
Though still notes it's committed to further $80 billion investment in AI infrastructure, saying: 'We will continue to grow strongly in all regions.'
Microsoft has pulled out of deals to lease its data centres for additional training of OpenAI’s language model ChatGPT. This news seems surprising given the perceived popularity of the model, but the field of AI technology is a contentious one, for a lot of good reasons. The combination of high running cost, relatively low returns, and increasing competition—plus working on it’s own sickening AI-made Quake 2 demo—have proven enough reason for Microsoft to bow out of two gigawatt worth of projects across the US and Europe.
Sometimes AI can refer to clever machine learning that helps researchers find plastic eating enzymes, or teach robot dogs how to walk using its own nervous system. However, more often than not AI refers to power guzzling language models and image generators, usually trained on stolen uncredited data to provide shonky results. ChatGPT falls into the latter of these categories, and according to Reuters, a large portion of Microsoft’s cut down on leasing is due to its decision to not support additional training workloads specifically for its creator, OpenAI.
Models such as ChatGPT require huge data centres, usually leased by tech giants like Microsoft, to train and function. Using these data centres doesn’t come cheap from both a financial and power perspective, especially with the current boom. Plus, companies could be leasing these centres to other uses that may have greater payoffs. While large language models like ChatGPT can absolutely be useful when used correctly, and not overly relied on, there’s greater hidden costs the end users don’t necessarily see.
Microsoft, as well as its investors, have witnessed this relatively slow payoff alongside the rise of competitor models such as China’s DeepSeek. Comparatively, DeepSeek is a much more cost effective AI that’s garnering more favour, including the attention of former Intel CEO’s faith based tech company. So Microsoft stepping back from leasing its data centres isn’t too surprising in the grand scheme of things.
In the United States Facebook’s Meta has stepped up in leasing its data centres to fill the gap left by Microsoft, and Google’s parent company Alphabet has done the same in Europe. Reuters also states Alphabet has committed to a $75 billion spend on its AI buildout this year, which is a 29% hike over Wall Street’s expectations. Meta, is set to spend up to $65 billion on AI. All three companies, including Microsoft, have defended AI spending as necessary to keep up with competition from the likes of DeepSeek.
Despite pulling out of hosting further data centre training for ChatGPT, Microsoft says its still on track to meet it’s own pledged $80 billion investment into AI infrastructure for the year. The company says while it may “strategically pace or adjust our infrastructure in some areas, we will continue to grow strongly in all regions”.
Best SSD for gaming: The best speedy storage today.
Best NVMe SSD: Compact M.2 drives.
Best external hard drive: Huge capacities for less.
Best external SSD: Plug-in storage upgrades.