Intel bets on on-device AI and US fabs to power the next generation of PCs

Intel unveiled its Core Ultra series 3 processors on Thursday, the first client chips built on its 18A process node and manufactured in Arizona, as enterprises gear up for a wave of PC refreshes driven by Microsoft’s October 2025 end-of-support deadline for Windows 10.

The Panther Lake platform delivers up to 180 platform TOPS for AI workloads through a balanced design that distributes processing across CPU, GPU, and neural processing units. Systems are expected to reach broad availability in January 2026, Intel said in a statement.

But while Intel is pitching AI acceleration as a key differentiator, analysts say most enterprise buyers remain unconvinced that on-device AI processing justifies the investment — even as survey data suggests they’re buying AI-capable hardware anyway.

Made in America, bought for other reasons

Intel emphasized its domestic manufacturing capability during the announcement. CEO Lip-Bu Tan said the company is “entering an exciting new era of computing” enabled by semiconductor advances, highlighting US production at Fab 52 in Chandler, Arizona, as central to Intel’s strategy.

The 18A node incorporates RibbonFET transistor architecture and PowerVia backside power delivery, with Intel claiming approximately 15% better performance per watt and 30% improved chip density compared to its Intel 3 process.

However, from an enterprise procurement perspective, manufacturing location carries limited weight for most buyers, according to Manish Rawat, semiconductor analyst at TechInsights.

“Intel’s ‘18A’ and ‘Made in America’ positioning primarily resonates with government, defense, and regulated industries where supply chain assurance and data sovereignty are critical,” Rawat said. “For the broader enterprise market, purchasing decisions remain driven more by performance-per-watt, TCO, platform stability, and manageability than by fab location.”

Maciek Gornicki, senior research manager for IDC’s Client Devices Group in Asia/Pacific, agreed. “The product’s ability to support business operations is the much bigger factor,” he said, though noting the 18A node “looks to deliver improvements in performance and efficiency.”

While US chip fabrication could benefit government procurement under Buy American provisions, it’s unlikely to affect tariffs on finished laptops, which are classified based on final assembly location rather than component sourcing.

180 TOPS in search of a killer app

Panther Lake features up to 16 cores and a new Intel Arc GPU with up to 12 Xe cores, with Intel claiming more than 50% performance improvements over previous generations. The company also previewed Xeon 6+ (codenamed Clearwater Forest), its first 18A-based server processor scheduled for first-half 2026 launch with up to 288 efficiency cores.

However, enterprises continue to view AI acceleration on client devices as a niche requirement rather than a universal need, according to Rawat.

“While Intel’s 180 TOPS figure is impressive on paper, most organizations still rely on cloud-based AI for scalability, central model management, and consistent performance,” he said. For sentiment to shift, enterprises need proven productivity gains from on-device inference, software ecosystems that fully exploit local AI hardware, and quantifiable cost savings compared to cloud AI usage fees. “Without that clear ROI, IT departments are unlikely to pay a premium for AI-capable endpoints.”

Gornicki echoed this, noting AI PC use cases remain works in progress. “Macroeconomic uncertainty also means that some organizations feel compelled to buy lower-priced systems, sometimes without NPUs,” he said.

Intel has been enabling developers to leverage local AI processing for advantages in privacy, personalization, and potential cost savings, Gornicki noted. “But it will still take a few years before we start to see bigger use cases.”

Futureproofing trumps immediate need

According to Gornicki, PC refresh activity stems from three factors: replacement of pandemic-era devices, Windows 11 migration, and AI adoption considerations.

“Key challenge around the AI PC adoption is the lack of game-changing killer applications which would make PC hardware refresh absolutely necessary,” Gornicki said.

Yet according to IDC survey data cited by Gornicki, 70% of organizations perceive AI capabilities as a top, critical, or very important factor for next-wave deployments, with 39% planning to upgrade the majority of their devices within 12 months to benefit from AI PC capabilities. Gornicki said IDC forecasts AI PCs will account for just over 50% of commercial shipments globally by 2026.

The apparent contradiction — high stated interest without clear business cases — suggests enterprises are hedging their bets, purchasing AI-capable hardware for futureproofing rather than immediate requirements. IDC projects overall PC shipments will grow 4.3% in 2025, driven primarily by the Windows refresh requirement.

Intel faces intensified competition

Intel’s launch comes as AMD has gained market share, and Qualcomm’s Snapdragon X series targets Windows on Arm deployments.

“Panther Lake signals Intel’s attempt to reclaim leadership after ceding ground to AMD in multi-core efficiency and to Arm-based systems in battery life and AI optimization,” Rawat said. “If Intel delivers on both 18A efficiency gains and integrated AI acceleration without thermal compromises, it could narrow the gap and stabilize market share in 2025-2026 refresh cycles.”

However, many enterprises have diversified vendors to avoid single-source dependency, Rawat noted, and AMD’s competitive pricing remains appealing. Arm’s penetration into enterprise laptops with growing Windows-on-Arm optimization adds additional pressure.

“Panther Lake helps Intel re-enter the leadership narrative but does not yet guarantee a decisive shift in enterprise buying behavior until real-world benchmarks validate its advantages,” Rawat said.

Gornicki cited IDC survey data that shows more than half of businesses are open to considering AMD or Qualcomm-based devices. “Panther Lake will undoubtedly ensure that Intel remains relevant and very often first choice among enterprises that seek to remain on Intel Architecture,” he said, though characterizing the launch as an expected development rather than a game-changer.

More Intel news:

AMD could be Intel’s next foundry customer

Who wins/loses with the Intel-Nvidia union?

Intel will design CPUs with Nvidia NVLink in return for $5 billion investment