Intel Unveils Crescent Island: A New AI-Focused GPU for Data Centers

Intel Announces "Crescent Island" Inference-Optimized Xe3P Graphics Card With 160GB vRAM

Back during the Intel Tech Tour in Arizona, Intel teased a new inference-optimized enterprise GPU would be announced soon. This new product would feature enhanced memory, bandwidth, and enterprise-level AI inference capabilities. Today the embargo expires on talking about this new GPU offering. When Intel was teasing this new inference-optimized GPU a few weeks back in Arizona it sounded like Intel may have had an unexpected trick up its sleeves. What's being announced today is indeed a new enterprise GPU for AI that is interesting from a technology perspective, but it's not shipping until at least H2'2026. So while there was hope that perhaps Intel had managed to innovate some interesting Battlemage / BMG-G31 part for AI or the like with lots of vRAM, what's being announced is a next-gen part but one that is at least one year away still. This new graphics card is codenamed Crescent Island and is built on their next-gen Xe3P Celestial micro-architecture. Xe3P will be optimized around performance-per-Watt and Crescent Island will feature 160GB of LPDDR5x memory to allow for plenty of space for large language models (LLMs). Intel's embargoed announcement also notes that Crescent Island will feature support for a variety of different data types and be an "ideal" solution for tokens-as-a-service providers and inference use cases. In addition to being optimized around performance-per-Watt, Crescent Island will also be air cooled and cost-optimized. Intel is currently working on refining their open-source software stack for Crescent Island via using current-generation Arc Pro B-Series GPUs. Intel's announcement notes that customer sampling of this new data center GPU will begin in the second half of 2026. No official release timeframe was provided if they also hope to squeeze it out next year or if (more than likely) it will actually ship more broadly in 2027 but just noting their customer sampling for H2'2026 in the embargoed news release. No slides or prototype images or anything else to share today on Intel's Crescent Island. Long story short, Intel is announcing Crescent Island today as a Xe3LP + 160GB LPDDR5X offering for H2'2026 or later that will be AI inference optimized around power efficiency and cost. It sounds interesting but technical details beyond those basics were light and it's going to be a long while before we see Crescent Island. Given the timing this will be going up against the AMD Instinct MI450 series and NVIDIA Vera Rubin. It seems like Intel wanted to have something to announce now given the ongoing AI rush albeit not many details today and no short term AI solution. At least this does lead to more weight for the ongoing Project Battlematrix Linux driver improvements and other ongoing Intel Compute Runtime and Intel Xe Linux driver enhancements that are currently ongoing for the Arc Pro B-Series. With confirming Crescent Island now it also opens the door to them beginning to push open-source hardware enablement patches without otherwise spilling the beans on this forthcoming enterprise AI product. Intel is using the OCP Global Summit to announce some additional Gaudi 3 rack-scale reference designs. These new Gaudi 3 rack-scale reference designs will allow up to 64 accelerators per rack with liquid cooling and 8.2TB of high bandwidth memory. Intel hasn't aggressively promoted Gaudi 3 in recent quarters after its launch last year. Gaudi 3 has enjoyed some reprieve since Intel canceled their Falcon Shores AI accelerator chip but still appears to be the end of the road for Gaudi especially with Jaguar Shores still expected and now Crescent Island too. The Gaudi 3 software support has been neglected over the past year with losing multiple rounds of the Habana Labs Linux driver maintainers and only recently seeing new activity to return to working on this AI accelerator Linux driver albeit as of writing for Linux 6.18 there still is no mainline kernel driver support for Gaudi 3.

Intel customers to test new Crescent Island GPU in second half of next year

Oct 14 (Reuters) - Intel (INTC.O), opens new tab said on Tuesday its data center graphics processing unit, code-named Crescent Island, was scheduled for customer sampling in the second half of 2026, signaling a renewed push to secure a spot in the fast-growing AI chip race. This marks the chipmaker's second attempt to break into the AI accelerator arena, after its earlier Gaudi chips failed to gain traction against market leaders Nvidia (NVDA.O), opens new tab and AMD (AMD.O), opens new tab. Intel's GPU roadmap trails behind competitors, highlighting the significant challenges it faces to establish itself as a serious contender in AI computing. Since the generative AI boom with the launch of OpenAI's ChatGPT in November 2022, startups and large cloud operators have rushed to grab GPUs that help run AI workloads on data center servers. The demand explosion has led to supply crunch and sky-high prices for AI-focused chips. Intel has said it would launch AI data chips on an annual basis, signaling a shift from irregular launch gaps to more predictable yearly updates. "Instead of trying to build for every workload out there, our focus is increasingly going to be on inference," said Chief Technology Officer Sachin Katti, referring to the phase of AI computation where models generate answers rather than train on data. The company argues that an open and modular approach in which customers can mix and match chips from different vendors is essential to scale AI systems efficiently, even as it struggles to catch up with Nvidia's dominance in AI computing. Nvidia said last month it would invest $5 billion, opens new tab in Intel, taking a roughly 4% stake and becoming one of its largest shareholders as part of a partnership to co-develop future PC and data center chips. Reporting by Akash Sriram in Bengaluru and Max Cherney in San Francisco; Editing by Shilpi Majumdar Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Technology Max A. Cherney Thomson Reuters Max A. Cherney is a correspondent for Reuters based in San Francisco, where he reports on the semiconductor industry and artificial intelligence. He joined Reuters in 2023 and has previously worked for Barron's magazine and its sister publication, MarketWatch. Cherney graduated from Trent University with a degree in history.

The plot thickens as Intel announces a new data center GPU that could also preview its next-gen gaming graphics cards, but there's still no mention of Celestial

Just as we're deep in thought pondering the true character of the Xe3 iGPU in Intel's new Panther Lake laptop chip and what the heck the "Xe3P" thing is that popped up in the marketing slides, here comes a left-field announcement. Intel will launch a new GPU later next year that's purpose built for AI inferencing. And guess what? It's using that Xe3P graphics architecture. Specs-wise, other than the simple fact that the new GPU, codenamed Crescent Island, is using that next-gen Xe3P technology, there's little to go on. The only other detail is 160 GB of LPDDR5X memory. Actually, that's a relatively revealing spec as it implies this GPU will be a value play, at least in the context of megabucks AI chips. Were it aiming at top-tier AI performance, it would more likely use something along the lines of HBM or High Bandwidth Memory over a huge memory bus. Notably, there's no mention at all of Celestial, the graphics architecture that was supposed to follow Intel's current Battlemage tech, which was previously presented under the Xe3 brand and was equally absent from the Panther Lake launch, despite that chip getting Xe3 graphics. Anyway, Intel says that Crescent Island won't be out until the second half of next year, so this is an uncharacteristically early mention for a product so far off in the future. However, given that Intel doesn't really have an existing product in this part of the market, if you assume that the existing Intel Arc Pro B-Series GPUs are probably much lower performing, there's no spoiler effect of announcing Crescent Island now. And as Phoronix suggests here, confirming Crescent Island now allows Intel to keep working on its software support for professional GPUs, without the need to worry that some or other patch that mentions the new GPU is spotted and causes speculation. Of course, the speculation that won't stop is the question of how Crescent Island might relate to future Intel gaming GPUs. It absolutely isn't hyperbole to suggest a possible link. After all, Intel has said Crescent Island is based on Xe3P and also placed Xe3P on the roadmap as the graphics architecture for PCs that will follow the new Xe3 technology in Panther Lake. In truth, we just can't know for now. It's possible Crescent Island could be just like an Nvidia RTX 5090 -- a powerful GPU that can turn its hand to almost anything, including AI inferencing and gaming. Or it could be much more purpose-built and not even able to run a full 3D pipeline, again like Nvidia's GB200 AI megachip, albeit probably smaller and less expensive. After all, GB200 is Blackwell, just like every RTX 50 graphics card. It's just that version of the Blackwell architecture lacks features like texture and ray-tracing units. So, it's hard to say with confidence what Xe3P will be like in data centers as opposed to in the PC. All that said, there is at least one reason to hope that Crescent Island could also do duty as a gaming GPU or at least signals that Intel is keeping faith with high-performance PC graphics. Xe3P has now been attached to both this new Crescent Island GPU and Intel's next generation of PC graphics, albeit the latter in only the vaguest terms. Now, given Intel's broader problems at the moment, one might suggest that it's unlikely the chip maker has enough resource to design lots of very bespoke graphics architectures. It might be more reasonable to expect that Intel is going for a more unified approach. And if that's the case, building a powerful GPU for inferencing could result in a ready-baked and very capable gaming chip. Or maybe not, we'll have to see. But we are allowed to hope. And we'll make do with this tidbit Intel has thrown us, for now. And also conclude that whatever Intel's Celestial graphics tech is or was, Intel has decided to ditch the name, at the very least.

Intel takes another crack at the AI chip market with its new Crescent Island GPUs - SiliconANGLE

Intel takes another crack at the AI chip market with its new Crescent Island GPUs Intel Corp. has announced a new artificial intelligence processor aimed at inference workloads, as it renews its efforts to break into the lucrative market for AI accelerators. The new graphics processing unit, codenamed Crescent Island, is optimized for energy efficiency and will be able to support a wide range of AI applications, said Intel Chief Technology Office Sachin Katti. His comments came at the Open Compute Summit today, where he announced the new chip. "It emphasizes that focus that I talked about earlier, inference, optimized for AI, optimized for delivering the best token economics out there, the best performance per dollar out there," Katti said. The Crescent Island GPU is based on Intel's new Xe3P architecture, which was announced just one week earlier at a special event where it revealed its upcoming Panther Lake Intel Core Ultra series 3 processors for laptops. Xe3P is an upgrade over Intel's existing Xe3 architecture, optimized for both power and cost, designed for air-cooled data centers and targeted at AI inference workloads, the company said. It added that the architecture will be able to support a broad range of data types, making it ideal for "token-as-a-service" providers and inference applications. The new GPU will feature a massive 160 gigabytes of memory capacity, but interestingly Intel has decided to base this on the LPDDR5X standard, rather than the higher-end High-Bandwidth Memory or HBM that's used in the GPUs of rivals such as Nvidia Corp. and Advanced Micro Devices Inc. Those rivals currently utilize HBM3E in their most current chips, and are already talking about upgrading to HBM4 for their next-generation processors, such as Vera Rubin and the MI400. But Intel's choice reflects the difficulties those companies have in sourcing HBM memory chips, which have been in short supply due to the increased demand. HBM prices have also skyrocketed in recent months. By leveraging LPDDR5X memory in its new GPUs instead, Intel may well gain an edge in terms of cost efficiency. Intel has not yet released any detailed specifications, but said it's currently targeting customer sampling for Crescent Island for the second half of 2026, so we can expect more details to emerge in the coming months. What is clear now is that Crescent Island is going to be a key part of its renewed attempt to capitalize on an AI frenzy that has already generated billions of dollars in revenue for Nvidia and other chipmakers. The company completely missed the boat on generative AI and has fallen far behind its competitors, and faces a huge struggle to capture a meaningful share of the market for AI accelerators, but it's not something it can just ignore. Indeed, Intel Chief Executive Lip-Bu Tan, who took over the chipmaker earlier this year, has vowed to revitalize its fortunes in the AI market. He's taken some drastic steps already, effectively mothballing the company's earlier efforts, such as the Gaudi GPUs. At the OCP Summit, Katti said Intel plans to release new versions of the GPUs every year, matching the annual release cadence of Nvidia and AMD, as well as cloud infrastructure providers such as Amazon Web Services Inc. and Google LLC, which design their own processors for AI. Intel faces a tough task in dislodging Nvidia from its dominant position in the AI chip market, but Katti said the company will focus its efforts on creating chips specifically for AI inference, which means the silicon that runs AI models in production, as opposed to training AI models. "Instead of trying to build for every workload out there, our focus is increasingly going to be on inference," he stated. Interestingly, Nvidia itself could yet help boost Intel's fortunes in the AI market, since it now has a vested interest in doing so after investing $5 billion into its ailing rival last month. That deal saw Nvidia take a roughly 4% stake in its rival, meaning it has become one of its largest shareholders alongside the U.S. government. The two companies have also agreed on a partnership that will see them co-develop chips for data center servers and personal computers. Speaking about that deal, Katti said it will help to ensure that Intel's central processing units will be installed in every AI server and PC that's sold in future.

Intel Unveils Xe3P GPU for Data Centers with Up to 160GB LPDDR5x Memory

Intel has shared new details about its next big data center GPU, code-named Crescent Island, and it's bringing something new to the table: the Xe3P architecture. Announced at the OCP Global Summit, Crescent Island will be Intel's first GPU to use this design, which will later show up in Celestial and Nova Lake-S products. It's built to handle modern AI inference workloads, offering better performance and power efficiency compared to the current Battlemage-based Xe3 architecture used in Panther Lake. What really stands out about Crescent Island is its memory setup. Intel says the GPU will support up to 160GB of LPDDR5x memory -- a big shift from the usual GDDR6 or GDDR7 that most GPUs rely on. LPDDR5x isn't as fast in raw bandwidth terms, but it's far more efficient and compact, making it a great fit for servers where power use and thermal limits matter more than maximum frame rates. This approach isn't entirely new for Intel; its early Xe DG1 graphics card also used low-power LPDDR memory before the Arc series launched. The GPU is being designed with flexibility in mind. It supports multiple data formats like FP8, INT8, and BF16, which are common in AI and machine learning tasks. This versatility makes it ideal for inference workloads or even Tokens-as-a-Service (TaaS) models that need to process large streams of generative AI data quickly and efficiently. Combined with the Xe3P architecture's improved compute density, Crescent Island aims to deliver stronger AI inference performance at lower power costs. Intel has made it clear that Crescent Island is mainly aimed at inference, not full-scale AI training, which differentiates it from GPUs like NVIDIA's Blackwell series. Instead, it's meant to work alongside CPUs and NPUs in Intel's broader data center ecosystem, forming part of a balanced compute platform. The company expects to start shipping samples in the second half of 2026, suggesting a full launch sometime in 2027. Crescent Island also represents an important step for Intel's GPU roadmap as the company unifies its architectures across desktop, mobile, and data center platforms. With Xe3P, Intel wants to streamline software development and optimize performance across multiple environments, from consumer PCs to high-performance inference clusters. Source: intel

Intel's new Crescent Island GPU is designed to take on Nvidia and AMD in AI

The card features up to 160 GB of LPDDR5X memory to handle large language models and heavy inference workloads. Intel has announced its Crescent Island graphics card, an inference-optimized GPU based on the Xe3P architecture. The device is designed to accelerate artificial intelligence workloads and is scheduled to begin shipping in the second half of 2026. The Crescent Island GPU utilizes the next-generation Xe3P Celestial micro-architecture, a design heavily optimized around performance-per-watt to maximize efficiency. The card will be equipped with up to 160 GB of LPDDR5X VRAM. This large memory capacity provides substantial scope for supporting large language models and other memory-intensive AI inference tasks. The focus on efficiency aims to extract maximum value from the AI workloads the GPU will process. The emphasis on performance-per-watt extends to the GPU's physical design and operational requirements. According to a report from Phoronix, the Crescent Island cards will be air-cooled. This cooling method contrasts with more resource-intensive liquid cooling systems, contributing to a reduction in both power and water usage for data center operations. To support the hardware at launch, Intel is also actively developing an optimized open-source software stack that will be available when the GPU debuts. Intel's announcement enters a market where Nvidia's GPUs, particularly the latest Blackwell generation, have been a primary choice for building powerful AI super-clusters for training models. While Nvidia maintains a strong position in the training sector, the market for inference hardware is becoming more diversified. A number of companies are now developing leaner and more efficient hardware, including both GPUs and ASICs, creating a competitive space for inference-specific solutions. The late 2026 release schedule places Crescent Island in direct competition with next-generation products from other major hardware manufacturers. Nvidia is set to launch its Vera Rubin architecture in the second half of 2026, which is anticipated to deliver significant performance and efficiency improvements. Concurrently, AMD has announced its MI450 range of GPUs. Projections suggest this series will offer capabilities that are comparable to, or potentially greater than, those of the Vera Rubin architecture.

Intel Bets Big On AI Revival With New 'Crescent Island' Chip Built To Optimize 'Performance Per Dollar' - Intel (NASDAQ:INTC)

Intel Corp. (NASDAQ:INTC) is making a renewed push into the artificial intelligence chip market with the launch of its new "Crescent Island" GPU, a data center processor set to debut next year. Intel Unveils Crescent Island GPU To Boost AI Inference Efficiency On Tuesday at the Open Compute Summit in San Jose, Intel announced that the Crescent Island GPU is designed for inference workloads. Intel CTO Sachin Katti said the chip is built to optimize "performance per dollar" and deliver "the best token economics," reported Reuters. The chip will feature 160 gigabytes of slower memory instead of high-bandwidth memory used by rivals Nvidia Corp. (NASDAQ:NVDA) and Advanced Micro Devices Inc. (NASDAQ:AMD). Katti added that Intel aims to release new AI data center GPUs every year to keep pace with its competitors and cloud providers developing in-house AI chips. "Instead of trying to build for every workload, our focus is increasingly going to be on inference," Katti said. See Also: Former Intel CEO Pat Gelsinger Slams CHIPS Act Rollout, Says Trump Administration's Stake Only Matters If It Builds And Fills Fabs Nvidia's $5 Billion Vote of Confidence The announcement comes weeks after Nvidia disclosed a $5 billion investment in Intel, giving it a roughly 4% stake and making it one of Intel's largest shareholders. Intel CEO Lip-Bu Tan has vowed to restart the company's stalled AI efforts after it paused earlier projects such as the Gaudi and Falcon Shores chips. Bank Of America Downgrades Intel To Underperform Earlier this month, Bank of America Securities analyst Vivek Arya downgraded Intel to Underperform with a $34 price target, citing weak competitiveness in AI and server CPUs. Arya noted that Intel's market cap recovery reflects improved finances and foundry potential, not real product traction. Intel shares also traded lower Tuesday as broader semiconductor stocks fell amid intensifying U.S.-China trade tensions and new tariff threats. Price Action: Intel shares fell 4.27% on Tuesday but remain up over 76% year to date, according to Benzinga Pro. Intel ranks among the top stocks in Benzinga's Edge Stock Rankings for Momentum, reflecting strong price performance across short, medium and long-term periods. A full analysis, including comparisons with peers and competitors, is available here. Read More: Nvidia CEO Jensen Huang Says Intel Spent 33 Years 'Trying To Kill Us' But Now Calls The Chip Rival A Partner: 'We're Lovers, Not Fighters' Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Photo: Tada Images / Shutterstock INTCIntel Corp$35.71-4.06%OverviewAMDAdvanced Micro Devices Inc$218.961.17%NVDANVIDIA Corp$181.21-3.78%Market News and Data brought to you by Benzinga APIs

Intel Reveals 160-GB, Energy-Efficient Inference GPU As Part Of New Yearly Cadence

Revealed this week at the 2025 OCP Global Summit, 'Crescent Island' marks the beginning of Intel's annual cadence of GPU releases that follows similar pushes by Nvidia and AMD after dealing with more than 15 years of missteps in the accelerator chip segment. Intel on Tuesday revealed a 160-GB, energy-efficient data center GPU that is part of a new annual GPU release cadence to deliver on the chipmaker's new strategy of providing open systems and software architecture for AI systems. Code-named "Crescent Island," the GPU is "power- and cost-optimized" for inference workloads running on air-cooled enterprise servers, according to Intel. The GPU features Intel's Xe3P microarchitecture that is optimized for performance-per-watt, 160 GB of LPDDR5X memory and support for a broad range of data types. [Related: Analysis: How Two Big Decisions Helped AMD Win The OpenAI Deal] The semiconductor giant said it plans to start sampling Crescent Island with customers in the second half of 2026. Until then, Intel is developing and testing what it's calling an open and unified software stack for heterogeneous AI systems on the company's Arc Pro B-Series GPUs to "enable early optimizations and iterations." The announcement confirms CRN's previous reporting last month that Intel had an unannounced GPU design with a lower power requirement for servers on its road map that could arrive next year at some point. Intel did not provide an update on "Jaguar Shores," the next-generation GPU announced earlier this year that the company is designing for rack-scale platforms. In a briefing with journalists last month, Sachin Katti, Intel's chief AI and technology officer, said that Crescent Island has "enhanced memory bandwidth" and "lots of memory capacity," making it a "fantastic product for token clouds and enterprise-level inference." Revealed this week at the 2025 OCP Global Summit, Crescent Island marks the beginning of a new annual cadence of GPU releases that Intel announced last week during its major press push around its forthcoming "Panther Lake" and "Clearwater Forest" CPUs. Following similar pushes by Nvidia and AMD over the last two years to move to annual release strategies, this represents Intel's latest push to gain a foothold in the Nvidia-dominated AI infrastructure market after more than 15 years of missteps in the accelerator chip segment that spanned across the company's last four CEOs. Katti, who was appointed by Intel CEO Lip-Bu Tan in April to lead the company's new AI strategy, said the chipmaker is building its new vision for the AI hardware market around an open systems and software architecture that will deliver the "right-sized" and "right-priced" compute needed to power future agentic AI workloads. "We will be building scalable heterogeneous systems that deliver that zero friction experience to agentic AI workloads and can deliver on the best performance-per-dollar for these workloads by leveraging this open heterogeneous architecture," he said. Katti (pictured above) said this open approach will give customers and partners "choice at the systems and the hardware layer," with opportunities for multiple vendors to partake. "As we bring more and more disruptive technologies to the table, we can insert it into this open, heterogeneous architecture," he said last month.

Intel Crescent Island GPU Unveiled: Features Next-Gen Xe3P Graphics Architecture, 160 GB LPDDR5X For AI Inference

Intel has announced its brand new AI inference GPU solution for data centers, codenamed Crescent Island, which features the Xe3P architecture. After falling short of its AI-centric goals on the data center front, Intel is now revising its strategy a bit. The company seems to be going the efficiency and higher memory capacity route with a brand new offering that we have never heard of before, and it's called Crescent Island. This is the same GPU that was teased at the Intel Tech Tour 2025. 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 last week during its Panther Lake and Xe3 deep dives. The new architecture will be a further upgrade over the Xe3 architecture, and for clients, the architecture will be featured on a 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. The new data center GPU code-named Crescent Island is being designed to be power and cost-optimized for air-cooled enterprise servers and to incorporate large amounts of memory capacity and bandwidth, optimized for inference workflows. Key features include: Intel's open and unified software stack for heterogeneous AI systems is currently being developed and tested on Arc Pro B-Series GPUs to enable early optimizations and iterations. Customer sampling of the new data center GPU code-named Crescent Island is expected in the second half of 2026. via Intel Intel Crescent Island will be both power- and cost-optimized. It will be targeted at air-cooled data center solutions 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. The card itself will feature a massive 160 GB memory capacity based on the LPDDR5X standard. Interestingly, Intel is going with LP5X. Competitors such as NVIDIA and AMD are offering their data center AI solutions with top-grade HBM memory, such as HBM3E, and already talking about HBM4 for next-gen parts such as Rubin and MI400. But at the same time, sourcing HBM has become difficult due to increased demand, and that has also led to higher prices. 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. 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.

Intel Takes Another Stab at AI GPUs | The Motley Fool

Intel will focus on artificial intelligence (AI) inference with its new Crescent Island GPU. Intel (INTC 4.24%) has largely missed out on the booming market for AI accelerators. The company's Gaudi line of AI chips, which came from its acquisition of Habana Labs in 2019, failed to gain any real traction, partly due to immature software and an unfamiliar architecture. Falcon Shores, which was meant to be a more traditional data center AI GPU, was scrapped as a commercial product earlier this year. While Intel is likely too late to compete in the market for AI accelerators used for training models, AI inference is another story. While AI training workloads require massive computational horsepower and data throughput, AI inference workloads aren't nearly as demanding. As AI models and AI agents are deployed to tackle a growing number of use cases, the need for efficient and affordable AI inference chips is on the rise. Intel announced a brand-new AI GPU on Tuesday at the 2025 OCP Global Summit. Unlike its previous efforts, Intel's upcoming GPU will be solely focused on AI inference. "AI is shifting from static training to real-time, everywhere inference -- driven by agentic AI," said Intel CTO Sachin Katti. "Intel's Xe architecture data center GPU will provide the efficient headroom customers need -- and more value -- as token volumes surge." Code-named Crescent Island, Intel's new AI GPU will be based on the Xe3P architecture, which is an enhancement of the Xe3 architecture that will be used in the company's Panther Lake PC CPUs. The GPUs will be optimized for performance-per-watt and feature 160GB of LPDDR5X memory. Intel says that Crescent Island is ideal for "tokens-as-a-service" providers that charge by the token for AI models, as well as for general AI inference workloads. Crescent Island is still in the works, and the company doesn't expect to sample the product with customers until the second half of 2026. While Nvidia dominates the market for GPUs used for AI training, the AI inference market is more competitive because the latest-and-greatest chips aren't necessarily required. For many agentic AI applications, smaller models that are cheap and fast to run are useful for certain types of tasks. Performance is still important for the GPU running these models, but efficiency is as well. Cloudflare represents an example of this concept in action. The company offers AI inference services through its Workers platform, and because it focuses solely on inference and smaller AI models, it can get away with using older GPUs that are less expensive to install. By 2030, the AI inference market is expected to more than double in size to over $250 billion, according to estimates from MarketsandMarkets. While Intel can't compete with Nvidia in the AI training market, it still has an opportunity to emerge as a winner in the AI inference market by focusing on efficiency. For AI inference providers that charge by the token, GPUs that maximize performance per watt should be appealing as a way to bring down costs. While Intel's new AI GPU looks promising, a lot could change over the next 18 months. Sampling to customers is still about a year away, so real revenue probably isn't coming until 2027. Given how fast the AI industry is evolving, and the risk that a bubble may be brewing in the AI infrastructure market, Intel could end up late for the party once again.

Intel customers to test new Crescent Island GPU in second half of next year

(Reuters) -Intel said on Tuesday its data center graphics processing unit, code-named Crescent Island, was scheduled for customer sampling in the second half of 2026, signaling a renewed push to secure a spot in the fast-growing AI chip race. This marks the chipmaker's second attempt to break into the AI accelerator arena, after its earlier Gaudi chips failed to gain traction against market leaders Nvidia and AMD. Intel's GPU roadmap trails behind competitors, highlighting the significant challenges it faces to establish itself as a serious contender in AI computing. Since the generative AI boom with the launch of OpenAI's ChatGPT in November 2022, startups and large cloud operators have rushed to grab GPUs that help run AI workloads on data center servers. The demand explosion has led to supply crunch and sky-high prices for AI-focused chips. Intel has said it would launch AI data chips on an annual basis, signaling a shift from irregular launch gaps to more predictable yearly updates. "Instead of trying to build for every workload out there, our focus is increasingly going to be on inference," said Chief Technology Officer Sachin Katti, referring to the phase of AI computation where models generate answers rather than train on data. The company argues that an open and modular approach in which customers can mix and match chips from different vendors is essential to scale AI systems efficiently, even as it struggles to catch up with Nvidia's dominance in AI computing. Nvidia said last month it would invest $5 billion in Intel, taking a roughly 4% stake and becoming one of its largest shareholders as part of a partnership to co-develop future PC and data center chips. (Reporting by Akash Sriram in Bengaluru and Max Cherney in San Francisco; Editing by Shilpi Majumdar)

Intel : to Expand AI Accelerator Portfolio with New GPU

What's New:Today at the 2025 OCP Global Summit, Intel announced a key addition to its AI accelerator portfolio, a new Intel Data Center GPU code-named Crescent Island is designed to meet the growing demands of AI inference workloads and will offer high memory capacity and energy-efficient performance. "AI is shifting from static training to real-time, everywhere inference-driven by agentic AI," said Sachin Katti, CTO of Intel. "Scaling these complex workloads requires heterogeneous systems that match the right silicon to the right task, powered by an open software stack. Intel's Xe architecture data center GPU will provide the efficient headroom customers need -and more value-as token volumes surge." Why It Matters: As inference becomes the dominant AI workload, success depends on more than powerful chips-it requires systems-level innovation. From hardware to orchestration, inference needs a workload-centric, open approach that integrates diverse compute types with an open, developer-first software stack-delivered as systems that are easy to deploy and scale. Intel is positioned to deliver this end-to-end - from the AI PC to the data center and industrial edge-with solutions built on Intel Xeon 6 processors and Intel GPUs. By co-designing systems for performance, energy efficiency, and developer continuity-and collaborating with communities like the Open Compute Project (OCP)-Intel is enabling AI inference to run everywhere it's needed most. About the GPU: The new data center GPU code-named Crescent Island is being designed to be power and cost-optimized for air-cooled enterprise servers and to incorporate large amounts of memory capacity and bandwidth, optimized for inference workflows. Key features include: * Xe3P microarchitecture with optimized performance-per-watt * 160GB of LPDDR5X memory * Support for a broad range of data types, ideal for "tokens-as-a-service" providers and inference use cases Intel's open and unified software stack for heterogeneous AI systems is currently being developed and tested on Arc Pro B-Series GPUs to enable early optimizations and iterations. Customer sampling of the new data center GPU code-named Crescent Island is expected in the second half of 2026. What's Next: Discover more at the Intel booth on the show floor in the Expo Hall #B3.

Crescent Island: Intel's bold bet to rival Nvidia in AI chips

Intel has officially thrown down the gauntlet in the AI accelerator race. With its newly announced Crescent Island GPU, the chipmaker is positioning itself as a serious contender in the growing market for inference-focused AI hardware - a segment currently dominated by Nvidia. Announced on October 14, 2025, Crescent Island represents Intel's most ambitious push yet to reclaim relevance in the data center AI space and to build a unified ecosystem that spans CPUs, GPUs, and AI accelerators. Also read: ChatGPT's flirting with the future when it stops being just an AI assistant: Here's why Unlike GPUs optimized for training massive AI models, Crescent Island is designed squarely for AI inference, the process of running trained models efficiently at scale. Built on Intel's Xe3P microarchitecture, the chip delivers high performance per watt and is tailored to handle the growing demands of real-time inference tasks like generative AI chatbots, video synthesis, and edge-based analytics. Crescent Island features 160 GB of LPDDR5X memory, providing the capacity and bandwidth needed for large language models (LLMs) and multimodal workloads that rely heavily on memory throughput. Intel emphasizes that this GPU will excel in "tokens-as-a-service" scenarios, where inference speed and energy efficiency matter more than raw training horsepower. In a deliberate move away from the increasingly power-hungry designs of its competitors, Crescent Island is air-cooled, optimized for deployment in standard enterprise servers. This design choice could significantly reduce total cost of ownership for data centers that want AI acceleration without overhauling their infrastructure for liquid cooling. Also read: Panther Lake: 2026 Intel laptops to have faster GPU, better AI and battery Intel claims Crescent Island will deliver industry-leading performance-per-watt, a critical metric as data centers face mounting energy and environmental pressures. Inference workloads, which now consume a majority of AI compute cycles globally, demand hardware that can scale sustainably and Intel wants to be the provider of that efficiency. Crescent Island isn't an isolated product; it's a cornerstone of Intel's broader AI strategy. The company is building what it calls a "unified AI platform", integrating CPUs, GPUs, and accelerators under a single software and development stack. This approach leverages Intel's open-source AI ecosystem, including tools like OpenVINO and frameworks for Xeon and Arc Pro GPUs, to make it easier for developers to deploy and optimize models across Intel hardware. Customer sampling for Crescent Island is expected in the second half of 2026, giving software partners time to fine-tune compatibility and performance. Intel is also working with members of the Open Compute Project (OCP) to standardize deployment frameworks for open AI infrastructure, a move that could make its chips more appealing to cloud providers wary of proprietary ecosystems like Nvidia's CUDA. Nvidia's dominance in AI is built on a decade of software maturity, ecosystem control, and hardware leadership. Its new Blackwell architecture GPUs promise record-breaking throughput and efficiency, setting a high bar for any competitor. Intel, however, is betting on a different path, open software, broader accessibility, and energy-conscious design. By targeting inference workloads, where flexibility, latency, and cost are paramount, Intel aims to find a foothold that Nvidia's training-centric solutions have not fully optimized for. If Crescent Island can deliver on its claims, it could appeal to enterprises running large inference deployments or hosting smaller, distributed AI models in edge and private cloud environments. Intel's comeback in the AI chip race won't be easy. The company is still recovering from years of delayed product launches and lost ground in the data center segment. But Crescent Island, with its focus on open ecosystems, energy efficiency, and inference-first performance, signals that Intel is finally playing to its strengths - integration, scalability, and cost-efficiency. Whether this is enough to challenge Nvidia's grip on the AI market remains to be seen. But for the first time in years, Intel's AI roadmap looks cohesive and ambitious, and Crescent Island could be the inflection point that defines the company's next chapter.