$5.6T: software value with limited public access

Microsoft listed in 1986 at roughly $500 million, and an individual investor could buy its shares. Twenty years later, a share was worth more than fifty times its IPO price. Salesforce, founded in 1999, listed in 2004 at $1.1 billion. Workday, ServiceNow, Shopify, Atlassian, HubSpot, Okta, Twilio, Veeva, Dropbox, Box, and Zendesk also spent most of their growth in public markets. Those twelve SaaS companies are now collectively worth roughly $794 billion, with that growth accessible through ordinary brokerage accounts.

The current AI cohort has followed a different path.

OpenAI, Anthropic, and xAI (now part of the merged X/SpaceX entity) have a combined value of roughly $2.1 trillion. Investors using ordinary 401(k), brokerage, or software ETF accounts cannot buy those shares directly. Access to private rounds is concentrated among repeat venture investors and sovereign wealth funds capable of ten-billion-dollar commitments; many pension funds and advisers also lack access. The largest data platform used by much of the AI cohort and the payments platform moving a trillion dollars a year remain private. In 2026, the companies setting software's productivity benchmark are largely inaccessible to the savers who funded the previous cycle.

Across the 2,398 companies in the register, $5.6 trillion of value sits in VC-private companies: 1,177 companies valued at financing rounds, with little predictable secondary liquidity for employees, early investors, or the limited partners behind their funds.

That differs from prior cycles and raises questions about who can participate.

Accreditation rules protect investors from risks that are difficult to assess. Retail access to OpenAI at a $500 billion mark could expose buyers to a valuation wrong by a factor of two. Companies also argue that quarterly public-market pressure can disrupt R&D whose value develops over longer periods. Disclosure rules designed in 1986 may fit those businesses poorly.

Those concerns leave liquidity and access unresolved.

For two years, software finance reporting has focused on $1.1 trillion of value inside private-equity-held software portfolios, including Veritas, Thoma Bravo, and Vista roll-ups and the private credit supporting their leverage. Our companion PE Software Reckoning report examines that exposure.

The VC-private pool with limited liquidity is nearly five times larger.

Of the register's 2,398 companies, 413 were founded in 2019 or later. Almost all are VC-private. The top ten account for roughly $2.7 trillion, about half the cohort's value concentrated in fewer than a dozen companies.

For fifty years, companies commonly developed privately, reached revenue and product-market-fit thresholds, filed an S-1, and let public markets price their next stage. Listed SaaS pure-plays took a median of about eight years from founding to IPO. None of the three leading AI companies has made that transition. The report records OpenAI at nine years old, with younger companies behind it showing little movement toward listing.

Remaining private is a choice these companies can finance, but it limits liquidity for other stakeholders.

Limited partners face delayed capital returns.

Venture portfolios traditionally return capital through IPOs and acquisitions. LPs in the 2018, 2019, and 2020 vintages expected meaningful exits by 2026. The marquee companies that could justify those vintages have stayed private, leaving high paper marks and few cash distributions.

Large institutional LPs can draw on other liquidity and negotiate secondaries. Smaller endowments, regional foundations, and mid-sized public pension funds have fewer options. Some sell venture fund stakes at discounts of twenty to forty percent to stated NAV to meet commitments when distributions do not arrive.

Employees also face limited liquidity. Few of the 1,177 VC-private companies offer predictable secondary sales; option holders may depend on company-controlled tenders, prices, and windows. A Microsoft engineer in 1996 could sell vested shares on a public trading day. An engineer with five years vested at a leading AI lab in 2026 has less flexibility. That affects the competitiveness of equity compensation.

Public investors face restricted access to these companies' growth.

Earlier mainframe, client-server, on-prem, and SaaS waves produced public companies including Microsoft, Oracle, SAP, Salesforce, Workday, ServiceNow, Adobe, Shopify, Atlassian, Snowflake, Datadog, and MongoDB. The report's examples trace $10,000 invested at IPO and held: roughly $55 million for Microsoft after nine splits and forty years, $260,000 for Salesforce, and $13,000 for Snowflake. Returns differed greatly, but participation was available.

Investors cannot make an equivalent direct public-market investment in the leading AI labs.

Twelve SaaS pure-plays reached $794 billion of value over twenty years. The three AI labs reached roughly $2.1 trillion in under a decade, with revenue trajectories suggesting further growth. Their financing marks are set among a small group of accredited and institutional investors.

An investor who spent the past decade buying public software ETFs might have seen SaaS reach twenty percent of large-cap exposure and expected to participate similarly in AI. The private AI companies competing with those public holdings remain unavailable through the same accounts.

Valuation and access are separate questions. If the cohort lists through S-1s in 2027 and 2028, retail investors may enter after private investors captured much of its growth. If it stays private indefinitely, direct retail participation may remain unavailable.

Public markets require disclosures such as quarterly 10-Qs, giving operators and investors information on revenue growth, retention, customer concentration, margins, and free cash flow. Private AI leaders do not provide equivalent reporting, leaving many valuable enterprise tools difficult to assess.

Deployment records provide evidence of what organizations install, at what scale, and with what outcomes. Airframe tracks 24,000+ software products across 232,000+ organizations and 174,000+ deployment case studies. This shows adoption of AI-native vendors relative to public incumbents. It cannot establish OpenAI's net retention, but it can show which public vendors are losing net-new projects.

AI labs have raised substantial capital at their chosen valuations. A public-market route would let individual investors participate in the $5.6 trillion cohort. Earlier mainframe, client-server, SaaS, and cloud waves produced public companies roughly three to five years after their categories became standard procurement choices. The report places the current cohort two years past that threshold, without an equivalent public venue.