It is being widely said that Open AI will not build any datacentres of its own and will rely for computing resource on getting deals with companies which do have data centres like AWS, Oracle, Google and Microsoft.
This is a huge change if it is actually going to happen because at one time OpenAI said it would spend $1.4 trillion in eight years on building datacentres which was recently downscaled to $600 billion.
Now, with a slowdown in the building of datacentre projects like the Stargate datacentre in Abilene Texas, it is being realised that a slowdown in building computing resources will slow the growth of OpenAI’s revenues and therefore OpenAI should revise its spending plans to more closely track its revenue growth.
The company, which does not own any datacentres itself, has changed course in the past to deal with inconvenient realities and this could be the next pivot.
Open AI remains keen on raising money from investors and is looking at an IPO which might value it at $1 trillion.
Cutting back on its spending plans by renting computing resource rather than building it might well seem like a more plausible pitch to potential investors.
FYI: Very interesting (long) reads on this topic, including detailed financial analysis, have been (and are being) written by Ed Zitron.