Inside Paradox's massive — and wild — IPO prospectus

An IPO prospectus is a detailed filing prepared before a company sells shares to the public. It gives potential investors a common factual basis for judging the business. It usually covers finances, customers, contracts, management, and risks. That information matters because buying shares involves accepting possible losses.

Paradox’s filing describes its revenue, cash, customer concentration, computing obligations, and unusual governance arrangements. It also says the company has $637 billion in computing commitments, including contracts that cannot be canceled easily. Those obligations may require payment even if Paradox uses less capacity. The filing also identifies policy violations, cloud dependence, and uncertain customer loyalty.

This disclosure lets investors compare Paradox’s rapid growth with its liabilities. The company has nearly 2,000 customers generating at least $1 million, but its five largest customers produced about 25% of revenue during the first nine months of 2026. The IPO will test whether investors see that balance as attractive or dangerous.

Paradox is already operating at unusual scale, although the filing details available here do not give total revenue or total employee count. It reports nearly 2,000 customers with at least $1 million of revenue and 250 with at least $10 million. Half of the Fortune 500 pays it $100,000 a year. These figures show broad enterprise reach.

The company’s financial position is also striking. It had more than $130 billion in cash and short-term investments as of June 30, and its filing lists more than $134 billion in cash and similar securities. Revenue per employee reached $6.6 million at the end of the third quarter, slightly below Nvidia’s figure. The article does not provide headcount or total revenue.

Its largest obligation is computing. Paradox reports $637 billion in computing commitments. A SpaceX arrangement could reach $120.5 billion through 2029. These commitments create significant future capacity, payment, and execution risk even alongside the company’s unusually large cash balance.

Paradox makes money by providing access to its AI models. Customers pay for usage, directly or through cloud marketplaces. The filing says nearly 50% of first-half revenue came through marketplaces operated by Amazon, Google, and Microsoft. Half of the Fortune 500 pays Paradox $100,000 a year, while larger customers generate millions.

The business can produce high revenue per employee because one model can serve many customers at once. Employees build, improve, sell, and support the system, while software delivers the same underlying capability repeatedly. The article reports $6.6 million in revenue per employee at the end of the third quarter, only slightly below Nvidia’s figure.

That measure is impressive but does not equal profit. Every customer interaction can require expensive computing. Paradox also has massive capacity commitments and pays cloud providers that may collect both computing revenue and platform fees. High revenue per employee therefore coexists with potentially heavy infrastructure costs.

Paradox needs computing capacity to train and operate powerful AI models. Securing it years in advance can support rapid growth and reduce the risk of being unable to serve customers. It also reflects management’s expectation that demand for advanced AI will remain strong. The article frames this as a central wager behind the company’s IPO.

The commitment is unusually large: $637 billion in computing obligations. Contracts extend for years and include minimum payments, regardless of how much capacity Paradox actually uses. Its SpaceX arrangement alone could involve $120.5 billion in computing through 2029, with payments potentially reaching $3.1 billion per month. Either side can cancel that deal with 90 days’ notice.

This structure creates operating leverage in both directions. Strong demand could help Paradox use capacity productively and justify premium prices. Weaker usage would leave it paying for unused capacity, pressuring profits and making its valuation harder to defend. Customer growth must keep pace with infrastructure commitments.

Paradox’s economics depend on customers continuing to use its models at high rates and pay prices that cover computing. If usage declines, revenue can fall quickly because many customers pay according to usage. If customers switch to cheaper alternatives, Paradox may need to reduce prices or accept slower growth. Either outcome would challenge the premium its IPO may seek.

The key mechanism is a mismatch between flexible customer demand and rigid supplier obligations. Many customers do not make multiyear commitments, but Paradox’s computing contracts extend for years and include minimum payments. Its five largest customers already represent about 25% of revenue. Losing or shrinking a few major accounts could therefore matter significantly.

Lower revenue would make those fixed commitments more burdensome and could compress operating margins. Investors might also question whether Paradox can sustain 50% growth toward a $2 trillion valuation. The filing’s central test is whether demand remains strong enough, long enough, to support the cost structure.

The three hyperscalers occupy several positions in Paradox’s business at once. They are investors, computing suppliers, and distribution platforms. They also compete with Paradox through their own AI products. This mix is unusual because a company can support Paradox’s growth while simultaneously strengthening its own commercial position.

The mechanism is visible in the revenue flow. Nearly 50% of Paradox’s first-half revenue came through marketplaces run by Amazon, Google, and Microsoft. Paradox can reach enterprise buyers there, while the platforms collect fees. They also earn money by supplying the computing needed to run the models. The article calls hyperscalers huge winners because they can collect revenue in multiple ways.

For Paradox, this arrangement expands distribution and infrastructure access. It also creates dependence on companies that may have different incentives. They can benefit when Paradox succeeds, but cheaper alternatives or rival models could help their own businesses. That makes cloud relationships both a growth engine and a competitive risk.

Traditional software can often serve additional customers at relatively low delivery cost once the product is built. AI models still have software-like scalability, but each training run and customer interaction requires substantial computing. That makes infrastructure a continuing operating cost, not merely a one-time development expense. The difference matters most when usage grows rapidly.

Paradox’s numbers show the tradeoff. It reports $6.6 million in revenue per employee, suggesting efficient human operations. Yet it also carries $637 billion in computing commitments, with minimum payments even if capacity goes unused. Customers often pay by usage and may avoid multiyear contracts. Revenue can rise quickly, but computing obligations may remain fixed for years.

Competition adds another pressure. Paradox must keep models strong enough to command premium prices while cheaper alternatives improve. Cloud providers can distribute its models, supply compute, invest in it, and compete against it. The result is a business with enormous upside, but costs, pricing, and customer loyalty are unusually sensitive to market change.