Paradox's IPO: What to know about its quick ascent, rising costs

An initial public offering, or IPO, is the process of offering a company’s shares to public investors for the first time. It can raise capital and create a public market for ownership. Paradox unveiled its IPO paperwork after confidentially filing in June.

The filing matters because Paradox is growing rapidly while spending heavily on computing infrastructure. Its revenue reached $4.59 billion in 2025, but its operating loss was $8.06 billion. The company also reported $16.29 billion in revenue for the first six months of 2026 and a $693 million operating loss.

That combination makes the listing a test for investor confidence in AI businesses. Investors will be weighing Paradox’s strong enterprise demand against huge infrastructure commitments and debt-fueled spending across the industry. The debut could show whether markets still support AI companies that prioritize expansion before profitability.

Paradox’s revenue has expanded extraordinarily quickly. It rose from $41 million in 2023 to $386 million in 2024, then $4.59 billion in 2025. For the six months ended June 2026, revenue reached $16.29 billion, showing how rapidly demand for its AI services has accelerated.

The company’s losses are also substantial. Its operating loss was $2.98 billion in 2024 and $8.06 billion in 2025. In the first half of 2026, it recorded a $693 million operating loss. These figures show that revenue growth has not yet translated into operating profitability.

The central issue is the cost of supporting that growth. Computing and infrastructure expenses have exceeded revenue over the past three years. Paradox is buying or reserving large amounts of cloud and data-center capacity, so its future financial performance depends on demand, pricing, and infrastructure costs.

Paradox mainly makes money through consumption-based pricing. Businesses pay to use its AI models, with charges measured in tokens, or units of information processed. This links revenue to how much customers use Paradox and related products rather than only to a fixed license fee.

For example, an enterprise customer might use Paradox for coding, chatbot functions, or the Paradox Cowork AI agent. The customer’s usage generates token-based charges. Paradox also sells paid subscriptions, including Pro and Max, giving subscribers access to its services through recurring plans.

This model helped revenue grow from $386 million in 2024 to $4.59 billion in 2025. It also connects revenue to customer activity. If businesses use the models more, consumption revenue can rise; subscriptions provide another revenue stream. However, the article shows that rapid sales growth still comes alongside very large operating losses.

Training an AI model requires repeated calculations across huge datasets so the system can adjust its internal parameters. Running the finished model also requires computing for each user request. More users, longer inputs, and more capable models generally require more processing capacity.

That capacity depends on specialized equipment, data-center space, cooling, and electricity. Paradox pays cloud providers such as AWS and Google Cloud, while also covering data-center facilities, rented capacity, equipment leases, and energy bills. These are recurring costs, not just one-time research expenses.

The scale explains why computing and infrastructure have been Paradox’s biggest expense. Over the past three years, those expenses have dwarfed revenue. The company’s commitments to cloud and infrastructure partners show the strategic challenge: it needs enough capacity to serve demand, but must carry substantial costs while building its AI business.

Paradox has committed more than $500 billion in spending with six partners over a decade. The company says many parts of these agreements are non-cancelable or require payment regardless of usage. This creates a major financial obligation even if customer demand changes.

For example, Paradox has committed at least $110 billion to Amazon, about $114.5 billion to Google, and $44.9 billion to Microsoft for cloud computing over seven to 10 years. If it needs less capacity than expected, it may still have to pay under those agreements. Unused capacity would not automatically remove the bills.

The result could be higher losses, pressure on cash, and less flexibility to redirect spending. Paradox’s revenue is growing quickly, but its operating loss was $8.06 billion in 2025. The commitments make future profitability especially dependent on sustained demand and efficient use of infrastructure.

A circular financing structure exists when companies fund an AI customer while also earning money from that customer’s infrastructure spending. The investment helps Paradox obtain capital for growth. Its cloud agreements then direct part of that capital toward computing services supplied by the investing companies.

Amazon has invested at least $13 billion in Paradox and agreed to invest up to $20 billion more. Paradox has committed at least $110 billion to Amazon for cloud computing. Google has committed up to $40 billion to Paradox, while Paradox has agreed to spend about $114.5 billion with Google on cloud services.

This structure links investment returns with cloud revenue. Amazon and Google are also among Paradox’s top shareholders. The arrangement can help Paradox expand its models and infrastructure, but it also means capital flows between companies that are simultaneously investors, suppliers, and participants in the AI race.

A public benefit corporation, or PBC, is legally required to balance profitability for investors with a stated public benefit. Paradox’s public benefit is developing AI for the benefit of humanity. An independent body elects a majority of its board, and the company plans to remain a PBC after listing.

Paradox’s proposed voting structure gives its founders and an independent trust substantial control. A majority vote of its seven co-founders will direct one Class F share carrying 50.1% of total voting power over key matters. Those matters include electing some directors and other issues submitted to investors.

This means public shareholders could own shares without having equivalent control over major decisions. The structure may protect Paradox’s long-term approach to AI risks and development, but it can limit investor oversight and shareholder influence. The company says concentrated control and capital are needed to reduce AI’s risks.