A Guardian investigation about Microsoft internal documents and its AI expansion has raised questions whether the company has enough computing chips for its data centers, or sufficient power and finished facilities to use the hardware it already owns effectively.
Since 2022, Microsoft spent $280 billion on AI infrastructure, revealing it also added five gigawatts (GW) of data center capacity. However, according to The Guardian investigative documents, it showed 2.2 million AI chips installed, fewer than some experts expected.
The gap does not prove to have a chip shortage, but that announced capacity is not always ready to run.
Is Power the Real Bottleneck than Chip Supply?
Microsoft CEO, Satya Nadella, has suggested electricity and construction are the problems. That point shifts the debate away from a simple shortage of computing chips but suggests the Azure-parent may already own hardware waiting for powered data center space.
“If you can’t do that, you may actually have a bunch of chips sitting in inventory that I can’t plug in. In fact, that is my problem today. It’s not a supply issue of chips. It’s the fact that I don’t have warm shells to plug into,” said Nadella on the All Things AI podcast.
Microsoft has rejected the Guardian’s calculations.
“Microsoft does not report on the volume of specific chips in its AI infrastructure. The estimates the Guardian has shared with us are inaccurate, drawing the wrong conclusions from incorrect assumptions,” a spokesperson said.
The company says its data centers use Nvidia, AMD, Intel, and custom silicon. That mix makes exact estimates difficult and adds complexity to chip designing.
Chip Numbers and Microsoft’s AI Build-out
Microsoft reportedly aimed for 1.8 million AI chips by the end of 2024. Nearly two years later, its internal count stood at 2.2 million.
While it does look modest besides its spending and expansion plans, especially as demand for computing chips grows, for companies considering NVidia alternatives, Microsoft proves that buying accelerators is only one part of AI expansion.
Servers need power, cooling, and completed buildings. Meanwhile, the existing uncertainty also affects chip engineering, because a company can secure processors but leave them unused if a data center cannot connect to the grid.
One Nvidia analyst told the Guardian they expected more chips.
“They’re low to me. They’re less than I expected Microsoft would have,” the analyst said.
The issue matters for Nvidia competitors in AI, because customers may judge suppliers on how easily hardware can be deployed, not only on speed. On the other hand, Shaolei Ren, a professor at the University of California, Riverside, questioned Microsoft’s capacity claims.
“According to their own metrics, Microsoft could be correct. But it isn’t clear what they mean when they say they have added data center capacity. They are giving insufficient context,” Ren said.
For chip designing teams, that difference matters. A processor may be ready while the power or cooling system needed to operate it is unfinished.
Is Big Tech Power Slowing Down Chip Supply?
Such aspect matters for processor makers because AI accelerators in storage create little value and must be installed and connected before customers can use them. Microsoft’s Fairwater project in Wisconsin highlights the issue.
According to Nadella, it “is going live,” but Microsoft later acknowledged the site was not yet online. That creates another challenge for chip engineering, as newer processors can demand more electricity and cooling than earlier systems.
It also gives Nvidia AI competitors another opening. Hardware that fits available power and cooling limits may appeal even if it is not the fastest option.
For companies studying Nvidia alternatives, Microsoft shows that large chip orders do not automatically become working with AI capacity. Teams that architect chips must think about the entire data center.
Power, cooling, and construction can decide whether hardware reaches full use.
The same lesson applies to chip designing. Faster hardware matters, but only when the infrastructure around it is ready.
Comin to Microsoft, it may have more chips waiting to be installed, but the evidence does not clearly show a simple shortage. The question remains whether its infrastructure can keep pace with the computing chips it is buying.
If Microsoft’s AI capacity grows faster than usable facilities, investors may struggle to judge what is operational. The debate is not only about how many computing chips Microsoft owns, but how many it can switch on.
Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Intelligent Tech sections to stay informed and up-to-date with our daily articles.
