Biotech IndustryAdvanced~20 min

Drug Development Pipeline

Why a drug costs billions and probably fails anyway

Set success probabilities, costs, and timelines for each clinical phase, then compute the true cost per approved drug and the risk-adjusted NPV that decides go/no-go.

The takeaway

The billion-dollar price tag is not the cost of the drug that worked. It is the cost of the drug that worked plus the nine that didn't, and that division is the whole business model.

Probability of successAttritionPhase II failureCapitalized costRisk-adjusted NPVPatent cliff

Read the theory: Gene Therapy Delivery

A generic small-molecule programme. Phase I → approval lands near the famous ~7–10%.

Value the programme as of

Investors do not value a drug from the beginning of time, they value it from where it stands today. Sunk cost is gone; only the remaining odds, spend and time count.

Phase I → approval

7.4%

the number everyone quotes

Discovery → approval

1.45%

1 in 69 programmes

Out-of-pocket cost

$460M

one clean run, no failures

Cost per approved drug

$2.83B

including the failures

rNPV at Ph I entry

$38M

GO, value creating

Stage parameters

Click a stage to tune its probability of success, cost and duration. Defaults are approximate published ballparks (Tufts/DiMasi, BIO–Informa success-rate reports) rounded for teaching.

~100–500 patients. The first honest test of whether the drug does anything. This is where the industry's dreams go to die, the lowest probability of success of any stage.

Probability of success29%

Cost70M

Duration2.5 yr

reach this stage: 10.1% · from here to approval: 14.3% · 7.0 entries per approval

Commercial assumptions

Peak annual sales1500M

Years of exclusivity after launch11 yr

Years to reach peak sales3 yr

Operating margin50%

Discount rate (cost of capital)11%

Standing at Phase I, Safety

Remaining PoS → approval7.4%

Remaining spend / time$420M · 8.7y

Peak-year operating profit$750M/yr

NPV if it works (no risk adj.)$1.32B

rNPV (risk-adjusted)$38M

Break-even peak sales$1.01B/yr

rNPV = Σ (cash flow × probability of reaching it) ÷ (1 + 11%)^t, with t measured from Phase I, Safety and mid-year discounting. Break-even peak sales is solved, not guessed: rNPV is linear in peak sales, so the peak revenue at which rNPV = 0 is just the risk-adjusted PV of costs divided by the risk-adjusted PV of revenue per $1M of peak. At industry-average odds it lands near $1.01B, which is precisely why the industry only chases blockbusters.

Attrition compounds

5,000 programmes in → 72.5 approved drugs out

Programmes entering5000

Target

−1,750 killed

3,250

Lead

−1,625 killed

1,625

Preclin

−650 killed

975

Ph I

−468 killed

507

Ph II

−360 killed

147

Ph III

−62 killed

Review

−13 killed

Launch

−0.0 killed

survivors leaving each stage · log-scaled widths · faded bar = programmes that entered

Stage	PoS	Enter	Survive	Killed	Entries / approval	Cost carried
Target Discovery	65%	5,000	3,250	−1,750	69.0	$345M
Hit / Lead Discovery	50%	3,250	1,625	−1,625	44.8	$448M
Preclinical	60%	1,625	975	−650	22.4	$560M
Phase I, Safety	52%	975	507	−468	13.5	$404M
Phase II, Efficacy	29%	507	147	−360	7.0	$490M
Phase III, Confirmatory	58%	147	85	−62	2.0	$507M
Regulatory Review	85%	85	72	−13	1.2	$24M
Approval & Launch	-	72	72	−0.0	1.0	$50M
True cost per approved drug	Σ (stage cost × entries needed per approval)					$2.83B

Where the billion dollars comes from

A single successful programme only spends $460M out of pocket. But to get one drug through, you must start 69 discovery programmes, 13.5 Phase I trials and 2.0Phase III trials, and the survivor's price has to pay for all of them. Charge each stage its cost times the number of attempts it takes, and the true cost per approval is $2.83B - 6.1× the direct cost. Published headline figures ($1–3B) additionally capitalise this at a cost of capital over 10+ years, which pushes it higher still.

The three lessons

Attrition multiplies

No single stage has hopeless odds, 52%, 29%, 58%, 85%, but you must pass all of them. Multiply and Phase I → approval is 7.4%. Probability chains are why the industry looks irrational from the outside and merely expensive from the inside.

The valley of death

Move the valuation anchor to Target Discovery: rNPV goes −$11M. Move it to Phase III: $938M. Same drug, same science, only the resolved risk differs. Early assets are worth almost nothing on paper, which is exactly why good biology starves between the lab and the first patient.

The exclusivity clock

The patent starts running years before the first patient is dosed. Every year of delay is a year sliced off the selling window, not added to the end. Drop exclusivity from 11 to 8 years and watch the rNPV fall, that is the same arithmetic that drives accelerated approval, priority review vouchers, and the fight over patent cliffs.

Default probabilities, costs and durations are approximate published ballparks(e.g. Tufts CSDD / DiMasi et al.; BIO–Informa–QLS clinical success-rate analyses; Wong, Siah & Lo 2019), rounded for teaching. Real figures vary enormously by indication, modality and sponsor. Costs are out-of-pocket in constant dollars; the headline "$2.6B per drug" figures additionally capitalise spend at a cost of capital, which this model does not do, so treat the cost-per-approval here as a floor.

Circular diagram of the drug discovery pipeline from target identification through clinical trials to approval — The drug pipeline: of ~10,000 screened compounds, roughly one survives discovery, preclinical work, and three trial phases.
Image: Ben Mills, Wikimedia Commons (public domain)

← Previous lab

Molecular Cloning Designer