Intel Cancels Falcon Shores AI Chip, Pivots to Rack-Scale Solutions

Intel Just Gutted Its AI Chip Ambitions | The Motley Fool

Intel (INTC -2.90%) has largely failed to tap into soaring demand for artificial intelligence (AI) accelerators. The company's acquisition of AI chip start-up Habana Labs back in 2019, three years before the AI boom kicked off in earnest, certainly seemed like a prescient move. Habana's Gaudi family of AI accelerators, which were tailor-made for AI workloads compared to more general-purpose GPUs, had all the makings of a big winner. Intel is now on the third generation of its Gaudi AI chips, and progress has been excruciatingly slow. The company set a goal of selling $500 million worth of AI accelerators last year, a target that it ended up missing. Meanwhile, Nvidia is churning up tens of billions of dollars in AI chip revenue each quarter, and AMD has guided for about $5 billion in AI chip sales for 2024. One big problem for Intel has been software. Gaudi is not a GPU, so the architecture is very different from the AI accelerators that currently dominate the market. Despite aggressive pricing from Intel, an immature software ecosystem has been holding Gaudi back. Intel has had some wins, including a deal with IBM to put Gaudi 3 chips in IBM's cloud data centers. Unfortunately for Intel, these wins just haven't been big enough to really move the needle. Intel's AI accelerator roadmap has been a bit scattered for quite a while. Until last week, the plan was to launch Falcon Shores in late 2025. Falcon Shores is a more traditional GPU that was expected to incorporate some features from the Gaudi family of chips. The planned end of the Gaudi chip family and the switch to an entirely new architecture may have been factors that kept potential Gaudi customers away. During its fourth-quarter earnings call on Thursday, Intel disclosed that Falcon Shores was being scrapped as a commercial product. The new plan is to use Falcon Shores as an internal test chip as the company focuses on rack-level AI solutions built around Jaguar Shores, which was supposed to be Falcon Shores' successor. This effectively marks the end of Intel's efforts to become a major supplier of AI accelerator chips. Instead, the company is pivoting to broader AI solutions for data centers. Jaguar Shores can be paired with the company's Xeon server CPUs and other Intel technologies, a strategy that certainly makes sense, given the difficulties Intel has had selling its Gaudi chips. "Falcon Shores will help us in that process of working on the system, networking memory, all the component functions of that, but what customers really want is that full-scale rack solution. And so, we're able to get to that with Jaguar Shores," said Intel interim co-CEO Michelle Holthaus. Jaguar Shores isn't expected to launch until sometime in 2026, and Intel's shift away from focusing on stand-alone AI accelerators may further depress interest in the company's Gaudi family of chips. 2024 was a disappointing year for Intel's AI business, and 2025 isn't looking much better. In the long run, Intel's new focus on AI systems rather than AI chips might be the best way for the company to finally break into the data center AI market. Gaudi certainly had potential, but the company bungled its opportunity to become a major player in the AI accelerator market. Intel can still turn itself around, but AI chips won't be a factor. Instead, the company will need to rely on its foundry business and its upcoming PC and server CPUs to carry it to the finish line.

Intel won't bring its Falcon Shores AI chip to market | TechCrunch

Intel is effectively killing Falcon Shores, its next-generation GPU for high-performance computing and AI workloads. The move comes as Intel tries to correct course after a number of disappointing product launches and historic losses, while competitors like AMD and Nvidia gain ground. Intel co-CEO Michelle Johnston Holthaus said during the company's fourth-quarter earnings call Thursday, the company will "leverage Falcon Shores as an internal test chip, without bringing it to market." "AI data center [...] is an attractive market for us," Holthaus said during an earnings call Thursday. "[B]ut I am not happy with where we are today. We're not yet participating in the cloud-based AI data center market in a meaningful way [...] One of the immediate actions I have taken is to simplify our roadmap and concentrate our resources." The focus instead will be on Jaguar Shores, which Holthaus called Intel's opportunity to "develop a system-level solution at rack scale [...] to address the AI data center more broadly." Holthaus tempered expectations for Falcon Shores last month, when she implied that it was an "iterative" step over the company's previous dedicated AI data center chip, Gaudi 3. "We really need to think about how we go from Gaudi to our first generation of Falcon Shores, which is a GPU," Holthaus said in an appearance at the Barclays Global Technology Conference. "And I'll tell you right now, is it going to be wonderful? No, but it is a good first step." Gaudi 3 is was widely seen as a miss for Intel. In November, the company said it would be unable to meet its goal of $500 million in Gaudi 3 sales due to software issues. Today, few major service providers beyond IBM have committed to using the chip. Intel faces an uphill battle in the AI data center chip space. Rival AMD expects to make around $7 billion in AI chip revenue in 2025, while Nvidia, the AI chip firm to beat, could hit $195 billion in revenue in fiscal 2026, according to some analysts. "As I think about our AI opportunity, my focus is on the problems our customers are trying to solve, most notably the need to lower the cost and increase the efficiency of compute," Holthaus said. "As such, a one-size-fits-all approach will not work, and I can see clear opportunities to leverage our core assets in new ways to drive the most compelling total cost of ownership across the continuum."

Intel Cancels Falcon Shores AI Chip To Focus On 'Rack-Scale Solution'

Interim Intel co-CEO Michelle Johnston Holthaus discloses the pivot in its AI data center strategy as Nvidia forges ahead with rack-scale solutions based on the rival's Blackwell GPU architecture. Intel said it no longer plans to sell its next-generation Falcon Shores AI accelerator chip that was due later this year so that it can focus on developing a "system-level solution at rack scale" with a successor chip it's calling Jaguar Shores. The news was delivered by Michelle Johnston Holthaus, interim co-CEO and CEO of Intel Products, during the semiconductor giant's fourth-quarter earnings call on Thursday. [Related: Outrun By Nvidia, Intel Pitches Gaudi 3 Chips For Cost-Effective AI Systems] Intel had previously teased Falcon Shores for a late 2025 launch as the successor to Gaudi 3, the AI accelerator chip it launched last fall with OEMs such as Dell Technologies and Hewlett-Packard Enterprise. But while Gaudi 3 and previous generations were application-specific integrated circuits (ASICs), Falcon Shores was slated by the company as a programmable GPU that incorporated Gaudi technology. However, Holthaus gave a blunt assessment of Intel's AI accelerator chip strategy at an investor conference in December, saying that Gaudi will not help Intel "get to the masses" and suggesting that Falcon Shores wouldn't be a game-changer. "Many of you heard me temper expectations on Falcon Shores last month," she said on Thursday's earnings call. Now, Holthaus said, Intel plans to use Falcon Shores for internal testing only. "This will support our efforts to develop a system-level solution at rack scale with Jaguar Shores to address the AI data center," she said. "More broadly, as I think about our AI opportunity, my focus is on the problems our customers are trying to solve, most notably lower the cost and increase the efficiency of compute." Holthaus said she arrived at the decision -- which she called one of her first major moves as the CEO of Intel Products and interim co-CEO after former CEO Pat Gelsinger abruptly departed in early December -- after receiving feedback from customers on what Intel needs to "be competitive and to deliver the right product." She said the decision was also influenced by discussions with Intel's AI accelerator chip team on where they were "from a competitive perspective and from an execution perspective." "One of the things that we've learned from Gaudi is it's not enough to just deliver the silicon. We need to be able to deliver a complete rack-scale solution, and that's what we're going to be able to be able to do with Jaguar Shores," Holthaus said. Intel disclosed the pivot in its AI data center strategy as Nvidia forges ahead with rack-scale solutions based on its Blackwell GPU architecture. Revealed at Nvidia's GTC event last year, the Grace Blackwell GB200 NVL72 rack-scale server platform has been pitched as the foundation for Nvidia AI offerings from major OEM and cloud computing partners, including Dell Technologies, Amazon Web Services, Microsoft Azure, Google Cloud and Oracle Cloud. Nvidia has previously said that the GB200 NVL72 platform as a "multi-node, liquid-cooled, rack-scale system for the most compute-intensive workloads," with each rack consisting of 36 Grace CPUs and 72 Blackwell GPUs connected over the NVLink chip-to-chip interconnect. In the company's November earnings call, Nvidia CEO Jensen Huang said the engineering does with OEM and cloud computing partners is "rather complicated," responding to a question about a report on Blackwell GPUs overheating in GB200 NVL72 systems. "The reason for that is because although we build full stack and full infrastructure, we disaggregate all of the AI supercomputer and we integrate it into all of the custom data centers and architectures around the world," he said at the time. "That integration process is something we've done several generations now. We've very good at it, but there's still a lot of engineering that happens at this point," Huang added. The Nvidia CEO and founder was referring to the fact that the AI computing giant has been selling complete AI systems powered by its GPUs and other components for several years, with the company moving from smaller appliances to supercomputer clusters more recently. Intel's AI strategy pivot was made after some of the chipmaker's channel partners told CRN earlier this month that it would take the company a long time to create formidable competition to Nvidia's AI chip dominance in data centers. Before Gelsinger abruptly retired from Intel in early December, the former CEO said during the company's third-quarter earnings call last October that it would "not achieve our target of $500 million in revenue" for Gaudi chips in 2024, blaming slower-than-expected sales on the "product transition" from Gaudi 2 to Gaudi 3 as well as "software ease of use." Meanwhile, Nvidia is expected to finish its 2025 fiscal year, which mostly lines up with the 2024 calendar year, with $128.6 billion in revenue, most of which is expected to come from its data center GPUs and associated systems. "Nvidia has got such a head start that it's just a long road," said Chris Bogan, vice president of sales at Houston-based systems integrator Mark III Systems. The lack of Intel's leadership in AI chips for data centers led one solution provider with a U.S. presence to choose AMD and not Intel for its AI center of excellence (COE) program, according to the partner alliance manager who manages vendors for the program. AMD has made more aggressive claims against Nvidia's data center GPUs with its Instinct chips, saying in October that its forthcoming Instinct MI325X can outperform Nvidia's H200 that launched last year. The rival has also announced an accelerated road map that will see it release new GPUs every year, similar to Nvidia's revamped strategy. "I don't have customers asking for [Gaudi]. What I do see is customers asking [to kick] tires on a GPU-as-a-service provider that might have AMD MI300X GPUs [or the like]," said the partner alliance manager, who asked to not be identified because he was not authorized to speak on behalf of his company regarding this matter. For Intel to gain ground in the AI chip market, the partner alliance manager said the company will need to find ways to convince partners to invest time into new products like Gaudi. "It requires extensive knowledge and training. That is literally competing with the same hours in the day that we're leveraging to actually train and build up knowledge and teams around Nvidia. So it's not just that you're competing for market share and awareness, but you're also competing for our time," he said.

Intel is canceling Falcon Shores, its next big AI chip.

"I am not happy with where we are today," Intel interim co-CEO Michelle Johnston Holthaus says of the company's AI server business on today's earnings call, a business which has not met its promises with Gaudi. So, says Holthaus, Intel will "simplify our roadmap and concentrate our resources" by canceling Falcon Shores. We plan to leverage Falcon Shores as an internal test chip only without bringing it to market." It'll focus on Jaguar Shores, a "system-level solution at rack scale," instead.

Intel cancels Falcon Shores GPU for AI workloads -- Jaguar Shores to be successor

In a surprising twist, Intel announced on Thursday that its Falcon Shores GPU for AI and HPC applications will not be released to the market but will remain an internal test processor used to develop hardware and software foundations for its successor, codenamed Jaguar Shores. This decision makes Intel's struggling Gaudi 3 processor, which has suffered from limited uptake amid the company's disclosed software issues, the company's only viable solution for AI applications for the next two years. "Many of you heard me temper expectations on Falcon Shores last month," interim co-CEO Michelle Johnston Holthaus said during the company's earnings call on Thursday. "Based on industry feedback, we have decided to leverage Falcon Shores as an internal test chip. Without bringing it to market, we will support our efforts to develop a system-level solution at rack scale with Jaguar Shores to address the AI data center more broadly." The company originally expected Falcon Shores to serve the AI and HPC markets currently covered by Gaudi processors. However, a quarter ago, Intel's interim co-CEO told analysts and investors to temper their expectations, as Falcon Shores would mostly serve as a vehicle to develop a hardware and software ecosystem around its hybrid processors. Apparently, after careful consideration, Intel decided not to launch Falcon Shores commercially but to use it solely for internal development to perfect hardware and software rather than launching a half-baked product commercially that could potentially damage the company's reputation. Intel described Falcon Shores as its first multi-chiplet design featuring Xe-HPC (or Xe3-HPC) GPU chiplets for highly parallel AI and HPC workloads, as well as x86 chiplets for general-purpose workloads with unified HBM memory. The product was meant to greatly increase performance and performance-per-watt efficiency compared to Intel's AI and HPC processors, though the company refrained from giving actual numbers. Oddly, Intel stated that Falcon Shores would be an "internal test chip." When developing an ecosystem, some processors that are not launched commercially are still shipped to external partners, including independent hardware vendors (IHVs) and independent software vendors (ISVs). For example, Intel's first-generation Xeon Phi processor (derived from the company's Larrabee GPU project), codenamed Knights Ferry, was not offered as a mainstream, fully supported commercial product. It was provided in limited quantities to select developers and research partners to begin porting and optimizing code for Intel's Many Integrated Core (MIC) architecture, which later became the commercial Xeon Phi product family. Since Falcon Shores and its successor, Jaguar Shores, have entirely new GPU microarchitectures, Intel would typically send samples of these processors to ISVs to ensure their software can work efficiently with the hardware. Also, since Intel is focusing on rack-scale solutions, it would stand to reason that it would also supply samples of Falcon Shores to its IHV partners. However, with Falcon's new designation as an internal test chip only, that strategy doesn't appear to be planned.