The AI Founders Simple Guide to Understanding Term Sheets

The AI Founders Simple Guide to Understanding Term Sheets - Deconstructing the Financial Terms: Valuation, Liquidation Preferences, and Participation

Look, when you're an AI founder, the number everyone talks about is the valuation, right? But honestly, focusing just on that big number is a massive mistake because the real risk lurks in the fine print: the financial terms that dictate who actually gets paid when the music stops. Think about it: for pre-revenue generative AI firms, valuation hinged heavily on the defensibility of your proprietary data moat. If you couldn't prove sub-$0.01 per-query cost efficiency compared to the leading foundation models, top-tier funds slapped a 15–20% discount on you, easy. This is where the liquidation preference comes in, acting as the investor’s insurance policy against a moderate exit. We saw over 35% of deep-tech Series B and C rounds locking in a 2x or higher non-participating preference, reflecting investor demand for heightened capital return protection in volatile sectors. True "full participation" rights, where they get their preference *and* a second bite at the common stock proceeds, appeared in less than 7% of seed sheets in Q3, signaling a high perceived risk profile for the underlying technology. And even when participation is included, you must find the cap, typically 3x or 4x the investment, which limits their upside and converts the security back to non-participating status unless the exit is a true blockbuster. Crucially, the legal structure of the liquidation waterfall dictates that accrued dividends often take priority over the principal return of junior preferred series. That means an early investor's smaller preference amount could be entirely wiped out by a later investor’s accrued dividend claim in a sub-optimal exit scenario—it’s brutal. Plus, when convertible notes or SAFEs use a valuation cap, the effective liquidation preference on those instruments can be artificially higher than the subsequent equity round's stated preference. Understanding these linkages—like how a high preference multiple weakly correlates with less flexibility on founder acceleration terms—isn't just theoretical; it’s about finally sleeping through the night knowing the math works for everyone.

The AI Founders Simple Guide to Understanding Term Sheets - Navigating Governance: Board Seats, Protective Provisions, and Investor Control Rights

We just talked about the money, but let’s pause for a moment and reflect on the control—because governance is really the subtle, slow burn that keeps founders up at night, and here’s where the power shifts. You know that traditional 2:2:1 board split? That’s changing fast; by Q3, 42% of Series B sheets for generative AI mandated five seats right out the gate, demanding two independent directors immediately to handle the regulatory risk exposure, shifting that old balance significantly. And honestly, protective provisions aren't just boilerplate anymore; look at the employee option pool veto threshold, which dropped hard from 18% down to 15% in most Series A deals, giving investors a much tighter leash on future dilution. That’s a Super Majority Protective Provision doing some heavy lifting. Maybe it's just me, but I thought dual-class stock was dead in private markets, yet nearly 18% of those high-growth B rounds actually included specialized Class B shares with 10:1 voting rights solely for board composition, letting founders maintain operational control even as their ownership shrinks. You also need to watch out for mandatory stock redemption rights, which surged in Series C this year, allowing preferred shareholders to demand a buyback at cost plus five percent compounded return if a liquidity event doesn't happen within seven years. Brutal. But we’re seeing some silver linings, like the "Founder Proportionality" carve-out now included in modern drag-along clauses—a critical detail ensuring founders get the same average consideration per share as the big investors, so you can't be forced into an economically inferior transaction. Here’s what I mean by AI-specific controls: a new beast called the "Operational Data Audit Provision" popped up in over 30% of $50 million-plus rounds, mandating annual third-party technical audits just to verify your training data’s defensibility. Think about that level of scrutiny. And finally, the decision to approve founder termination "for cause" is now almost entirely removed from a standard board majority vote; instead, it often requires the specific affirmative vote of that investor-designated director. That completely neutralizes the protection you thought you had in your change-of-control acceleration clauses. We need to check every single one of these governance terms—they are the hinges on the cage.

The AI Founders Simple Guide to Understanding Term Sheets - Securing Founder Equity: Understanding Vesting Schedules and Repurchase Rights

We’ve talked about the money the investors get, but let's pivot and talk about the money *you* keep, because founder equity protection often gets treated like an afterthought until it’s too late, and the risk of procedural failure is surprisingly high. Honestly, look at the data: the failure rate among first-time founders to correctly file that crucial 83(b) election within the 30-day statutory window reached nearly 14% last year—a purely procedural error that can cost you millions later. But vesting isn't just about time anymore; in highly capitalized AI infrastructure deals exceeding $50 million, we’re seeing up to 20% of the founder's common stock increasingly subject to milestone-based performance vesting. I mean, you’re not just clocking hours; you might actually need to hit deliverables like achieving sub-50ms latency on core model inference or securing necessary regulatory clearances just to vest those shares. Now, everyone assumes double-trigger acceleration is the standard protection, right? Yes, but about 45% of sophisticated Series A term sheets now include a "Material Negative Change" carve-out. Think about it: this provision lets investors nullify your double-trigger protection if your division sees a significant revenue drop post-acquisition, essentially taking away your safety net just when you need it most. And that classic, founder-friendly full single-trigger acceleration? It’s dropped below 15% in late Series B sheets, largely replaced by a highly negotiated "Partial Single Trigger" which only accelerates 50% of your unvested equity upon an acquisition—not the whole damn thing. Even on the flip side, when the company exercises its standard repurchase rights for unvested shares, sophisticated seed funds are successfully pushing for a repurchase floor. This floor is often set at 80% of the last preferred price per share, even for unvested equity, which is a big financial hit if you're involuntarily terminated early. For founders dealing with complex frontier AI and proprietary data moats, look closely at the cliff; some term sheets mandate a six-month extension of the standard one-year cliff if you fail to confirm international data rights assignments by that initial date. Finally, don’t forget the collateral damage: many modern AI companies now embed non-compete clauses tied directly to your remaining vesting schedule, forcing adherence for up to 12 months post-termination just to retain the right to exercise your already vested shares.

The AI Founders Simple Guide to Understanding Term Sheets - Spotting the Red Flags: Unpacking Common Anti-Founder Provisions (e.g., Pay-to-Play and Drag-Along Rights)

We just covered the money investors get, but honestly, the nastiest stuff—the provisions designed to put maximum pressure on you—often hides in the sections covering participation and control. Take Pay-to-Play: you know the standard clause forces you to convert to non-participating preferred if you skip a pro-rata investment, but data from Q3 shows 68% of Series B sheets are far harsher, demanding conversion straight into *junior common stock*. That’s a brutal downgrade, completely stripping the non-participating investor of almost all voting power and priority in a small exit. But look for the "Good Faith Exemption"—it’s a small trend, but it can protect your shares from converting if you can prove your inability to participate was due to verifiable personal financial hardship or some unexpected regulatory snag. Then you hit the Drag-Along rights, and we’re seeing them morph into something far more aggressive than just compelling a whole-company sale. In 22% of Series C deals, these rights explicitly include the power to compel the sale of specific, non-core Intellectual Property assets, letting investors de-risk their thesis by carving out parts of the business that aren’t performing. And speaking of pressure, I’m seeing full ratchet anti-dilution sneak back into 11% of high-capitalization AI rounds, but with a new trigger: failure to meet a defined "Technical Performance Baseline" on a key benchmark task within eighteen months. You also need to watch the new control mechanisms specific to AI, like the "Model Licensing Veto," which requires specific investor consent before you can change the core licensing model—say, switching your foundation model from proprietary to open-source. Also, don’t assume your vested equity is fully liquid; over 40% of Q4 term sheets grant investors an expanded Right of First Refusal, specifically blocking secondary sales to named competitive entities regardless of the price. And here’s the kicker: 15% of late-stage sheets now include "Founder Triggered Redemption Rights," allowing investors to demand back 25% of their original capital if all key named founders voluntarily walk away within the first three years post-investment. That’s the kind of contractual leverage that completely changes the risk calculus for you, the person building the thing.