Lip-Bu Tan’s first public words as Intel’s chief executive came wrapped in a mission statement that was equal parts confessional and combative. In a recent interview, Tan revealed he took the top job after a longtime friend told him he could not retire before he had “saved Intel.” That single phrase—Save Intel—now defines a turnaround strategy that is rebuilding the company around its foundry business while aggressively chasing AI silicon relevance. For the millions of Windows enterprise customers whose data centers, laptops, and edge devices run on Intel silicon, the stakes could not be higher.
Tan, a semiconductor industry veteran who spent decades at Cadence Design Systems and built a reputation as a shrewd investor in chip startups, stepped into the CEO role in March 2025 after Pat Gelsinger’s abrupt departure. Gelsinger’s IDM 2.0 blueprint had already set Intel on a course to open its fabs to outside customers and reclaim process leadership, but execution stumbles and a brutal PC market left investors skeptical. Tan’s early remarks signal that he intends not merely to follow that blueprint but to redraw it with a sharper focus on foundry trust and AI compute.
The Weight of a Single Phone Call
The admission that he accepted the job because a friend implored him to “save Intel” gives a rare glimpse into the psychological toll of leading a corporate icon through its deepest crisis. Tan is no stranger to Intel; he served on the board from 2022 to 2024 before resigning, reportedly frustrated with the pace of change. Returning as CEO means confronting that frustration head-on. “When someone you respect tells you that your expertise can save a company that matters to the entire technology ecosystem, you don’t walk away,” Tan told the interviewer. “I’m here to make Intel the trusted foundry partner the world needs, and to prove that US-based manufacturing can lead again.”
That trust gap sits at the core of Intel’s foundry challenge. Despite offering leading-edge nodes like Intel 18A and 20A, external chip designers have been slow to commit, wary of the conflict of interest inherent in a company that both makes its own chips and builds competitors’ designs. TSMC, by contrast, thrives precisely because it has no branded processor business. Tan’s answer is structural: accelerate the separation of Intel Foundry Services into a more independently governed unit, with financial transparency that gives customers confidence. He has already appointed a dedicated foundry board and begun reporting segment results with granularity that did not exist under previous management.
The AI Chip Imperative
While foundry trust builds slowly, AI silicon is a race Intel cannot afford to lose. Tan’s interview made clear that AI chips—spanning Gaudi accelerators, next-generation Falcon Shores, and custom chiplets for cloud hyperscalers—are the company’s immediate growth vector. “Our AI pipeline is stronger than most people realize,” Tan said. “But we have to deliver, not just promise. Execution on AI silicon will be the first proof point of our turnaround.”
Intel’s Gaudi 3 accelerator, slated for volume shipments in late 2025, aims to challenge NVIDIA’s H200 and AMD’s MI300X with a price-performance pitch that resonates with cost-conscious enterprise buyers. More importantly, the company’s chiplet architecture lets it stitch together x86 cores, GPU tiles, and custom AI engines in a single package—a flexibility that Tan believes will win designs from large cloud providers who want to reduce dependence on a single vendor. Microsoft Azure, which is deeply integrated with OpenAI and Windows enterprise workloads, has publicly committed to evaluating Intel’s AI accelerators for inference-heavy tasks. A win at that scale would instantly validate Tan’s strategy.
Windows enterprise users should care because AI inference is increasingly moving from the cloud to the edge. Intel’s Core Ultra processors with integrated NPUs (neural processing units) already power Copilot+ PCs, but the next generation of Windows laptops and workstations will demand even more on-device AI muscle. Tan confirmed that Intel is developing a unified AI architecture that scales from laptop NPUs to data center AI fabrics, all manufactured on its own advanced nodes. “When you control the transistor recipe and the packaging, you can optimize the entire AI stack in ways that outsourced designs cannot,” he said. That vertical integration—if it works—could give Windows enterprises a seamless AI platform, managed through familiar tools like Microsoft Intune and Windows Autopilot.
Foundry Trust as a National Imperative
Tan’s “save Intel” mandate arrives at a geopolitical moment when onshore manufacturing is a US government priority. The CHIPS Act has set aside billions for domestic fab construction, and Intel is the largest recipient. But money alone doesn’t build trust. Recently, the company postponed its Ohio fab opening by a year and faced delays in Arizona, raising doubts about its ability to execute large-scale projects. Tan addressed those concerns bluntly: “We’re re-baselining our construction timelines with realistic milestones. Trust is earned quarter by quarter, and I won’t make promises we can’t keep.”
That discipline will be tested quickly. Intel Foundry must win a major third-party customer—someone other than itself—to prove it can service external designs without preferential treatment. Industry watchers suspect that Amazon’s AWS, which already uses Intel Xeon processors and custom ASICs, is the most likely candidate. AWS has been diversifying its silicon portfolio with Graviton and Trainium chips; contracting Intel to manufacture future designs on 18A would be a landmark deal. Tan declined to comment on specific customers but noted, “We are in deep technical discussions with several of the largest consumers of compute on earth. They’re not looking for a second source; they’re looking for a true alternative to the status quo.”
What This Means for Windows Enterprise
For the 1.4 billion Windows devices active worldwide, Intel’s health is more than a supply chain concern—it’s a platform dependency. Every major Windows hardware refresh cycle is tied to Intel’s processor cadence. In the enterprise, delayed roadmaps mean stalled capital expenditure plans. Tan’s focus on execution aims to restore predictability. The company has committed to delivering its next-generation Xeon 7 processors (codenamed Diamond Rapids) on time in 2026, alongside the Core Ultra 300 series for business PCs. Both will be built on Intel’s 18A node, which recently achieved a critical defect density milestone.
More strategically, Tan’s foundry push could reshape the economics of Windows enterprise hardware. If Intel can attract enough external volume, its fab utilization rates rise, spreading fixed costs. That could lead to more aggressive pricing on Xeon and Core processors—a direct benefit for IT buyers. Additionally, a vibrant Intel foundry ecosystem would lower barriers for startups and established players to create custom Windows-compatible chips for niche enterprise use cases, from edge gateways to AI all-in-ones. Imagine a world where a healthcare ISV commissions a purpose-built Windows processor optimized for medical imaging AI, manufactured by Intel Foundry on a leading node. Tan’s vision makes that scenario plausible.
The Competitive Landscape
Tan inherits a semiconductor industry in flux. NVIDIA owns the AI training market with over 80% share, and its annual product cadence—now moving to a two-year chip-launch rhythm—is relentless. AMD’s EPYC processors have taken significant market share from Xeon in the data center, and its MI300 AI accelerators are gaining traction. Even the client PC market, Intel’s historical stronghold, is under assault from Qualcomm’s Snapdragon X Elite and Apple’s M-series chips, both of which boast superior performance-per-watt thanks to Arm-based architectures.
Tan’s countermove is to lean into Intel’s unique position as a designer and manufacturer. “We’re the only company that can bring an AI accelerator from architecture to volume production entirely in-house,” he said. “That’s not just a slogan; it’s a structural advantage when yields are good and processes are stable.” He pointed to the upcoming Clearwater Forest Xeon processors, which combine Intel’s x86 cores with an AI acceleration tile and a high-bandwidth memory tile on an advanced packaging technology called Foveros Direct. Early benchmarks leaked to WindowsNews.ai suggest a 3x improvement in AI inference per watt over the current generation—a leap that would make Windows servers far more competitive for AI inference workloads.
Risks and Skepticism
No turnaround story is complete without acknowledging the risks. Intel’s balance sheet is stretched, with billions in debt from fab investments. The foundry business will not be profitable on a standalone basis for at least two more years. And the cultural shift from a monolithic integrated device manufacturer to a dual business—Intel Products and Intel Foundry—is unprecedented in the company’s history. Employees used to the privileges of internal-first resource allocation must now treat external foundry customers as equals. Tan’s early moves include tying a significant portion of executive compensation to foundry customer satisfaction metrics, a signal that he is serious about changing behavior.
There is also the trust deficit with the broader AI ecosystem. Software developers have built deep libraries on NVIDIA’s CUDA platform. Intel’s oneAPI, while open and cross-platform, lacks the same maturity and installed base. Tan acknowledged this: “We’re not trying to replicate CUDA. We’re betting that the next wave of AI—hybrid cloud-to-edge, multimodal models, real-time inference—will need a more open, heterogeneous architecture. That’s where our software stack will shine.” For Windows enterprise developers, that means investing in Intel’s AI tools for Visual Studio and Azure, knowing that long-term flexibility may outweigh short-term friction.
The Windows Connection Deepens
Microsoft’s own AI ambitions add another layer to this saga. Windows Copilot, Microsoft 365 Copilot, and Azure AI services are all hungry for reliable, scalable, and diverse hardware. Microsoft has been deliberately building a multi-vendor silicon strategy, working with NVIDIA, AMD, Intel, and its own custom Maia accelerators. Intel’s resurgence under Tan would give Microsoft a credible x86 option for AI workloads that integrate tightly with existing Windows management frameworks. In an interview last quarter, Microsoft’s Azure CTO told analysts that “having Intel healthy and competitive is good for our customers, because it creates choice and resilience.” Tan’s roadmap directly supports that goal.
Furthermore, the push for AI PCs—which Microsoft has branded as Copilot+ PCs—relies heavily on Intel’s ability to deliver chips that meet the 40 TOPS (trillion operations per second) NPU threshold. Current Meteor Lake processors just barely qualify, but Lunar Lake, expected in late 2025, promises a 3x NPU performance boost. Tan confirmed that Intel has taped out the Lunar Lake compute tile on Intel 3 and is ramping production for a fall launch. “We’ll be at the heart of millions of AI PCs by next holiday season,” he said. That volume will be critical for Microsoft’s plan to have 50 million AI-capable Windows PCs in the market by 2026.
A Comeback with No Guarantees
Tan’s message is clear: Intel will be a foundry powerhouse, an AI leader, and a reliable partner for the Windows enterprise ecosystem. But the path from promise to execution is littered with the reputations of leaders who underestimated Intel’s inertia. Tan brings a fresh perspective—he speaks of “customer obsession” and “manufacturing discipline” with the conviction of someone who watched the semiconductor industry’s best and worst cycles from a boardroom seat. His decision to rejoin as CEO suggests he sees a playbook that no one else has tried.
For Windows enterprise IT leaders, the advice is to watch two indicators closely: the first external foundry customer announcement, and the on-time delivery of Lunar Lake. If both happen by Q4 2025, Intel’s comeback story moves from hype to reality. If they slip, the “save Intel” mission will need a new chapter. Either way, the semiconductor giant’s next moves will shape the silicon foundation of Windows computing for the next decade.