16. December 2025 06:06 Samir Bashir
Intel took the stage at the Barclays Global Technology Conference 2025 to dive deep into the company’s progress, challenges and strategic shifts. John Pitzer, Corporate Vice President of Corporate Planning and Investor Relations, provided a remarkably candid update on Intel’s manufacturing technologies, capacity planning and market positioning. The main focus was on the progress of 18A process technology, the beacon of hope for Intel’s internal manufacturing strategy. Panther Lake, the first PC product based on Intel 18A, was delivered before the end of the year as promised, with a planned launch at CES in January 2026. Although no specific yield figures were given, Pitzer emphasized that the monthly improvement is progressing steadily and the goal is to reach industry-standard yields by the end of 2026 or early 2027. Behind the scenes, it is clear how much the new management team under Lip-Bu Tan has changed course: Yield data is being shared with external suppliers for the first time, a strategic break with Intel’s previous compartmentalization policy. The aim: more efficient error identification and closer supplier integration.
The relationship between internal production and outsourcing was also a key topic. While large parts of the logic tiles for Meteor Lake, Arrow Lake and Lunar Lake were manufactured externally, around 70 percent of Panther Lake is to be returned to in-house production. This trend is set to continue with Nova Lake. At the same time, Intel is actively shifting internal capacities from the PC sector to server production, a clear sign of where margins and demand are currently strongest. The dynamics in server demand have surprised even Intel. While the PC business is stable at over 290 million units in 2025, hyperscalers have significantly raised their server forecasts in the second half of the year. Intel is openly talking about undersupply in both segments, with demand peaking and a supply bottleneck in the first quarter of 2026.
The reasons for the strong demand for servers are complex: after years of investment backlog, hyperscalers are now catching up, particularly in order to replace old, inefficient servers with new, energy-saving CPUs. In addition, the growing use of generative AI is creating additional hardware requirements, not only for GPUs, but also for traditional CPUs, for example for RAG-based architectures and data access. The market for nearline HDDs is also booming. According to Intel, customers are showing interest in long-term supply contracts beyond 2026 – an indication of sustained demand.
As far as the next generation of 14A processors is concerned, Intel is much more confident than with 18A. The development processes have been improved: earlier customer feedback in the definition phase, more mature PDKs and the fact that 14A only contains incremental, not fundamental technology disruptions such as gate-all-around and backside power. Nevertheless, Intel expects the first external customers to see production volumes in the second half of 2026 to early 2027 at the earliest. The investment planning for 2026 also shows flexibility. While originally signaling declining CapEx, Intel is now keeping a ” /- 1 billion” range open, depending on whether external customers are won for the foundry and additional capacity is required in Arizona or later Ohio.
In the area of advanced packaging, Intel is relying on EMIB and EMIB-T, technologies that should enable higher density and larger reticle sizes than CoWoS. However, Pitzer admits that the expectations for the packaging business in 2025 were too optimistic. Competitors’ CoWoS capacities have developed faster than expected, offering Intel customers cost-effective alternatives. The new argument: EMIB is technologically superior and will be converted into sales for the first time in the course of 2026.
Another key topic was the cooperation with NVIDIA, also in connection with NVLink server architectures. Intel is positioning itself here as an alternative provider to Grace and Vera and sees the cooperation with NVIDIA not only as capital relief, but also as confirmation of the future viability of the x86 platform. Particularly exciting: the cooperation with NVIDIA is multi-year and covers several product generations, which implies deep trust in Intel’s long-term roadmap. Despite supply bottlenecks for Intel 7/10nm processes, the company emphasizes that the demand not only affects older products, but also current generations such as Sapphire and Emerald Rapids. The new Granite Rapids chips on Intel 3 are technically on schedule, but are still under construction in terms of volume.
In terms of margins, Intel expects a range of 40 to 60 percent for 2026, depending on the product mix, prices and production costs. Positive effects will result from price increases for Raptor Lake and a withdrawal from the low-margin PC base segment. Negative influences come from higher costs for externally manufactured Arrow and Lunar Lake chips, in particular due to embedded memory, as well as the start-up costs for 18A. However, this is expected to turn positive in the second half of the year.
Finally, there were also insights into Intel’s long-term AI strategy. In the area of GPU accelerators, the focus will be on energy-efficient inference solutions, not on training LLMs. At the same time, the development of a dedicated ASIC business will be accelerated under the new head of engineering Srini. Intel already has several ASIC customers in the networking sector (e.g. SmartNICs) that are benefiting strongly from AI-driven growth. The aim is to penetrate an XPU-like market à la Broadcom or Marvell – with the advantage of also being able to work directly with hyperscalers as a foundry.
Intel therefore wants to deliver on several levels at the same time: better chips of its own, a credible foundry alternative, new AI products and the recapture of manufacturing sovereignty. This is ambitious, but necessary, because the market remains merciless.
www.igorslab.de
Intel took the stage at the Barclays Global Technology Conference 2025 to dive deep into the company’s progress, challenges and strategic shifts. John Pitzer, Corporate Vice President of Corporate Planning and Investor Relations, provided a remarkably candid update on Intel’s manufacturing technologies, capacity planning and market positioning. The main focus was on the progress of 18A process technology, the beacon of hope for Intel’s internal manufacturing strategy. Panther Lake, the first PC product based on Intel 18A, was delivered before the end of the year as promised, with a planned launch at CES in January 2026. Although no specific yield figures were given, Pitzer emphasized that the monthly improvement is progressing steadily and the goal is to reach industry-standard yields by the end of 2026 or early 2027. Behind the scenes, it is clear how much the new management team under Lip-Bu Tan has changed course: Yield data is being shared with external suppliers for the first time, a strategic break with Intel’s previous compartmentalization policy. The aim: more efficient error identification and closer supplier integration.
The relationship between internal production and outsourcing was also a key topic. While large parts of the logic tiles for Meteor Lake, Arrow Lake and Lunar Lake were manufactured externally, around 70 percent of Panther Lake is to be returned to in-house production. This trend is set to continue with Nova Lake. At the same time, Intel is actively shifting internal capacities from the PC sector to server production, a clear sign of where margins and demand are currently strongest. The dynamics in server demand have surprised even Intel. While the PC business is stable at over 290 million units in 2025, hyperscalers have significantly raised their server forecasts in the second half of the year. Intel is openly talking about undersupply in both segments, with demand peaking and a supply bottleneck in the first quarter of 2026.
The reasons for the strong demand for servers are complex: after years of investment backlog, hyperscalers are now catching up, particularly in order to replace old, inefficient servers with new, energy-saving CPUs. In addition, the growing use of generative AI is creating additional hardware requirements, not only for GPUs, but also for traditional CPUs, for example for RAG-based architectures and data access. The market for nearline HDDs is also booming. According to Intel, customers are showing interest in long-term supply contracts beyond 2026 – an indication of sustained demand.
As far as the next generation of 14A processors is concerned, Intel is much more confident than with 18A. The development processes have been improved: earlier customer feedback in the definition phase, more mature PDKs and the fact that 14A only contains incremental, not fundamental technology disruptions such as gate-all-around and backside power. Nevertheless, Intel expects the first external customers to see production volumes in the second half of 2026 to early 2027 at the earliest. The investment planning for 2026 also shows flexibility. While originally signaling declining CapEx, Intel is now keeping a ” /- 1 billion” range open, depending on whether external customers are won for the foundry and additional capacity is required in Arizona or later Ohio.
In the area of advanced packaging, Intel is relying on EMIB and EMIB-T, technologies that should enable higher density and larger reticle sizes than CoWoS. However, Pitzer admits that the expectations for the packaging business in 2025 were too optimistic. Competitors’ CoWoS capacities have developed faster than expected, offering Intel customers cost-effective alternatives. The new argument: EMIB is technologically superior and will be converted into sales for the first time in the course of 2026.
Another key topic was the cooperation with NVIDIA, also in connection with NVLink server architectures. Intel is positioning itself here as an alternative provider to Grace and Vera and sees the cooperation with NVIDIA not only as capital relief, but also as confirmation of the future viability of the x86 platform. Particularly exciting: the cooperation with NVIDIA is multi-year and covers several product generations, which implies deep trust in Intel’s long-term roadmap. Despite supply bottlenecks for Intel 7/10nm processes, the company emphasizes that the demand not only affects older products, but also current generations such as Sapphire and Emerald Rapids. The new Granite Rapids chips on Intel 3 are technically on schedule, but are still under construction in terms of volume.
In terms of margins, Intel expects a range of 40 to 60 percent for 2026, depending on the product mix, prices and production costs. Positive effects will result from price increases for Raptor Lake and a withdrawal from the low-margin PC base segment. Negative influences come from higher costs for externally manufactured Arrow and Lunar Lake chips, in particular due to embedded memory, as well as the start-up costs for 18A. However, this is expected to turn positive in the second half of the year.
Finally, there were also insights into Intel’s long-term AI strategy. In the area of GPU accelerators, the focus will be on energy-efficient inference solutions, not on training LLMs. At the same time, the development of a dedicated ASIC business will be accelerated under the new head of engineering Srini. Intel already has several ASIC customers in the networking sector (e.g. SmartNICs) that are benefiting strongly from AI-driven growth. The aim is to penetrate an XPU-like market à la Broadcom or Marvell – with the advantage of also being able to work directly with hyperscalers as a foundry.
Intel therefore wants to deliver on several levels at the same time: better chips of its own, a credible foundry alternative, new AI products and the recapture of manufacturing sovereignty. This is ambitious, but necessary, because the market remains merciless.
Intel at Barclays 2025: Roadmap, manufacturing, demand and margins under pressure | igor´sLAB
Intel took the stage at the Barclays Global Technology Conference 2025 to dive deep into the company’s progress, challenges and strategic shifts. John Pitzer, Corporate Vice President of Corporate…
www.igorslab.de