Q2 2024. Eight vendor quotes spread across my desk—inverters, a 215 kWh battery system, four 60A EV chargers. Three weeks of TCO analysis. Every line item compared. I picked the lowest price for each component and saved $8,200.
Six months later, I was $12,400 over budget on integration costs. The savings didn't survive the first year.
The Quote That Looked Clean
I'm a procurement manager at a 340-person facilities services company. I've managed our energy equipment budget ($180,000 annually) for six years, negotiated with 40+ vendors, and tracked every order in a spreadsheet my team calls "the bible."
For this project—a commercial facility in the Midwest—I ran the numbers on Huawei's Sun2000 against Tesla's String Inverter, plus four other brands. Both had similar specs on paper. The Huawei unit came in 6% lower, and the local distributor had better stock.
The battery was a different brand (cheaper per kWh). The chargers were a third (lowest of the four quotes). Total savings from splitting the order: $8,200. That's 11% of the project budget.
In my opinion, that's a clean win. It wasn't.
The Problem Isn't the Hardware. It's the Conversations It Won't Have.
By November, I had three monitoring apps on my phone. Three.
The Huawei app store listing for FusionSolar looked solid—4.7 rating, updated three times in the past year. And it worked well for the inverter. String-level data, clean interface, responsive. But it only saw Huawei devices. The battery had its own portal. The chargers had a third dashboard that hadn't been updated since 2022.
It's tempting to think you can just pick the best price per component. But inverters, batteries, and chargers aren't appliances—they're network devices with opinions about protocols. Mixed-vendor systems rarely fail because of the hardware. They fail because of the gaps between devices.
Here's what the quotes didn't show:
- The inverter and battery used different Modbus register maps. Getting them on speaking terms required a third-party gateway ($2,800—which, honestly, felt excessive for a box that just translates ones and zeros).
- The EV chargers' load management couldn't read the battery's state of charge. Peak shaving and charging schedules operated in separate silos.
- Every firmware update risked breaking a handshake that nobody had tested together.
Note to self: add "protocol compatibility" to the RFP template. I keep meaning to do that.
The Continuous-Load Math on 60A Chargers
This part trips up a lot of facility managers, so let me be precise.
A 60A EV charger doesn't draw 60A continuously. It draws 60A when a vehicle is actively fast-charging—and then it tapers. So it's tempting to size the circuit at 60A. That's what the box says. The "charging only" scenario is what most people picture.
But per NEC Article 625 (NFPA 70, 2023 edition), EV supply equipment is classified as a continuous load. For branch circuit sizing, that means a 125% multiplier: a 60A charger requires a 75A circuit. Code doesn't care how you use it—it cares about the worst-case sustained draw.
If your panel schedule was built around 60A per charger—which mine was, because that's the number on the spec sheet—you're looking at a service upgrade or a load management system. In our case: $3,100 in extra panel work that appeared on no equipment quote.
I still kick myself for not asking for the load calculation upfront. "60A charger" is not the same as "75A circuit." The difference lives in code, not in the sales sheet.
A 3V Lithium Battery Almost Cost Us Five Figures
This one is embarrassing, so I'll keep it short.
Our bill of materials had a line item for "lithium battery—3V." That's a standard CR2032 coin cell, used in the inverter for RTC backup. Costs about $0.40.
Our purchasing system categorized it under "batteries—lithium," which is also where the 215 kWh storage system was filed. Different line item, same category. A junior buyer almost issued a PO for the wrong one—a five-figure mistake hiding behind a three-character voltage rating.
Not a vendor problem. A taxonomy problem. But it's the kind of thing that surfaces when you're managing 40+ line items across four vendors and every spec sheet uses slightly different language.
What Three Days of Downtime Actually Costs
Month four. The battery stopped responding to a grid signal. The inverter's firmware couldn't interpret the battery's response. Each vendor's support team pointed at the other's device.
Huawei Support was responsive on the inverter side—they provided logs within 24 hours. But they couldn't diagnose a battery they didn't make. To be fair, the battery vendor was also helpful once we escalated. But first-line tech support had no context for a mixed system. Why would they?
We lost three days. In a commercial facility, that's not just an operations problem. Tenants notice. Our facility manager fielded six complaint calls in the first week. In my opinion, that's the cost nobody calculates—the perception hit when your building goes dark for reasons nobody can explain.
Tally for year one:
- Gateway hardware: $2,800
- Integration labor: $1,900
- Extra monitoring: 12 hours/month (about $4,300 annually)
- Panel upgrade: $3,100
- Downtime and tenant complaints: hard to price, harder to forget
Total: over $12,000. Against $8,200 in component savings.
The Policy I Now Follow
Not every project needs a single-vendor stack—that's an oversimplification. Some components genuinely don't need to talk to each other. But for energy systems, the dependencies are real.
I now split procurement into two buckets:
- Interdependent systems—inverter, battery, and management platform. One vendor. One protocol. One support line.
- Independent loads—EV chargers, lighting, HVAC. Can be sourced separately if the load management layer is open and documented.
Two questions before I sign anything:
- "What protocol does this speak, and what does it need to talk to?"
- "Show me the continuous-load calculation, not just the rated capacity."
On our next facility—Q3 2025—I'm pricing Huawei's full stack: Sun2000 inverters, Luna2000 storage, FusionSolar for a single dashboard. Not because it's the cheapest line by line. Because when something breaks at 2 a.m., I want one number to call.
Most of these problems are avoidable—not with a bigger budget, but with a better RFP.
Simple. But it took a $12,400 lesson to learn.
Ask for engineering context