Intel's AI Everywhere event, held in New York on 14 December 2023, was where the company launched its Core Ultra processors for AI PCs and its 5th Gen Xeon processors for servers, and gave a first look at the Gaudi 3 AI accelerator due in 2024. This post carried the event replay. The headline message was that AI would run across every class of Intel hardware, from the laptop to the data centre and the network edge.
Event replay as embedded in December 2023. If the player does not load, Intel's newsroom and its YouTube channel carry the event material.
Key takeaways
- Core Ultra (code name Meteor Lake) brought a built-in NPU to Intel laptop chips, built on the Intel 4 process. Laptops went on sale the same day.
- 5th Gen Xeon Scalable (code name Emerald Rapids) put AI acceleration in every core and was positioned for inference in the data centre.
- Gaudi 3 was previewed, not launched, and was aimed at Nvidia's H100 and AMD's MI300X.
- Pat Gelsinger, then Intel CEO, led the keynote.
The three announcements
| Product | Where it runs | What was new |
|---|---|---|
| Core Ultra | Thin and light laptops | CPU, GPU and NPU on one package built from tiles (chiplets), Intel 4 for the compute tile, Foveros 3D packaging |
| 5th Gen Xeon Scalable | Servers and cloud | Advanced Matrix Extensions (AMX) in every core, more cache than the 4th generation, claimed gains in AI inference per watt |
| Gaudi 3 | Data centre AI clusters | Next-generation Habana accelerator for training and inference of large models, shown on stage for 2024 |
Summarised from Intel's 14 December 2023 press release and CNBC's report of the same day. Performance claims are Intel's own and were not independently measured here.
Why the "AI everywhere" pitch mattered
Most AI spending in 2023 went to large GPU clusters for training. Intel's argument was that inference, the day-to-day running of trained models, would spread to every device, and that much of it could run on hardware customers already buy: a laptop with an NPU, or a server CPU with matrix units. That plays to Intel's installed base in PCs and servers, where it was strongest, while Gaudi was its entry into the accelerator market led by Nvidia.
For chip designers, the event was also a showcase of disaggregation. Meteor Lake was Intel's first high-volume client product built from separate tiles joined with Foveros packaging, which is a useful case study for anyone working on chiplets and advanced packaging.
How it was received
Reaction was mixed. Laptops with Core Ultra went on sale immediately, and Intel's share price rose about 1 percent on the day, as CNBC reported. Some analysts were less impressed by the AI accelerator news, because Gaudi 3 was only a preview. See our post on Stacy Rasgon's comments on Intel's AI chip.
What happened next
- Gaudi 3 launched at Intel Vision in April 2024.
- Core Ultra was followed by Lunar Lake and Arrow Lake in 2024.
- Pat Gelsinger left Intel in December 2024, and Lip-Bu Tan became CEO in March 2025.
FAQ
When was Intel's AI Everywhere event?
Thursday, 14 December 2023, in New York City.
Was Gaudi 3 launched at the event?
No. It was shown as a preview, with availability promised for 2024.
What is the difference between an NPU and AMX?
An NPU is a separate low-power AI engine on a client chip. AMX is a set of matrix instructions and hardware inside each Xeon CPU core, so AI work runs on the CPU itself.


Your comments will be moderated before it appears here.