Microsoft AI chief says AI limits still far off, computing power to jump 1,000-fold by 2028
Microsoft AI chief says AI limits still far off, computing power to jump 1,000-fold by 2028
Publish Date: 2026-04-09 01:46:00
Source Domain: www.digitaltoday.co.kr
- Mustafa Suleyman, CEO of Microsoft’s AI group, asserts that artificial intelligence development is not facing near-term limits.
- Significant growth in computing power has been a primary factor influencing recent advancements in AI performance.
- The computation used to train leading AI models has risen dramatically since 2010, increasing from 10 to the power of 14 flops to over 10 to the power of 26.
- Improvements in areas beyond semiconductors, including memory, networks, and software, have contributed to AI improvements, alongside Moore’s Law.
- Hardware advancements, such as the eightfold increase in the raw performance of Nvidia chips and the introduction of Microsoft’s AI chip Maia 200, have significantly enhanced computational abilities.
- Dramatic reductions in training times and increased efficiency have been achieved through advancements in connectivity technologies and software.
- Forecasts indicate an accelerating pace of growth; AI research organizations’ capabilities are expanding fourfold annual, with computing capacity projected to increase significantly by 2027.
- Suleyman predicts AI will evolve beyond basic assistants to human-level agents capable of managing long-term projects and logistics across industries.
