Rohit Mittal @rohitdotmittal
vertical AI startups are trading at insane multiples right now
i dug into the numbers - ARR, valuations, and the math on whether these deals make sense
here's what i found
high level multiples: (caveat is that revenues are growing rapidly, so the multiples may already be lower)
- Cursor: $1B ARR → $29B valuation (29x)
- Sierra: $100M ARR → $10B valuation (100x)
- Harvey: $100M ARR → $8B valuation (80x) - raised $160M today
- Glean: $100M ARR → $7.2B valuation (72x)
- Abridge: $100M ARR → $5.3B valuation (53x)
For comparison, public vertical SaaS:
Veeva: $2.45B rev → $35B (14x)
Toast: $1.5B rev → $12.5B (8x)
Procore: $780M rev → $11.2B (14x)
ServiceTitan: $772M rev → $7.5B IPO (10x)
55x (for fast growth AI in private) vs 12x (lower growth public markets)
the gap is 4-5x
at 50x entry on $100M ARR with 100% annual growth:
year 1: $200M
year 3: $800M
year 5: $3.2B
exit at 10x (public comp) = $32B
entry cost = $5B
MOIC = 6.4x
IRR = ~45%
if the growth doesn't stay at 100%
realistic scenario:
year 1: 100% growth
year 3: 50% growth
year 5: 30% growth
adjusted year 5 ARR: ~$1B (not $3.2B)
new exit at 10x = $10B MOIC = 2x IRR = ~15%
still works. but not venture-style returns.
the magic number for 25%+ IRR with 5yr hold:
need 3x MOIC minimum
which means:
50x entry → need 15x+ exit with 80% CAGR
30x entry → need 10x exit with 60% CAGR
multiple compression could be the killer
by vertical:
Healthcare AI (Abridge, Ambience): 40-55x quite rational - clear ROI, replacing $15B/yr in scribe costs
Legal AI (Harvey, EvenUp): 50-80x fast adoption but crowded
Coding (Cursor): 30x fastest growth - most reasonably priced in a way - but competition from the big AI labs is a risk
valuations are pricing "winner take most" dynamics
but theres 5+ well-funded players per vertical
all assuming they'll be the winner
AI vertical SaaS is real - the revenue growth is unprecedented
but paying 50-100x ARR requires:
sustained 60%+ growth for 5+ years
no multiple compression
being the category winner
