the exit #001: BGIS and the consequence premium

BGIS did not stop being a facilities company. It spent years moving closer to the failures its customers cannot afford.

The deal

Some companies get paid for the work they do.

Others get paid for what never happens.

A lobby can stay dirty until Tuesday. A data center cannot stay dark until lunch.

Both problems fall under the enormous umbrella called facilities management. Both involve buildings, equipment, technicians, schedules, and maintenance. But economically, they are different species.

In the first business, the customer is buying a task.

In the second, the customer is buying the absence of a catastrophe.

That difference is the story of BGIS.

On July 15, Veritas Capital agreed to acquire a majority stake in BGIS from CCMP Capital and AIMCo. At first glance, the match looks odd. Veritas manages more than $54 billion and has spent decades investing where technology meets government: healthcare systems, defense software, national-security infrastructure, and other things a country would prefer not to discover are broken.

BGIS manages buildings.

It fixes HVAC systems. It runs maintenance programs. Its work includes the deeply uncinematic business of keeping thousands of offices, hospitals, retail sites, government facilities, and data centers operating.

Look at the company that way and the deal is confusing.

Look at the cost of the failures BGIS prevents and it starts to feel almost obvious.

CCMP acquired BGIS from Brookfield in 2019 for approximately US$1 billion (about C$1.3 billion).

At the time, BGIS managed more than 320 million square feet across roughly 30,000 locations. Today, it manages more than 620 million square feet across 65,000 facilities.

That is a lot of growth. It is not, by itself, the interesting part.

If you double a conventional services business, you often end up with a larger conventional services business. More customers and more employees are useful, but scale alone does not change what the market believes the company is.

The more important question is what kind of square footage BGIS added, what it was hired to do inside those buildings, and how painful it would be for a customer if BGIS failed.

This is where the public record gets interesting.

BGIS had been building critical-environments capabilities before CCMP arrived. It expanded its US data-center operations through McKinstry FMS in 2016. In 2018, it acquired Critical Solutions Group, adding commissioning, testing, and consulting capabilities for data centers.

CCMP did not invent that strategy. It inherited one with a meaningful head start.

Then it kept going.

In 2020, BGIS acquired Schneider Electric's Critical Facility Operations division, expanding from the mechanical systems around a data center into the white space where the servers and network equipment live. In 2021, it acquired Cormant, a data-center infrastructure management software company.

That sequence matters.

BGIS was no longer just maintaining the building around the technology. It was moving closer to the technology itself, from commissioning to monitoring to day-to-day operations.

Still facilities management, technically.

Just with much more expensive consequences.

Customers negotiate hard over work they believe is interchangeable.

They behave differently around work they cannot afford to get wrong.

That creates the consequence premium.

When failure is annoying, you are a vendor. Your customer can rebid the contract, push on price, or try someone new.

When failure shuts down a hospital, compromises a government facility, or takes a data center offline, the purchasing decision changes. Track record matters more. Technical depth matters more. Trust matters more. The theoretical savings from switching providers get compared with the very real cost of a bad handoff.

The work may still be performed by people carrying tools into a building. But the customer is no longer paying only for the labor.

It is paying for continuity.

That does not magically turn a services company into a software company. BGIS still has a large technical workforce, and Veritas has not disclosed what it is paying. There is no public evidence that this deal received a particular EBITDA multiple.

Public comps are imperfect, but they show the size of the bet. In KPMG's Fall 2025 facilities-services update, facility-management companies traded at a median 9.2x EBITDA as of August 2025. Houlihan Lokey's data-center and managed-services comps traded at 21.3x at the end of 2025. BGIS fits neatly into neither bucket, but moving even partway between them can change the economics of a hold.

The direction of travel is clear.

BGIS has spent years moving its mix toward technical, mission-critical environments and surrounding that work with software, monitoring, engineering, and data. The more important uptime becomes, the harder it is to judge the company by labor hours alone.

There are usually two clocks running inside a private equity hold.

The first is the capability clock.

It runs slowly. Companies hire specialists, earn trust, integrate acquisitions, win contracts, make mistakes, and build a reputation one customer at a time. You cannot announce your way into being trusted with a data center. The building either stays online or it does not.

At BGIS, that clock started before CCMP and continued throughout its ownership. The McKinstry and Critical Solutions Group acquisitions established the base. Schneider's Critical Facility Operations business expanded the operating footprint. Cormant added software and visibility into the infrastructure.

The second is the narrative clock.

It runs faster, especially as a sale approaches. Capabilities that took years to build get compressed into phrases such as "proprietary technology," "AI-enabled solutions," and "mission-critical environments."

Those phrases are not necessarily empty. But they are summaries, not explanations.

The best exit story is usually not invented at exit. It is the shortest possible description of a much longer build.

The terms are private, so the public record cannot tell us whether BGIS earned a specific re-rating. The buyer is still a useful clue.

Veritas describes itself as an investor at the intersection of technology and government. Its portfolio is full of companies serving healthcare, education, national security, and other environments where reliability is not a nice-to-have.

In announcing the deal, Veritas said it plans to accelerate BGIS's digital and AI roadmap, deepen its service capabilities, and pursue additional M&A.

That wording suggests two things.

First, Veritas sees enough real technology and specialized human capital inside BGIS to fit its existing playbook.

Second, it does not think the transformation is finished.

The software is not the whole thesis. It may be the next layer of it. BGIS has already built the relationships, operating footprint, technical workforce, and access to the facilities. Veritas now gets to ask what happens when more of the knowledge inside that workforce becomes software, more failures are predicted remotely, and more of the operation is managed through a common digital layer.

That is more useful than asking whether BGIS is "really" a technology company. It is also probably closer to what Veritas is underwriting.

The broad shape of the hold is visible. The most important details are not.

Only the people involved can answer:

Those answers would tell us whether BGIS merely became larger or whether it became a meaningfully different economic asset.

The public record points to the latter.

Not because facilities management disappeared. It did not.

Because BGIS moved closer to the part of the building where failure becomes intolerable. It added the people, systems, and software required to take responsibility for that risk. And it found a buyer whose entire investment strategy is built around operations that cannot be allowed to fail.

The lesson is not to put an AI label on a labor business.

It is to change what the customer trusts you with.

The closer your work gets to the consequence, the less interchangeable you become.

And that is often where the multiple is hiding.