Phase Space AI

06 Model Notes

BillionToOne [BLLN]

BillionToOne, Inc. [BLLN] — Financial Model Notes (Task 2)

Workbook: BLLN_model.xlsx — six tabs, live formulas throughout. Every figure below was read back out of Microsoft Excel after a forced recalculation, not computed in Python and pasted.

Tabs

  1. Assumptions — market data, discount rate, operating drivers, and a block of sourced facts. All hard-coded inputs are shaded; everything downstream is a formula referencing them.
  2. Income Statement — FY2023A–FY2035E. Actuals transcribed from filed statements; forecast years driven by revenue × cost percentages.
  3. Balance Sheet — FY2024A, FY2025A, Q1-2026A, with a live balance check.
  4. Cash Flow — FY2023A–FY2030E, linked to the Income Statement.
  5. DCF — 10-year explicit forecast, terminal value on both Gordon and exit-multiple bases.
  6. Scenarios — Bull/Base/Bear, the Gate 4 expected-return calculation, the E[R] range and flip points, and the required base-rate check.

Verification — what the recalculation actually returned

Tie-out to the filed FY2025 10-K (Income Statement tab, rows 22–28). This is the check that matters, because a zero balance check verifies internal consistency and not input accuracy:

Line Filed Model Difference
Revenue 305.112 305.112 0.00
Gross profit 208.458 208.458 0.00
R&D 49.384 49.384 0.00
SG&A 143.051 143.051 0.00
Operating income 16.023 16.023 0.00
Net income 7.454 7.454 0.00

Balance check: 0.00 at FY2024, FY2025 and Q1-2026.

Key computed outputs: WACC 11.45%; EV $6,639.27m; FY2035E EBIT $517.0m; DCF value/share $53.35 (Gordon) and $79.90 (4.5x exit); scenario targets $74.63 / $136.01 / $203.62; E[R] +0.64%, hurdle 4.7%, verdict FAIL.

An independent Python implementation produced $53.20 and $80.25 for the same two DCF cases — agreement to within 0.5%, which is the cross-check that the workbook is doing what it claims.

Two real errors found by opening the workbook — recorded because they changed the answer

The skill requires verification by actually opening the file rather than trusting that a formula string was written without error. Both of the following produced a workbook that looked fine and was badly wrong:

  1. Assumption row references drifted. The DCF discount factor referenced Assumptions!$B$19, which is the beta cell (1.35), not WACC — i.e. the model was discounting at 135%. D&A, capex, ΔNWC, terminal growth and the exit multiple were all similarly off by three rows, and the cost-of-equity formula itself multiplied the wrong two cells. Fixed and re-verified.
  2. The forecast is ten years (columns E–N) but the cost-percentage lists held only nine entries. FY2035 therefore carried revenue of $1,880m with no COGS and no operating expenses, so EBIT equalled revenue. Because the Gordon terminal value keys off the final year, this inflated TV from $4,297m to $17,960m and produced a "fair value" of $147/share — above spot. The exit-multiple branch, which keys off revenue rather than FCF, was unaffected and disagreed by $60/share; that disagreement is what exposed the bug. Fixed and re-verified.

Neither error would have been caught by the balance check, which read 0.00 throughout.

Forecast drivers (base case)

FY2026E FY2027E FY2030E FY2035E
Revenue ($m) 462 615 1,105 1,880
Gross margin 73.0% 72.5% 71.0% 71.0%
R&D % of revenue 12.5% 12.0% 10.5% 10.0%
SG&A % of revenue 44.0% 42.0% 35.5% 33.5%
Operating margin 16.5% 18.5% 25.0% 27.5%

D&A 1.6% of revenue; capex 4.5%; ΔNWC 10% of incremental revenue. Cash tax steps 3% → 23% as the $256.8m accumulated deficit and its full valuation allowance are absorbed.

Base-rate check (required)

The Base case takes revenue from $305m (FY2025) to $1,105m (FY2030) — a 29.4% five-year CAGR. Against the reference class (Chan, Karceski & Lakonishok 2003), growth persistence beyond chance is close to nonexistent and sustained >20% growth for five-plus years is a top-decile outcome. The Base case is therefore explicitly a top-decile assumption, and the memo names that prior rather than assuming it away.

The sector counter-evidence is real and is stated alongside: Natera compounded revenue from $391m (2020) to $2,337m (TTM) at roughly 35%, and Exact Sciences did likewise. Molecular diagnostics with expanding reimbursement is one of the few reference classes where the base rate genuinely beats the all-industry prior — but only for the winners, and survivorship is exactly what a base rate is supposed to correct for.

The Bull case (37.0% five-year CAGR) is labelled as optionality, not expectation. The reverse-DCF path implied by the current price — FY2035 revenue of ~$3,515m, or 11.5x FY2025 — sits above even the Bull case, which is the core valuation finding.