This is v0.1 of the Modern Hardware Development Stack map. Living document, updated as the ground moves. The goal is not encyclopedic coverage; it is a defensible, opinionated shape of the space that a builder can use to decide where to build and an operator can use to decide what to buy.
How to read this map
Seven layers, bottom up. Each layer is in a different state of code-nativeness -- meaning how much of that layer can be represented in version control, automated in CI, and diffed like code. That is the single dimension that matters most for where the stack is going.
For each layer we name the category, the incumbents, the insurgents, and the gap worth building into. This is a map, not a buyer's guide. It is opinionated on purpose.
Layer 1 -- Requirements and specs
The top of the stack is also the least code-native. Requirements still live mostly in Word, Confluence, and PDF. Traceability is usually a spreadsheet.
- Incumbents: Jama, DOORS, Polarion, Confluence.
- Insurgents: ReqIF-as-code tools, open-source traceability DSLs, git-hosted spec formats.
- Gap: A requirements layer that is actually code -- diff-able, reviewable, machine-linked to tests and manufacturing artifacts. The tool that wins this layer owns the trace graph of the entire product realization stack.
Layer 2 -- Mechanical and electrical design (CAD / EDA)
The design layer is code-native in patches. Hardware description languages for silicon have been code-native for decades; PCB capture and mechanical CAD are still mostly binary files in vendor formats.
- Incumbents: Synopsys, Cadence, Siemens EDA, Altium, SolidWorks, Autodesk, PTC.
- Insurgents: KiCad, OpenROAD, Build123d and CadQuery, Fusion Electronics APIs, Quilter, JITX, Yosys-based flows.
- Gap: A code-first PCB capture tool that is actually adopted by production teams, not just hobbyists. Also: design-as-code for mechanical with the ergonomic parity of modern software build systems.
Layer 3 -- Simulation, verification, validation
Verification has been code-native the longest, because verification is software. SystemVerilog, UVM, cocotb, Verilator. This layer is ahead of every layer above it.
- Incumbents: VCS, Xcelium, Questa, Calibre, PrimeTime.
- Insurgents: Verilator, open-source formal, AI-assisted DV stimulus generation, cloud simulation farms.
- Gap: Verification-as-code extending cleanly into the mechanical and thermal and EMI domains, where most teams still rely on a quarterly physical validation run.
Layer 4 -- Test
Test is the control plane of modern hardware development. It belongs in CI on every commit, not on a lab bench once a week. The stack is split four ways -- unit (simulator), integration (hardware-in-the-loop), system (on-target), regulatory (compliance) -- and each sub-layer is in a different state.
- Incumbents: NI TestStand, Keysight, Rohde & Schwarz, Tektronix, Teradyne, Advantest.
- Insurgents: Software-defined instruments, HIL frameworks in CI, open ATE toolchains, AI-assisted test authoring, cloud-hosted lab benches.
- Gap: A "CI for hardware" primitive that orchestrates HIL runs across a fleet of instruments and returns a pass/fail in minutes. This is probably the single biggest greenfield in the stack.
This sub-space will likely get its own Stack Map: Shift-Left Testing Stack.
Layer 5 -- Build and release (supply chain, BOM, sourcing)
The supply layer runs on spreadsheets and distributor portals. Every team that ships hardware wrote their own script against Octopart or Digi-Key at some point. That alone says the layer is unfinished.
- Incumbents: SAP, Oracle, Arena, Agile, Netsuite, Epicor. Distributor portals.
- Insurgents: Octopart, Nexar, Cofactr, Z2Data, direct distributor APIs, open BOM formats, AI-assisted substitution search.
- Gap: A BOM management layer with first-class code-native BOM formats, real-time availability and compliance data, and substitution suggestions that respect electrical compatibility rather than just drop-in parametric match.
This sub-space will likely get its own Stack Map: BOM / Supply Chain Intelligence Stack.
Layer 6 -- Manufacture
Manufacturing is where the incumbents are strongest and the shift is subtlest. MFG instructions are becoming code; contract manufacturers are starting to expose APIs; traveler-as-code is real in spots.
- Incumbents: Tulip, Aegis, iBase-t, FactoryLogix, legacy MES deployments.
- Insurgents: API-first contract manufacturers (in the startup-friendly tier), MFG-instructions-as-code, open MES stacks, factory-in-a-container frameworks.
- Gap: A full code-native path from design release to first-unit build, where every MFG artifact -- AOI programs, ICT programs, functional test binaries, traveler steps -- is generated from source and diffs cleanly across revisions.
Layer 7 -- Field and lifecycle
Once the unit ships, the software side has won. Telemetry, OTA, field analytics, device management -- this layer operates at software cadence already.
- Incumbents: Cloud IoT platforms (AWS, Azure, GCP), Memfault, Golioth, established PLM field modules.
- Insurgents: Open-source device management, edge-to-cloud pipelines with code-native schemas, AI-assisted failure mode attribution.
- Gap: Closing the loop from field failure back into the requirements and test layers at the top of the stack. Today that loop is mostly manual.
What is load-bearing
The layers that ship the most surprise in the next three years, in order:
- Test -- if it becomes CI-native, entire business models rewrite.
- Requirements and specs -- the trace graph is the substrate for every other automation; whoever owns it owns a lot.
- Supply and BOM -- the data is opening, the workflow is not yet, the gap is productizable.
- PCB capture -- the most visible GUI-bound workflow left in mainstream hardware; code-first is coming.
Changelog
- 2026-04-24 -- v0.1 published. Opening layer model, named incumbents and insurgents per layer, named the three biggest load-bearing gaps. To refine: per-layer detail in follow-up Stack Maps, concrete examples of layer-wins in the wild.