Bitcoin Node Cost Tracker: 2009

2009 era desktop PC running a Bitcoin node

2013 era laptop with external hard drive

2035 era projected mini PC with 4-8 TB NVMe node

—

Block Chain Size ⓘRaw block data (~753 GB as of Jul 2026; ~760 GB Aug). Core also keeps ~12% extra undo data — counted in Min Storage Needed. UTXO/chainstate (~12.5 GB model) and optional Electrum-style indexes are tracked separately below.

—

Min Storage Needed ⓘArchival: (blocks + undo data ~12% + UTXO set) ×1.10 headroom, plus optional indexes. Matches Bitcoin Core's own disk guideline.

—

One-Time Hardware

—

Annual Running Cost

—

Hobbyist Affordability ⓘMonthly equivalent: one-time setup amortized over 3 years + annual internet & electricity, divided by 12. Tuned for homelab/hobby budgets, not enterprise hosting.

—

Hardware Recommendation

—

Storage $ = Buy TB × $/TB + enclosure (enclosure only 2012–2023). Need GB is headroom; Buy TB is the whole-drive SKU.

Swipe horizontally for more columns · header stays pinned while scrolling

Year	Blocks (GB)	UTXO (GB)	Need (GB)	Buy (TB)	Media	$/TB	Encl. $	Storage $	Index $	CPU/RAM	Internet/mo	Total Setup	With Index	Annual OpEx

Use the clipboard or download icon on any chart to copy or save it as a PNG screenshot.

One-Time Hardware Cost Over Time

Storage drives + CPU/RAM kit to buy and run a node that year — the core affordability story

Block Chain Size Growth (GB)

Storage $/TB Over Time

Market media price — separate from how many TB you buy. Bold line = effective (Auto/HDD/SSD setting).

Cost Breakdown Over Time

Total Annualized Cost (3-yr amortized setup + OpEx)

Current Year Breakdown

Technology deflation, not a straight line

Over decades, the unit cost of storage, compute, and bandwidth has collapsed by orders of magnitude. Floods, pandemics, and AI demand create periodic spikes — real, painful, and temporary relative to the multi-decade envelope. Bitcoin node affordability rides the same curve: chain growth is roughly linear; tech unit costs have historically fallen much faster.

Storage

~10⁹×

Cheaper per TB

Since the mid-20th century (McCallum / Our World in Data class). Log scale is essential.

Bandwidth

~−80%

Real $/Mbps (home)

USTelecom BPI-Speed ~2015→2023. Plan dollars stay ~flat; megabits per dollar explode. Wholesale transit now cents/Mbps.

Compute

~10× / 4–8 yr

FLOPS per dollar

Long-run price-performance (Moore era / AI Impacts). Single-thread gains slowed after ~2005; multi-core & efficiency continued.

Unit costs · log index (1980–2026)

Three long-run series as an index (2009 = 100). Lower = cheaper per unit. Storage uses era media (HDD → SSD from 2021, matching Auto mode). Click a legend item to hide/show a series.

Straight-ish lines on a log chart ≈ exponential decline. Spikes and 2027–2040 paths: see Storage $/TB below.

So what for Bitcoin nodes? From 2009→2026 the storage unit-cost index fell far more than the chain grew — that’s why hobbyist archival nodes stayed affordable for most of history. The 2026 flash spike is a real hiccup (see linear chart); long-run tech deflation is still the main story.

Storage $/TB · linear (2009–2040)

Node-era media prices + dashed NVMe Baseline vs AI-Shock. 2011 flood & 2026 NAND spikes are intentional.

Moments on the curve

Click a year to jump the dashboard year (2009+) and highlight the linear storage chart.

What ~$100 bought in storage

Illustrative capacity for roughly $100 of magnetic disk (HDD-class) at that year’s price — order-of-magnitude, not a quote.

Bridge to the node cost model

In this dashboard’s archival model, bulk HDD fell from ~$110/TB (2009) to ~$10/TB (2026). Consumer flash fell for years, then spiked to ~$145/TB in 2026. Chain size grew from megabytes to ~753 GB — still often cheaper to store because $/TB fell faster than the chain grew (until temporary spikes reverse that).

Sources, units & methodology

Storage long-run: John C. McCallum historical disk/memory tables (via Our World in Data); Matt Komorowski hard-drive cost series (1980–2009); Backblaze cost-per-GB blogs (2009–2022+). Pre-2009 points are representative retail anchors, not a matched-model PPI.

2009–2026 node series: Same RAW.hddPerTB / RAW.ssdPerTB as the Charts tab (see ASSUMPTIONS.md). 2011 flood is annotated; the series is smoothed yearly averages, so the bump is modest vs retail day-to-day noise.

Index chart (2009 = 100): Three series only — Storage, Bandwidth, Compute — on a shared log axis (colors match the KPI cards). Storage = era media (HDD through 2020, SSD/NVMe from 2021). History only through 2026; click the legend to hide a series. No scenario fan on this chart.

Compute series: Illustrative cost of a fixed unit of performance (inverse of FLOPS/$ trends; AI Impacts / Sandberg–Bostrom class, ~10× every 4–8 years). Sparse anchors, log-interpolated — not a BLS PPI.

Bandwidth series: Approximate U.S. residential $/Mbps/mo (nominal) from early broadband → BPI-class 2015–2023 (~$0.87 → ~$0.18) → mid-2020s. Plan dollars stay sticky; Mbps explode. Separate from wholesale IP transit (~$0.05/Mbps floors).

Projections 2027–2040: Only on the linear Storage $/TB chart: dashed NVMe Baseline vs AI-Shock via the same scenario math as the node model. The Charts tab follows the active Projections panel scenario.

Nominal USD: Underlying storage/bandwidth dollars are nominal (year of sale), not CPI-adjusted. Inflation adjustment would steepen early declines further in real terms for many eras.

Not financial advice. Order-of-magnitude illustration for educators. Spike years prove the thesis is not “prices only go down every month.”

Archival Node

Runs the full 750+ GB blockchain plus indexes (~753 GB as of mid-2026).
Requires 2 TB storage — today that means spending roughly $500–$550 on a Pi 5 or N100 mini PC plus 2 TB SSD (memory and NAND shortages pushed both up in 2026). Longer term, as drive prices normalize, this cost should come down.
Best if you want to run an Electrum server or scan many different wallets quickly. Fast and convenient, but more expensive.

Pruned Node

Only keeps recent blocks — total size around 15–20 GB.
Can usually run on your existing laptop or desktop at almost no extra cost.
Wallet scanning is slower, and switching to a new wallet often requires a full reindex, which is time-consuming.

Simple trade-off: pay more for convenience and flexibility, or keep it cheap and simple if you're only using one or two wallets regularly.

Why It Got Cheaper — FAQ

2009–2011 spare-PC era: Early nodes reused desktops you already owned — kit is ~$50 marginal (cables/wear), not a new PC. Setup totals ~$100–110 because the model still prices a small data drive. Dedicated kit pricing starts in 2012 as the chain left the MB scale.

Storage price history: Bulk HDD $/TB fell from ~$110 (2009) to ~$10 (2026). Consumer SSD/NVMe was ~$50–70/TB through 2024–2025, then spiked to ~$145/TB in 2026. Auto mode prices HDD through 2020, SSD from 2021+ (usual archival recommendation). You always buy whole drive SKUs: min 0.5 TB → 1 TB (2015+) → 2 TB from 2019 (future-proofing before need alone forces it). USB enclosure is charged only for external-drive eras (2012–2023), not spare-PC internals or modern NVMe HAT kits.

2020→2021 step: Auto storage flips HDD→SSD while Pi-era kits were scalped — modeled one-time setup roughly doubles. That is policy (recommended media + kit market), not a sudden chain-size shock.

Efficient hardware: Raspberry Pi 4/5 runs Bitcoin Core with 4–8 GB RAM, replacing $500+ desktops from 2009–2012. Hero eras match the cost model: spare PC → laptop → NUC → Pi+HDD → Pi+SSD → N100/NVMe.

Bandwidth abundance: Unlimited home internet became standard; node traffic is a tiny fraction of monthly caps after initial sync.

OpEx assumptions (USA): Internet plan cost follows USTelecom Broadband Pricing Index anchors — ~$38/mo entry broadband (2009), ~$66 (2015 popular tier), ~$52 (2020), ~$70 (2025–26 BPI-Speed 100–940 Mbps). Default attributes 15% of the plan to the node. Electricity uses EIA residential averages: 11.5¢/kWh (2009) → ~18.8¢/kWh (2026, Apr EIA). Default watts are era-aware (always-on desktop ~40–55 W → NUC ~25 W → Pi ~12 W → modern mini PC ~15 W); override with the slider.

Software optimization: Compact blocks, assumevalid, and better I/O let Core run on modest hardware while still validating the full chain.

The 2026 NAND spike: Consumer 2 TB NVMe jumped from ~$120–150 (2025 lows) to ~$290+ in 2026 as AI datacenters absorbed NAND supply. The same squeeze hit DRAM: Pi 5 8GB went from $80 to $175, and 16 GB N100 mini PCs from ~$160 to ~$250. A 2026 archival build is ~$530 — roughly double 2024.

The catch: Absolute chain size still grows ~80–85 GB/year (2024–2026; Bitcoin Core's own disk guideline rose 810 → 897 GiB from Sep 2025 to Sep 2026). Archival nodes always need more disk — historically $/TB fell faster than the chain grew; the 2026 spike is a temporary reversal.

Pruned vs archival storage: Pruned nodes validate the full chain but discard old block files. Steady-state disk grew with the chainstate (UTXO set): <1 GB (2010–12), ~2–3 GB (2015), ~5 GB (2017–18), ~8–10 GB (2020–22), ~15–20 GB today (2025–26, ~12.5 GB chainstate + headroom) — still far below the ~940 GB archival minimum. First sync still downloads the full chain (~753 GB bandwidth today) either way.

Pruned cost model: Pruned mode assumes Core on a laptop/desktop you already own. One-time setup is $0 by default — you use existing free disk space (~18 GB today), not a new Pi or USB drive. The real pruned cost is annual OpEx (electricity + internet share). Toggle “USB/SD reference” in Assumptions if your disk is full and you’d buy a small stick.

2031–2040 (projected): Hardware class stays similar (efficient mini PC / ARM, 16 GB RAM) — bigger change is buying 4–8 TB NVMe instead of 2–4 TB. Baseline assumes NAND supply recovers and $/TB resumes deflation.