Microsoft launched the second-generation self-developed AI chip Maia 200 in January 2026, using TSMC’s 3-nanometer process. Alistair Speirs, General Manager of Microsoft Azure global infrastructure, visited Taiwan and gave his first interview to the Taiwanese media Central News Agency on Tuesday (September 1). He stated that Taiwan’s supply chain has made Microsoft’s independent chip development strategy a reality, saying, “Without Taiwan, there would be no Maia 200 chip,” highlighting the collaborative achievement of Microsoft Taiwan and the global hardware and software teams.
Speirs is primarily responsible for cloud data centers and other infrastructure and took the opportunity during the SEMICON Taiwan 2026 International Semiconductor Exhibition in Taipei this week to visit Taiwan for the first time. During the interview, he personally showcased samples of wafers, chips, and circuit boards to the Central News Agency interview team, expressing his fascination with the intricate structures and the technology contained within.
Regarding one particular circuit board, Speirs remarked, “Looking at these layered structures, I find it incredible. How much technology is embedded inside?” The samples are small enough to fit in a travel case, allowing him to showcase them to clients around the world.
He noted that Taiwanese people understand the differences between CPUs and GPUs, know about ASICs (Application-Specific Integrated Circuits), as well as HBM (High Bandwidth Memory) and SRAM (Static Random-Access Memory), which are topics that are difficult to discuss on the streets of Silicon Valley but can be openly conversed about in Taipei, making it an interesting aspect. Speirs believes that Taiwan is one of the most mature and technically proficient markets for cloud computing and AI applications globally, possessing unique advantages.
He emphasized the crucial role Taiwan plays in Microsoft’s infrastructure and partner relationships, stating that “collaboration is not just a supply chain; it is an ecosystem.” With AI technology evolving rapidly, true collaboration, partnerships, and co-design involving chips, boards, memory, components, and software layers are required to be integrated.
On the topic of Nvidia announcing a $3.5 billion investment in MediaTek with overseas convertible bonds (ECBs) to deepen AI infrastructure collaboration, Speirs mentioned the significant growth in industry AI capabilities and chip designs along with the emergence of various methods. Microsoft’s focus is on establishing deep collaborative relationships with Taiwan and the global industry to create differentiated systems optimized for Microsoft’s cloud platform workloads.
Regarding whether Microsoft will adopt a similar strategy in investing in companies in Taiwan, Speirs did not provide a direct response.
Addressing concerns in the market about rapid growth in capital expenditure for super-scale AI data centers and worries about an “AI bubble,” Speirs stated that Microsoft’s business covers office efficiency, cloud services, consumer services, and gaming, all of which see the demand for new features. Hence, they will adopt new models, features, and hardware, including a higher reliance on CPUs with new AI models, to maintain a balance between global infrastructure operational capacity and investment.
“Microsoft’s mission is to build the necessary infrastructure to meet the demands we have already seen for the coming years,” he mentioned, highlighting that customers do not buy chips but long-term business outcomes. Therefore, it is not just hardware or cloud infrastructure investment but also software investment and how technology is transformed into a driving force for business outcomes. The Microsoft team is focused on integrating new features into existing products like Windows, Office, and Microsoft 365, while developing new products like Copilot specially designed for native AI functionality.
Speirs commented that Microsoft has established deep collaborations with Nvidia and also cooperates with AMD, Intel, and ARM to understand product roadmaps, share insights on technological developments, and customer requirements. When existing commercial chips and systems fail to meet demands, Microsoft seeks to apply its own innovation, as exemplified by Maia.
“Maia is a great example,” he said. The development goal of Maia is to create a platform focused on reasoning and unique AI applications, which do not require the functionalities necessary for GPU or training chips. This allows Microsoft to build chips highly optimized for various cloud workloads, supporting Windows, Office, and AI functionalities used by customers.
He highlighted that as chip power consumption and array density increase, simply enhancing chip transmission performance is no longer sufficient to meet demands. Microsoft sees investment opportunities in everything from chips, servers, network architectures, cooling systems to software optimization. With AI models rapidly evolving, even with the same hardware, performance can be improved by 30% to 40% within just one product generation through optimization.
(Source: Central News Agency)