How Intel’s CEO brought the storied company back from the brink
When Pat Gelsinger returned to Intel as chief executive in early 2021, the company that once set the cadence for the entire semiconductor industry had lost its rhythm. Years of manufacturing delays, resurgent competitors, and the rise of outsourced chipmaking had left Intel struggling with missed commitments and a fading aura of inevitability. The “brink” was not bankruptcy; it was irrelevance—a slow slide from leadership to legacy.
Gelsinger’s turnaround playbook fused an old Intel ethos with new realities. He bet that manufacturing could still be a strategic weapon in a geopolitically fraught world, that design leadership would come only with brutal execution, and that a humbled Intel could partner as well as it once dominated. The work is still in progress, but the contours of a revival are visible: a rebuilt process roadmap, a foundry business with marquee customers, a sharper product line, and a government-backed manufacturing footprint aligned with national security priorities.
Telling the truth, then raising the bar
Gelsinger’s first act was cultural. He named the problems out loud—Intel’s process slips had cost it a decade of leadership—and revived the Grove-era insistence on facts over spin, open confrontation of issues, and schedule discipline. He set an audacious north star: “five nodes in four years,” a pledge to compress what had taken competitors nearly a decade into half the time. That roadmap—Intel 7, Intel 4, Intel 3, 20A, and 18A—was more than branding. It reintroduced extreme ultraviolet lithography across Intel’s fabs, pulled forward two breakthrough technologies (RibbonFET gate-all-around transistors and PowerVia backside power delivery), and restored a tick-tock-like cadence of steady improvements instead of perpetual slips.
He also made engineering the center of gravity again. Veteran technologists were elevated, “constructive confrontation” was encouraged in design and manufacturing reviews, and the company’s leadership ranks were repopulated with operators known for shipping on time. The message internally and externally was the same: Intel would be judged by delivery.
Betting the company on manufacturing—and opening it to the world
The strategic linchpin was IDM 2.0, a reimagined integrated device manufacturer model with three legs:
– Build: reinvest in Intel’s own leading-edge fabs.
– Buy: use external foundries where it makes sense.
– Sell: turn Intel’s manufacturing into an open commercial foundry, Intel Foundry.
This was a break from the past. It acknowledged that disaggregated chips built from multiple “tiles” could mix Intel process technology with outside silicon, and that Intel’s crown jewels—packaging, design enablement, and process expertise—could be valuable to customers that once would have been competitors. It also required a mental shift: a foundry must compete for customers on merit, openness, and price.
The company moved quickly. It embraced advanced packaging—EMIB for 2.5D and Foveros for 3D stacking—as a differentiator, opening those capabilities to external customers. It co-founded an industry standard, UCIe, for chiplet interconnects, signaling that the future would be modular and multi-vendor. And it accelerated relationships with EDA firms and IP providers so that customers could target Intel 18A with familiar design flows.
Proof points followed. Intel announced design collaborations and wins for its foundry services, including a headline deal to manufacture a custom chip for Microsoft on Intel 18A and engagements tied to U.S. defense programs. It also secured a partnership with MediaTek and worked with Arm to optimize cores for Intel’s leading-edge nodes. None of these erased years of lost ground, but they mattered: they showed that customers believed Intel could again be a credible manufacturer beyond its own products.
Rebuilding the process machine
If IDM 2.0 was the strategy, the process roadmap was the test. Under Gelsinger, Intel hit several long-missed milestones:
– Intel 4, the company’s first EUV-based node, reached manufacturing readiness and powered new client processors.
– Intel 3 followed with performance and power gains aimed at data center and foundry customers.
– 20A introduced two firsts in high-volume logic—RibbonFET transistors and PowerVia—designed to recapture transistor density and performance-per-watt leadership.
– 18A, the capstone of the “five nodes in four years” sprint, was positioned for customer tape-outs with early high-volume manufacturing targeted around 2024–2025.
Behind the scenes, Intel retooled its fab network, pushed yield-learning earlier in development, and tightened cross-functional lines between design and manufacturing. Perhaps the most symbolic milestone: Intel took delivery of the industry’s first high-NA EUV tool from ASML, an investment in the post-18A era signaling that the company intended not just to catch up, but to lead.
Products that show, not tell
Turnarounds gain credibility when products stabilize. For Intel, that meant simpler roadmaps, fewer slips, and clearer positioning.
On the client side, Intel pivoted to disaggregated architectures—“tiles” connected with advanced packaging—to mix internal and external silicon. It leaned into an “AI PC” narrative with processors that integrated NPUs for on-device inference, aligning with a wave of AI-enabled software from major platforms. On the server side, after painful delays, Intel returned to a steadier cadence: mainstream Xeons, dense-core variants optimized for cloud scale-out, and accelerators targeted at AI training and inference. The Gaudi line, acquired via Habana Labs, offered a cost-effective alternative to the dominant AI accelerators, supported by an open software stack that appealed to customers seeking supplier diversity.
Importantly, Intel adopted a more pragmatic stance on where to build. It was willing to use TSMC for specific tiles while manufacturing others in-house, choosing the best tool for the job and letting packaging knit it all together. That pragmatism helped products ship while Intel’s own leading-edge nodes ramped.
Balance sheet triage and capital discipline
Gelsinger’s Intel also restructured financially to buy time and focus. The company:
– Streamlined the portfolio, exiting lower-synergy businesses and reducing spend in areas that weren’t differentiating.
– Monetized stakes in non-core assets and took Mobileye public to unlock value and fund capital needs.
– Reorganized reporting to give investors visibility into the foundry business as a distinct unit, forcing accountability on costs and pricing.
– Targeted multi-billion-dollar cost reductions while protecting critical R&D and fab investments, including a painful dividend cut to preserve cash.
Perhaps the most consequential financial move was aligning Intel’s expansion with public policy. In the United States, Intel secured a landmark preliminary award under the CHIPS and Science Act—billions in grants and access to loans—alongside a 25% investment tax credit, supporting new and modernized fabs in Arizona, New Mexico, Ohio, and Oregon. In Europe, it pursued a flagship site in Germany with substantial local support. These packages didn’t erase execution risk, but they made the capital math for advanced manufacturing more survivable and tied Intel’s future to national priorities around secure supply chains.
Opening the tent: ecosystem, software, and trust
A foundry cannot succeed without an ecosystem. Intel moved from a closed world to an open courtship:
– Deepened partnerships with Arm, Synopsys, Cadence, and Siemens to ready design kits and IP on Intel 18A.
– Backed open standards (UCIe) and open software stacks (oneAPI, OpenVINO) to reduce friction for customers and developers.
– Emphasized security and traceability in manufacturing—features attractive to governments and enterprises with regulated workloads.
Trust, though, is earned in small steps: predictable roadmaps, transparent yield curves, competitive pricing, and the humility to admit and fix issues. By separating the foundry P&L and courting anchor customers, Intel signaled it understood how a modern contract manufacturer must behave.
Why this turnaround is different
Intel has attempted resets before. What makes this effort more credible is its coherence across five fronts:
– Clarity of mission: make manufacturing a competitive advantage again.
– Cultural reset: restore engineering-first decision-making and execution rigor.
– Technological bets: EUV everywhere, advanced packaging, RibbonFET, PowerVia, and high-NA EUV for the next decade.
– Market pragmatism: use external foundries when needed; win foundry customers on merit.
– Policy alignment: anchor U.S. and European industrial strategies to Intel’s expansion.
Early indicators support the trajectory: more on-time product launches, a visible pipeline of foundry engagements, tangible progress on 18A, and large-scale government backing. Intel also regained a measure of narrative momentum—no small asset in recruiting talent and persuading cautious customers to place bets.
What could still go wrong
Turnarounds are fragile. Intel faces real hazards:
– Execution risk: any slip on 18A or its successors could erode customer confidence and reset the clock.
– Competitive pressure: TSMC’s consistency, AMD’s design velocity, and NVIDIA’s dominance in AI accelerators leave little room for error.
– Foundry economics: building a leading-edge foundry at scale requires sustained utilization; without a diversified customer base, the math is punishing.
– Geopolitics and supply chains: export controls, material constraints, and global tensions can disrupt even the best-laid plans.
– Software and ecosystem: winning in AI and heterogeneous compute demands not just silicon, but end-to-end platforms that developers love.
The lesson: leadership is a system, not a slogan
Gelsinger’s rescue of Intel is not a single bold move but a system of reinforcing changes. Tell the truth. Set an ambitious but testable plan. Marry design with manufacturing discipline. Open the business to partners and standards. Use packaging and process to differentiate in a chiplet world. Align with national priorities to underwrite the capital intensity of leading-edge fabs. And then, deliver, deliver, deliver.
Intel is not “back” in the sense of having completed a comeback; the work of converting process milestones and pilot customers into durable market share will take years. But the company is no longer drifting. The music has a beat again, and the industry is listening—for the first time in a long while—not to eulogize a legend, but to see if a storied giant can write a modern refrain.
