The race wasn’t won by the fastest. It was won by the one who knew when to stop pretending the old model still worked. Lightspeed’s $600 million secondary fund—targeting OpenAI and Anthropic equity—isn’t a bet on AGI. It’s a bet on the death of the classic VC pipeline.
Context: Why Now
By August 2024, the AI IPO window was a ghost. OpenAI and Anthropic burned billions annually, their valuations hit nine figures, and the public markets weren’t ready. Early employees and angel investors held paper worth millions but no liquidity. Enter the GP-led continuation fund: a structure where a VC firm raises new capital to buy out existing LPs, effectively extending the fund’s life while offering an exit for those who want out. Lightspeed christened this “Project Mercury”—fast, flexible, transactional. The name fits.
But the real story isn’t the $600M. It’s the signal embedded in the capital architecture.
Core: The Mechanics of a Structured Exit
Lightspeed didn’t just raise a single fund. It combined three vehicles: Select V Fund, Opportunity II Fund, and a Separately Managed Account (SMA). The SMA is the tell. A single large institutional LP—likely a sovereign wealth fund or university endowment—demanded a dedicated AI allocation, separate from the common pool. This isn’t venture capital. It’s bespoke asset management disguised as VC.
Based on my own experience reverse-engineering the 0x protocol v2 in 2017, I learned that speed of capital deployment matters more than the underlying asset. Here, Lightspeed deploys institutional-grade speed into a secondary market that has no central exchange, no price oracle, and no settlement layer. The friction is human: the sellers are early employees and strategic investors (Microsoft, Amazon) who want to lock in gains without triggering a public signal. The buyers are LPs who can’t get into the primary rounds. Lightspeed is the middleman, charging management fees and carried interest on both sides.
The hidden leverage is the discount. Secondary trades typically transact at a 20-50% premium to the latest round valuation. That means Lightspeed is paying $600M for equity that was worth $400M a year ago. The implied bet: OpenAI’s revenue—already at $30B annualized—will grow at 40% CAGR for the next three years. Anthropic’s enterprise-heavy model, with its safety-first branding, will command sticky contracts. The math works only if the AI hype cycle extends longer than the patent cliff on current models.
Contrarian: The Real Blind Spot
Sustainability is just a loan from the future. The conventional take is that Lightspeed is securing rare AI equity. The contrarian take: this secondary fund is an admission that the VC model has failed to generate exits through IPOs, and that the only way to return capital to LPs is to sell to each other. This is a circular flow. The same capital that exited through the secondary fund will likely be reinvested into the next VC fund, perpetuating a system where valuations are set by internal LP demand rather than market fundamentals.
More importantly, the structure creates a feedback loop of overconfidence. Lightspeed now has a financial incentive to talk up OpenAI and Anthropic—not because they are great companies, but because the fund’s carry depends on the secondary market price staying high. The fund is a levered bet on the narrative, not the technology. And when the narrative shifts—when an open-source model like Llama 4 or DeepSeek v3 erodes the commercial moat—the secondary price will crater before the primary round adjusts. Liquidity didn’t disappear; it just moved to a new venue where the spreads are wider and the information asymmetry deeper.
Takeaway: What to Watch Next
Trust is a variable, not a constant. The next move will be an arms race: every major VC firm will launch a secondary fund for AI stakes. As these funds pile in, the premium will compress, and the risk of a liquidity event—a forced sale at a discount—will rise. The question isn’t whether OpenAI or Anthropic will succeed. It’s whether the secondary market can sustain a pricing model that depends on everyone believing the same story at the same time. First in, first served, or first to flee. The choice is yours.