# Assumptions & sources — Delta Starlink Cost

**Audience:** operators / researchers (the public UI is shareholder-facing and does not lead with this file).
**Status:** All dollar figures on the site are **model estimates**, not audited airline fuel accounting.
**Last reviewed:** 2026-09-30
**Config file:** [`js/config.js`](js/config.js) — three scenarios (`best` / `average` / `worst`); default = `average`.

---

## What the counter measures

Extra jet-fuel **cost** attributable to flying older, higher-drag Ku/Ka / radome satcom antennas instead of a Starlink Terminal for Aviation–class low-profile ESA, on the share of Delta **mainline** flying hours that **could already have been** on Starlink if Delta had matched a United-like install pace.

\[
\text{extra \$/day} = N_{\text{aircraft}} \times h_{\text{day}} \times f_{\text{Starlink}} \times g_{\text{extra}} \times p_{\text{JetA}}
\]

\[
g_{\text{extra}} = (\text{legacyPct} - \text{starlinkPct}) / 100 \times \text{baselineBurnGalPerHour}
\]

The displayed total is the integral of that constant daily rate from `startDateISO` through now (tick ≈ 50 ms; whole dollars).

**Out of scope:** Delta Connection / regional fleet (325 aircraft in the same 10-Q), passenger revenue / NPS, install labor & STC, Leo vs Starlink commercial terms, carbon.

### Future v2 (not on the public homepage)

A later model could track actual Delta flights and daily Jet A prices; v1 intentionally uses a constant daily rate.

---

## Scenario table (defaults)

| | **Best** | **Average (default)** | **Worst** |
|--|----------|------------------------|-----------|
| \(N\) aircraft | 900 | **1,004** | 1,004 |
| \(h\) hours/aircraft/day | 9 | 10 | 11 |
| \(f\) counterfactual Starlink share | 0.20 | 0.30 | 0.45 |
| legacy / Starlink penalty % | 1.3 / 0.3 | 2.0 / 0.3 | 2.55 / 0.3 |
| baseline burn (gal/hr) | 800 | 800 | 800 |
| \(g\) extra gal/FH | **8.0** | **13.6** | **18.0** |
| \(p\) Jet A $/gal | 4.00 | 4.50 | 5.00 |
| **≈ $/day** | **~$51.8k** | **~$184.3k** | **~$447k** |
| Span vs best | 1× | ~3.6× | ~8.6× (within ~10×) |

Start window: `2026-03-30T00:00:00.000Z` (~6 months before 2026-09-30).

---

## Known vs assumed

| Input | Average value | Classification | Why |
|-------|---------------|----------------|-----|
| \(N\) mainline aircraft | 1,004 | **Known** | SEC Form 10-Q, period ended 2026-06-30 — mainline *operating* fleet = 1,004 (not the 1,314 combined figure that includes regional). |
| Jet A $/gal | 4.50 | **Known** (spot index) | Argus / Airlines for America U.S. Jet Fuel Index late Sep 2026 ~$4.42–4.53; default rounded to $4.50. |
| Hours / aircraft / day | 10 | **Assumed** | Order-of-magnitude U.S. mainline utilization; not Delta-reported for this model. |
| Counterfactual Starlink fraction \(f\) | 0.30 | **Assumed** | Peer-pace proxy from United tracker (~607/1,669 ≈ 36% as of 2026-09-30); mid case uses 30%. **Not** a claim Delta contracted Starlink. |
| Drag delta \(g\) | 13.6 gal/FH | **Assumed** (labeled estimate) | Derived from public Nicolls / Ryanair–Starlink Jan 2026 talking points, applied to Delta’s Viasat/Hughes mix as an estimate. |

Best/worst move the assumed knobs (and soften/firm Jet A); average \(N\) stays on the SEC figure for mid and worst.

---

## Fleet & satcom context

### Mainline count — 1,004 (known for average/worst)

- Delta Air Lines Form 10-Q for the quarterly period ended **2026-06-30**: mainline operating fleet **1,004** aircraft. Regional (~325) is **out of scope**.
- Source: [SEC EDGAR — dal-20260630.htm](https://www.sec.gov/Archives/edgar/data/27904/000002790426000031/dal-20260630.htm)

### Connectivity path

- Delta today: **Viasat / Hughes** (legacy GEO-style / radome path) — **no Starlink contract**.
- Amazon Leo (formerly Project Kuiper): public plan to begin installs on ~**500** aircraft from **~2028**.
- Implication for the counter window (2026): nearly the entire mainline fleet remains on pre-Leo hardware.

Sources:

- [Delta / Amazon Leo announcement (Delta News Hub)](https://news.delta.com/delta-amazon-leo-sign-agreement-deliver-next-era-connected-travel-and-digital-experiences)
- [Amazon Leo × Delta (About Amazon)](https://www.aboutamazon.com/news/amazon-leo/amazon-leo-delta-in-flight-wifi-2028)

### Counterfactual fraction — 30% mid (assumed)

- United began Starlink installs in **2025**. Independent tracker as of **2026-09-30**: roughly **607 / 1,669 ≈ 36%** equipped.
- Average scenario uses **30%** as a conservative mid-window peer-pace proxy (“if Delta mainline had matched a United-like cadence, what share of hours would already avoid legacy-antenna drag?”).
- Best 20% / worst 45% bracket that assumption.

Sources:

- [United Starlink install rate index](https://unitedstarlinktracker.com/install-rate)
- [United accelerates Starlink — PR Newswire, 2026-06-22](https://www.prnewswire.com/news-releases/united-accelerates-starlink-wi-fi-rollout-with-first-widebody-transatlantic-flight-302806746.html)

---

## Drag / fuel burn delta

### Average — 13.6 gal/FH (assumed, derived)

Public January 2026 exchange (Ryanair vs Starlink / Michael Nicolls, Starlink VP Engineering):

| System | Stated fuel impact | Notes |
|--------|--------------------|-------|
| Legacy high-profile radome terminals | **~2%** | Public figure in the dispute |
| Starlink aviation terminal | **~0.3%** on 737-800 | Nicolls analysis |
| Baseline 737-class burn | **~800 gal/hr** | Same commentary |

\[
g = (2.0\% - 0.3\%) \times 800\ \text{gal/hr} = 13.6\ \text{gal/FH}
\]

Best uses 1.3% legacy → **8.0** gal/FH; worst uses 2.55% legacy → **18.0** gal/FH (same 0.3% Starlink and 800 gph baseline).

**Caveats:** public talking points / vendor analysis, not FAA certification drag polars for Delta’s exact Viasat/Hughes mix. Widebody vs narrowbody, aperture count, and fairing shape all move the real number. Treat \(g\) as a **labeled estimate**.

Sources:

- [Aerospace Global News — Musk vs O’Leary / fuel economics](https://aerospaceglobalnews.com/news/musk-vs-oleary-starlink-ryanair-wifi-costs/)
- [Benzinga — Starlink VP 0.3% claim](https://www.benzinga.com/news/space/26/01/49926178/starlink-adds-only-0-3-to-aircraft-fuel-burn-starlink-vp-says-elon-musk-eyes-further-cuts)
- [Aviation A2Z — Starlink / Ryanair fuel dispute](https://aviationa2z.com/index.php/2026/01/18/starlink-ryanair-spar-over-fuel-impact-of-in-flight-internet/)

---

## Jet A price

- **Airlines for America / Argus U.S. Jet Fuel Index** (wholesale airline-relevant): about **$4.42–4.53/gal** in late September 2026.
- Average default **$4.50**. Best $4.00 / worst $5.00 bracket the spot.
- Not FBO retail Jet-A. v1 uses a **constant** price per scenario.

Source: [Airlines for America — Argus U.S. Jet Fuel Index](https://www.airlines.org/dataset/argus-us-jet-fuel-index/)

---

## Worked examples

**Average (default):**

- Exposed FH/day = \(1004 \times 10 \times 0.30 = 3{,}012\)
- Extra gal/day = \(3{,}012 \times 13.6 = 40{,}963\)
- Cost/day ≈ \(40{,}963 \times 4.50 = \$184{,}334\)

**Best:** \(900 \times 9 \times 0.20 \times 8.0 \times 4.00 = \$51{,}840\)/day

**Worst:** \(1004 \times 11 \times 0.45 \times 18.0 \times 5.00 = \$447{,}282\)/day

Ratio worst/best ≈ **8.6×** (within ~1 order of magnitude).

---

## How to revise

1. Edit scenario objects in [`js/config.js`](js/config.js).
2. Keep `knownVsAssumed` flags honest when a primary source replaces a placeholder.
3. Update this file’s tables and links in the same change.
4. Bump `meta.lastReviewed`.
5. **Local only until Jason approves a push/deploy** — do not `git push` or `wrangler pages deploy` without explicit go-ahead.