Intel Crescent Island targets AI inference with cheaper memory, air-cooling to challenge Nvidia

Intel: Our upcoming AI chip will be cheaper, run cooler than Nvidia, AMD options

Intel plans to ship an AI chip by the end of this year that uses cheaper memory and cooling technology than rival offerings from Nvidia and AMD, as the US chipmaker seeks to capitalize on a sharp turnaround in its fortunes. Kevork Kechichian, who leads Intel's data center group, told the FT that the company is "starting with the basics" as it tries to challenge its rivals in the booming market for semiconductors that power AI. Its new "Crescent Island" graphics processing unit is designed to speed up "inference" tasks, the stage when a user makes their request, rather than the training of models, an area where Nvidia's processors are dominant. An earlier attempt at building a GPU for training AI models called "Gaudi" saw poor sales, and its planned successor was cancelled last year. "We decided to start rebuilding our muscles in AI... [but] we are not particularly aiming for [the training market] based on past experience," said Kechichian, who joined Intel last year from chip designer Arm. He added the new chip would start shipping in limited quantities to customers by the end of this year, following an 18-month development process. Intel is also looking to take advantage of two constraints encountered by Nvidia and AMD: the need to incorporate expensive high-bandwidth memory and liquid-cooling infrastructure. Crescent Island is an air-cooled chip that uses LPDDR5 memory, a significantly cheaper type of memory than the HBM used in chips such as Nvidia's Blackwell. The effort is Intel's first push into the lucrative AI infrastructure market under chief executive Lip-Bu Tan, who took over last year after Pat Gelsinger was ousted amid concerns that his turnaround strategy was failing. Intel's new GPU was first unveiled in October as part of Tan's broader effort to revive a product line-up that had allowed Nvidia to dominate the market for chips used to train models such as OpenAI's ChatGPT. Investors have welcomed the leadership change, after which Tan moved to cut costs and rein in spending on some manufacturing projects. Intel's shares are up more than 200 percent since the start of this year, part of a broader rally in semiconductor stocks driven by enthusiasm for AI. Kechichian said Intel was assessing whether a version of the chip could potentially be sold in China in compliance with US export controls. Nvidia and AMD's AI chip sales to the Asian nation have been blocked by trade tensions between Washington and Beijing. "There are tiers of [the chip] that might be OK there... and we'll confirm that over time: clearly there is demand for that particular price point in that particular market," Kechichian said. In August the US government announced it would take a 10 percent stake in the company over time, as Donald Trump's administration sought to deter the chipmaker from selling its foundry manufacturing business. Intel subsequently launched its own advanced PC and server chips built in its own factories this year, after a long period of having them made by Taiwan Semiconductor Manufacturing Company. Kechichian said Intel hoped to build its new chip in-house, another move that would ultimately make it cheaper than those offered by rivals who rely on TSMC. "For all data center products we are moving aggressively into our own foundry," he said. "That's the intent in general."

Intel details long-awaited Crescent Island AI GPU at Computex, boasts up to 480 GB of LPDDR5X to combat memory shortages -- company shares more details of its Xe3P inference accelerator at Computex

Unusual memory choice brings lots of AI data closer to the chip for efficiency At Computex 2026, Intel is offering a few more details and updates for its next-generation Data Center GPU product, code-named Crescent Island. The Crescent Island GPU will be built on Intel's Xe3P GPU architecture. Intel says this architecture is "built for agentic AI," and it supports a broad range of potential data types, from FP4 for high-performance AI inference all the way up to FP64, potentially for scientific computing applications. Intel isn't providing any raw throughput specs at this stage of Crescent Island's development, so we can't make any guesses about its compute performance. Crescent Island will be a PCI Express add-in card with a 350W power target, placing its power and thermal requirements close to products like Nvidia's RTX Pro 5000 Blackwell card. But Crescent Island's architecture is quite different from anything else on the market. It forgoes GDDR or HBM memory for LPDDR5X. Intel says its Crescent Island reference design will include 160GB of LPDDR5X, but that the chip is designed to allow partners the flexibility to build accelerators with up to 480GB of memory. Recent leaks and past analysis have suggested that Crescent Island will take a wide-and-slow approach with LPDDR5X, potentially using a 640-bit bus connecting 20 LPDDR5X devices, to achieve these high capacities. Some basic math suggests that partners would need to employ 24GB LPDDR5X modules to fully realize that memory capacity, and those modules are already available from sources like Samsung. With 10.7 Gbps LPDDR5X, Crescent Island would offer 684 GB/s of memory bandwidth. From a design standpoint, maximizing memory capacity while maintaining adequate bandwidth will help keep more AI data close to the GPU and require less data movement, potentially making Crescent Island a more efficient inference engine compared to GPUs built with lower-capacity GDDR devices. Going with LPDDR5X also doesn't put pressure on valuable advanced packaging capacity or compete with higher-end accelerators for scarce HBM, making it potentially easier for Intel to produce these accelerators economically and in volume. There's no word on how or where the Crescent Island package itself will be fabricated, however. Because Crescent Island is an air-cooled card with relatively modest power requirements, it's likely ready to drop into traditional 4U or 5U GPU servers, potentially making it appealing for companies trying to develop on-premise inferencing solutions. Eight of these accelerators with a full 480GB of RAM each would produce an impressively dense server with 3.8 TB of local GPU memory, allowing for massive models or swarms of smaller AI agents to reside within one box. Of course, orchestrating AI work across multiple GPUs requires a capable software stack to manage the entire show, and Intel touts its oneAPI stack for use with Crescent Island. oneAPI is far less widely adopted than CUDA or ROCm, but those blazing the AI inference trail on Crescent Island will find software that the company calls "open, upstreamed, and Day 0 ready" for the product. Intel describes Crescent Island as "coming soon" and has touted a second-half 2026 launch for the platform, so we'll presumably learn more about the product and the ecosystem building around it as we progress further into the year. Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Intel's attempting to break into the AI market once more, but this time avoiding Nvidia's dominance in training by going for inference

Gaudi was practically a flop. Second time's a charm with Crescent Island? Under the leadership of Pat Gelsinger, Intel tried to compete against Nvidia and AMD in the AI market with its Gaudi series of GPUs. However, with little in the way of sales, it looked like the chip giant would just give up. It turns out that this is not the case, because it's trying again, this time targeting the world of inference instead. That's according to a report by the Financial Times, which spoke with Kevork Kechichian, Intel's general manager of its data centre group. At this year's Computex event, Crescent Island was given some more details on top of those given in last year's announcement, but the general gist of it all is that it has a very different approach to the whole machine learning shebang than Intel's previous attempt, Gaudi. Marketed as AI accelerators, the previous GPUs looked good on paper and were apparently being sold at an enticing price, but Nvidia's dominance of the AI training market with its Hopper and Blackwell chips led to somewhat underwhelming sales of Gaudi. So much so that Intel cancelled its successor, Falcon Shores. Now it's trying again with Crescent Island (via Tom's Hardware), and once again, everything looks great on paper. However, Intel is being far more savvy with this new GPU, in that rather than tackling Nvidia and AMD head-on, it's looking at picking up sales for data centers that handle AI inference. Such systems run models that have already been trained; essentially, they're just processing questions and tasks that users are demanding of the model. Because of this, Crescent Island GPUs don't need to have copious amounts of super-expensive High Bandwidth Memory (HBM), and they can be simply air-cooled, rather than using complex and costly liquid systems. If the AI chip market shifts significantly towards inference, then it might help to relieve the global memory crisis somewhat, because it uses cheaper LPDDR5X memory. Nvidia is also targeting inference and has teamed up with Groq (note: not Grok) to produce a new chip that blends a language accelerator with its Rubin platform. The question is, who will get to market first? None of the big three chip giants has given any kind of official launch date for their AI inference products, with Intel just generally hinting at some point later this year. If you've been holding off on a gaming PC upgrade for an end to the RAMpocalypse, I'm afraid that ray of sunshine isn't likely to be seen for a good while yet.

Intel Crescent Island "Xe3P" GPU Scales To 480 GB of "Cost-Optimized" LPDDR5X Memory, Beating NVIDIA Rubin & AMD MI450X With Highest Capacity (June 1 11 AM Taiwan)

Intel's Crescent Island GPU, featuring the Xe3P architecture, will accelerate Agentic AI with its massive memory pool of up to 480 GB. Intel Throws Away HBM & Bundles Almost Twice The Memory As NVIDIA Rubin & More Than AMD's MI450X On Its Crescent Island "Xe3P" GPUs For Agentic AI The Intel Crescent Island GPU is based on the brand-new Xe3P architecture, which is the same graphics architecture that was teased by the company during its Panther Lake and Xe3 deep dives. The new GPU architecture will be a further upgrade over the Xe3 architecture, and for clients, the architecture will be featured on the next-gen Arc family, the Arc C-Series. But Xe3P is going to be even more scalable, from client iGPUs to data center AI GPUs. * Xe3P microarchitecture with optimized performance-per-watt * 480GB of LPDDR5X memory * Support for a broad range of data types, ideal for "tokens-as-a-service" providers and * inference use cases Intel Crescent Island will be both power- and cost-optimized. It will be targeted at air-cooled data center solutions and workstations and will be aimed at AI inference workloads. According to Intel, the Xe3P graphics architecture used for Crescent Island will be optimized for performance per watt. Competitors such as NVIDIA and AMD are offering their data center AI solutions with top-grade HBM memory, such as HBM3E, and are already shipping the first HBM4 solutions for next-gen parts such as Rubin and MI450X. But at the same time, sourcing HBM has become difficult due to increased demand, and that has also led to higher prices. Just for comparison: * Intel Crescent Island - Up To 480 GB (LPDDR5X) * AMD Instinct MI450X - Up To 432 GB (HBM4) * NVIDIA Vera Rubin - Up To 288 GB (HBM4) LPDDR5X Saves Power & Costs For Crescent Island GPUs Leveraging LPDDR5X memory can give Intel a big edge in the cost/performance segment. Furthermore, the architecture will support a broad range of data types that are ideal for "Tokens-as-a-service" providers and inference use cases. This is why Intel is going big with Crescent Island. Starting with the specifications, the Intel Crescent Island Data Center GPU will feature up to 480 GB of LPDDR5X capacity. The reference Intel PCIe card will feature 160 GB of LP5X memory, but partners will be given the freedom to build ODM-branded cards with flexible options up to 480 GB. The use of LP5X memory also cuts down power significantly, with just 350W of TDP on the air-cooled PCIe version. Intel states that LP5X memory enables a densely packed channel design, which enables a significant bandwidth increase. We have already seen initial PCIe boards with support for up to 160 GB of memory, so the partner board should be even more interesting. Another thing about Crescent Island is that traditional GPU blocks, such as graphics or 3D support. The chip solely focuses on GPGPU, and that freed more die area for additional AI compute. Plus, the chip is optimized for Perf/Watt and Perf/TCO. The GPU will be backed by a wide-range data format support from FP4 up to FP64, and will also support a fully Open and Robust software stack. Intel is already evaluating its open and unified software stack for heterogeneous AI systems with its existing Arc Pro B-series lineup, so future iterations will be able to access these optimizations early on. Intel is currently targeting customer sampling for its Crescent Island GPU for the 2H of 2026, so we'll definitely learn more about the GPU in the coming months. Follow Wccftech on Google to get more of our news coverage in your feeds.