AI startups that raised before ChatGPT are getting crushed in the secondary market

Venture capital has split into two AI markets that no longer talk to each other. Startups that last raised a round before ChatGPT launched in November 2022 are now trading at steep discounts on secondary markets — down 52% for 2022 vintages and as much as 68% for 2021 vintages, according to recent secondary market data. Meanwhile, companies built natively around generative AI are raising seed and Series A rounds at valuations 42% above non-AI baselines. Same asset class, same investors, two completely different pricing regimes.

Why the split happened

The mechanism is straightforward: a company that raised in 2021 built its pitch, its roadmap, and often its entire product architecture around a pre-transformer, pre-LLM world. If that company didn't pivot fast enough — or pivoted into a crowded generative AI feature set that a foundation model vendor could replicate in a product update — its technical moat evaporated. Investors who backed the 2021 vintage are now facing a choice: hold through years of dilution and uncertain exits, or sell on the secondary market at whatever price a buyer will pay, which today means a 50-plus percent haircut.

Compare that to a company that raised its first round in 2024 or 2025 with an AI-native thesis. It never had to unwind legacy architecture. Its metrics, if it has any revenue at all, get read against a growth curve investors assume will look like OpenAI's or Anthropic's. That assumption alone can justify a premium multiple even before the product ships.

The secondary market's role

Secondary markets — where existing shareholders sell stakes to new investors without the company raising a primary round — used to be a quiet corner of venture finance, mostly used by early employees cashing out or funds needing liquidity before a fund's term ends. In 2026, they've become the pricing mechanism of record for an entire cohort of startups. When primary investors won't fund a follow-on round at the last valuation, secondary trades set the real price, and that price is now public enough that other cap table holders use it as a mark.

There's a perverse wrinkle here too: some pre-2022 companies with genuinely weak fundamentals are still seeing active secondary demand, not because buyers believe in the business, but because AI-sector deal flow is so hot that speculative buyers will take a flyer on anything with "AI" in the pitch deck, hoping to flip the position again before anyone checks the underlying numbers. Traders have started calling these "pump and dump" rounds — a label borrowed from public markets, now applied to what used to be a staid private-equity-adjacent process.

What this means for founders and employees

If you hold equity in a company that last raised before late 2022, don't assume your last 409A valuation reflects what the stock is actually worth today. Ask your company directly whether it has seen secondary market activity, and at what price. If you're negotiating an offer with equity from a pre-ChatGPT-era startup, price the equity at a discount to the last preferred round, not at face value — a 50%+ discount is now a reasonable working assumption unless you have direct evidence otherwise.

If you're evaluating whether to join or invest in an AI-native startup instead, remember the flip side of this dynamic: valuations 42% above baseline are pricing in growth assumptions that most companies won't hit. The same repricing correction that hit 2021-vintage startups will eventually hit today's premium-priced AI-native startups once the market re-rates growth expectations against actual revenue. Ask what happens to the current valuation if revenue growth comes in at half the assumed rate — if the answer is "down round," treat the equity accordingly.

Takeaways

The AI funding market isn't one market — it's two, priced on almost entirely different logic. Vintage matters more than product quality right now. Anyone holding equity, negotiating an offer, or writing a check in this space needs to know which market they're actually in before trusting any headline valuation number.