to run SES, MongoDB and Redis themselves.

An illustrative total-cost comparison (made-up numbers)

This is a made-up example to show the math, not data from any real company. Plug in your own numbers.

Say a tool costs $3,000 per year and you think an AI-assisted rebuild would take 40 hours, plus 4 hours a month of upkeep. If you value your time at $75 an hour:

Year one cost of building: 40 hours × $75 = $3,000 to build, plus 48 hours × $75 = $3,600 to maintain, plus maybe $200 in hosting. That's about $6,800.
Year one cost of buying: $3,000.
Year two: building costs about $3,800 (maintenance plus hosting) vs $3,000 to keep buying, or more if the vendor raises prices again.

Under these assumptions, building only wins if the vendor's price climbs sharply (as Marc Lou's did), if upkeep turns out much lower than 4 hours a month, or if the custom version earns you something the SaaS can't, like a feature your business needs. Change one input, for example a $12,000-a-year tool, and the answer flips. That's the point: do the math with your numbers, including your time.

When building tends to win

The job is simple and stable. A form that writes to a database, an internal dashboard, a status page, a link shortener, a changelog. The requirements won't change much next year.
You're paying for 5% of a big product. If you use one feature of a tool priced for teams that use all of them, a focused replacement can be small.
Pricing scales with something that isn't the vendor's cost. Marc Lou's argument was exactly this: newsletter platforms charge per subscriber, "when in fact the real cost is the number of emails sent."
The vendor keeps raising prices or forces a sales call. In September, a founder vibe-coded a B2B tool instead of buying it because the salesperson insisted on a meeting before sharing the price.
It's core to how you differentiate. If the workflow is your competitive edge, owning it can be worth the upkeep.
There's a solid open-source base. Levels described taking an open-source project that does the same thing and customizing it with Claude. Starting from maintained code beats starting from a blank prompt.
You already have the operational skills. Someone who runs servers daily pays less for "now I run this too" than someone who doesn't.

When buying still wins

The domain is regulated or high-stakes. Payroll, tax, payments, identity, health data, anything with legal liability. The vendor's real product is its compliance work.
The hard part is operations, not code. Email deliverability, fraud detection, uptime across regions, spam filtering. AI can write the app in an afternoon, but it can't build sender reputation or a 24/7 on-call rotation.
Many people depend on it. A CRM used by five salespeople, or a help desk used by support, needs permissions, audit history, onboarding and reliability that a weekend build won't have.
Integrations are the value. If a tool's worth is its 200 connectors, rebuilding the core doesn't replace that.
Your time is worth more on the product. If you're pre-product-market fit, the cheapest tool is the one you never think about.
The price is small and stable. A $10-a-month tool that just works rarely justifies even a day of building.

There's a middle path too. Before cancelling, downgrade, renegotiate, or switch to a cheaper vendor or a hosted version of an open-source tool. You often get most of the savings with none of the upkeep.

Build-or-buy checklist

Answer these for each tool you're thinking of replacing:

What does it really cost per year, including seats and add-ons, and what's the trend over the last two years?
What exactly do you use? List the features. If it's one or two, building gets more attractive.
Is the hard part code or operations? Deliverability, compliance and uptime are operations.
What breaks if it's down for a day? If the answer is "revenue" or "legal exposure", be careful.
Who will maintain it, and how many hours a month will it realistically need? Put a dollar value on that time.
Is there a maintained open-source option you could self-host or extend instead of starting from scratch?
What data is involved? Personal data, payment data or health data raises the bar for security and compliance.
Can you get the data out? Before cancelling, export everything and test the import into your new tool.
What's the exit plan if your build fails? Could you go back to the vendor, or to a competitor, without losing data?
Is there a cheaper option short of building, like a lower plan, an annual discount, a competitor or a hosted open-source alternative?

If most answers point to "simple, stable, mostly code, low stakes, someone owns it", build. If they point to "operations-heavy, regulated, many users, nobody owns it", keep buying, and maybe negotiate.

What to do this week

Monday: export your card and invoice history and list every subscription with its annual cost and the last price change.
Tuesday: mark each one core, convenience or unused. Cancel the unused ones today.
Wednesday: for the top three by cost, run the checklist above and the total-cost math with your real hourly value.
Thursday: for anything that looks like a build candidate, look for a maintained open-source project first, and read its README for support, security and setup notes.
Friday: pick one low-risk tool to replace as a pilot. Set a time budget, and keep the subscription running until the replacement has worked for a full month.
Before any renewal: ask the vendor for a lower tier or a discount. It costs one email.

Soft close

The cancellation wave is real, and AI coding agents have changed the math for a lot of simple tools. But the winners won't be the people who cancel the most subscriptions. They'll be the ones who replace the right ones and leave the operations-heavy, regulated work to vendors whose whole business is doing it well.

If you're weighing a custom internal tool against a subscription and want a second opinion on scope, cost or maintenance, get in touch.

About the author

Muhammad Tayyab is a full-stack and mobile developer based in Lahore. He builds DripScore, an AI outfit rating app live on the App Store, under dawnapps.co. Find him on GitHub, LinkedIn, and X, or say hello at /#contact.

Sources

Back to all posts

Thanks for reading 🙏