Are We Rewriting the WeWork Story with OpenAI?

How the world’s most influential AI company is repeating the same structural mistakes that brought down the last great unicorn.

In 2019, WeWork was the future.

It promised to reinvent the way people lived and worked, replacing cubicles and leases with collaboration and community. Its co-founder Adam Neumann spoke in the language of spirituality and belonging. Investors saw a scalable model that blended technology and real estate. The company’s valuation reached forty-seven billion dollars. Within a year, it had collapsed.

The downfall of WeWork was not a failure of demand. It was a failure of structure, discipline, and truth. A business built on long-term leases and short-term tenants finally met the arithmetic of cash flow.

Today, OpenAI occupies a similar place in the collective imagination a company so central to the narrative of progress that scrutiny feels almost disloyal. Its systems power a global industry, shape government policy, and command a valuation above half a trillion dollars. Yet its structure, funding, and dependencies echo many of the same patterns that doomed WeWork.

The parallel is not about product. It is about power, ownership, and the illusion of inevitability.

---

1. Leased Foundations

WeWork’s model was simple. It signed long-term leases with landlords, spent heavily to convert those spaces into sleek offices, and then sublet them on flexible monthly terms. The spread between those two horizons was its profit engine until it wasn’t. When growth slowed, it was locked into billions in obligations that outlived its tenants [1].

OpenAI’s foundation is equally dependent on assets it does not own.

The company rents vast compute capacity from Microsoft, Amazon Web Services, CoreWeave and Oracle under multi-year contracts that already exceed three hundred billion dollars in aggregate commitments [2]. It controls access to this infrastructure but not the infrastructure itself.

This model turns control into dependency. WeWork’s leases became its liability when demand fell. OpenAI’s compute contracts could do the same if training costs rise faster than revenue or if external partners decide to change terms.

Both companies presented their dependency as strategic vision. WeWork called its leases a technology platform. OpenAI calls its infrastructure contracts a route to national capability. The underlying reality is the same: leverage without ownership.

---

2. Valuation as Belief

WeWork’s revenue in 2018 was around $4.4 billion, yet its private valuation reached more than ten times that amount [3]. The figure was justified through unconventional metrics such as “community-adjusted EBITDA”, which excluded most of the company’s expenses.

OpenAI’s reported revenue run-rate sits between $10 billion and $12 billion [4]. Microsoft’s 2025 filings record its 27 percent economic exposure to OpenAI at a value of $135 billion, implying a total enterprise valuation of roughly $500 billion [5]. Analysts estimate that the company remains loss-making at scale, with Microsoft’s third-quarter report attributing $11.3 billion in quarterly losses to OpenAI-related lines [6].

Valuation, in both cases, functions as narrative. It is the market’s expression of faith in what a company might one day control rather than what it currently earns.

WeWork sold the idea of infinite demand for collaboration.

OpenAI sells the idea of infinite demand for intelligence.

The danger is not fraud but asymmetry: investors price the future, but the liabilities exist in the present.

---

3. The Myth of Control

WeWork collapsed under a governance model that blurred the line between founder and company. Adam Neumann held super-voting shares that gave him twenty times the voting power of any other shareholder [7]. The board existed largely to ratify his decisions.