Lightspeed is raising $600 million for a secondary fund. Code name: Project Mercury. The target? OpenAI and Anthropic. The thesis? Chasing alpha until the trail goes cold.
This isn't a fresh injection of capital into the labs. It's a GP-led continuation fund — a structured exit for early investors, a new entry for later-stage LPs, and a management fee stream for Lightspeed. The companies themselves don't see a dime. The real story is the liquidity machine: a $600M bet that the AI frontier is still worth betting on, even when the IPO window is sealed shut.
Let me break this down. Because if you're not watching the secondary market, you're missing the single most important signal in AI finance right now.
The Hook: Project Mercury Launches
A $600 million secondary fund. Select V Fund, Opportunity II Fund, and a separate managed account (SMA) — all pooled into one buying spree. The SMA is the tell: some sovereign wealth fund or mega-endowment demanded a dedicated AI sleeve, no co-mingling. That's the kind of institutional hunger that moves markets.
Lightspeed is buying equity from early employees and early investors — not from OpenAI or Anthropic. They're paying a premium to the last round valuation (likely 20-50% more). Why? Because the supply of top-tier AI equity is vanishingly small. Sequoia, Thrive, Khosla — they're not selling. The only way in is secondary.
The Context: Why Now?
We're in a bull market for AI, but a bear market for exits. The IPO window for high-burn, high-valuation tech companies hasn't fully opened since 2021. OpenAI is burning billions. Anthropic is burning hundreds of millions. Both are generating revenue — OpenAI's annualized run rate hit $30-35B by mid-2024, Anthropic's was in the single-digit billions but growing at 50%+ quarter-over-quarter — but they're not profitable. The public markets demand a path to profit. The private markets demand a story.
Lightspeed's story is simple: AI capability curves are still steep. Model iterations every 6-12 months. GPT-4o vs. Claude 3.5 Sonnet — two different philosophies, one commoditization risk. But Lightspeed is hedging by buying both. They're betting that the moat (compute, data, talent, brand) will hold for at least 3-5 years. That's a bold call in a world where open-source models (Llama 3, Qwen, Mistral, DeepSeek) are closing the gap.
The Core: What the $600M Really Buys
Lightspeed isn't just buying equity. It's buying positioning. The firm wasn't a lead investor in either OpenAI or Anthropic — Sequoia, Thrive, Menlo, Khosla have the deep relationships. So Lightspeed is using capital as a wedge: "We'll pay more than anyone else for your shares." That's a valid strategy, but it's a price war, not a diligence advantage.
The fund's structure is telling. Select V is a classic VC fund. Opportunity II is a growth-stage vehicle. The SMA is a bespoke mandate. Together, they allow Lightspeed to offer custom AI exposure to LPs with different risk appetites. That's asset management creeping into venture capital. It's a sign that AI is becoming an asset class, not just a sector.
But here's the catch: secondary funds don't give you board seats or information rights. Lightspeed is a passive financial investor in these companies. They're betting on the narrative, not on governance. If the narrative shifts — if model commoditization crushes margins, if regulatory hammer falls, if key talent walks — they have no levers to pull.
The Contrarian Angle: The Liquidity Trap
Everyone is celebrating this as a vote of confidence in AI. I see it differently. This is a liquidity trap — a mechanism that lets early investors cash out without waiting for an IPO. It's a band-aid for a broken exit market.
Think about it: Lightspeed's LPs get to sell their OpenAI/Anthropic stakes at a premium. The new LPs get access to scarce assets. Lightspeed gets management fees and carry. Everyone wins — except the actual companies. The capital doesn't fund R&D. It doesn't hire researchers. It just changes hands.
This is DeFi Summer all over again. Remember 2020? Liquidity mining subsidized TVL numbers. Stop the incentives, and the users vanish. Here, the incentive is the premium. Stop the premium, and the secondary market dries up. The underlying value of the equity depends on the narrative holding. And narratives are fragile.
There's also a structural risk: if secondary funds become the norm, the IPO becomes optional. That means companies stay private longer, with less public scrutiny. Governance becomes opaque. Valuation becomes a negotiated number, not a market-clearing price. We saw this in crypto with private token sales — it created a two-tier market where insiders got rich while retail got left holding the bag.
Lightspeed's move also accelerates capital concentration. The top 5 AI companies (OpenAI, Anthropic, Google, Meta, Microsoft) already vacuum up the vast majority of AI funding. By pouring $600M into secondary purchases, Lightspeed is reinforcing that star system. The middle class of AI startups — the ones building on top of these models — will struggle to get VC attention. That's a structural drag on innovation.
The Takeaway: Watch for the Cascade
Project Mercury is a signal. If it succeeds, expect other VCs (a16z, Sequoia, Benchmark) to launch similar secondary funds. The late-stage AI market is becoming a secondary market in disguise. The real question is: will the IPO window ever open, or will AI become a permanent private asset class?
I've been in this game long enough to know that liquidity isn't always a blessing. When the music stops — when the next model iteration disappoints, or when regulation bites — the secondary market will be the first to freeze. The alpha will turn to beta, and then to zero.
Chasing the alpha until the trail goes cold. That's the mantra. But when the trail turns to ice, you better have a way out.
— William Jackson, Exchange Market Lead, Zurich
