Doc #134 — Computing Self-Sufficiency Roadmap

---

RECOMMENDED ACTIONS (BY ACTUAL URGENCY)

Phase 1 (Months 0–12)

1. Include all computing devices in the national asset census (Doc #8). Establish the actual installed base: servers, desktops, laptops, tablets, smartphones, industrial controllers, embedded systems, networking equipment, test and measurement instruments. Capture model, age, condition, and battery status. Cost of delay: low — devices are not being consumed rapidly in the first months, and other census categories (fuel, food, medical) take priority.

2. Identify and secure specialist electronics repair capability. NZ has a small but real population of electronics technicians, IT professionals, and hobbyist repairers. The skills census should capture: soldering and board-level repair, component testing, operating system installation and maintenance, embedded systems programming, and radio/electronics engineering. These skills become irreplaceable as imported training materials deplete.

3. Preserve the AI inference facility (Doc #129) as NZ’s highest-value computing asset. Whatever hardware is running AI inference models is orders of magnitude more capable than any computing NZ will build for decades. Every day of operation produces knowledge that can be printed and distributed permanently.

4. Begin printing computing reference materials. While printers and toner last (Doc #5, Doc #29), print: programming manuals, circuit design references, semiconductor physics texts, algorithm textbooks, repair schematics for common hardware, and the complete technical content of Doc #5. This material must exist in durable physical form before the last printer fails.

Phase 2–3 (Years 1–7)

5. Establish a national device repair and cannibalisation programme (Doc #88). Train repair technicians. Centralise spare parts from non-functional devices. Develop standardised triage: which devices are worth repairing, which are parts donors, which are beyond salvage. This is the single highest-impact action for extending the computing bridge.

6. Develop a device allocation framework. As the functional device count declines, allocate remaining computers to highest-value uses: national logistics optimisation, medical records, agricultural planning, engineering calculations, navigation, education. Personal computing becomes a lower priority as devices become scarce.

7. Establish the human computation bureau. Trained clerks performing systematic arithmetic using printed algorithm guides. This capability must be functional before the last computers fail, not after. See Doc #157, Part I for staffing and throughput analysis.

8. Begin electronics fundamentals education. University and trade training programmes (Doc #157) should include: basic circuit theory, semiconductor physics, analogue and digital electronics, and computer architecture. NZ’s existing electrical engineering programmes (University of Auckland, University of Canterbury, Victoria University of Wellington) and polytechnic electronics courses (Ara Institute of Canterbury, Waikato Institute of Technology) provide the curriculum foundation. This builds the workforce for eventual local manufacturing and maintains the knowledge base for interpreting and using imported computing equipment.

Phase 4–5 (Years 7–30)

9. Develop NZ’s position as a knowledge provider. As maritime trade develops (Doc #140), NZ’s most valuable computing-related exports are intellectual: printed technical manuals, trained personnel, algorithm designs, and consulting expertise. NZ’s Recovery Library and its preserved technical knowledge base are assets that larger manufacturing regions need but may not have preserved as systematically.

10. Initiate discrete transistor research programme. When prerequisite industries (acid production, metallurgy, glasswork, wire drawing) are sufficiently developed, begin the germanium transistor programme described in Doc #115. Target: proof-of-concept computer by year 40–50.

11. Negotiate computing hardware import through trade. As trans-Tasman and wider maritime trade develops, NZ should seek to import computing hardware — from discrete component systems to whatever level of IC manufacturing other regions achieve. NZ’s agricultural exports and intellectual property are the trade goods.

Phase 6–7 (Years 30–100+)

12. Build the first NZ-manufactured discrete transistor computer per Doc #115.

13. Assess integrated circuit feasibility based on the actual state of the regional industrial base at that time. This assessment cannot be made today with any confidence.

---

ECONOMIC JUSTIFICATION

3.1 The value of computing to the recovery economy

Computing capability is an economic multiplier. Doc #135 estimates that a single discrete transistor computer solving 80–100 resource allocation problems per year replaces approximately 30–40 person-years of human computation bureau labour annually, once operational.² The value of optimised resource allocation — better logistics, better agricultural planning, better medical supply distribution — compounds across the entire economy.

But the economic analysis must be honest about timing. NZ does not need to build computers in Phase 1 or Phase 2. The stockpiled device fleet handles all computing needs during these phases. The economic question is about the transition: when do stockpiled devices become scarce enough that investment in alternatives is justified?

3.2 Stockpile bridge: economic cost of maintenance

Maintaining the stockpiled device fleet requires:

• Repair technicians: Estimated 20–50 trained electronics repair technicians nationally, working full-time on device maintenance and cannibalisation. At recovery wage rates, this represents 20–50 person-years per year.³
• Facilities: Workshop space with bench tools, soldering equipment, test instruments, and parts storage. These facilities exist in NZ already — the investment is in maintaining and staffing them, not building them.
• Parts: Drawn from cannibalised non-functional devices. No import cost during isolation; limited by the shrinking pool of donor devices.

The annual cost of maintaining the stockpile bridge is modest — 20–50 person-years per year — and the computing output is enormous by comparison to any alternative. Every year the bridge holds is a year the economy benefits from modern computing capability at low marginal cost.

3.3 Human computation bureau: the interim solution

When the device fleet is exhausted, the fallback is a human computation bureau (Doc #135, Part I). Cost:

• Staffing: 20 trained clerks working in duplicate-checked pairs produces approximately 4,000–5,500 arithmetic operations per day, depending on problem complexity and clerk proficiency. This handles approximately 15–20 national-level optimisation problems per year.⁴
• Annual cost: 12–16 person-years of skilled labour for this output level, depending on problem mix and required checking intensity.⁵
• Limitations: Slow (weeks per problem), error-prone (0.1–0.5% systematic error rate even with duplicate checking, based on historical human computation bureau performance),⁶ and does not scale without proportional staff increases.

3.4 Discrete transistor computer: the long-term investment

From Doc #115:

• Construction cost: 30–55 person-years over 5–7 years, depending on prerequisite industry readiness.⁷
• Annual operating cost: 1–2 person-years, depending on machine reliability and maintenance burden.⁸
• Output: 80–120 optimisation problems per year at 1–5 Hz clock speed; substantially more with a second-generation machine. This assumes problems of comparable complexity to those the bureau handles; simple tabulations would be faster, multi-variable optimisations slower.
• Breakeven vs. bureau: Year 7–10 after project initiation, depending on construction duration and bureau staffing costs.⁹

3.5 Trade-sourced computing hardware: uncertain but potentially cheaper

If maritime trade produces access to computing hardware manufactured elsewhere, the cost to NZ is:

• Trade goods: Agricultural exports, intellectual property, trained personnel.
• Shipping: Sail-based trade capacity.
• Maintenance: Local repair and operation, which NZ can support.

The cost comparison between importing a computer and building one domestically is impossible to specify today, because it depends on what trade relationships exist 30–50 years from now. But the general principle is clear: NZ’s comparative advantage lies in agriculture and knowledge, not in semiconductor manufacturing. If trade is available, importing hardware and exporting knowledge is almost certainly more efficient than domestic manufacturing.

3.6 Economic summary

Strategy	Cost (person-years)	Computing output	When viable
Stockpile maintenance	20–50/year	Modern-level	Years 0–25
Human computation bureau	12–16/year	15–20 problems/year	Years 15+ (built before needed)
Discrete transistor computer	30–55 construction + 1–2/year	80–120 problems/year	Years 40–60
Trade-sourced hardware	Variable (trade goods)	Depends on what is available	Years 15–30+ (trade-dependent)

The optimal strategy is sequential: maintain the stockpile as long as possible, establish the bureau before the stockpile is exhausted, pursue trade for hardware when maritime routes allow, and invest in domestic manufacturing only if trade is insufficient or unreliable.

---

4. THE STOCKPILE BRIDGE (YEARS 0–25)

4.1 NZ’s installed computing base

NZ’s computing device population as of the mid-2020s, estimated:

Device category	Estimated NZ quantity	Typical useful life (years)	Notes
Smartphones	4.5–5.5 million	3–5 (battery-limited)	High per-capita ownership; battery degradation is primary failure mode
Laptops	1.5–2.5 million	5–8 (battery, hinge, display)	Widely distributed in households and businesses
Desktop PCs	500,000–1 million	7–12 (capacitor, fan, storage)	Concentrated in businesses, schools, government
Tablets	500,000–1 million	4–6 (battery-limited)	Battery replacement harder than laptops
Servers	20,000–50,000	5–10 (designed for continuous operation)	Data centres, government, businesses
Industrial controllers (PLCs, SCADA)	10,000–30,000	10–20	Grid, water, manufacturing — critical infrastructure
Networking equipment	100,000–300,000	7–15	Routers, switches, wireless access points
Printers/copiers	500,000–1 million	5–10	Multi-function devices with embedded computing

Total computing devices: Roughly 7–11 million, of which 5–7 million are general-purpose computers (smartphones, laptops, desktops, tablets).¹⁰

These figures are estimates based on NZ’s population (~5.2 million), household composition, and business density. The actual inventory should be established through the national asset census (Doc #8). NZ’s per-capita device ownership is broadly comparable to other developed nations — approximately 1.5–2 devices per person.¹¹

4.2 Device degradation patterns

Not all devices fail at the same rate. Understanding failure modes determines which devices to prioritise and how to extend fleet life.

Batteries are the most common failure mode for portable devices. Lithium-ion batteries degrade through calendar ageing (chemical degradation even when unused) and cycle ageing (degradation from charge/discharge cycles). A battery stored at room temperature loses approximately 20% of its capacity per year; stored at 15°C, this slows to roughly 10–15% per year.¹² Batteries stored fully charged degrade faster than those stored at 40–60% charge.

Implication: Smartphones and tablets have the shortest useful lives. Within 3–5 years, most will have severely degraded batteries. Battery replacement extends life but requires compatible replacement batteries from donor devices — a diminishing supply. Laptops are somewhat better because their batteries are larger (more capacity to lose before becoming unusable) and many can operate on mains power without a battery.

Capacitors (particularly electrolytic capacitors) are the primary failure mode for desktops, servers, and power supplies. Electrolytic capacitors have a rated life of 2,000–10,000 hours at rated temperature, with life approximately doubling for every 10°C below rated temperature.¹³ In practice, well-built desktop PCs and servers in temperate environments (NZ qualifies) can last 10–15 years before capacitor failure becomes common. Failed capacitors are replaceable by a trained technician with soldering equipment and compatible replacements.

Storage media: Solid-state drives (SSDs) have a finite number of write cycles (typically 500–3,000 programme/erase cycles for consumer NAND flash) but last decades under moderate use. Hard disk drives (HDDs) fail mechanically — bearings, head crashes — typically after 5–10 years of continuous operation, longer for intermittent use.¹⁴ Data preservation (ensuring critical information survives storage media failure) is a separate concern from device functionality.

Displays: LCD displays fail through backlight degradation (LEDs last 30,000–50,000 hours, but driver circuits fail sooner) and liquid crystal degradation.¹⁵ A failed display does not mean a failed computer — external displays can substitute.

Connectors and cables: Mechanical wear on USB ports, power connectors, and hinges. These are repairable with basic tools and skills.

4.3 Managed depletion strategy

The goal is to maximise the total useful computing-years extracted from the stockpile. Key principles:

Triage and consolidation. Not all devices are worth maintaining. Prioritise: servers and desktops (longest intrinsic life, no battery dependency, most capable), laptops (versatile, moderate life), industrial controllers (critical infrastructure). Deprioritise: smartphones (short battery life, difficult to repair, limited utility without cellular network), old or damaged devices beyond economic repair.

Cannibalisation. Every non-functional device is a parts donor. Standardise cannibalisation procedures: which components to harvest (RAM, storage, power supplies, displays, batteries, capacitors, fans, cables), how to test harvested components, how to store them. This extends the functional fleet by years.

Environmental management. Store inactive devices in cool (15–20°C), dry, dark conditions. Batteries at 40–60% charge. Remove batteries from devices in long-term storage where possible. This slows all degradation mechanisms.

Operating system and software standardisation. As devices age, modern software becomes a barrier — operating system updates consume resources that ageing hardware cannot spare. Establish standard lightweight software configurations for recovery-relevant tasks: spreadsheet computation, database management, text processing, scientific calculation. Linux-based operating systems are ideal for this: they run on minimal hardware, are freely available, and do not require internet activation.¹⁶ NZ’s existing Linux user community and the computer science departments at the University of Auckland, Victoria University of Wellington, and the University of Canterbury provide the expertise to develop and maintain a standard NZ recovery computing distribution.

Power management. NZ’s grid (85%+ renewable) provides reliable mains power, eliminating the power constraint that would otherwise shorten device life. Devices should be operated on mains power wherever possible, preserving battery capacity for genuine portable use.

4.4 Projected stockpile timeline

This timeline is an estimate with wide uncertainty bounds.

Years 0–5: Minimal computing constraints. The vast majority of NZ’s device fleet remains functional. Normal computing operations continue for government, business, education, and personal use. The primary change is loss of international internet (Doc #127) and gradual contraction of cellular networks.

Years 5–10: Smartphone fleet largely non-functional (battery exhaustion). Laptop fleet declining — perhaps 50–70% of original devices still operational. Desktops and servers mostly functional. Computing remains adequate for all critical national functions. Repair and cannibalisation programmes become important.

Years 10–15: Laptop fleet significantly reduced (30–50% of original). Desktop and server fleet beginning to decline. Capacitor failures become the dominant failure mode. Allocation becomes necessary — personal computing contracts, institutional computing is prioritised.

Years 15–25: The tail end of the stockpile. A diminishing fleet of the most durable devices — well-maintained desktops and servers — provides computing for the highest-priority national functions. Some devices may last well beyond 25 years if maintained and operated conservatively, but the functional fleet is a small fraction of the original. At some point in this range, the last general-purpose computer fails.¹⁷

The transition must be managed, not reactive. The human computation bureau and any trade-sourced or locally manufactured computing capability must be operational before the last stockpiled computer fails. A gap — even a short one — in which NZ has no systematic computation capability would degrade logistics, planning, and administration across the entire economy.

---

5. THE GAP: HUMAN COMPUTATION (YEARS 15–50+)

5.1 When the computers stop

There will be a period — possibly decades — during which NZ has no electronic computing. Stockpiled devices are exhausted. Discrete transistor computers are not yet built (Doc #115 targets this for years 40–60 after the event). Trade-sourced hardware may or may not be available.

The performance gap is severe. A modern laptop performs billions of operations per second. A human computation bureau of 20 clerks performs thousands per day — a throughput reduction of roughly 10 million to one. Problems that a computer solves in seconds take the bureau weeks. Many classes of computation (real-time control, large-scale simulation, database queries across thousands of records) become entirely infeasible. NZ’s planning and logistics capability during this period regresses to approximately 1930s levels.

During this period, NZ relies on:

• Printed reference tables (Docs #10–28): Navigation, engineering, mathematical, medical, and agricultural data precomputed and printed during the AI facility’s operational life and the stockpile bridge period.
• Human computation bureau (Doc #135, Part I): Trained clerks performing systematic arithmetic for national resource allocation.
• Slide rules and mechanical calculators: If any survive or can be manufactured. Slide rules are achievable with NZ materials — a precision-engraved logarithmic scale on native timber (tōtara or rimu for dimensional stability) or aluminium from NZ’s Tiwai Point smelter stockpile. Manufacturing requires accurate linear engraving capability, which NZ’s existing machine shops (Doc #91) can develop. Mechanical calculators (e.g., Curta-type or Brunsviga-type) require high-precision gear cutting, turning, and assembly — at the upper end of what NZ’s machine shop capability (Doc #91) can achieve — plus months of skilled labour per unit.¹⁸
• Analogue computers: Specialised devices for specific calculations — tide predictors for NZ’s major ports (Auckland, Lyttelton, Wellington, Bluff), agricultural yield estimators, harmonic analysers. These use mechanical or electrical analogue circuits rather than digital logic and are achievable with NZ’s machining and electrical capability (Doc #91). Performance gap: analogue computers achieve 0.1–1% accuracy on their target problems — adequate for tide prediction and engineering estimation, but insufficient for the precise resource allocation calculations that digital computers or the human computation bureau handle. Their utility is narrow but real for specific recurring problems where approximate answers are acceptable.¹⁹

5.2 Preserving computing knowledge through the gap

The most dangerous aspect of the gap is not the loss of computation per se — human civilisation managed without electronic computers until the 1940s — but the potential loss of the knowledge needed to rebuild computing. If the people who understand semiconductor physics, circuit design, and computer architecture die or retire during a 30-year gap without passing their knowledge on, the rebuilding timeline extends further.

Mitigation:

• University curriculum (Doc #115): Maintain electrical engineering, semiconductor physics, and computer science education even when no computers exist. This is counterintuitive but essential — the purpose is to preserve the knowledge base for eventual rebuilding, not to train people for immediate employment.
• Printed documentation: The complete technical content of Doc #115, plus semiconductor physics texts, circuit design manuals, and programming references, must be printed and distributed to libraries, universities, and technical schools.
• Hands-on training with analogue and mechanical systems: Students who build slide rules, analogue computers, and telegraph systems develop practical electronics and precision fabrication skills that transfer to transistor manufacturing.
• Heritage skills preservation (Doc #160): Identify and document the knowledge of NZ’s remaining electronics engineers and computer scientists while they are alive.

---

6. DISCRETE TRANSISTOR COMPUTING (YEARS 40–60+)

6.1 What Doc #115 specifies

Doc #115 provides a complete guide for building a germanium point-contact transistor computer from NZ-available raw materials. The key specifications:

• Technology level: Equivalent to 1947–1955 era computing. Point-contact germanium transistors.
• Machine specification: 12-bit word, 256 words of magnetic core memory, ~1,500 transistors, 14-instruction instruction set.
• Clock speed: 1–5 Hz conservatively for a first-generation machine; up to 100 Hz for a second-generation junction transistor machine.
• Construction cost: 30–55 person-years over 5–7 years, depending on prerequisite industry readiness.
• Capability: Solves national resource allocation problems — logistics optimisation, agricultural planning, fisheries management. Replaces and augments the human computation bureau.

6.2 The prerequisite industrial chain

Building a discrete transistor computer requires prior development of multiple industries, each with its own multi-year development timeline. This is the full dependency chain:

Germanium source material. NZ’s sphalerite (zinc sulfide) deposits contain trace germanium, typically 1–300 parts per million.²⁰ Extracting germanium from zinc ore processing residues requires a functioning zinc extraction operation — which does not currently exist in NZ. Alternative sources: coal ash (NZ coal contains trace germanium, but concentration is very low — extraction requires processing large volumes) or imported germanium concentrate via maritime trade. The germanium sourcing question is one of the largest uncertainties in the entire programme.

Acid production (Doc #113). Germanium extraction via the chloride route requires hydrochloric acid. Hydrochloric acid production requires sulfuric acid and salt. Sulfuric acid production from geothermal sulfur or pyrite requires acid-resistant containment vessels, which require either ceramic or lead-lined steel — each with its own supply chain. This is a [C]-rated capability requiring years of development.

Glasswork (Doc #115). Fractional distillation of germanium tetrachloride requires precision glassware — distillation columns, condensers, collection vessels. NZ can produce basic glass from Parengarenga silica sand in electric furnaces, but laboratory-quality borosilicate glassware (resistant to thermal shock) is significantly harder. Whether NZ can produce adequate distillation apparatus from soda-lime glass with careful technique, or whether borosilicate is essential, is an open question.

Copper wire (Docs #54, #108). Magnetic core memory requires fine copper wire. The memory matrix described in Doc #70 uses 3,072 ferrite cores, each threaded with three wires. Wire drawing to the required gauge (~0.3–0.5 mm) is achievable with NZ steel dies and recycled copper — NZ has substantial copper in its electrical grid wiring, building stock, and vehicle wiring harnesses, totalling an estimated 100,000–200,000 tonnes recoverable nationally. Wire drawing to the required gauge requires a dedicated drawing operation with progressively smaller dies, annealing capability between draws, and quality control for conductivity and tensile strength.

Ferrite cores. Magnetic core memory requires ferrite (iron oxide ceramic) toroidal cores. Production requires: iron oxide (available from ironsand or rust), manganese or zinc oxide (as dopants), ceramic processing capability (mixing, pressing, sintering at 1,200–1,400°C in NZ’s electric furnaces), and quality control for magnetic properties. This is a novel manufacturing process for NZ, requiring experimentation and development.²¹

Precision metalwork (Doc #91). Circuit boards, switch contacts, relay components, and the physical structure of the computer require machine shop capability at a level above general blacksmithing — filing, drilling, tapping, and turning to tolerances of 0.1–0.5 mm. NZ’s existing engineering workshops (particularly those supporting the dairy and agricultural equipment industries) have some of this capability, but achieving the consistency required for computer assembly will require dedicated training and tooling.

Electrical test equipment. Building and debugging the computer requires, at minimum: a multimeter (achievable with a galvanometer, precision resistors, and a rectifier — NZ’s instrument repair workshops can produce these), a signal generator (achievable with a simple oscillator circuit), and ideally an oscilloscope (significantly harder — requires a cathode ray tube, which requires glassblowing and vacuum technology). Without an oscilloscope, debugging is slower and more error-prone but not impossible — early 1940s computer projects such as the Colossus relied heavily on basic test equipment (multimeters and signal lamps) for much of their circuit debugging, even though oscilloscopes were available in limited quantities.²²

6.3 Realistic timeline

The dependency chain described above means that the computer project cannot begin until substantial prerequisite infrastructure exists. A realistic staging:

Years after event	Milestone	Prerequisites
0–15	Stockpile bridge period; begin industrial development	Grid power, machine shops, basic chemistry
15–25	Acid production capability; glass production; wire drawing	Doc #113, #98, #70
25–35	Germanium sourcing resolved (domestic extraction or trade); ferrite core development begun	Mining or trade infrastructure
35–45	Germanium purification and first transistor fabrication attempts	All of the above
40–50	Proof-of-concept computer (Doc #115, Phase 1)	Working transistors, basic circuit assembly
45–55	Functional computer (Doc #115, Phase 2)	Proven transistors, magnetic core memory
50–60+	Second-generation machine; junction transistors	Experience from first machine

This timeline assumes continuous, deliberate development. If industrial development stalls, or if the prerequisite industries take longer than estimated, or if germanium sourcing proves more difficult than hoped, the timeline extends. If trade provides germanium concentrate or other critical inputs earlier than expected, it compresses.

Honest assessment: NZ may not build its first electronic computer until 50–60 years after the event. This reflects the genuine industrial prerequisites. For comparison, the original development from the vacuum tube to the stored-program computer (ENIAC, operational 1945, to EDSAC, operational 1949) took the combined resources of the United States and United Kingdom, with their full industrial bases, wartime urgency, and thousands of engineers, approximately 4–5 years.²³ NZ will be doing this with a fraction of the workforce, from a lower industrial starting point, and without the benefit of an existing electronics industry to draw on.

---

7. INTEGRATED CIRCUITS: THE DISTANT HORIZON

7.1 What integrated circuits require

Moving from discrete transistors to integrated circuits — which is the step from 1955-level to 1970s-level computing — requires capabilities that are qualitatively different from transistor fabrication:

Photolithography. ICs are made by projecting patterns of light through a mask onto a photosensitive layer on a semiconductor wafer, then chemically etching the exposed pattern. This requires: a UV light source of controlled intensity and wavelength, precision optics (lenses ground to sub-micrometre accuracy), photoresist chemicals (light-sensitive polymers), and mask fabrication capability (creating the circuit pattern at high resolution on glass or quartz).

Clean room environment. IC fabrication requires particle-free environments — a single dust particle can ruin a circuit. Clean rooms require HEPA filtration (which requires glass microfibre filter media — producible from NZ silica but requiring specialised fibre-drawing capability), positive pressure air handling (blowers, ductwork, seals), specialised gowning (lint-free synthetic fabric — NZ would need to substitute with tightly woven natural fibre, at reduced performance), and contamination monitoring (particle counters, which are themselves precision instruments requiring optical components). Building a basic clean room is achievable with NZ materials (concrete, steel, glass, fabric filters), but the resulting environment would likely achieve ISO Class 6–7 at best — adequate for 1960s-era large-geometry ICs but not for anything approaching modern feature sizes. Maintaining it requires ongoing filter replacement and environmental control.²⁴

Semiconductor-grade silicon. While discrete transistors can use germanium at relatively modest purity (99.9% — “three nines”), ICs require silicon at 99.9999% purity or higher (“six nines” to “nine nines”). Producing this requires the Siemens process or equivalent: converting metallurgical-grade silicon to trichlorosilane gas, purifying the gas by fractional distillation, and depositing pure silicon by chemical vapour deposition. This is a complex chemical engineering process requiring high-purity gases, precise temperature control, and corrosion-resistant reactor vessels.²⁵

Wafer processing. Slicing silicon ingots into thin wafers requires diamond or silicon carbide wire saws. Polishing wafers to optical flatness requires colloidal silica slurries and precision lapping equipment. Each step requires materials and equipment that must be developed or imported.

Chemical processing. Etching, doping, oxidation, and metallisation each require specific chemicals: hydrofluoric acid (extremely dangerous and difficult to produce), phosphine and borane gases (for doping), and high-purity metals (aluminium, gold) for interconnects.

7.2 Why NZ is unlikely to manufacture ICs

The integrated circuit manufacturing chain described above requires:

• A chemical industry producing high-purity specialty gases and acids
• An optical industry producing precision lenses
• A materials industry producing semiconductor-grade silicon
• A clean room infrastructure that must be built and continuously maintained
• A trained workforce of hundreds of specialists in dozens of sub-disciplines

This is not a capability that a nation of 5 million people, rebuilding from industrial disruption, can realistically develop in isolation. Modern semiconductor fabrication is the most complex manufacturing process humanity has ever created. Even in the pre-war world, only a handful of nations (US, Taiwan, South Korea, Japan, Netherlands, China) maintained cutting-edge fabrication capability, each supported by a global supply chain of thousands of specialised suppliers.²⁶

NZ might, over multiple generations, develop the capability to produce very basic ICs — equivalent to the small-scale integration of the 1960s (tens of transistors per chip).²⁷ This would be useful: a single SSI chip replaces dozens of discrete transistors, improving reliability by a factor of 5–10 (fewer solder joints and interconnections) and reducing physical size by roughly two orders of magnitude. But even this level requires photolithography, clean rooms, and semiconductor-grade materials — and the resulting chips would perform at roughly 1/1,000,000th the capability of a pre-war smartphone processor.

The realistic assessment: IC manufacturing is a [D]-rated capability for NZ. It might emerge 60–100+ years after the event, or it might never develop domestically at a useful scale. NZ should plan for a future where IC manufacturing happens elsewhere and NZ imports the products, rather than a future where NZ manufactures its own chips.

7.3 Doc #115 and the semiconductor processing roadmap

Doc #115 addresses the multi-generational project of developing semiconductor processing capability. The key takeaway from that document, relevant here: semiconductor processing will likely emerge first in a region with a larger population and industrial base — Australia (26 million people, mining industry, existing chemical industry), Brazil, India, or a reconstituted industrial centre in another hemisphere. NZ’s role is more plausibly as a contributor of knowledge, trained personnel, and designs than as a manufacturer.

---

8. NZ’S REALISTIC ROLE: KNOWLEDGE PROVIDER AND DESIGN CENTRE

8.1 NZ’s comparative advantage

In a post-catastrophe world, NZ has several attributes relevant to computing:

Preserved knowledge base. The Recovery Library, the AI facility’s output, the national printing programme, and NZ’s university system (eight universities, sixteen polytechnics and institutes of technology) collectively represent one of the world’s best-preserved repositories of technical knowledge. This includes computing knowledge: algorithms, programming, circuit design, semiconductor physics, and the complete technical content of Doc #135.

Educated workforce. NZ’s pre-war education system, maintained and adapted post-event (Doc #162), produces graduates who understand mathematics, physics, and engineering. These people can design computing systems even if NZ cannot manufacture them.

Operating experience. NZ’s extended stockpile bridge period means that NZ retains practical computing experience longer than most nations. NZ technicians who maintained and operated the last pre-war devices have institutional knowledge that is directly valuable.

Trade position. NZ’s agricultural surplus (Doc #74) is the primary trade good. But knowledge — printed manuals, trained personnel, consulting expertise, algorithm designs — is a secondary export that is light, durable, and high-value. A single trained computer engineer sent to an Australian semiconductor facility as a consultant produces more value for NZ than tonnes of exported cheese.

8.2 What NZ can export

Printed technical documentation. Complete computing reference libraries: from basic circuit theory through semiconductor physics to computer architecture and programming. These are produced during the stockpile bridge period using NZ’s printing capability and the AI facility’s knowledge.

Trained personnel. Electronics engineers, computer scientists, programmers, and repair technicians trained in NZ’s universities and trade schools. These individuals can work in manufacturing facilities elsewhere and bring knowledge back.

Algorithm and software designs. The mathematical techniques for optimising logistics, agriculture, fisheries, and resource allocation that NZ develops through its human computation bureau and early computer operations. These are directly applicable to any computing platform.

Hardware designs. Circuit schematics, architecture specifications, and construction guides for discrete transistor computers. NZ’s experience building the Doc #115 machine — if it reaches that point — produces practical construction knowledge that other regions can use.

8.3 What NZ should import

Computing hardware. If any region develops IC manufacturing capability, NZ should trade for finished chips, assembled systems, or even partially assembled computing kits. The cost of importing a single IC is far lower than the cost of building the domestic manufacturing capability to produce it.

Specialist materials. Semiconductor-grade silicon, germanium concentrate, photoresist chemicals, high-purity acids — materials that NZ cannot produce domestically but that are needed for whatever level of local electronics work NZ undertakes.

Test and measurement equipment. Oscilloscopes, spectrum analysers, and precision instruments that NZ cannot manufacture but that dramatically improve the quality and efficiency of any local electronics work.

---

9. CRITICAL UNCERTAINTIES

Uncertainty	Range	Impact	How to resolve
Actual NZ device inventory	7–11 million total devices	Determines stockpile bridge duration	National asset census (Doc #8)
Battery degradation rate under NZ storage conditions	10–20% capacity loss per year at ambient storage	Determines portable device lifespan	Empirical testing of stored devices
Device repair/cannibalisation effectiveness	Could extend fleet life by 3–10 years	Major impact on bridge duration	Establish repair programme and track results
Trade development timeline	Maritime trade functional years 10–30+	Determines when imported hardware available	Dependent on Doc #140, #151
Whether any region achieves IC manufacturing within 100 years	Unknown	Determines whether import is possible	Cannot be resolved from NZ; monitor via trade contacts
NZ germanium ore viability	Sphalerite deposits exist; germanium concentration uncertain	Determines whether Doc #22 is feasible without imports	Geological survey and assay (Doc #22)
Prerequisite industry development timeline	15–35 years for acid, glass, wire at required quality	Directly gates discrete transistor programme	Track progress of Docs #113, #98, #70, #135
Nuclear winter effects on electronics	UV and temperature fluctuations affect stored devices	Could accelerate or decelerate stockpile degradation	Monitor and test
Knowledge preservation through the gap	Depends on education system continuity	Determines whether NZ has the workforce to rebuild	Track university output, Doc #162

---

10. CROSS-REFERENCES

Document	Relationship to this document
Doc #135 — Computer Construction	The construction guide for discrete transistor computing; this document provides the strategic context
Doc #115 — Semiconductor Processing Roadmap	Longer-term IC manufacturing pathway; this document explains why NZ likely imports rather than manufactures
Doc #127 — Telecommunications Maintenance	Parallel degradation trajectory for network equipment; shared repair workforce
Doc #128 — HF Radio Network	Communications backbone after telecom fails; radio equipment shares electronics repair skills
Doc #129 — AI Inference Facility Operations	NZ’s highest-value computing asset; operations during stockpile bridge period
Doc #130 — Device Life Extension	Detailed technical guide for the repair/cannibalisation programme described here
Doc #91 — Machine Shop Operations	Prerequisite for both device repair and eventual transistor manufacturing
Doc #98 — Glass Production	Prerequisite for distillation apparatus needed in germanium purification
Doc #105 — Wire and Fencing Nails	Wire drawing capability; prerequisite for magnetic core memory
Doc #113 — Sulfuric Acid	Prerequisite for acid production chain needed for germanium extraction
Doc #138 — Sailing Vessel Design	Enables the maritime trade through which NZ imports computing hardware
Doc #151 — Trans-Tasman Relations and Trade	NZ–Australia trade relationship; Australia is the most likely computing hardware trade partner
Doc #157 — Trade Training	Training pipeline for electronics repair technicians
Doc #160 — Heritage Skills Preservation	Preserving electronics knowledge from ageing pre-war technicians
Doc #162 — University and Research Priorities	Maintaining computing education through the gap

---

FOOTNOTES

---

1. NZ device ownership estimated from population (~5.2 million) and per-capita device penetration rates for developed nations. The International Telecommunication Union estimated NZ had approximately 6.7 million mobile connections in 2023. Desktop and laptop penetration estimated from household broadband penetration (approximately 90% of NZ households, per Stats NZ Household Use of ICT Survey). Figures are approximate and should be verified through the national asset census.↩︎

2. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

3. The 20–50 person-year estimate for repair technicians is an assumption based on the scale of the device fleet and the complexity of board-level repair. NZ’s actual pool of capable electronics repair technicians is uncertain and should be established through the skills census. The figure could be lower if the repair programme focuses on the highest-value devices and accepts higher attrition in lower-priority categories.↩︎

4. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

5. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

6. Historical human computation error rates: pre-electronic computation bureaus (e.g., the WPA Mathematical Tables Project in the US, 1938–1943, and the UK Nautical Almanac Office) achieved error rates of approximately 0.1–1% in final published results through systematic duplicate and triplicate checking. The 0.1–0.5% range assumes well-trained clerks with systematic checking protocols. Source: Grier, D.A., “When Computers Were Human,” Princeton University Press, 2005.↩︎

7. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

8. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

9. Computation bureau staffing, throughput, and computer construction cost estimates are from Doc #135 (Computer Construction: From Raw Materials to Stored-Program Computer), Parts I and VII. These estimates involve significant uncertainty; see Doc #135 for detailed assumptions and ranges.↩︎

10. NZ device ownership estimated from population (~5.2 million) and per-capita device penetration rates for developed nations. The International Telecommunication Union estimated NZ had approximately 6.7 million mobile connections in 2023. Desktop and laptop penetration estimated from household broadband penetration (approximately 90% of NZ households, per Stats NZ Household Use of ICT Survey). Figures are approximate and should be verified through the national asset census.↩︎

11. Per-capita device ownership in developed nations: the OECD reports approximately 1.2–2.5 connected devices per person across member nations as of 2023. NZ is in the middle of this range. Source: OECD Digital Economy Outlook, various editions. https://www.oecd.org/digital/↩︎

12. Lithium-ion battery calendar ageing: Barré, A. et al., “A review on lithium-ion battery ageing mechanisms and estimations for automotive applications,” Journal of Power Sources, 2013. Calendar ageing rates vary significantly by cell chemistry (NMC, LFP, NCA), state of charge, and temperature. The 20% per year at room temperature figure is approximate for typical consumer electronics lithium-ion cells stored at full charge. Stored at partial charge and lower temperature, degradation is significantly slower.↩︎

13. Electrolytic capacitor lifetime follows the Arrhenius equation: life approximately doubles for every 10°C reduction below rated temperature. This is well-established in capacitor reliability engineering. See: Parler, S.G., “Deriving Life Multipliers for Electrolytic Capacitors,” IEEE Power Electronics Society Newsletter, 2004.↩︎

14. HDD failure rates: the Backblaze annual drive statistics reports (2013–2024) provide large-scale empirical data on HDD failure. Annualised failure rates for consumer drives typically range from 1–5% per year, with failure rates increasing after 3–5 years of continuous operation. For intermittent use (as in a managed recovery stockpile), mechanical wear is reduced and lifespans extend significantly. Source: Backblaze Hard Drive Stats, https://www.backblaze.com/cloud-storage/resources/hard-dr...↩︎

15. LED backlight lifetime ratings are typically specified at L70 (time to 70% of initial brightness). Manufacturer specifications for LED backlights typically claim 30,000–50,000 hours to L70, but driver circuit failures (capacitor ageing, solder joint fatigue) often cause backlight failure well before the LEDs themselves degrade. Source: general LED reliability literature; specific figures vary widely by manufacturer and should be treated as approximate.↩︎

16. Linux operating system distributions suitable for low-resource hardware include Debian, Alpine Linux, and various embedded Linux distributions. These run functional computing environments on hardware with as little as 256 MB RAM and 1 GHz processors — specifications well below most devices manufactured after 2010. No internet activation or ongoing license verification is required.↩︎

17. The “last general-purpose computer” timeline is one of the most uncertain estimates in this document. Individual devices have been documented operating for 20–30+ years in favourable conditions (stable power, temperature-controlled environment, minimal physical stress). The NZ fleet includes a distribution of ages and conditions — the last survivors will be the youngest, best-maintained, most robust devices in the fleet. Whether this extends to 20, 25, or 30+ years depends on factors that cannot be predicted today.↩︎

18. Mechanical calculators: the Curta calculator (designed by Curt Herzstark, produced 1948–1972) is an example of a precision mechanical calculator achievable with advanced machining capability. It performs four-function arithmetic to 11 digits. Manufacturing one from scratch requires high-precision turning, gear cutting, and assembly — within the capability described in Doc #91 but at the upper end of difficulty. Simpler adding machines (e.g., the Pascaline or Leibniz wheel designs) are easier to produce but less capable.↩︎

19. Analogue computers for specific applications have a long history. Lord Kelvin’s tide-predicting machine (1876) used mechanical analogue circuits to predict tides from harmonic components. The Differential Analyser (Vannevar Bush, 1931) solved differential equations mechanically. These devices require precision machining but no electronics. Their utility is limited to the specific class of problem they are designed to solve.↩︎

20. Germanium occurrence in sphalerite: germanium concentrations in zinc ores vary widely, from negligible to over 300 ppm in some deposits. NZ’s sphalerite deposits (e.g., in the Coromandel Peninsula and West Coast) have not been systematically assayed for germanium content. This is a critical data gap that should be addressed through geological survey (Doc #22). Source: Höll, R. et al., “Origin and Occurrence of Germanium,” in Germanium-Based Technologies, Elsevier, 2007.↩︎

21. Ferrite core manufacturing: ferrite materials for magnetic cores were developed at Philips Research Laboratories in the 1940s–1950s. The basic process (mixing iron oxide with manganese or zinc oxide, pressing into toroids, sintering at high temperature) is well-documented. The challenge is achieving consistent magnetic properties — specifically the “square loop” hysteresis required for memory applications. This requires precise control of composition and sintering conditions. Source: Smit, J. and Wijn, H.P.J., “Ferrites,” Philips Technical Library, 1959.↩︎

22. Oscilloscopes were available as laboratory instruments from the 1930s onward (the DuMont 164 was commercially available from 1937), but wartime demand meant supply was limited. The Colossus machines at Bletchley Park (1943–1945) were debugged primarily using multimeters, signal lamps, and output printer analysis. Oscilloscopes were used when available but were not the primary debugging tool for early digital machines. Source: Copeland, B.J., “Colossus: The Secrets of Bletchley Park’s Codebreaking Computers,” Oxford University Press, 2006.↩︎

23. ENIAC (Electronic Numerical Integrator and Computer) became operational at the University of Pennsylvania in February 1946. EDSAC (Electronic Delay Storage Automatic Calculator) ran its first programme at the University of Cambridge in May 1949 — approximately 3 years later. The broader development from the first electronic computing concepts (Atanasoff-Berry Computer, 1942; Colossus, 1943) to practical stored-program computers spans approximately 5–7 years. Source: Ceruzzi, P.E., “A History of Modern Computing,” MIT Press, 2003.↩︎

24. Clean room construction: modern semiconductor clean rooms are rated by ISO 14644 standards (e.g., ISO Class 5 allows fewer than 3,520 particles per cubic metre at 0.5 micrometre size). The earliest IC fabrication (1960s) used clean rooms that would be rated approximately ISO Class 6–7 by modern standards — less stringent than modern fabs but still requiring HEPA filtration, controlled airflow, and gowning protocols. Building such a facility from NZ materials is feasible in principle but requires significant engineering effort.↩︎

25. Siemens process for semiconductor-grade silicon: the process converts metallurgical-grade silicon (98–99% purity, producible in electric arc furnaces from quartzite and carbon) to trichlorosilane (SiHCl₃) gas by reaction with hydrochloric acid at 300°C, purifies the gas by repeated fractional distillation, and deposits pure polycrystalline silicon by chemical vapour deposition at ~1,100°C. The process was developed by Siemens AG in the 1950s and remains the dominant silicon purification method. Source: Ceccaroli, B. and Lohne, O., “Solar Grade Silicon Feedstock,” in Handbook of Photovoltaic Science and Engineering, Wiley, 2011.↩︎

26. The concentration of semiconductor manufacturing: as of 2024, TSMC (Taiwan) alone produces over 50% of the world’s advanced semiconductors. The total global semiconductor supply chain involves thousands of specialised companies across dozens of countries. The CHIPS and Science Act (US, 2022) and similar programmes in the EU and Japan reflect the difficulty even wealthy nations face in establishing domestic semiconductor manufacturing. This context underscores why NZ — with 0.07% of the pre-war world’s population — should not plan to replicate this capability independently.↩︎

27. Small-scale integration (SSI): the first commercial integrated circuits (e.g., Fairchild 900 series, Texas Instruments SN51x series, early 1960s) contained 1–20 transistors per chip, implementing basic logic gates (NAND, NOR, flip-flops). By 1965, medium-scale integration (MSI) reached 20–200 transistors per chip. For comparison, a 2020s smartphone processor contains approximately 10–15 billion transistors. Source: Moore, G.E., “Cramming more components onto integrated circuits,” Electronics, 1965.↩︎