OpenAI isn’t just burning through cash. It’s lighting an entire mountain of money on fire.
But since it’s not a publicly traded company, the extent of that mountain remains difficult to gauge. But clues periodically emerge: as the Financial Times reports, for instance, the company recently signed a staggering $250 billion rental agreement with Microsoft — as well as a $38 billion contract with Amazon, less than a week later.
According to HSBC, whose software and services team issued an update to its financial model of OpenAI, the company will be spending a nauseating $620 billion per year on renting data center capacity to power its AI models alone. That’s despite only a third of the total contracted amount of 36 gigawatts actually scheduled to come online before 2030.
It’s a precarious moment, with lenders expecting a blockbuster growth in revenue — an enormous ask that will require OpenAI to accomplish things unprecedented in business history.
Meanwhile, experts are warning of an AI bubble that has grown so much that it’s proping up the entire US economy. If it were to pop, the consequences could be disastrous, with the discussion shifting towards whether firms can even survive the next couple of years as spending on data centers continues to rise precipitously.
And that’s not the only looming challenge. OpenAI also has to contend with increased competition as well, with Google’s Gemini 3 throwing down the gauntlet earlier this month. That’s more naysaying from HSBC, by the way, which expects OpenAI’s consumer market share to drop substantially by the end of the decade.
Whether OpenAI will be able to pay its bills in the upcoming years remains hazy at best. According to HSBC, the company will need to reach three billion ChatGPT users by 2030, a steep goal considering userbase growth is already plateauing. According to CEO Sam Altman, the company has around 800 million weekly active users.
As the FT reported last month, paying ChatGPT subscribers represent roughly 70 percent of the outfit’s annual recurring revenue — yet only a pitiful five percent of users are currently willing to pay for a subscription.
New revenues will also have to be poured into further data center build-outs to support new users, as the FT explains, in a vicious cycle that could heavily eat into OpenAI’s margins.
If things were to go south, HSBC suggests that OpenAI will end up changing its data center commitments.
“Less capacity would always be better than a liquidity crisis,” the broker wrote.
While HSBC remained bullish about the prospect of enormous “productivity-driven” economic growth that could “dwarf what is often seen as unreasonable [capital expenditure] spending at present,” OpenAI still has to pull it off.
And investors are already asking some tough questions over AI companies’ sky-high valuations and meager revenues, with AI chipmaker Nvidia’s shares falling for several weeks now — despite posting better-than-expected quarterly earnings.
And Nvidia is the one selling shovels during the ongoing AI gold rush. OpenAI has a much steeper hill to climb, since it has to create a profitable service while paying for all those shovels.
In a telling moment earlier this month, Altman became irate after being asked by podcaster and OpenAI investor Brad Gerstner how a company “with $13 billion in revenues” can “make $1.4 trillion of spend commitments.”
“If you want to sell your shares, I’ll find you a buyer,” Altman shot back. “Enough.”
More on OpenAI: OpenAI Locks Down Office After Violent Threat