"Software-defined video walls have lower TCO" is the marketing line of the decade. It is sometimes true, sometimes false, and almost never true for the reasons the marketing deck gives. This article does the BOM math for a concrete 16-display NOC over five years on both architecture stacks, lists the costs neither vendor side puts in the slide, and finishes with the conditions under which each approach actually wins on total cost.
The mistake most buyers make
The standard comparison line is "hardware controllers cost $X per display, software costs $Y per display, Y is smaller, ergo software wins on TCO". This frames the choice as a licensing-price problem. It is not. The real-world TCO breakdown across a five-year horizon looks roughly like this:
- 30-45% initial hardware CAPEX (controllers, capture cards, servers, displays themselves)
- 15-25% software licensing and support contracts over five years
- 20-30% integrator labour for design, commissioning, training, and first-year hand-holding
- 10-20% operational costs: spares, power, cooling, IT staff time per incident
- 5-15% refresh / EOL / migration cost at year-3 to year-5
Software licensing is in there, but it is a minority share. The architectural choice moves money between the other categories more than it moves the licence column itself. That is where the actual TCO story lives.
What a hardware stack actually costs
A representative hardware video-wall stack for a 16-display NOC with 12 active sources builds out roughly like this. Numbers come from public reseller listings (B&H, PCNation, OfficeWonderland, NetworkHardwares) and tendered procurement docs where prices are public.
- Controller hardware — Datapath WallControl 10 server with one VSN1192 chassis, or two cascaded Fx4 units, or a Matrox MuraControl-IPX equivalent. Mid-range BOM: €15,000-25,000 depending on capture-card mix.
- Capture cards — for 12 sources, typically two to four cards (DVI, HDMI, IP at various counts). €4,000 to €9,000 in cards.
- Wall control software — Datapath WallControl 10 Pro, RGB Spectrum Zio GUI, or equivalent. Some are included with the controller (the "free" claim); others run €2,000-4,000 separately.
- Operator workstation — Dedicated control PC the operator uses to drive the wall. €1,500-3,000 with the necessary peripherals.
- Annual support — Vendor support contract, typically 18% to 22% of hardware CAPEX per year. €4,000-7,000 per year.
- Year-3 refresh component — Most hardware controllers see EOL announcements at the 3-5 year mark (Christie Phoenix Dec 2025, RGB Spectrum Galileo Dec 2025 are recent examples). Plan a refresh cost equal to roughly 40-60% of the original controller BOM.
Sum the 5-year line items: roughly €55,000-90,000 total cost of ownership for the hardware stack, not counting the 16 displays themselves (€32,000-80,000 in panels — same on both stacks).
What a software stack actually costs
The equivalent software stack for the same 16-display NOC with 12 sources looks like this:
- Server — Commodity workstation with an AMD Ryzen 7 / Threadripper or Intel Xeon, an NVIDIA RTX 4070 / A5000- class GPU, 64-128 GB RAM, NVMe storage. €3,000-6,000 depending on redundancy.
- OS — Ubuntu / Debian / RHEL. €0 to ~€800/year depending on whether the buyer wants a paid support contract from Canonical or Red Hat.
- Video wall software licence — Three concrete points on the curve: Craft Wall €2,500 perpetual per canvas, Hiperwall Essentials < €9,000 bundle, Userful Enterprise subscription ≈ €500 per display per year (verified by Austin TX contract CT-2200-24121300113 at $22,737/year for 40+ displays — normalises to ≈ $470/display/year).
- Operator workstations — Optional. Browser-controlled software walls let any existing PC be the operator interface; the dedicated-PC line item disappears.
- Annual support — Optional for perpetual-licence products (Craft Wall, Hiperwall); built into the subscription for Userful. €0 to ≈ 20% of licence depending on vendor.
- Year-3 refresh — Commodity-hardware refresh is optional and incremental. GPU upgrade for €1,500-3,000 buys three more years of performance. No mandatory full-stack rebuild.
Sum the 5-year line items for a Craft Wall-style perpetual-licence stack: roughly €12,000-22,000 total cost of ownership. For a Userful- subscription stack on the same wall: roughly €110,000-140,000 — because the per-display annual fee runs for five years and adds up faster than the hardware controller depreciates.
The Userful subscription number deserves a closer look
The Austin TX city contract reference cited in the Craft Wall vs Userful comparison prices a 40-display deployment at $22,737 per year for the Enterprise Software Subscription. That normalises to roughly $470 per display per year before any add-ons. Over five years on a 16-display wall: 16 × $470 × 5 = $37,600 in licence alone. Add the platform server (Userful runs on x86 Linux too, so the hardware story is similar to the perpetual-licence software stacks), and the total Userful 5-year TCO sits north of the perpetual- licence software stacks by a wide margin — even though Userful is itself a software stack on commodity hardware.
The point: "software" is not a TCO category. It depends on the pricing model. Perpetual licensing reads differently from per-display subscription even when both run on the same commodity Linux box.
Hidden costs that neither marketing deck mentions
Hardware stack hidden costs
- Forced refresh on vendor EOL. When the vendor announces end-of-life on the controller family (Christie Phoenix EOP Dec 2025, RGB Spectrum Galileo EOL Dec 2025), the customer's choice is upgrade now or live without firmware patches. This is a real BOM line, not a "maybe".
- Spare-parts inventory. Critical sites carry a spare controller on the shelf. €5,000-15,000 of capital tied up in spares for the project life.
- Training tied to the vendor UI. Operator and integrator training is specific to the controller's management application. Switching vendors means retraining.
- Capture-card scaling. Adding sources beyond the initial design usually means new capture cards, sometimes a controller chassis swap. Linear cost growth, not amortised.
Software stack hidden costs
- GPU sizing miscalibration. A software wall on an undersized GPU drops frames at peak source count. Fixing means a workstation swap, which is cheaper than a controller swap but still €1,500-3,000.
- Operator training on a new paradigm. Browser-based control is different from a dedicated console. Existing operators need a half-day to be productive on a software wall; integrators charge for that.
- Network engineering for AV-over-IP. If the source mix moves from baseband HDMI to NDI / IPMX / SDVoE, the network team owns more of the deployment than they used to. Adds 10-15% to the integrator labour line.
- Vendor lock-in via SaaS cloud control plane. Some "software-defined" vendors require a cloud-side control plane that holds the customer's config. If the SaaS contract lapses or the vendor pivots, the wall becomes much harder to reconfigure. On-prem software stacks (Craft Wall, Hiperwall) avoid this; cloud-anchored ones (Userful's newer tier) inherit it.
Where the TCO inversion is real
The honest summary, after stripping the marketing on both sides:
- Perpetual-licence software on commodity Linux beats every hardware stack on 5-year TCO for the 16-display NOC scenario above — usually by 60-80%, often by more once year-3 hardware refresh costs land. This is the real "software- defined wins" case.
- Per-display subscription software does not necessarily beat a hardware stack on TCO past three years — the annual fee keeps running while the hardware depreciates. Whether the subscription wins depends heavily on operator- count, support intensity, and how often the buyer would have refreshed the hardware anyway.
- Hardware stacks still win on specific workloads — most often sub-frame latency requirements (trading floors, broadcast OB), mission-critical deployments with 15-20 year support horizons (WEY, Barco), and tenders that explicitly require FPGA-based AV switching for cybersecurity reasons.
Decision tree for your own project
Five questions to ask before signing the BOM:
- What is the source count today and the projected count at year 3? If it more than doubles, a software stack absorbs that growth far cheaper than a hardware stack with capture cards.
- What is the operator-seat count today and at year 3? Per-display subscription software (Userful tier) gets expensive past a few dozen seats; perpetual licensing (Craft Wall, Hiperwall) stays flat regardless.
- What is the latency budget for source → display? Sub-frame requirements still favour hardware. Single-frame tolerance is fine on software walls.
- What is the support horizon the buyer actually needs? Year-20 support commitments are real value when they exist; pay for hardware if the use case justifies the premium.
- What does the IT team look like? A site with a real IT operations team absorbs software walls cleanly. A site without one is better served by a hardware vendor's annual support contract — the operational labour shifts off the customer.
Where Craft Wall sits in this picture
Craft Wall is the perpetual-licence software stack option. €2,500 per canvas flat, on commodity Linux, runs the same AV-over-IP sources as the rest of the software-defined category (NDI, IP-KVM, RTSP, HTML5, IPMX on the roadmap). For the 16-display NOC case above, this is the lowest-TCO option on the table by a wide margin — typically 60-80% below hardware stacks and a fraction of the per-display subscription alternatives.
We are not the right answer for every deployment. Sub-frame latency control rooms, broadcast facilities with ST 2110 requirements, and tenders specifying FPGA hardware all push you toward the Tier 1 hardware vendors. The honest framing is "Craft Wall is the right answer for the workloads where software- defined wins on TCO — not the right answer for every wall in the building".
Run the numbers for your specific case in the interactive TCO calculator — it implements the model from this article with your own source count, display count, and operator count. Read next: IPMX vs ST 2110 vs SDVoE for the transport question underneath, Craft Wall vs Userful for the subscription-vs-perpetual angle, and Craft Wall vs Datapath for the canonical hardware-vs-software comparison.