|
Over the past few weeks, headlines have reported that OpenAI has entered a trillion-dollar partnership with NVIDIA, AMD and Oracle – a number so big it sounds more like a typo than a transaction.
Cue a mix of excitement and head-scratching. How can a company that still loses money afford something bigger than the GDP of most countries? And what’s in it for the world’s chipmakers and data giants?
The truth is, nobody’s writing a trillion-dollar cheque. It’s more like OpenAI has found a way to put the entire AI revolution on a very long-term payment plan. The company gets the vast computing power it needs now, while its partners build it, fund it and wait patiently for the instalments to roll in.
In return, those partners get something even more valuable than cash up front: a guaranteed customer, steady income, and in some cases, a small stake in OpenAI’s future success.
It’s a bold bit of deal making – and a reminder that when it comes to creative finance, humans are still running the show.
Teaching machines to “think” doesn’t come cheap. Training an AI model like ChatGPT needs staggering amounts of computing power – banks of servers stretching across areas the size of football pitches, humming day and night. We’re talking thousands of chips, endless cooling fans, and enough data-centre capacity to make a small country blush.
But despite the hype, OpenAI still isn’t profitable. In fact, its cashflow looks more like a student overdraft. Those enormous computing costs, research expenses, and sky-high salaries for some of the world’s sharpest minds all add up. And even with Microsoft’s backing, the money going out still has a head start on the money coming in.
So OpenAI has done what any sensible company does when faced with a colossal bill: it’s paying over time. The servers, the chips, the data centres – even the buildings themselves – are being built and financed by others. OpenAI just signs the lease, moves in, and carries on building the future.
Of course, paying over time only works if someone’s willing to front the bill. That’s where NVIDIA, AMD and Oracle come in – three companies with the cash, the chips and the confidence to bankroll OpenAI’s ambitions, and to make a tidy return while they’re at it.
NVIDIA designs the processors that power most of the world’s AI systems. Demand is so intense that even billion-dollar companies have to join the queue. So NVIDIA has decided to skip the sales pitch and become both supplier and landlord.
It’s now building vast “AI campuses” filled with its own processors and renting them to OpenAI for twelve years. The rent is fixed, so even if OpenAI hits a slow patch, the payments keep rolling in. For NVIDIA, it’s as close to guaranteed income as business gets.
And there’s a cherry on top: every time a new site goes live, NVIDIA picks up a small ownership stake in OpenAI – predictable rent today, a potential jackpot tomorrow.
AMD, NVIDIA’s long-time rival, didn’t fancy sitting this one out. It’s struck its own six-year deal to supply around $20 billion worth of AI chips to OpenAI. And hidden inside the contract is a clever little bonus: OpenAI can buy a small portion of AMD’s shares cheaply if the company’s share price takes off.
It’s a win-win arrangement. AMD secures a high-profile customer for its new generation of chips, while OpenAI gets a bonus if the technology lives up to the sales pitch. It’s friendly rivalry, Silicon Valley style.
Once best known for its database software and 1990s nostalgia, Oracle has reinvented itself for the AI age. Under a venture called Stargate – a suitably sci-fi name for a data-centre empire – the company, backed by SoftBank and Abu Dhabi’s Mubadala fund, plans to spend up to $500 billion building enormous data-centre campuses across the US.
It’s a bold undertaking. Oracle buys the chips, borrows the construction money, builds the sites and leases them to OpenAI on “take-or-pay” terms – meaning OpenAI pays the rent whether it switches the lights on or not. That kind of certainty is catnip to investors. The company’s future order book has ballooned, and the once-steady software firm now finds itself the unlikely landlord of the AI age – a comeback even Hollywood would struggle to write.
It’s easy to see why some people say the money just goes round in circles. OpenAI pays Oracle, Oracle pays NVIDIA and AMD, and the chipmakers end up with small stakes in OpenAI. From the outside, it looks like a very expensive game of pass the parcel.
But in reality, each deal stands on its own. Oracle books rental income, NVIDIA and AMD log hardware sales, and OpenAI records long-term lease payments rather than one giant expense. No one’s spending a trillion dollars upfront – it’s a decade-long chain of contracts, each designed to keep the others standing.
So, who actually comes out on top? Well, everyone really.
OpenAI gets the world’s best computing power without spending like it. NVIDIA locks in years of chip sales and a small piece of OpenAI for good measure. AMD wins a high-profile client and a chance to benefit if its technology proves its worth. And as for Oracle, it’s gone from database dinosaur to AI landlord – possibly the comeback of the decade.
Of course, it’s not all smooth sailing. OpenAI now owes billions in rent, whether business booms or not. Oracle has borrowed eye-watering sums to build its new empire. And NVIDIA and AMD have tied themselves to a customer whose balance sheet depends on people still paying for chatbots in a few years’ time.
Throw in the fact that chips have a shorter shelf life than a tabloid headline, and you’ve got a structure that works brilliantly – until it doesn’t.
For now, though, everyone’s smiling, and the trillion-dollar arrangement works beautifully – if everyone keeps their nerve.
What makes this whole saga fascinating isn’t the scale of the numbers (which, frankly, are getting a bit silly now), but the simplicity of the idea. OpenAI wanted power, others wanted promise, and somehow everyone got what they wanted – for now. If AI really is the next revolution, this might be its first lesson: progress still depends on faith, financing and a very persuasive story.
Stay updated with the latest insights and articles delivered to your inbox weekly.
Subscribe to our newsletter for the latest insights and expert advice
on funding structures.
