In January 2024, I replaced 80 halogen downlights across three office locations with LED retrofits. Eight months later, 14 of 80 were dead. The LED chips were fine—some even carried Cree-branded diodes. The failure was in the driver, the power supply component most buyers never think about. That first batch, which I bought because it was 35% cheaper than the premium option, ended up costing more than the good batch that replaced it. If you're researching downlight halogen replacements for a commercial space, here's what I wish someone had told me before I started.
A lighting retrofit is only as good as its drivers. Cheap bulbs with good chips still fail if the driver can't handle the heat. The second batch, with properly designed drivers, cost more upfront but ran for over a year with zero failures. The expensive batch was cheaper in total cost. That's the whole lesson, and the math behind it is worth understanding.
What's an LED Driver, Anyway?
LED chips get the attention, but the driver is what dies. An LED driver converts building AC power into the steady DC current the chip needs. If the driver is poor quality—or poorly suited to an enclosed, recessed fixture—it fails early and quietly. The chip never gets a chance to show its rated 50,000-hour lifespan.
So, practically speaking: the driver is the brain of the bulb. The chip is the heart. You need both, and the cheapest bulbs usually economize on the brain. That's why I now check the driver before I check anything else.
Why I Know This
I'm the office administrator for a 120-person professional services company. I manage facilities purchasing—roughly $180,000 a year across 11 vendors. I report to both operations and finance, which means I see the full cost picture: purchase orders, expense rejections, and the hidden labor of dealing with things that break. When I took over purchasing in 2020, I inherited a lighting system that hadn't been touched since the building was fitted out in 2016. Our 2023 energy audit flagged lighting as nearly 20% of utility spend. The halogen downlights in the recessed ceiling grid were running 10+ hours a day in conference rooms and common areas. The math said retrofit. The execution taught me things the math didn't.
How the First Batch Failed
When I started planning the retrofit, I assumed LED bulbs were basically commodities. Same sockets, similar specs, comparable quality. The $6.40-per-unit price on a batch of 80 was attractive—about 35% below the mid-range options I was considering, and the spec sheet looked fine on paper: 10W, 85 CRI, 4000K, a five-year warranty. The supplier's catalog mentioned Cree LED components. I figured that settled it. Cree is a well-known name in LED chips, and if the chips were good, the bulbs were good.
That assumption cost me.
By September, 14 of 80 units had failed. The pattern was telling. Every failure was in an enclosed recessed fixture. Every survivor was in an open one. Heat was killing the bulbs, and the heat was killing the drivers, not the chips. I pulled one apart—so glad I kept a failed unit instead of mailing it back for warranty. Inside was a driver board roughly a third the size of the one in the quality batch, with no thermal protection worth mentioning. Components chosen for cost, not longevity. A lesson learned the hard way.
The warranty did cover replacements, technically. After a week of emails and photos. But it didn't cover the labor of climbing ladders, the color mismatch of the replacement batch (different production run, slightly different tint), or the awkward conversation with our VP of finance about why the five-year warranty bulbs failed in eight months.
The Second Round: What Different Looks Like
The second order was for 88 units at $9.80 each. Same wattage, similar CRI, same 4000K color temperature. On the spec sheet, barely different. Inside, a different world:
- Driver boards with proper thermal management for enclosed fixtures
- Higher-grade capacitors and inductors from known manufacturers
- Overload protection that shuts the driver down instead of letting it cook
- Dimmable driver versions for the two zones we needed them in
That last line cost us an extra $2 per unit in the meeting rooms. It saved us from trying to explain subtle flicker to a client during a board presentation. Worth every cent.
The second batch has run for 14 months (as of February 2025, at least)—roughly 3,500 hours per fixture—with zero failures. Not one. The surprise wasn't that the quality batch lasted. It was that the cheap batch's failure rate was so predictable. Same fixture types, same heat exposure, same failure pattern. The only variable was the driver.
The Total Cost That Mattered
Here's the math I put in front of finance in January 2025. This is the total cost of ownership (TCO) view that should drive any lighting purchase decision.
Halogen baseline:
- Energy: 80 fixtures × 50W = 10,000 kWh/year ≈ $1,500
- Lamp replacements: ~100 halogen bulbs/year ≈ $350, plus ladder time
- Total: roughly $2,150/year
Cheap LED batch (first order):
- Energy: 80 × 10W = 2,000 kWh/year ≈ $300
- Upfront: 80 × $6.40 = $512
- Failures in year one: 14 bulbs—warranty covered the units, but not the $230 in labor, color-matching rework, and shipping
- Year-one total: ≈ $1,040
Quality LED batch (second order):
- Energy: same $300
- Upfront: 88 × $9.80 = $862
- Failures: 0. Maintenance: 0
- Year-one total: $1,162
In year one, the cheap batch looks cheaper by $122. Year two, the story flips. If even half the remaining cheap units fail over the next 18 months—consistent with the first-year pattern—you're looking at more labor, more color matching, more disposal fees. The cheap batch's five-year cost lands above the quality batch, with worse light consistency and a harder conversation with finance. The quality batch costs $300/year after year one. Done.
The lesson isn't 'always buy premium.' The lesson is: compare the full cost curve, not the unit price. Marketing catalogs sell on price per unit. Reality bills you in failures per year.
Cree, Quality Signals, and the Spotlight Test
Let me be clear: I'm not blaming Cree for the first batch's failures. The chips in those bulbs were sourced from Cree, and they performed fine. The problem was the driver, which came from whoever the bulb manufacturer chose. Cree makes excellent LED chips—the XHP and XT-E families have been used reliably across commercial and industrial lighting for years. That's why I ordered spares with Cree chips in the second batch.
If you follow Cree LED news, you know the brand has been through changes—the Wolfspeed spinoff in 2021, the move under SMART Global Holdings. For a corporate buyer, the reassuring part is product stability. The chip families we specified are still available, and the quality reputation has held. That consistency is a real data point. The same chip class appears in premium downlights, in stage fixtures, and in a friend's custom bicycle headlights. He swears by the Cree Q5 LED bicycle lights he's been building for years—the color consistency and efficiency never let him down. When the same component is trusted by commercial fixture makers and hobbyist bicycle lighting enthusiasts, that's a reliability signal. It means the chip is trustworthy.
But a good chip in a bad bulb is like a good engine in a car with a junk transmission. Our tenant theater company reinforced this. Their technical director spent last spring relamping the stage spotlight system and encountered the same pattern: fixtures with quality drivers and known-brand chips stayed consistent through long rehearsals, while budget fixtures drifted in color as they warmed up. For a spotlight actor under a focused beam, that drift is a career-ruining distraction. In our conference room, it was a flicker during a client presentation. Same failure mode, different scale.
When I'd Do It Differently
Not every project needs the premium driver. Here's where I'd tell you to ignore my advice:
- If you're replacing fewer than ten bulbs in open, ventilated fixtures—the risk is lower and the payback window is too long to justify premium pricing. Buy mid-range and move on.
- If your space rents under a full-service lease—your landlord may own the lighting maintenance. Your incentive structure changes completely. (We're in a triple-net lease, so it's our bill. That changes everything.)
- If you're in a temporary space with a payback horizon under a year—the math might not favor any retrofit. Run the numbers first.
- If your fixtures are on dimmers, always check the driver compatibility list before you buy. We had to swap six units to dimmable drivers at the last minute. That was a $12 lesson in reading spec sheets thoroughly. (Honestly, it should have been a free one.)
Looking back, I should have bought the quality batch first. I didn't because the $3.40-per-unit delta looked like margin padding. Now I know it was insurance. But given what I knew then—that all LEDs looked the same on paper—the initial mistake was understandable. The second time around is when experience starts paying for itself. That's the TCO of knowledge, I suppose.
If your retrofit is bigger than a few bulbs, do the full math. Consider the driver. Check the dimmer compatibility. Order spares from the same batch. And remember: the cheapest quote is the beginning of the conversation, not the end. The total cost is what you actually pay.