The companies that don't make the next round are not edge cases. They are the substrate on which the next generation of value gets built — provided someone runs the close-out with the same discipline as the original raise. This is what we believe, why we built Graveyard.vc, and what AI-first wind-downs change about the math.
The venture industry is configured for the upside. Capital deployment, board governance, recruiting, term sheets, partner economics — every piece of infrastructure assumes that companies either work or get acquired. The wind-down — the structural reality of what happens to the 70%+ of venture-backed companies that do neither — is a footnote in the operating model.
That asymmetry is expensive. It shows up as capital that doesn't get returned to LPs because the wind-down compresses the residual estate; as founders who can't move on because the wind-down keeps them attached to a dead company for eighteen months; as creditors who lose their priority position because the wind-down practitioner extracts value before they ever see proceeds; as IP portfolios that get assigned to nothing because no one had the operating depth to monetize them.
The cost of running a bad wind-down is rarely visible at the time, and never visible to the people writing the engagement letter. It is visible later, in the recovery memo that gets distributed to creditors three years after the fact, by which point everyone involved has moved on to the next thing.
This is the gap Graveyard.vc was built to close.
The wind-down is the only point in the venture lifecycle where capital actually gets returned to the people who put it in. Treating it as a footnote is treating the return as a footnote.
The dominant ABC practice in Silicon Valley was built in the mid-1990s. The market that practice was built for has not just changed — it has been replaced. The 2026 wind-down is not the 2022 wind-down, and the 2022 wind-down is not the 1995 wind-down. The fastest, deepest break-point is the productionization of generative AI in 2023. Every assumption about what a venture-backed company is — what its assets are, what its moat is, what an acquirer will actually pay for — got rewritten over an eighteen-month window.
Three structural shifts define the post-AI wind-down:
In the pre-AI era, the codebase was the durable asset of a venture-backed software company. Engineers spent years building it; competitors couldn't easily replicate it; an acquirer paid for the team-built system itself. Post-2023, AI generates production-grade code at a pace that compresses years into days. Codebases retain some value — accumulated context, business logic, integrations — but they are no longer the moat.
What remains durable across a wind-down is what AI cannot generate:
Our practice is built around exactly these three cornerstones. Trademarkia runs the brand-protection registry that makes the trademark portfolio liquid across jurisdictions. The acqui-hire workstream runs the team-as-asset transactions where the people are what's being acquired. PatentVC runs the federal-court enforcement and IP-monetization workstream where the patent stack is the highest-recovery asset on the balance sheet. We focus on these three because they are what's left when the code goes commodity.
For brand, people, and patent assets, the highest-recovery buyer is rarely a local PE firm or a regional competitor. It is a Fortune 500 corporate-development team that already operates at global scale, has the integration capacity to absorb the asset cleanly, and pays a strategic premium that financial buyers cannot justify. The wind-down practice that wins in 2026 is the one with direct relationships to those teams — not the one with a thirty-year rolodex of regional contacts.
This is where Graveyard.vc's parent firm differentiates structurally. Across fifteen years of operating Trademarkia, PatentVC, and LegalForce RAPC, our team has worked directly with the multinationals that buy these assets:
Patent prosecution, licensing, and brand-protection mandates across enterprise infrastructure, semiconductors, networking, and consumer hardware.
IP transactions, federal-court enforcement, and corporate-development work across search, social, storage controllers, and accelerated computing.
The relationships are not a marketing claim. They are operational, current, and live. We have closed exits to several of these acquirers across patent transactions, trademark transactions, and acqui-hire deals. The buyer-development infrastructure that produced those exits is the same infrastructure that runs every Graveyard.vc wind-down today — and it sits inside a firm whose Trademarkia network has been tracking corporate-development patterns at 125,000+ companies across 80+ countries since 2009. The Buyer Graph is computed against that data, and the relationships are the channels through which the auctions actually close.
A rolodex of regional contacts cannot find a Fortune 500 buyer for an AI-tooling company's patent portfolio. A graph trained on fifteen years of operating data — combined with relationships at the corporate-development teams that actually buy — can. This is the structural advantage.
The 2010s funded growth on narrative. The 2020s rate cycle priced narrative back to numbers. A meaningful share of the venture-backed companies funded between 2018 and 2022 will not raise the next round, and the wind-down funnel that was a niche practice with a few hundred matters per year nationwide is now a structural feature of the venture market with thousands of distressed companies in pipeline. The convergence of AI commoditizing code and capital tightening the runway is not coincidental. It is the same trend line: a forced re-pricing of where durable value in a venture-backed company actually sits — and a forced acceleration of the timeline on which the answer has to be found.
The legacy ABC practice — the dominant incumbent in Silicon Valley wind-downs — runs an operating model that has not materially changed since 1995. It is built around three load-bearing assumptions: that buyer outreach is a manual rolodex, that document drafting is a manual templated process, and that creditor reporting is a quarterly PDF.
Each of those assumptions creates a specific structural failure in 2024:
A rolodex of 30 corporate-development contacts curated over decades produces a 30-buyer auction. The right buyer for a distressed AI-tooling company is rarely in that rolodex. They're in a buyer-graph computed across patent-citation networks, M&A history, and corporate-development pattern data — a database, not a relationship.
Every additional week of operating cost in a wind-down comes out of the creditor estate. A practice that takes six weeks to issue creditor notice, four weeks to set up an auction, and weeks per legal document is extracting from the recovery, not preserving it.
Creditors deserve real-time auditable visibility into distributions. Quarterly PDF statements are an artifact of paper-era accounting that creates information asymmetry between the assignee and the parties in interest — and that asymmetry creates room for the assignee to extract more value than they would if the books were live and visible.
The legacy engagement letter contains a $35K up-front fee earned at signing, a 12% transaction bonus, a 24-month tail, one-way indemnification, and a "may express full opinions" clause. None of these terms are necessary to do the work. They exist because no one has ever competed with the incumbent on legal terms.
The result, on a typical $5M asset sale, is roughly $635,000 in fees extracted from the estate on terms that would not survive scrutiny in any other professional service category. Lower-ed costs are not a marketing claim — they are a structural consequence of the legacy operating model being older than the internet.
"AI-first" is overused as a marketing term. It is also load-bearing as a structural claim about how the wind-down practice should be built. We use it to mean a specific thing: that the workstreams in a wind-down are run by data infrastructure rather than by individual practitioners, and that the practitioners are deployed at the human-judgment layer rather than the document-production layer.
Concretely, that's six systems running on every Graveyard.vc engagement:
None of these are speculative. All six are running. Each one materially compresses the operating cost of the engagement, which translates directly into more recovery flowing to creditors and equityholders rather than to the assignee.
The Buyer Graph is the most differentiated infrastructure in the practice. It is the answer to a single question: who is most likely to close on these specific assets, in this specific time window, at the highest recovery?
It works because Graveyard.vc operates inside the same firm that runs Trademarkia — the largest brand-protection registry in the United States. Trademarkia has been tracking, since 2009, the brand-protection filings, IP transfers, M&A activity, and corporate-development patterns of 125,000+ companies across 80+ countries. That data, queried against the asset profile of a distressed company, produces a ranked list of likely acquirers — not 500 we haven't talked to since 2018, but the 30 buyers who have actually closed on similar assets, in similar sectors, at similar valuations, in the last 24 months.
This is not a CRM. It is a buyer-matching engine. It surfaces:
The recovery math turns on this. A rolodex auction with 30 buyers produces a single bid above floor. A Buyer-Graph auction with the same 30 buyers — identified differently — produces three competitive bids. Three competitive bids vs one is the difference between 60% and 95% of fair-market recovery. The auction structure is identical. The buyer identification is the entire game.
The single most consequential failure mode in venture wind-downs is delay before the wind-down conversation actually happens. Founders avoid the topic for months because the conversation feels terminal. Boards avoid raising it because the politics are difficult. By the time the wind-down conversation gets onto the agenda, the runway has compressed to weeks, the team has started to leave on its own, and the residual estate has shrunk to almost nothing.
The Triage AI memo addresses this. It is a one-business-day diagnostic, generated from a confidential intake form, that models the company's specific facts — cap table, debt stack, IP inventory, headcount, runway, asset profile — against the available paths forward. It produces:
The memo does not commit the company to anything. It does not require an NDA to look at the analysis. It does not require board approval to receive. It is decision-support infrastructure — the document that lets a CFO or a lead investor have an informed conversation with the board before a vote, rather than after.
Most wind-downs run badly because the conversation started two months too late. Compress the time-to-decision, and you compress the time-to-close, and you compress the cost. The memo exists to compress the first leg.
For a venture-backed company that has run out of capital, the asset that most often survives — and that most often funds the highest distribution to creditors — is the IP portfolio. Patents, trademarks, copyrights, trade secrets, and domain assets routinely outlive the operating business by years. They are also the assets the legacy practice handles least well.
This is partly a structural disadvantage of being a stand-alone wind-down firm. When the legacy practice needs to file a Patent Assignment with the USPTO, it goes to outside counsel. When it needs to enforce a patent license to extract value, it goes to outside litigators. When it needs an international trademark recordal, it goes to a network of foreign agents. Each handoff adds cost, time, and information loss.
Graveyard.vc operates inside a three-firm IP and litigation platform: Trademarkia (the brand-protection registry), PatentVC (federal-court patent litigation), and LegalForce RAPC Worldwide (the operating law firm). Every IP-related workstream in a Graveyard.vc engagement runs in-house. Patent recordals, trademark assignments, license enforcement, international filings, and federal-court enforcement leverage are all available without subcontracting and at no incremental engagement cost.
The numerical impact is meaningful. Patents authored or invented by our founder, Raj Abhyanker, have been licensed or sold for over $1 billion over the last fifteen years. That is not a marketing number. It is the kind of operating record that lets the IP workstream actually produce a competitive auction for the patent portfolio rather than a single-line residual entry on a creditor distribution sheet.
The single most important cultural difference between Graveyard.vc and the legacy practice is the disposition toward transparency. We publish the engagement letter on the open web. The pricing tiers are listed before the first conversation. The team-page bios disclose individual practitioners' credentials and prior firms. The thesis page — this page — names the structural problems with the legacy operating model in print.
This posture is not virtue-signalling. It is competitive positioning. The legacy practice cannot publish its engagement letter on the open web because the terms in that letter would be uncompetitive in the light. The 12% bonus, the 24-month tail, the one-way indemnification, the "may express full opinions" clause — these are terms that survive only in conditions where the prospective client cannot see what they are signing until after the conversation is committed.
Transparency, in this context, is a strategic moat. Founders, boards, and lead investors who can read the comparison before the first call are inoculated against the hard sell. They walk into the conversation having already done the math, and the math favors the modern practice. The legacy industry standard cannot match transparency without rewriting the engagement letter, which would compress its own economics.
There is a comfortable assumption in the venture community that hiring the incumbent is the safe, conservative, fiduciary-defensible choice. The 2026 reality is closer to the opposite. The fiduciary case for hiring the incumbent in 2026 is hard to make on the math:
An engagement structure that pays the assignee less, aligns the assignee with creditor recovery, runs on transparent published terms, and uses a mutual-indemnification clause is the more defensible choice on the fiduciary record. Not the riskier one. A board memo recording the comparison is straightforward to draft. We have drafted it for boards before.
The structural opportunity for an AI-first wind-down practice is the convergence of three independent trend lines:
The 2010s funded growth on narrative. The 2020s rate cycle priced narrative back to numbers. A meaningful share of the venture-backed companies funded between 2018 and 2022 will not raise the next round. The wind-down funnel is now a structural feature of the market, not a niche.
The companies winding down in 2024+ are IP-heavy in a way the 1995 cohort was not. AI-tooling companies, SaaS companies, fintech companies, consumer-brand companies — all of them have asset bases dominated by trademarks, patents, customer-data assets, and brand equity rather than physical inventory and AR. The right wind-down practitioner for an IP-heavy estate is fundamentally different from the right wind-down practitioner for a hardware estate.
The buyer-matching engines, the document-production pipelines, the live-ledger systems, and the forensic-AI workflows that AI-first wind-downs run on were not deployable at operating-grade reliability in 2018. They are deployable in 2024+. The structural advantage of the modern practice over the legacy practice is not theoretical — it is measurable in days-to-close, in dollars-of-recovery, and in legal-terms transparency.
The wind-down market does not move on first-mover advantage. It moves on credibility, infrastructure, and proof of recovery. Graveyard.vc was built to be the practice that combines all three.
Some companies don't make the next round. We help them land. That is the entire thesis.
The Confidential Intake routes to the small team that wrote this thesis. One business day to a Triage AI memo. No engagement, no NDA required to receive the analysis.