SpaceX is apparently set to make its own GPUs, based on insights gleaned from its S-1 form ahead of the company’s IPO, although these won’t be products you’ll be installing in your PC.
Tom’s Hardware picked up the report by Reuters, with the news site claiming to have seen the S-1 form, which is filed by companies going public. SpaceX is expected to be worth in excess of $1 trillion – and probably closer to $2 trillion – when its shares are available on the stock market (next month, in theory).
Part of the form explains major spending plans to develop AI, which includes a mention of “manufacturing our own GPUs” in the listed ‘substantial’ capital expenditures.
Article continues below
As noted, these wouldn’t be a rival for the likes of Nvidia‘s GeForce GPUs, but rather graphics boards built for AI tasks, commonly known as AI accelerators.
Nvidia makes those solutions, as well, of course – as does AMD – but SpaceX would seemingly rather have its own homegrown AI accelerators, because as noted in the S-1 filing, it does not have “long-term contracts with many of our direct chip suppliers”. And clearly, that’s going to be problematic given the current RAM crisis and its knock-on effects.
What isn’t clear is whether SpaceX’s mention of ‘GPUs’ in the filing could more broadly refer to Tesla’s AI processors, which are currently on the AI4 generation, with the next-gen AI5 set to provide a big leap in performance (of a factor of 40x, it’s claimed). It’s possible that the company may be referring to this series of AI chips, rather than an entirely separate line of GPUs for AI tasks.
When it comes to making these chips, presumably the ‘Terafab’ project – a huge chip manufacturing center in Texas which involves SpaceX, xAI and Tesla – is going to be involved, even though GPUs haven’t been mentioned directly in connection with this facility.
Analysis: AI5 or something else entirely, it won’t be in your PC
At least to me, this does read more like a play for a new product, as opposed to a mention of existing AI chips and calling them ‘GPUs’ informally. That said, looking at the bigger picture, as Tom’s points out, Elon Musk has referred to Tesla’s AI5 chips as GPUs, and even noted that it “basically is a GPU” – and so you can see where the confusion comes in. All this could simply boil down to a liberal use of the term GPU, and mean precisely nothing.
At any rate, for the average consumer, the point to bear in mind is that whatever SpaceX is up to on the GPU front, if anything, it won’t be relevant to the everyday computer user. This is all about silicon designed to drive AI performance specifically, and whether it’s technically an AI chip in the existing line, or something different in terms of a new GPU accelerator designed for AI tasks, is all semantics – it won’t be a graphics card as such (as in one designed to process 3D graphics).
Any confusion aside, the overall aim for Musk is clear enough, and that’s to ensure AI chip supply by producing them himself, given the mentioned lack of ‘long-term contracts’ with chip makers.
How comforting this leak will be to would-be investors in SpaceX remains to be seen, but Musk appears to have some grand plans here, and not for the first time.
The best laptops for all budgets
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course, you can also follow TechRadar on YouTube and TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.