OpenAI says Stargate, its huge AI infrastructure buildout for data centres, chips, cloud capacity and power, has already passed its original target of securing 10GW of US AI capacity by 2029. The milestone lands as OpenAI explores a possible IPO that Reuters has reported could value the ChatGPT maker at up to $1tn.
For investors, the timing creates a sharper question around the business model. OpenAI is increasing the physical cost of running and improving ChatGPT before public markets can fully see whether revenue will cover the bill.
OpenAI said it added more than 3GW of capacity in the past 90 days alone, a pace that shows how quickly the company believes AI demand is growing. The announcement followed reports that OpenAI had missed recent revenue and user-growth targets, with concerns inside the company over whether future data-centre costs can be funded if revenue does not grow quickly enough.
Stargate is the infrastructure machine behind OpenAI’s next phase. In plain English, it means more data centres, more chips, more cloud capacity and more power so ChatGPT and future AI tools can train larger models, answer more users, serve companies and support developers. OpenAI needs the capacity now to keep products fast, reliable and competitive. The revenue proof comes later. That is uncomfortable for any IPO story because public investors will eventually ask how much of each dollar of ChatGPT revenue is left after paying for the infrastructure behind it.
A normal software company can often add users at high margins once the platform is built. Generative AI has a heavier cost curve. Every prompt, model upgrade, business tool and developer workload uses expensive computing power. More usage can increase revenue, but it can also push up the cost of delivery.
OpenAI’s earlier IPO debate was already centred on whether ChatGPT revenue could outrun the cost of running the technology. Stargate raises the stakes because the infrastructure target has been hit years ahead of schedule. Capacity can create growth, but only if enough users and customers pay at levels that justify the buildout.
The reported missed targets make the timing harder to sell. OpenAI still has to keep building so it does not fall behind Anthropic, Google, Meta and other AI rivals. Waiting too long could hurt performance and limit enterprise sales. Building too fast risks putting spending ahead of revenue.
The Microsoft relationship adds another layer to the IPO case. OpenAI says Microsoft remains its primary cloud partner, while OpenAI can now serve products across any cloud provider. Microsoft’s licence to OpenAI models and products continues through 2032, but it is now non-exclusive. That gives OpenAI more room to secure capacity beyond one cloud provider. It also gives future investors more to examine: who supplies the infrastructure, who gets paid first, and how much revenue remains after data-centre and cloud partners take their share. The cleanest financial test for Stargate is utilisation. Securing 10GW of AI capacity sounds powerful, but unused or weakly monetised data-centre capacity would weigh on margins. Capacity sold into high-value business, developer and government demand would make the spending easier to defend. Enterprise customers may decide whether the numbers work. Consumer use gives OpenAI scale and brand power, but businesses usually bring larger contracts, clearer budgets and stronger retention. If Stargate helps OpenAI sell reliable AI capacity to companies and governments, the data-centre bill becomes part of a larger revenue engine.
Consumer demand is less certain. ChatGPT remains one of the best-known products in technology, yet reported user-growth misses suggest the app may not convert every wave of public attention into predictable paid revenue. A future IPO will need to show that free usage, paid subscriptions and enterprise contracts are moving in the right balance. The wider market is already exposed to the same question. Reuters reported that shares in AI-linked companies fell after the Wall Street Journal report on OpenAI’s missed targets, with Oracle and CoreWeave among the names hit by concern over OpenAI’s growth prospects. OpenAI’s spending plans now reach chipmakers, cloud providers, data-centre operators and power suppliers.
Stargate has also been linked to a much larger AI infrastructure buildout involving OpenAI, SoftBank, Oracle and other partners, with potential investment reaching up to $500bn. That scale explains why investors are watching OpenAI’s revenue path so closely. If AI demand keeps rising and customers pay for it, the buildout supports a powerful growth story. If revenue disappoints, the sector is left with a larger cost base and weaker visibility on returns.
A trillion-dollar valuation would make the margin for error even smaller. Public investors can accept heavy spending when it creates durable advantage, but they will expect evidence that the spending turns into revenue quality, customer retention and a path to cash generation. OpenAI still has strong arguments - ChatGPT is a global consumer brand, AI demand is still growing, and more infrastructure can improve speed, reliability and model performance. Those strengths could support higher-value products and larger enterprise deals.
The pressure sits in the sequencing in that OpenAI is expanding the machine before the market can fully see the profit model. That may be necessary in AI, where capacity shapes the product itself, but it also makes the IPO case more demanding.
OpenAI has shown it can secure the infrastructure for the next wave of AI demand. The next proof point is financial: whether ChatGPT and enterprise AI revenue can justify the data-centre bill before the company reaches public markets.
More from from Finance Monthly - Anthropic Amazon Compute Deal: What the $100 Billion Commitment Says About AI Infrastructure Power
