← All autopsies

// STARTUP COMPARISON

Lemonade (valuation crisis) vs Silicon Valley Bank

Lemonade (valuation crisis) failed in 2022 due to Unit Economics. Silicon Valley Bank failed in 2023 due to Unit Economics. Both failed for the same reason — Unit Economics.

METRIC🔥 Lemonade (valuation crisis)🔥 Silicon Valley Bank
SectorFintechFintech
CountryIsraelUSA
Founded20151983
Died20222023
Raised$480MPublic company (SIVB)
Peak$7B+ market cap$209B assets
Primary CauseUnit EconomicsUnit Economics

// WHY EACH FAILED

🔥 Lemonade (valuation crisis)
Unit Economics
Lemonade, an Israeli AI-powered insurance startup, IPO'd in 2020 and reached a $7B+ market cap. The company's promise: AI would reduce loss ratios and adverse selection. By 2022, loss ratios remained above 90% (industry standard is 60-70%), the company was burning $150M+ annually, and the stock had fallen 90%+ from its peak. The AI advantage in insurance underwriting proved harder to achieve than marketed.
// LESSON
AI-powered insurance requires the same years of proprietary claims data as traditional actuarial methods before loss ratios improve. The AI is not a shortcut to accurate risk pricing — it is a better tool for processing the same data incumbents already have.
🔥 Silicon Valley Bank
Unit Economics
Silicon Valley Bank collapsed in March 2023 after a bank run driven by duration mismatch. SVB had invested deposits in long-duration bonds during low-rate periods. When rates rose, those bonds lost value. SVB announced a $1.8B loss on bond sales and a capital raise — triggering a $42B bank run in 24 hours. The FDIC seized SVB on March 10, 2023 — the second-largest bank failure in US history.
// LESSON
Asset-liability duration matching is not optional for banks. Investing short-term deposits in long-term bonds is a structural bet against rising rates. SVB had $80B in long-duration bonds when the Fed began the fastest rate rise cycle in 40 years.

// IN THE SIMULATION

Lemonade triggers AI_UNDERWRITING_PROMISE_FAILURE — the simulation models insurtech AI claims as requiring 7+ years of proprietary claims data to outperform actuarial tables. Without that data, loss ratios match or exceed industry averages regardless of AI sophistication.

SVB triggers DURATION_MISMATCH_BANK_RUN — the simulation models banks with long-duration bond portfolios as having existential rate sensitivity. A 400bps rate rise on a long-duration portfolio creates mark-to-market losses that exceed capital when forced to sell.

// EXPLORE FURTHER

Full Lemonade (valuation crisis) autopsy →Full Silicon Valley Bank autopsy →More Unit Economics failures →All Fintech autopsies →