When the Build Decision Makes Sense — and When It Doesn’t

In 1927, iron ore arrived by freighter at Ford's River Rouge complex, in Dearborn, Michigan, and left the other end as a finished automobile. The plant stretched a mile and a half wide. It had its own steel mill, glass factory, rubber plant, power station, and 100 miles of internal railroad track.

At peak production, 100,000 workers turned raw materials into cars every 49 seconds. Henry Ford's vision was total self-sufficiency: own every input, control every process, depend on no one. And for a while, it worked.

Then the maintenance burden started compounding. The company found itself running a steel operation, a glass operation, a rubber operation, a power utility, and a railroad, all to support what it actually existed to do: design and sell cars. By the 1950s, Ford's grandson had begun unwinding the model. The steel mill was eventually sold off. The glass plant shut down. Today the Rouge occupies less than half its original footprint, fed by a global network of specialized suppliers.

Ford learned something that cost decades and billions to internalize: that building a capability does not mean your organization is built to sustain it.

The same pattern is playing out in financial services

That same impulse shows up every time a financial institution decides to build infrastructure internally. The business case looks compelling at the outset: cost to build versus cost to integrate, estimated delivery time, and projected ROI in year three. And the spreadsheet almost always points toward building, because they capture the cost of creation but rarely models what sustained ownership actually requires.

The real cost surfaces about 18 months later.

The internal team maintaining the build can't attract senior talent because the best infrastructure engineers want to work where infrastructure is the mission.

The quarterly roadmap review allocates 80% of development resources to the core business, with whatever remains going to the internal build. McKinsey estimates that companies pay an additional 10 to 20 percent to address tech debt on top of every project, and 30% of CIOs report that more than a fifth of their new-product budget gets diverted to resolving legacy issues.

Ford could afford to run a steel mill for thirty years before acknowledging the mismatch. Financial institutions on compressed margins have considerably less runway.

What organizations are actually meant to do

Every company is optimized to do a small number of things exceptionally well. For financial institutions, that list usually includes risk management, capital allocation, regulatory compliance, and yield optimization. For infrastructure companies, it includes uptime, observability, API reliability, and activation speed.

That shapes incentive structures, hiring profiles, promotion criteria, and where executive attention concentrates. A bank CEO who prioritizes API latency over yield has misaligned priorities. An infrastructure company trying to manage loan portfolios or make lending decisions is equally out of position.

Those organizational realities are also the reason why non-core capabilities become second-class citizens after they're built. The FI's infrastructure team will always lose the budget fight to the lending team, because lending is what the institution exists to do. And that team should lose that fight. The mistake is greenlighting the build while assuming it will somehow receive first-class attention inside a company that structurally cannot give it that.

Pesh Patel, VP of Personal & Business Banking at CIBC, described this tension from the inside at a closed-door panel in Toronto earlier this year:
‍

"We often think everything that any fintech can do, we can do ourselves. The challenge is: do we really want to? And even if we did, it would take so long that everybody would have moved on by then."

The honest answer to that question, in most infrastructure scenarios, is already embedded in how the organization is built.

When building makes sense

Partnering is the right answer in most infrastructure scenarios, but building internally makes sense when a capability is genuinely differentiating—when the organization would restructure around it and promote executives based on excellence in that domain.

The test is this: would this organization hire a C-level executive whose entire mandate is excellence in this domain? If yes, the capability likely belongs in-house. If that role sits two or three levels below the executive team, the organization has already answered the question about how much long-term attention it will receive.

Goldman Sachs offers a clear illustration of when the answer is yes. In the early 1990s, the firm built SecDB, a proprietary risk management platform capable of aggregating trading risks across the entire firm in real time. It took years to develop and required sustained organizational attention at the most senior levels. But risk modeling and capital deployment are core to what Goldman exists to do — the firm promotes executives based on excellence in exactly that domain. By 2014, Goldman had externalized SecDB's capabilities through its Marquee platform, giving institutional clients access to the same pricing models and risk analytics its own traders use. The internal build became a durable competitive asset precisely because the organization was structured to sustain it.

Where this pattern is heading

By late 2025, Gartner had predicted that at least 30% of generative AI projects would be abandoned after proof of concept, largely because the organizations building them lacked the data infrastructure, talent, and operational focus required to move from prototype to production. The pattern mirrors every previous wave of internal builds: initial feasibility is proven, and then sustained excellence demands more organizational attention than the company can justify directing away from its core mission.

Ford's Rouge complex proved that total self-sufficiency is technically achievable. It also proved that building a capability and sustaining it at excellence are two entirely different organizational commitments. The question worth sitting with is whether the spreadsheet showing year-one savings accounts for what the organization will look like at year five, when the build is competing for attention against everything it was created to do.

‍