Over four trillion dollars of enterprise value is sitting in the private stack. The exits are not coming on the old schedule.

For two years the anxious conversation in enterprise software finance has been about the $574 billion of value inside private-equity-held software portfolios. The number that matters is more than seven times larger. Across the 1,425 companies in the register, $4.17 trillion of enterprise value sits in the venture-capital-private stack, held inside 785 companies that offer almost no secondary liquidity. The three largest AI labs, the largest data platform, and the largest payments platform are all in it, and none are public. This is not a bubble question. It is a disclosure question, and nothing about the current equilibrium is stable.

For the last two years, the anxious conversation in enterprise software finance has been about the $500 billion of value sitting inside private-equity-held software portfolios: the Veritas, Thoma Bravo, and Vista roll-ups, together with the private credit exposure underwriting the leverage on top of them. That conversation has now largely resolved. The PE stack is heavy, but it is not a systemic risk; it is levered, it is marked conservatively, and its assets generate real cash. The system can absorb it.

The conversation that has not happened is about the number that is more than seven times larger. Across the 1,425 companies in the register, roughly $4.17 trillion of enterprise value is trapped in the venture-capital-private stack: 785 companies whose valuations are set only at the moments someone chooses to lead a round, almost none of which offer meaningful secondary liquidity to the employees and early investors who hold their paper. It is an odd arrangement, and it has become a durable one. Many of the best software companies in the world are, in 2026, simply not listed.

When a private-market tender for OpenAI shares cleared at a valuation north of half a trillion dollars in late 2025, the coverage treated the headline figure as the story. It was not. The story is that a company of that size, with a consumer product rivaled in scale only by a handful of platforms and a revenue trajectory that is now measured in tens of billions of dollars a year, is not on a public exchange. Neither is the second-largest AI lab. Neither is the third. Neither is the company that makes the data platform most enterprises run their AI workloads on. Neither is the payments platform that moves a trillion dollars a year.

This is an arrangement that would have been unthinkable in any previous cycle in enterprise software. In 1986, Microsoft went public at a valuation of roughly $500 million. In 1999, Salesforce was founded. It went public five years later at $1.1 billion. The pattern for fifty years was: build the company in private, cross some threshold in revenue and product-market fit, file an S-1, let the public price-discover the rest. That pattern has broken, and the record shows where the cost of it has gone.

The mainframe era ended in the public markets. IBM, DEC, Wang, Data General, Burroughs, and Sperry all listed, competed, and consolidated in view of a public tape that marked them down in real time when the next wave arrived. The client-server era ended the same way: Oracle, Sybase, Informix, Novell, Lotus, and BEA were public, price-discovered, and visible when the air went out. The SaaS era is ending the same way too, more slowly, with Salesforce, Workday, and ServiceNow still trading while Adobe and Microsoft quietly absorb the layer beneath them.

What is different about the AI-native cycle is not the valuations; valuations run hot in every cycle and always have. What is different is that the companies setting the pace of this transition have chosen not to be on the tape, and for the first time in the history of enterprise software they have had the means to make that choice stick. Secondary markets, tender offers, and strategic investors writing ten-billion-dollar checks into private vehicles have given them the cash to stay off it. The result is a trillion-dollar class of assets whose true value the market has never been given the chance to set.

Three things break when a cohort this large stays private, and they are all downstream of the same mechanism. The first is capital access. A public company of meaningful size can raise equity on days of ordinary liquidity at a modestly discounted clearing price. A private company of equivalent size cannot. The substitutes it reaches for instead, convertible debt and structured preferred and the various flavors of synthetic secondary, all carry terms that compound the exposure for every earlier round in the stack: senior preferences, ratchets, and IPO thresholds that, once tripped, force the next round's valuation floor down through the rounds that came before.

The second is employee equity, and this is the quieter crisis inside the $4.17 trillion. Almost none of the 785 VC-private companies in the register offer meaningful secondary liquidity. A decade of option holders at the largest of them are sitting on paper gains that they can monetize only through company-sponsored tenders, at company-controlled prices, within company-controlled windows, and often not at all. That is a less liquid form of compensation than what the SaaS era offered, and the people doing the work have begun to price that difference into where they will work next.

The third is market intelligence. The public tape is not just a trading venue; it is a disclosure regime. A quarterly 10-Q on the largest software platform in the world tells every other operator in the market what unit economics look like at scale, what the mix of deployment footprints is, what the net retention is, what the customer concentration is. None of that exists for the AI-native leaders. Enterprises procuring them are procuring a category whose underlying economics have never been disclosed, from companies whose valuations have never been tested against a liquid market. The most valuable tools in the enterprise stack are, in 2026, the least observable ones.

Part 02 showed that the succession from one cohort to the next is running faster than the incumbents can adjust. The class taking their place has not yet submitted to the disclosure regime the incumbents did, and whether it eventually does is the single largest unresolved question in the enterprise software market in 2026. Either a wave of S-1s or a structural shift into permanent private-market listings will eventually reorganize where trillions of dollars of value sit. Nothing about the current equilibrium is stable.

The debate in the trade press has settled on "is this a bubble." It is the wrong debate. Secondary platforms like Forge, Hiive, EquityZen, and the various bank-run tender programs do print real marks on a handful of the largest private names, and those marks are not nothing. But the volumes are thin, the holders who transact are constrained, and the price is set without the disclosure regime that makes a public mark a disciplined one. The interesting question is not whether the four-trillion-dollar private stack is priced correctly. It is whether the companies inside it will ever submit to the disclosure that turns a mark into a price. A bubble is a pricing error that eventually corrects. What this cycle has produced instead is a pricing absence, and you cannot have a correction in a place where there has never been a fully observable mark.

The piece of infrastructure most conspicuously missing from the current cycle, then, is not capital. It is a venue. Every prior wave in the register, from mainframe to client-server to SaaS to cloud, produced its public-market cohort within roughly three to five years of the new category becoming the default answer in enterprise procurement. The current cohort passed that threshold two years ago. The venue has not followed.