Level 1 — Absolute Beginner
AMD, a company that makes computer chips, showed off new AI chips at an event in San Francisco. The event happened on July 22 and 23.
AMD's new chips are called the Instinct MI400 series. They are built to help computers learn and think faster, which is called artificial intelligence, or AI.
AMD also showed a new computer rack called Helios. It holds 72 of AMD's new chips together, working as one giant machine.
AMD wants to compete with Nvidia, another chip company that leads the AI chip market today.
- chip
- a small electronic part inside a computer that helps it think and do work
- event
- a planned public happening, such as a meeting or show
- artificial intelligence
- computer systems built to perform tasks that normally need human thinking
- rack
- a frame used to hold and organize computer equipment
- giant
- extremely large in size
- compete
- to try to be better than others in the same activity or market
- lead
- to be in first place or ahead of others
- market
- the buying and selling of a particular kind of product
Level 2 — Elementary
AMD unveiled a new lineup of artificial intelligence chips at its Advancing AI 2026 conference in San Francisco on July 22 and 23, marking the company's most direct challenge yet to Nvidia's dominance of the AI hardware market.
The centerpiece announcement was the Instinct MI400 series of graphics processing units, which AMD says can deliver up to ten times more performance for large AI models than its previous generation, along with 432 gigabytes of next-generation HBM4 memory.
AMD also introduced Helios, a rack-scale system that packs 72 of the new MI400-series chips into a single unit with a combined 31 terabytes of memory, designed to compete directly with Nvidia's flagship rack systems.
Rounding out the announcements were the next-generation Zen 6 'Venice' server processors, the first x86 chips built on a cutting-edge 2-nanometer manufacturing process, paired with a new high-speed networking chip to move data between servers.
- lineup
- a set of products or people arranged for a particular purpose
- conference
- a large meeting where people share information on a topic
- dominance
- the position of being the most powerful or influential
- graphics processing unit
- a specialized chip originally designed for images, now widely used for AI computing
- memory
- the part of a computer that stores data for quick access
- rack-scale
- describing computer hardware designed to fill and operate as an entire equipment rack
- manufacturing process
- the method used to produce something, especially in a factory
- networking
- connecting computers or devices so they can share data
Level 3 — Intermediate
AMD used its Advancing AI 2026 conference in San Francisco on July 22 and 23 to unveil its most comprehensive AI hardware lineup to date, positioning the company for its most direct challenge yet to Nvidia's dominance of data center AI computing.
At the center of the announcements were the Instinct MI400-series accelerators, including the MI450X and MI455X, which AMD claims deliver up to ten times the performance of the prior generation on frontier AI models, backed by 432 gigabytes of HBM4 memory and roughly 2.9 exaflops of FP4 compute throughput.
AMD paired the chips with Helios, a rack-scale system housing 72 MI400-series accelerators and a combined 31 terabytes of HBM4 memory, a configuration explicitly benchmarked against Nvidia's NVL72 rack systems that currently anchor much of the industry's AI infrastructure.
Completing the platform were next-generation Zen 6 'Venice' EPYC server processors, the first x86 chips fabricated on a 2-nanometer process with 256 cores, alongside an 800-gigabit-per-second networking chip and support for open interconnect standards such as UALink and the Ultra Ethernet Consortium.
- comprehensive
- complete and including everything necessary
- accelerator
- specialized computing hardware designed to speed up a particular type of task, such as AI processing
- frontier
- at the very forefront or cutting edge of a field
- throughput
- the amount of data or work a system can process in a given time
- configuration
- the particular arrangement or setup of a system's components
- benchmark
- to measure the performance of something against a standard or competitor
- fabricate
- to manufacture or construct, especially using an industrial process
- interconnect
- a system or standard that allows different devices to connect and communicate
Level 4 — Advanced
AMD leveraged its Advancing AI 2026 conference in San Francisco on July 22 and 23 to unveil its most comprehensive AI hardware portfolio to date, positioning the company for its most direct challenge yet to Nvidia's entrenched dominance of data center AI computing.
At the center of the announcements were the Instinct MI400-series accelerators, including the MI450X and MI455X, which AMD claims deliver up to a tenfold performance uplift over the prior generation on frontier AI workloads, underpinned by 432 gigabytes of HBM4 memory and roughly 2.9 exaflops of FP4 compute throughput.
AMD paired the accelerators with Helios, a rack-scale system integrating 72 MI400-series chips and a cumulative 31 terabytes of HBM4 memory, a configuration explicitly benchmarked against Nvidia's NVL72 rack systems that currently anchor much of the industry's AI infrastructure buildout.
Rounding out the platform were next-generation Zen 6 'Venice' EPYC server processors, the first x86 silicon fabricated on a 2-nanometer process with 256 cores apiece, complemented by an 800-gigabit-per-second networking chip and support for open interconnect standards such as UALink and the Ultra Ethernet Consortium, an architecture AMD frames as a counterweight to Nvidia's proprietary ecosystem.
- portfolio
- a collection of products or offerings from a company
- entrenched
- firmly established and difficult to change
- uplift
- an increase or improvement in something, especially performance
- underpin
- to support or form the basis of something
- cumulative
- increasing or growing by successive additions
- apiece
- for or by each one individually
- proprietary
- privately owned and controlled, often referring to technology exclusive to one company
- counterweight
- something that balances or offsets an opposing force or influence
Talk about this article with KatePractice this story's vocabulary in a real spoken conversation with your AI tutor