Microsoft AI: Scale Hurdles Stall Growth

• Microsoft AI chief says scaling is limited by computing power.
• Tech giants invested $650bn in AI, but transformer shortages delay data centres.
• Microsoft aims to build long-term computing capacity internally, impacting AI development.

While Microsoft makes headway into AI and expands its capabilities further, its AI branch CEO Mustafa Suleyman has said that the company still lacks the computing power needed to build systems at the largest scales possible. The aim behind the expansion has been to reduce reliance on its external partners.

Highlighting the efforts, Suleyman told Financial Times, “We are not able to build models at the very largest scale yet, although our computation ramp is coming to enable us to do that later this year.” He further added that they are competing in the mid-class range, describing the position as “optimal” in balancing the cost and performance at large-scale usage.

Suleyman shared these remarks as the company revealed a new speech transcription model as part of its broader push to strengthen its position in AI. Challenges like constraints of data centre capacity, equipment shortages, power availability, and labor are impacting its pace of AI development.

Suleyman detailed how Microsoft is focusing on building long-term computing capacity and internal capabilities as it develops its own advanced AI systems.

Data Centres Planned for 2026 in the US Are Likely to Be Delayed

Suleyman’s opinion about the lack of data centres comes at a time when nearly half of the data centres expected to come online in the United States by 2026 are likely to face cancellations or delays. The shortage of this key infrastructural element is impacting artificial intelligence. Leading tech brands such as Alphabet, Amazon, Meta, and Microsoft have collectively committed close to $650 billion this year to the industry, but the bottleneck does not lie in funding or computing technology. Instead, it is due to the lack of transformers, switchgear, and batteries required to power these facilities.

There is a stark difference between ambition and execution, as there exists a widening gap between the plans to enforce projects that consume 12 gigawatts of power; only about a third are currently under construction.

The biggest hindrance for development has been the shortage of electrical equipment needed to supply power to the centres and expand the network grid.