AI industrial revolution is about hardware and electricity 

Metaverse & AI

Tags:

AI is coming to your workplace, and it is promising drastic efficiency improvements – but at what cost? The story of efficiency in shared cloud hardware is changing.

At Sulava we have worked with Microsoft 365 and Azure from the very start. We have always understood that the increase in cloud service usage requires more servers and datacenters. Even though Azure has had some specialized workloads that require special hardware, for the most part the datacenters have been running quite standardized x86 servers. And because the hardware has been shared with other customers, it has so far been a beautiful story about the customers’ underused on-premises servers replaced with a smaller amount of shared cloud hardware.

The AI revolution changes this paradigm, with the new technology that was born in the cloud. Suddenly, Microsoft and other companies build massive new datacenters running new specialized hardware only for the AI needs. There is a new major factor in pricing of AI services: not just the total cost of service software R&D, but also the amount of hardware and electricity needed to run the solutions. We are seeing this in practice, some examples:

• Microsoft starting to build their own chips. Read the article Microsoft’s First In-House AI Chip Could Be Revealed Next Month
• Users pay $10 per month for Github Copilot, but a report says Microsoft still loses $20. Read the article GitHub Copilot Loses an Average of $20 Per User Per Month
• Microsoft is going into nuclear reactors to power its datacenters. Read the article Microsoft is going nuclear to power its AI ambitions

This is a new factor for the sustainability strategies of the organizations using the services. Using AI must bring tangible benefits, as this time the automation is not directly about reducing carbon emissions.

There is a lot of positive development that will hopefully even the situation over the years. Luckily Microsoft is aiming to be carbon negative by 2030, and the developments in AI algorithms and related services will surely start to focus on improvements in energy consumption. Microsoft is also improving its customer-specific reporting, and hopefully the Emissions Impact Dashboard for Microsoft 365 will include Microsoft Copilot services at some point.

AI has a huge promise of increasing end-user efficiency. But this time, the extra efficiency will surely come from electric current and silicon.