I felt the floor tilt when the Bloomberg alert hit. Lightspeed Venture Partners—a name more familiar in enterprise SaaS than frontier AI—was quietly assembling a $600 million secondary fund. Not for a new token. Not for a DeFi protocol. For OpenAI and Anthropic shares. The sprint to the ETF finish line in 2024 was a warm-up. Now, the race is for equity in the two companies that define the AI narrative. And the starting gun just fired.
Context: Why Now?
The IPO window for high-burn, high-valuation AI companies remains welded shut. OpenAI and Anthropic are burning cash at rates that would terrify public market investors, yet their private valuations have soared past $80 billion and $20 billion respectively. Early employees and angel investors are sitting on paper wealth they can't touch. Lightspeed's "Project Mercury" fund steps into that gap: a GP-led continuation vehicle that buys secondary shares from those desperate for liquidity, while offering new LPs—think sovereign wealth funds, family offices—a rare chance to buy into the hottest names in tech. The fund combines Select V, Opportunity II, and a separate managed account (SMA) for a single whale client. This isn't just a fund; it's a liquidity bridge between two eras of AI investment.
Core: The Data Behind the Hype
Let's strip away the narrative. Lightspeed is paying a premium—likely 20-50% above the latest primary round valuations—for shares of two companies with very different revenue curves. OpenAI's annualized revenue hit $30-35 billion by mid-2024, driven by a three-layer funnel: ChatGPT subscriptions (consumer), API credits (developer infrastructure), and enterprise deals (high-ticket B2B). Anthropic, by contrast, sits at a few billion in annualized revenue but grows at 50%+ quarter-over-quarter, targeting regulated industries with its "safety-first" Claude model. Lightspeed's internal due diligence must have concluded that both companies' model iteration cycles—6-12 months for a meaningful capability leap—can sustain these growth rates for at least 3-5 years. That's a bet on the steepness of the AI learning curve, not just on market share.
But here's the hidden detail the Bloomberg article missed: this secondary fund does not inject a single dollar into OpenAI or Anthropic's balance sheets. The $600 million goes to sellers—early employees, former investors like Khosla Ventures, and strategic partners like Microsoft. That means the fund is essentially a liquidity event masquerading as an investment. It's a signal that the AI primary market has become so crowded that the only way to get meaningful exposure is to buy from those who already hold. This is the financialization of AI equity, and it's happening faster than anyone anticipated.
Contrarian: The Unreported Angle
Everyone is celebrating Lightspeed's "smart bet." But I see a darker pattern. This fund is a bet that the AI IPO window will remain closed for years. If that's true, the secondary market becomes a liquidity trap: early investors cash out, but new money comes in at inflated prices with no exit path. Worse, Lightspeed is simultaneously betting on OpenAI (capability-first) and Anthropic (safety-first). That's a hedge, not a conviction. It's like buying both Bitcoin and Ethereum in 2021—you're not picking a winner, you're buying the index. But indices don't produce alpha; they track beta. And in a market where open-source models like Llama 3 and DeepSeek are closing the gap, that index could lose value fast.
There's also the regulatory sword hanging over both companies. The EU AI Act, U.S. Executive Order 14110, and ongoing copyright lawsuits (New York Times vs. OpenAI) could dramatically increase compliance costs. Lightspeed's fund doesn't discount this risk. In fact, the dual bet suggests they believe regulation will harm both companies equally, or that the "safety-first" narrative of Anthropic will actually command a premium in a regulated world. But history shows that regulation often favors incumbents with deep pockets—so both could survive. Yet the market's willingness to pay a premium for secondary shares implies that investors have already priced in a benign regulatory outcome. That's a fragile assumption.
Takeaway: The Next Watch
This fund is a test case. If Lightspeed successfully closes and deploys $600 million, expect a flood of copycat funds from a16z, Sequoia, and Tiger Global. The AI secondary market will evolve from a niche to a core asset class, complete with its own pricing benchmarks and liquidity providers. But the real question is one of timing: how long can the private market sustain these valuations without a public exit? The race isn't just about owning AI equity; it's about getting out before the music stops. And right now, the rhythm is accelerating.