AMD and Intel Reveal 16x AI Compute Boost for x86 Chips
AMD and Intel have introduced a joint AI Compute Extensions (ACE) architecture. As a result, this framework delivers a 16-fold increase in compute density compared to similar AVX10 operations. Moreover, the announcement marks a major step for the x86 Ecosystem Advisory Group, which both companies formed in 2024 to strengthen the future of x86 computing.
At the same time, the collaboration reflects a shared effort to stay competitive against alternative chip architectures. Therefore, ACE represents both a technical breakthrough and a strategic alignment. While the companies have long competed, they now cooperate to address growing AI demands.
What ACE Brings to AI Workloads
ACE introduces native matrix multiplication support for several AI-focused data types, including INT8, FP8, and BF16. Consequently, it directly improves the efficiency of tasks used in machine learning and large language models. Because matrix multiplication forms the backbone of AI computation, this enhancement targets a critical performance bottleneck.
In addition, the architecture complements AVX10 through an outer product operation. As a result, it achieves significantly higher compute density using the same number of input vectors. Meanwhile, developers will see growing software support, as integrations are planned for tools like NumPy, SciPy, PyTorch, and TensorFlow.
Therefore, ACE aims to become a unified matrix-acceleration standard across devices, ranging from laptops to data centers. As adoption expands, developers will be able to deploy AI workloads more consistently across systems.
Collaboration and Future AI Strategy
The ACE framework builds on broader cooperation within the x86 Ecosystem Advisory Group. Alongside AMD and Intel, members include Broadcom, Dell Technologies, Google, Meta Platforms, Microsoft, and Oracle. In addition, advisors such as Linus Torvalds and Tim Sweeney contribute to its direction.
Together, the group works to reduce fragmentation in x86 software development. Consequently, unified standards like ACE and AVX10 will simplify programming and improve compatibility. According to Robert Hormuth, new standards will begin appearing in chips from 2026.
Meanwhile, both companies continue expanding their AI ambitions. AMD is preparing for its “Advancing AI 2026” event, while Intel is developing hybrid processors that combine traditional cores with AI accelerators. As Jensen Huang has noted, this collaboration plays an important role in sustaining the relevance of x86.
