The Server in the Closet

There's a specific kind of technology leader who has become endangered: the one who builds things.

Not the kind who orchestrates vendors. Not the kind who manages integration roadmaps between Salesforce, HubSpot, and whatever AI wrapper launched last week. I'm talking about the CTO who looks at a problem and thinks, "We should own this."

David Heinemeier Hansson (DHH to most) is one of these people. When his company 37signals wanted to build an innovative email product, he didn't start by evaluating Gmail API limits or building a wrapper on top of existing email platforms. He built an email server. From scratch. His reasoning was elegantly simple: "If you want to do interesting things with email, you have to own the email server."

This sounds almost quaint in 2025, doesn't it? Like someone suggesting you raise your own chickens instead of buying eggs.

But here's what happened: 37signals pulled their entire infrastructure off AWS. They spent $700,000 on Dell servers (hardware you can actually touch) and saved $2 million in their first year. Over five years, they'll save more than $10 million. Their operations team didn't grow. Their product didn't slow down. They just stopped renting what they could own.

The math is almost offensive: a $350 consumer-grade mini PC provides the same computing power as $1,200 per month on Heroku. The cloud markup isn't a service fee. It's a tax on not thinking.

The Integration Theater

Walk into most tech companies today and you'll find an elaborate performance I call "integration theater."

Everyone's running Salesforce for CRM. HubSpot for marketing. AWS for infrastructure. OpenAI's API for their "proprietary AI." Snowflake for analytics. The technology stack looks identical to their competitors' because, well, it is. They bought it from the same catalog.

Then everyone sits around conference tables wondering why they have no competitive advantage.

The delusion is that excellence comes from picking the right items off the menu. It doesn't. It comes from owning the kitchen.

Netflix's recommendation algorithm drives 80% of viewing time and saves roughly $1 billion annually in reduced churn. Amazon's recommendation system generates 35% of the company's revenue. TikTok's algorithm is valued at over $100 billion (more than most Fortune 500 companies are worth in their entirety).

You can't rent that kind of advantage from a SaaS vendor. You have to build it.

The AI Acceleration

AI is accelerating the commoditization crisis, and most companies are sleepwalking into it.

Two years ago, having "AI-powered" anything was a differentiator. Today? There are companies whose entire business model is "ChatGPT with a nice interface." Industry analysts are openly asking: "Are you just an LLM wrapper? Because you're replaceable now."

The models themselves are becoming commodities. LLaMA is open source. The cost difference between AI providers is basically compute pricing. If your "proprietary AI solution" is just an API call to OpenAI with some prompt engineering, your competitor can replicate your entire value proposition by Tuesday.

But AI is also the great equalizer for building.

Five years ago, you needed a team of specialists to build custom infrastructure. Today, a talented engineer with Claude or Cursor can build in a weekend what used to take months. The barrier to creating proprietary technology is collapsing at the exact moment that renting commodity technology is becoming worthless.

This is the inflection point. The companies that realize they can build are going to pull away from the companies that keep renting. Fast.

Who Wins, Who Loses

The SaaS vendors see this coming. Why do you think there are 702 CRM solutions on G2? Not because the world needs 702 ways to track customer data. Because CRM is so commoditized that differentiation is nearly impossible. Everyone's selling the same thing with different logos.

The vendors are trapped in a feature-parity death spiral. You add a feature. Your competitor copies it in three weeks. Customers start choosing based on price. Margins compress. Everyone loses except the customer, who still doesn't have a competitive advantage because everyone else bought the same stuff.

Meanwhile, companies like 37signals are running on servers they bought six years ago. Still humming. Still paid off. Still creating compounding advantages.

The winners in the next decade will be companies that wake up and ask: "What are we paying millions for annually that we could own for a fraction of that?" The answer is usually shocking.

The losers will be the SaaS vendors who can't answer why anyone should keep paying them when AI makes building so much easier. Watch what happens when a CFO realizes their $2 million annual Salesforce bill could be a one-time $500K custom build that does exactly what they need (and nothing they don't).

The Five-Year Question

Building sucks for the first year. Maybe two years. It's slower. It's harder. You'll have bugs the SaaS vendor already fixed.

But ask a different question: What would you have if you'd spent five years building things only you have?

The answer is the only real moat that exists in 2025: proprietary technology so specific to your business that competitors can't buy it, can't rent it, and can't replicate it without years of their own work.

Modern technology leadership has forgotten patient capital. We think in quarters. In sprints. In OKRs that reset annually. Making a decision that pays off in year four feels almost irresponsible.

But year four is exactly where competitive advantage lives.

When 37signals bought those servers, they bought them. Past tense. Done. The servers keep running. The savings compound. The knowledge compounds. Meanwhile, that AWS bill would arrive every month, forever, growing as usage grew.

What You Actually Own

Could you explain to your board what technology you own that competitors don't?

If the answer involves "our unique Salesforce configuration" or "our sophisticated integration layer," you don't own anything. You're renting shelf space in someone else's store.

Real ownership sounds different: "We built our own recommendation engine because Algolia couldn't do real-time personalization at our scale, and now our conversion rates are 40% higher than category average." Or: "We built our own data pipeline because we needed sub-second latency, and that's why we can offer same-day delivery when competitors take three days."

This isn't romanticism about building everything yourself. Buy commodity stuff. Buy your email service and your calendar and your video conferencing. Buy anything where being different doesn't matter.

But when something is core to how you deliver value? When it's the reason customers choose you? Own it. Build it. Make it yours.

The leaders who get this (who can code, who understand infrastructure, who think in five-year horizons while everyone else thinks in quarters) are going to build companies that are genuinely difficult to compete with.

The ones managing vendor relationships are going to wake up one day and realize they're running the same company as everyone else, just with a different logo on top.

The question isn't whether you can afford to build.

It's whether you can afford to keep renting.