Intel launches Xeon 6+ processors as AI moves from chatbots to 'Agents'

Intel has announced its new range of data center technologies, including processors, designed for the next generation of AI software as the company aims to claim a prominent space in the artificial intelligence universe. The announcements, made at the Computex trade show, also include high-speed networking gear and a sneak peek at a new graphics chip code-named "Crescent Island." These chips are capable of handling autonomous “agents” that can execute multi-step tasks on their own even as the traditional CPU is re-emerging as the brain that coordinates it all.Intel’s headline announcement is the Xeon 6+ processor family, which focuses heavily on cutting down the massive energy bills that plaque modern AI data centres. It marks the first time Intel is using its cutting-edge “Intel 18A” manufacturing process for a data centre chip. The new chips are designed to keep applications responsive under real-world power constraints, the company claimed. Their key features include up to 288 “Efficient-cores”, delivering up to 2.5 times better performance than older models, while offering 45% better power efficiency per thread than rival chips. Intel also claims the new chips are so powerful that businesses can replace up to nine older servers with just one new Xeon 6+ server, shrinking a data center's physical footprint and lowering energy costs. Thirdly, a new built-in feature called Application Energy Telemetry allows companies to monitor exactly how much electricity individual software programs are consuming in real time.The new processors are already being built into systems by tech giants like Dell Technologies, Lenovo, HPE, Asus, and Supermicro. Intel also broadened the family for small and mid-sized businesses by introducing a smaller, budget-friendly 12-core model.As AI systems scale up, moving data between thousands of computer chips quickly has become one of the tech industry’s biggest hurdles.To tackle this, Intel expanded its 800 Series Ethernet portfolio with the new Intel Ethernet E835 controllers and network adapters. These devices can supercharge data speeds up to an ultra-fast 200 gigabits per second (200GbE).According to Intel, the E835 network adapter is up to 1.9 times more energy-efficient than Nvidia’s comparable ConnectX-6 DX card and 1.4 times more efficient than a similar offering from Broadcom.Looking to the future, Intel shared early technical details about its next-generation data center GPU, code-named Crescent Island.While standard graphics chips were built for heavy AI training, Crescent Island is being specifically optimized to run “agentic systems.” These are autonomous AI agents that require intense memory and quick thinking rather than just raw math power. Intel says the upcoming chip will directly solve the memory and power bottlenecks that currently cause AI tools to stall or lag.