Home Assistant Green, Home Assistant Yellow, and the Raspberry Pi ecosystem form the tertiary tier of the collective's hardware recommendation — appropriate for hobbyist deployments, very small operations, learning environments, and as spare or backup systems. These machines are designed specifically for running Home Assistant (in the case of Green and Yellow) or for general hobbyist computing with Home Assistant OS support as a standard path (in the case of the Pi). They offer the simplest path from purchase to working system. They also have real capacity limits that the collective is clear about: these are not the right starting point for deployments that expect to grow into computer vision, local AI, substantial history, or the graybox pattern. This page covers what each option offers, where each fits for agricultural use, and what the upgrade path looks like when an operation outgrows this tier.
When purpose-built hardware is the right choice.
Several specific situations fit the tier-3 path well.
Hobbyist greenhouses and small operations. A single greenhouse with a handful of sensors, a few automations, and a dashboard on a phone is within Raspberry Pi 4 or Home Assistant Green capacity comfortably. The grower who wants to monitor their home greenhouse or a small operation without building expertise around Linux administration is well-served by an appliance-like solution.
Learning environments. A grower new to Home Assistant who wants to explore the platform before committing. A Raspberry Pi or a Home Assistant Green is a low-stakes way to evaluate whether the approach fits, learn the concepts, and develop operational habits. If the operation grows beyond the tier-3 machine, migration to more capable hardware is straightforward.
Backup or redundant systems. Operations that already run on tier-1 or tier-2 hardware may keep a tier-3 machine as a hot or cold standby. A Home Assistant Green or Raspberry Pi kept on the shelf, ready to restore from backup if the primary machine fails, is cheap insurance against catastrophic hardware failure.
Remote or secondary sites. A multi-site operation with a main farm running full graybox hardware and a secondary site doing basic monitoring may reasonably use tier-3 hardware at the secondary site. The scope at each site drives the capacity decision separately.
Agricultural research or experimentation. A temporary deployment — a research plot, a trial greenhouse, a short-term monitoring project — that will be decommissioned in months to a year is appropriately served by inexpensive hardware that doesn't warrant the investment in tier-1 or tier-2 equipment.
Classroom and training use. For educators teaching agricultural technology, purpose-built hardware is more accessible to students, simpler to support at scale, and cheaper to replace when equipment fails.
These are the situations where tier-3 hardware is genuinely the right choice. For any agricultural deployment intended as working production infrastructure with a multi-year horizon, the primary and secondary recommendations usually fit better.
Home Assistant Green.
Home Assistant Green is the Home Assistant project's consumer-friendly appliance. It is a dedicated Rockchip RK3566-based device in a custom enclosure, sold by Nabu Casa with Home Assistant OS pre-installed and configured for the simplest possible setup — plug in power and ethernet, open a browser, complete the onboarding.
What it is technically. A Rockchip RK3566 SoC (quad-core Arm Cortex-A55) with 4 GB of RAM, 32 GB of eMMC storage (not a microSD card — more reliable for 24/7 operation), a gigabit ethernet port, a few USB ports, and a custom enclosure with appropriate ventilation. Pre-loaded with Home Assistant OS.
Pricing. Typically a modest amount varying with availability and Nabu Casa store pricing.
What it does well. The simplest possible Home Assistant experience. Onboarding takes about ten minutes from opening the box. Updates, backups, and add-on management all work through the Home Assistant interface with no command-line interaction required. The Home Assistant project supports this hardware directly.
What it does less well. The RK3566 with 4 GB RAM is adequate for a small deployment but limiting for larger ones. No room for local AI of any serious kind (LLMs through Ollama are out of scope). Frigate is possible with a USB Coral AI accelerator but capacity is modest. The 32 GB storage is plenty for Home Assistant itself but fills quickly if the deployment includes Frigate recordings or long retention of detailed history. No expansion ports for Zigbee or other radios beyond USB.
When Home Assistant Green fits. Readers who want an appliance experience, operations with small monitoring scope (a dozen or two entities, simple automations, basic dashboards), learning environments, and backup systems. Readers who value simplicity over capacity.
When Home Assistant Green does not fit. Deployments with serious growth plans, operations that want Frigate with multiple cameras, operations that want the graybox pattern, deployments that will include local AI. These cases argue for a more capable machine from the start.
Home Assistant Yellow.
Home Assistant Yellow is the Home Assistant project's more capable purpose-built hardware. Designed around the Raspberry Pi Compute Module 4, Yellow adds features that the standard Pi does not have and is intended for more serious deployments than the Green.
What it is technically. A carrier board that accepts a Raspberry Pi Compute Module 4 (the CM4, a variant of the Pi 4 designed for embedded applications) with configurations from 2 GB to 8 GB of RAM and 8 GB to 32 GB of eMMC storage. An integrated Zigbee and Thread radio (no USB stick required). Power over Ethernet Plus (PoE+) support (optional, with the PoE+ module installed). An M.2 NVMe slot for fast SSD storage. A custom enclosure.
Pricing. The base Yellow Kit is typically moderately priced depending on configuration (CM4 module is usually sold separately, though some bundles include it). Adding a CM4 with appropriate specifications adds a modest amount more. Adding the PoE+ module adds a small additional amount. A fully-configured Yellow typically lands in the mid-range.
What it does well. The integrated Zigbee and Thread radio removes one USB dependency and supports a more robust radio setup than most USB sticks offer. PoE+ power means the device can receive power over its ethernet cable, which is valuable in installations where running AC power is inconvenient. The NVMe slot provides faster storage than the microSD-style storage on most Pi hardware. The enclosure is well-engineered and looks professional.
What it does less well. Despite the polish, the underlying CM4 still has the capacity limits of a Pi 4 — 8 GB RAM maximum, modest CPU compared to a business mini PC, limited capability for heavy AI or Frigate duty. The total cost of a well-configured Yellow is close to what a refurbished business mini PC with substantially more capacity costs, which makes the value proposition questionable for growers prioritizing capability per dollar.
When Home Assistant Yellow fits. Readers who specifically want purpose-built Home Assistant hardware with the Zigbee/Thread radio integrated and PoE+ support. Installations where the PoE+ feature genuinely matters. Growers who prioritize the polish of an appliance-grade solution and are willing to pay the premium for it.
When Home Assistant Yellow does not fit. Growers looking for the most capacity per dollar (a refurbished mini PC wins). Growers with plans for heavy AI workloads. Growers who would value the graybox flexibility (the CM4 is not the right host for a substantial graybox).
Raspberry Pi.
A Raspberry Pi 4 or Raspberry Pi 5 with Home Assistant OS installed is the simplest generic-hardware path. The Pi ecosystem is enormous; the community knowledge around running Home Assistant on a Pi is substantial.
What it is technically. A single-board computer. The Pi 4 comes with 2, 4, or 8 GB of RAM; the Pi 5 comes with 4, 8, or 16 GB. Both have gigabit ethernet, WiFi, multiple USB ports, an HDMI output, and a general-purpose I/O header. Storage is typically a microSD card (not recommended for 24/7 use) or an SSD attached via USB or (on the Pi 5) NVMe via a hat.
Pricing. A Pi 4 or Pi 5 with appropriate RAM, plus a case, plus a proper power supply, plus storage is typically affordable depending on the configuration. Plus an SSD for reliable storage that most Pi deployments should have. Total realistic cost is a moderate amount for a properly configured Pi 4 or Pi 5 for Home Assistant duty.
What it does well. Flexibility. The Pi is a general-purpose computer, so the grower is not tied to Home Assistant — the same machine can run other things (though usually not all at once, at the Pi's modest capacity). Strong community support. Wide availability of accessories (Zigbee coordinators, AI accelerators, camera modules, sensor hats). The Pi 5 specifically is substantially more capable than earlier Pi generations and handles Home Assistant plus a few add-ons comfortably.
What it does less well. The microSD card storage on most Pi setups is a real reliability concern — microSD cards wear out, especially with Home Assistant's database write patterns. The mitigation (booting from an SSD instead) is straightforward but requires upfront thought. The Pi 4's CPU becomes limiting for Frigate at scale or for any serious local AI; the Pi 5 is better but still nowhere near a business mini PC. Heat management on the Pi 4 and Pi 5 requires active cooling for sustained loads (a decent case with a fan is essential).
When a Pi fits. Learning, very small deployments, backup systems, and situations where the grower specifically wants the Pi ecosystem's flexibility. Also a reasonable choice for experimental or research deployments where the Pi's extensive hat ecosystem matters.
When a Pi does not fit. Any serious agricultural deployment with a multi-year horizon and growth expectations. The Pi frequently becomes the limiting factor as the deployment matures, forcing hardware migration within a few years. Starting with more capable hardware avoids that rework.
The Pi's microSD card problem.
Worth calling out specifically because it catches many Pi deployments off-guard.
microSD cards are flash storage designed for cameras, phones, and similar devices — write-once-occasionally, read-many. Home Assistant's database writes state changes many times per minute, which is write-heavy enough to wear out consumer microSD cards within a year or two. The symptoms appear gradually — the system becomes slower, then unreliable, then unbootable — and often surprise growers who assumed the microSD card would last as long as the Pi.
Three mitigations.
Boot from SSD instead of microSD. The Pi 4 supports booting directly from a USB-connected SSD, and the Pi 5 adds native NVMe support through a hat. A a modest amount SSD as the main storage, with no microSD card in use, eliminates the wear problem entirely. This is the recommended configuration for any serious Pi-based deployment.
Use industrial-grade microSD cards if microSD is required. High-endurance cards (marketed for dashcams and continuous video recording) handle the write pattern better than consumer cards. Industrial-grade cards are more robust still. Neither is as reliable as an SSD, but both are better than the common consumer cards.
Aggressive external backups. Whatever the storage choice, external backups stored elsewhere are essential. When the storage fails (not if — when), the backup is what restores the operation. [Backup and Recovery](/home-assistant/operations/backup) covers the practices.
The Home Assistant Green's eMMC storage is not subject to the same wear problem as microSD, which is part of why the Green is a more reliable deployment than a Pi with a microSD card for the same nominal cost.
Setup considerations.
A few practical notes for tier-3 deployments.
Install Home Assistant OS. For all three tier-3 options, installing Home Assistant OS is the standard path. It is pre-installed on Green and Yellow; for a Pi, the Home Assistant project provides installer images that the grower writes to storage. The [Installing Home Assistant OS](/home-assistant/installation/haos) page walks through the Pi installation.
Use wired ethernet. Every tier-3 option supports wired gigabit ethernet, and every tier-3 deployment should use it. WiFi works but adds unnecessary points of failure for a 24/7 service.
Add a UPS or battery backup. Unlike laptops, tier-3 hardware has no built-in battery. A small consumer UPS protects against power events; the runtime can be substantial because these devices draw very little power (5 to 15 watts continuous). For a Pi or Home Assistant Green, a small UPS provides several hours of runtime during outages. For installations where cost matters more than runtime, a USB power bank with pass-through charging provides ride-through for brief outages at minimal cost.
Provide adequate cooling. All three options benefit from cases with effective cooling. The Pi 4 and Pi 5 specifically need active cooling (fan-equipped cases) for sustained loads; passive-only cooling causes thermal throttling that degrades performance and eventually storage reliability.
Budget for upgrade eventually. For hobbyist deployments this is not a concern. For deployments with any growth expectations, anticipate that the tier-3 hardware will be outgrown and budget the eventual migration to tier-1 or tier-2. Home Assistant's backup and restore makes the migration a few-hours operation rather than a reconfiguration project, but the hardware transition should be planned rather than forced by failure.
The upgrade path.
When tier-3 hardware is outgrown, the migration pattern is straightforward.
Take a full backup of the current deployment. Home Assistant's built-in backup captures configuration, history, automations, add-ons, and user data. Store the backup somewhere safe (cloud, network share, external drive).
Acquire new hardware. A repurposed business machine or a refurbished mini PC. The tier decision is the same one covered on [Choosing Your Hardware](/home-assistant/hardware/choosing).
Install the new hardware with Ubuntu Server plus Home Assistant Supervised (the recommended graybox path) or with Home Assistant OS if the grower wants a similarly simple experience on the new hardware.
Restore the backup on the new hardware. The restore process brings the new installation up in the same state as the old one.
Migrate any external dependencies. If the old installation was using integrations that reference the host's IP address, update those references. If the old installation had USB devices (Zigbee coordinator, Bluetooth adapter, AI accelerator), move them to the new hardware.
Decommission the old hardware. Keep it as a spare, or repurpose it elsewhere.
The migration typically takes an afternoon of focused work, assuming current backups and available new hardware. Operations that plan the migration before the tier-3 hardware becomes limiting generally have smooth transitions. Operations that wait until the tier-3 hardware is failing have harder migrations with time pressure.
Comparison to the primary and secondary recommendations.
A direct comparison helps clarify where tier-3 fits.
Capacity per dollar. Tier-1 (repurposed) is best — free or very cheap hardware with substantial capacity. Tier-2 (mini PC, refurbished) is second — a moderate amount for capable hardware. Tier-3 (purpose-built) is third — similar dollar amounts for substantially less capacity.
Ease of setup. Tier-3 wins clearly. Home Assistant Green is the easiest deployment possible. Tier-1 and tier-2 require Ubuntu installation and Docker setup for the graybox path, which takes a few hours of work.
Long-term capacity headroom. Tier-1 and tier-2 both have substantial headroom. Tier-3 has little headroom beyond its initial specifications.
Graybox compatibility. Tier-1 and tier-2 support the full graybox comfortably. Tier-3 does not; it runs Home Assistant effectively but not Home Assistant plus many companion services.
AI and computer vision readiness. Tier-1 and tier-2 can run Frigate with multiple cameras and local LLMs through Ollama. Tier-3 can run basic Frigate (one or two cameras, with a Coral accelerator) but not serious local AI.
Power consumption. Tier-3 is lowest (5 to 15 watts continuous for a Pi or Green), tier-2 is middle (10 to 30 watts), tier-1 is highest (15 to 60 watts for a desktop, 7 to 15 watts for a laptop). All are modest for 24/7 operation.
Upgrade difficulty. Tier-1 and tier-2 have in-place upgrade paths for RAM, storage, and eventually whole-machine replacement. Tier-3 typically requires whole-machine replacement when outgrown.
For a grower whose situation matches the tier-3 fit criteria above, tier-3 is genuinely the right choice. For a grower whose operation is likely to grow, starting with tier-1 or tier-2 avoids the eventual migration.