Data Centers in Decline: Microsoft Shifts Gears While Users Foot the AI Bill
The tech giant is cutting its data center plans, thanks to the rising costs of AI applications and processes.
The world is witnessing a rapid advancement of large-scale generative AI technology, and with it, rising costs of operations. Recently, Microsoft has implemented notable measures to offset its substantial operational costs across its products and services. The company has increased subscription prices for its Microsoft 365 software, with increases as high as 45%. It has also introduced ad-supported product versions and canceled some data center lease agreements.
Value-Defying Costs
According to Microsoft CEO, Satya Nadella, despite the company's massive investment in AI technology, the value generated so far has been insignificant. These measures seem to reflect Microsoft's efforts not to retreat from AI investment, but rather to explore new monetization models, essentially shifting the brunt of the costs to consumers.
The cost of generative AI is considerable and even market leader OpenAI has had to adjust. Despite generating $3.7 billion in revenue last year, the maker of ChatGPT saw its expenses reach $9 billion, resulting in a net loss of approximately $5 billion. OpenAI's operational costs primarily consisted of model training and inference. As the user base grows, inference costs continue to rise.
A Move to Reduce Expenses
As OpenAI's largest investor, Microsoft is also bearing the brunt of these cost pressures. As a result, Microsoft is seeking to reduce these expenses by shifting more AI computation tasks to users' devices. This means that users would bear the cost of the hardware and its operation, potentially reducing Microsoft's operational burden.
The move doesn't stop at the price increase. Microsoft also added dedicated Copilot buttons to its new devices, encouraging users to perform AI processing on their own devices. This "edge computing" strategy aims to reduce data center energy consumption and resource waste, while also enhancing user privacy.
However, this strategy is not without challenges. One is the shifting of AI computational costs, meaning consumers may need to frequently update their devices, leading to increased e-waste. Furthermore, hardware disparities could lead to uneven user experiences, particularly in education, potentially increasing inequalities.
