This is a continuation of sharing my ideas similar to pop-corn-brain and trillion dollar idea. This idea came from reading an excellent piece on Pure AI Plays in Biotech by Luis Pareras. Seeing that his pattern of thinking about AI and ventures can be applied across multiple verticals, and very closely follows some of my earlier work in the skincare industry.
The Problem
Cosmetic ingredient manufacturers are trapped in low-margin commodity positions (15–25% EBITDA) while beauty brands capture 70–80% gross margins. Ingredient suppliers hold rich but under-monetized datasets—efficacy, safety, stability, sensory profiles—locked in silos.
Meanwhile, brands and contract manufacturers suffer 12–18-month development cycles and high failure rates in a trend-driven market that shifts every season.
The result: wasted R&D spend, slow reaction to consumer trends, and a massive unclaimed value gap between raw ingredients and finished products.
The Analogy
AI-first biotech platforms (Recursion, Exscientia, Insilico) rewired drug discovery economics by combining proprietary data, generative models, and closed feedback loops—transforming molecule design into scalable software.
The same “intelligence-toward-zero” logic applies to cosmetics: as formulation intelligence becomes abundant, value migrates to those who own proprietary data, close the testing loop, and adapt faster than trend cycles.
Just as pure AI biotech platforms compress biology, an AI-formulation platform can compress beauty R&D—turning ingredients into IP and formulation into a learning system.
The Solution
A three-part acquisition + build strategy. Meaning this requires significant capital.
1. Platform Acquisition
Acquire a mid-sized ingredient manufacturer ($50–150 M revenue) with:
- Established brand relationships and proven ingredient portfolio
- Proprietary data on stability, efficacy, and safety (on ingredient level)
- Moderate R&D but no formulation capability
- EBITDA 15–25%, ripe for tech-driven margin expansion
2. CRO Bolt-On
Add a cosmetic testing CRO ($10–30 M revenue) with:
- Stability, microbiology, and efficacy labs
- ISO 17025 / GMP certification and regulatory coverage across US, EU, Asia
- Brand clients and data infrastructure for validation loops
- Some formulation experience, but not full CDMO
3. AI Integration Layer
Build an AI stack that fuses:
- Formulation Intelligence – generative models trained on ingredient + CRO data to design stable, efficacious formulations.
- Trend Intelligence – LLM-based analytics scanning social, search, and launch data to detect emerging claims, textures, and ingredient clusters.
- Brand Mimicry – private style embeddings that learn each brand’s sensory, linguistic, and packaging DNA, enabling “on-brand” formulations aligned with live market trends.
This effectively creates a similar closed-loop lab / commercial feedback engine that we see in a series of the AI biotechs. This allows for compressing the timeline in development and getting faster to market on trends. And each new formulation adds rich data to further optimize the engine.
4. Business model
Co-creation partner with faster to market execution. A full-stack formulation-as-a-service tool, effectively moving from smaller profit margins, into royalty agreements on full product. Moves from selling single ingredient components to selling the entire final formulation.
Market & Economics (Back-of-Envelope)
- Global skincare market: ~ $120 B (2024), growing ~6 % CAGR.
- Typical hero SKU: $50 M annual retail sales, ~70 % brand gross margin → $35 M gross profit.
- Platform share: 10 % of gross profit = $3.5 M per SKU in recurring royalties.
- With full-stack supply (ingredients + pack + fill), add another ~$10 M revenue per SKU at ~40 % margin = +$4 M EBITDA.
- Total per-SKU contribution ~ $7–8 M EBITDA on ~$0.5 M development cost.
Portfolio math (Year 5 steady state):
- 10 partner brands × 3 SKUs each = 30 live products.
- $900 M retail sales / $630 M brand profit.
- Platform royalties (~10 %) = $63 M + supply EBITDA (~$4 M × 30) = $120 M.
- ~$180 M total EBITDA potential.
Even at a conservative 8–9× multiple, that implies $1.4–1.6 B enterprise value on a ~$250 M entry. That is ~5–6× MOIC and 30 %+ IRR.
These numbers are before all the other usual PE tricks to leverage the business, such at FTE optimization, debt financing, back-office streamlining, cashflow optimization etc.
Challenges
The above assumes we can compress timelines from 18 months to 6 months – while not easy we did it at Revea and was able to move at this speed. The biggest challenge will be to get the right partners and get them fast enough. None of the bigger skincare conglomerates move fast, and there will be massive not invented here syndrom about taking on a full-stack external formulation service.
My notes
This is just one example of how the thinking laid out in the paper can be applied to a vertical. I am sure there are multiple others where the asymmetry of data and capability can create similar ventures. I would challenge you, the reader, to look at your own industry and see if you can spot them. I’d love to hear about similar takes.