One device set, three choices
This is the Acme synthetic sample estate's Cisco successor, sample 2029 end-of-support cohort in the Americas: sample 477 devices in every column. The three accepted runs use the same slice key and membership fingerprint. These are illustrative outputs. Synthetic successor hardware and third-party maintenance prices could change their order. Not a recommendation.
Replace now would win if the cost of waiting, including outage exposure and continued support, exceeded its earlier hardware spend. Replace at end of support would win if the devices could be kept safely until then and later replacement cost less in present value. Extension would win if third-party support, post-support escalation and outage exposure stayed low enough to offset replacement. The required price and risk bands have not been run, so the supported choice is Not modeled.
What the same devices cost
Each column is a refresh choice for the same sample devices. The five-year total is undiscounted. The NPV values discount spending at the end of each year in the sample 2026 through 2030 window; the rate is an assumption.
Replace now
Five-year total
Sample $2,284,902
- NPV at 5%
- Sample $2,170,333
- NPV at 8%
- Sample $2,106,969
- NPV at 12%
- Sample $2,028,046
Run 776 · eol:2029|region:Americas · Existing contracts retained to replacement
packages/demo/bcs-dataset-pricing (synthetic successor and TPM prices); End-of-support dates from the Acme dataset (vendor where matched, synthetic otherwise); cohort vendor share unverified
Replace at end of support
Five-year total
Sample $2,284,902
- NPV at 5%
- Sample $1,823,875
- NPV at 8%
- Sample $1,603,552
- NPV at 12%
- Sample $1,360,263
Run 774 · eol:2029|region:Americas · Existing contracts retained to replacement
packages/demo/bcs-dataset-pricing (synthetic successor and TPM prices); End-of-support dates from the Acme dataset (vendor where matched, synthetic otherwise); cohort vendor share unverified
Extend with third-party support
Five-year total
Sample $686,416
- NPV at 5%
- Sample $599,969
- NPV at 8%
- Sample $556,314
- NPV at 12%
- Sample $505,790
Run 873 · eol:2029|region:Americas · Third-party support extension
packages/demo/bcs-dataset-pricing (synthetic successor and TPM prices); End-of-support dates from the Acme dataset (vendor where matched, synthetic otherwise); cohort vendor share unverified
The displayed discount-rate band changes the present values, but it does not test the synthetic prices that could decide the choice. The supported decision remains Not modeled.
Sensitivity runs still needed
- Same-cohort low/base/high successor hardware price for both replacement timings
- Same-cohort low/base/high third-party maintenance price and post-support escalation
- Same-cohort low/base/high outage exposure under each refresh choice
- Lifecycle provenance join by device, PID and source, followed by a timing sensitivity for synthetic dates
Where the ripple stops
The other levers cannot be added to these totals from the accepted export. Each needs a run on the exact membership above.
Location: Not modeled
Energy, labor and carbon reference packs exist, but no country rerun uses this exact cohort.
Source: bcs-ref-energy-prices; bcs-ref-labour-rates; bcs-ref-carbon-intensity
Support tier: Not modeled
The branch Americas support output uses a different device set. Its total also contains contract costs already included in refresh.
Source: NASPO AR3227 ceiling; acme-corp-cases R-34; synthetic prices for other vendors
Optics: Not modeled
The device-power reference pack has optics watts, but Acme refresh is device hardware only. This cohort has no exact optics quantity and power ledger.
Source: bcs-ref-device-power; separate optics sweep
Automation: Not modeled
The labor reference pack feeds automation, but its accepted model uses a synthetic single-device anchor, not this cohort.
Source: bcs-ref-labour-rates; Acme automation case
What a spreadsheet would have to hold
This comparison rests on 3 model runs and 6 tracked input classes, not a count of individual source records or modeled interactions. A spreadsheet would have to retain exact device membership, end-of-support dates, successor mapping, contract state, prices, annual cash flows and discount treatment for each choice. A general AI tool cannot verify this comparison without those sourced inputs and run records.
Sources and method
Each refresh output already contains capex_new + opex_new + capex_retain + opex_retain for its own choice and five-year window. Sum its own annual spend once. Discount those same annual totals for NPV. Do not add a support, energy, stack or automation output: those use different device sets, and the refresh output already includes retained contracts and the extension support choice.
The refresh figures use Acme round-2 accepted refresh outputs, after !112; no affected old sweep variants. The source export is tools/bcs/acme/results/final/acme-results.json; slice fingerprint d86050cdb1bb091190501a4847bdf1fed1cedee8e2dd01cc2213d6817808a150. Every cost cell identifies its run, slice, contract choice and price package. This is aggregate lineage; a device-level contract and price ledger is still needed. This page uses no affected variant from the earlier refresh sweep.
- Device membership: slice-members.json (sample estate)
- End-of-support dates: packages/demo/bcs-dataset-eox (vendor where matched, synthetic otherwise; cohort share unverified)
- Cisco successor mapping: packages/demo/bcs-dataset-eox (mixed; mapping coverage limited)
- Successor hardware and third-party support price: packages/demo/bcs-dataset-pricing (synthetic)
- Existing contract state: nb04 Acme model outputs (sample estate)
- Discount rate: packages/demo/acme-corp-cases/RULINGS.md R-25 (synthetic sensitivity)
What remains untested
- Acme device membership, successor prices and third-party support price are synthetic.
- End-of-support dates come from the Acme dataset, vendor where matched and synthetic otherwise. The cohort vendor share has not been established by a device-to-PID-to-source join.
- Post-support escalation and outage exposure were not modeled for these choices.
- The discount rates are assumptions. The missing price, outage and lifecycle sensitivity runs prevent a supported choice.
A joint run on this cohort is unavailable. Until the same slice can run with sourced location, support and optics inputs and an automation workload band, the combined answer is Not modeled.