Intel unveils chip-to-rackscale AI lineup at Computex 2026

Intel is fighting to reclaim AI relevance with a chip-to-rackscale strategy spanning data center, edge, and PC platforms.

Intel Corp. unveiled a sweeping lineup of AI hardware at Computex 2026 in Taipei, spanning data center processors, edge inference platforms, and PC chips, as the company battles Nvidia Corp. for share in a market projected to exceed $400 billion by 2028.

"Customers want AI infrastructure that scales from the client to the cloud without rewriting their software stack," said Michelle Johnston Holthaus, interim co-CEO of Intel, in a statement. "Our chip-to-rackscale approach delivers that."

The announcements include new Xeon 6+ processors for AI inference workloads, edge AI platforms co-developed with partners including IEI Integration Corp., and a collaboration with Phison Electronics Corp. to run larger AI models on Intel Core Ultra PCs. Phison's aiDAPTIV memory extension technology enabled a 26-billion-parameter model to run on a system with 16 gigabytes of DRAM, compared with 32 gigabytes required without the technology, according to Phison's internal testing.

Intel shares have fallen 12 percent this year as Nvidia's push into PC chips threatens the company's last stronghold. The new lineup aims to convince investors that Intel can defend its $50 billion data center and client computing business against Nvidia's expanding portfolio, which now includes PC-focused Arm-based processors.

Xeon 6+ Targets the AI Inference Gap

The Xeon 6+ processors, built on an enhanced version of Intel's Intel 3 process node (equivalent to 3-nanometer-class), target the fastest-growing segment of the AI chip market: inference. While Nvidia's H100 and B200 GPUs dominate training workloads, Intel argues that Xeon's memory bandwidth and software ecosystem make it competitive for running already-trained models in production. The company did not disclose specific performance benchmarks for the Xeon 6+ against Nvidia's comparable products.

Edge and PC Partnerships Expand the Footprint

At the show, IEI Integration Corp. demonstrated edge AI platforms powered by Intel processors for industrial automation, including a TANK-XM813 system that consolidates mobile robot management and AI intrusion detection using Intel's OpenVINO toolkit. The system targets semiconductor wafer transport safety, a niche where reliability requirements are extreme.

Separately, Intel deepened its collaboration with Phison, the NAND flash controller leader, to bring aiDAPTIV memory extension to Intel Core Ultra PCs. The technology allows AI PCs to run larger Mixture-of-Experts models locally — a capability that could reduce reliance on cloud-based AI services for enterprise users. Phison's booth at Computex also featured partners including Ollama, LLMWare, and TurinTech AI, alongside hardware collaborations with ASUS, MSI, and Acer.

Intel trades at 22 times forward earnings, a discount to Nvidia's 35 times and AMD's 30 times, reflecting investor skepticism about the company's ability to execute its turnaround. The Computex announcements provide a near-term narrative for the stock, but analysts will be watching for independent benchmarks and customer adoption timelines. Intel's next earnings report, expected in July, will reveal whether the chip-to-rackscale strategy is translating into design wins.

This article is for informational purposes only and does not constitute investment advice.