On December 1, 2024, Intel's board of directors delivered Pat Gelsinger an ultimatum: retire or be removed. After a contentious board meeting the week prior, directors concluded that Gelsinger's costly and ambitious plan to turn Intel around was not working and the progress of change was not fast enough. Gelsinger chose to step down.
The announcement described it as retirement. The reality was forced resignation. Intel's stock had fallen 60% since Gelsinger took over as CEO in February 2021. The company had just posted a $16.6 billion loss in Q3 2024—its worst quarterly performance in decades. Intel's market capitalization had plummeted from over $500 billion in 2000 to just $89.78 billion by April 2025, a loss of nearly $413 billion in shareholder value.
CFO David Zinsner and executive Michelle Johnston Holthaus were named interim co-CEOs while the board conducted a search for a permanent replacement. Frank Yeary, independent chair of the board, became interim executive chair during the transition period. The search would focus on external candidates—a break from Intel's tradition of promoting from within—including Marvell Technology CEO Matt Murphy, former Cadence CEO Lip-Bu Tan, and potentially Apple's senior vice president of hardware technologies Johny Srouji.
For Gelsinger, the departure marked the end of a 40-year relationship with Intel. He had joined the company at 18 years old in 1979 as a quality assurance technician. He rose to become the youngest vice president in Intel's history at age 32. He served as chief technology officer from 2001 to 2009. He left for EMC and VMware, tripling the latter's revenue to nearly $12 billion during his eight years as CEO. He returned to Intel in 2021 as the chosen savior, the technical visionary who understood silicon and could restore Intel's manufacturing leadership.
Three years and nine months later, the board concluded he had failed. Intel remained years behind TSMC in advanced process nodes. The company's 18A manufacturing process—the cornerstone of Gelsinger's IDM 2.0 strategy—was stuck at 10% yields, making mass production impossible. Intel had missed the AI revolution entirely, ceding an estimated 80% market share to NVIDIA. The foundry business that Gelsinger launched with such fanfare had attracted only a handful of customers and none with the chip volumes needed to ensure profitability.
Patrick Paul Gelsinger was born on March 5, 1961, and raised on family farms by his parents, June and Paul Gelsinger, in rural Robesonia, Pennsylvania—an Amish and Mennonite part of the state. He joined Intel at 18 years old in October 1979, just after earning an associate degree from Lincoln Tech. He worked as a quality assurance technician for microprocessors while pursuing his bachelor's degree from Santa Clara University, which he earned in 1983, and his master's degree from Stanford University in electrical engineering, which he completed in 1985 magna cum laude.
Gelsinger was part of the design team for the Intel 80386 microprocessor, leading the effort to port Unix to the processor. In 1987 he co-authored his first book, about programming the 80386. His technical contributions extended to etching his initials ("PG") into the microcode of every Intel 80386 processor produced—a practice reflecting his direct involvement in verification and design processes and creating a permanent technical signature embedded in millions of chips.
At age 25, Gelsinger became the lead architect of the Intel 80486 processor, introduced in 1989. The 80486 represented a major architectural advance over the 80386, integrating the math coprocessor and cache memory onto a single die—a design decision that significantly improved performance. Gelsinger led the 486 development while simultaneously working on his Stanford master's degree, demonstrating the intense work ethic that would characterize his entire career.
Intel CEO Andrew Grove took notice. Grove, the legendary executive who transformed Intel from a memory company into the world's dominant microprocessor manufacturer, personally mentored Gelsinger. At age 32, Gelsinger was named the youngest vice president in Intel's history. Over the next three decades, he would lead 14 chip projects and contribute to key technology developments including Wi-Fi, USB, Intel Core processors, and Intel Xeon processors.
In 2001, Gelsinger became Intel's chief technology officer. The role positioned him at the center of Intel's technology strategy during a period of tremendous growth. Intel's dominance in PC and server processors appeared unassailable. AMD struggled to compete. Mobile computing was nascent. Cloud computing had not yet emerged as a major market. Intel's integrated device manufacturing (IDM) model—designing and fabricating its own chips—delivered both performance leadership and cost advantages that competitors could not match.
But in September 2009, Gelsinger left Intel. The reasons remain somewhat opaque. Publicly, the departure was described as an opportunity to join EMC as president and chief operating officer—a promotion to a more senior executive role than CTO. Privately, industry observers noted that Gelsinger had been passed over for the CEO position that went to Paul Otellini in 2005. Gelsinger's departure removed one of Intel's most prominent technical leaders and set the stage for a decade away from the company before his 2021 return.
After three years at EMC, Gelsinger became CEO of VMware in 2012. VMware was the dominant virtualization software company, enabling enterprises to run multiple virtual machines on single physical servers. The company had pioneered server virtualization in the early 2000s and rode the virtualization wave to billion-dollar revenues. But by 2012, VMware faced strategic challenges: cloud computing threatened to commoditize virtualization, Amazon Web Services was rapidly gaining enterprise adoption, and Microsoft was aggressively pushing Hyper-V as a lower-cost alternative to VMware's vSphere.
Gelsinger transformed VMware's strategic positioning. He expanded beyond server virtualization into cloud infrastructure, enterprise mobility, and cybersecurity. He repositioned VMware as a multi-cloud platform provider, enabling enterprises to build and operate applications across VMware infrastructure, Amazon AWS, Microsoft Azure, and Google Cloud Platform. The strategy acknowledged that customers wanted cloud flexibility rather than vendor lock-in to a single infrastructure provider.
The financial results validated the strategy. Under Gelsinger's leadership, VMware's revenues nearly tripled to almost $12 billion. The company maintained profitability while investing heavily in R&D and acquisitions to expand its product portfolio. Gelsinger received recognition as the best CEO in America in 2019, according to an annual survey by Glassdoor that measured employee satisfaction with leadership.
Gelsinger's success at VMware demonstrated capabilities beyond technical architecture. He had proven he could run a large public company, execute strategic pivots, manage through competitive disruption, and deliver consistent financial growth. When Intel's board began searching for a CEO to replace Bob Swan in early 2021, Gelsinger emerged as the leading candidate. He combined deep Intel technical knowledge from his 30-year career with proven general management skills from his eight years running VMware. He understood semiconductor manufacturing, chip architecture, and the foundry business model. He also understood enterprise customers, software ecosystems, and cloud computing dynamics.
On February 15, 2021, Gelsinger rejoined Intel as CEO. He stepped down from VMware after eight years in the role, with VMware CFO Zane Rowe assuming the position of interim CEO while the board conducted a permanent search. Gelsinger's return was framed as a homecoming—the prodigal son returning to save the company where he had built his career and reputation.
On March 23, 2021—just five weeks after becoming CEO—Gelsinger announced "IDM 2.0," a major evolution of Intel's integrated device manufacturing model. The strategy consisted of three core pillars that fundamentally reshaped Intel's business model and competitive positioning.
First, Intel reaffirmed its commitment to manufacturing the majority of its products internally. This represented a continuation of Intel's historic IDM model but with explicit acknowledgment that internal manufacturing would no longer suffice for all products. Second, Intel announced it would significantly expand its use of third-party foundries, particularly TSMC, to manufacture advanced chips that Intel's own fabs could not produce competitively. This represented a strategic retreat from pure vertical integration and an admission that Intel had fallen behind TSMC in process technology.
Third, and most ambitiously, Gelsinger announced Intel Foundry Services (IFS)—a standalone business unit that would manufacture chips for external customers, competing directly with TSMC and Samsung. IFS would offer leading-edge process technology, advanced packaging capabilities, committed capacity in the United States and Europe, and access to Intel's IP portfolio including x86 cores as well as ARM and RISC-V ecosystem IP. Intel would become a major provider of foundry capacity in the U.S. and Europe to serve customers globally.
The foundry strategy represented the most radical departure from Intel's historical business model. For decades, Intel had manufactured chips exclusively for itself, viewing manufacturing as a competitive advantage rather than a standalone business. Intel's fabs produced Intel-designed processors for PCs, servers, and data centers. The company did not manufacture chips for AMD, NVIDIA, Qualcomm, or other semiconductor designers. Vertical integration was Intel's moat.
Gelsinger's foundry pivot acknowledged that vertical integration had become a liability. TSMC's pure-play foundry model—manufacturing chips for any customer willing to pay—had enabled TSMC to achieve economies of scale and process technology leadership that Intel could not match while manufacturing only for itself. By 2021, TSMC's 5-nanometer process was more advanced than Intel's manufacturing nodes. Apple, AMD, NVIDIA, and Qualcomm were all fabbing their most advanced chips at TSMC. Intel had lost its process technology lead for the first time in decades.
Gelsinger announced massive manufacturing expansion to support IDM 2.0. Intel would invest an estimated $20 billion to build two new fabs in Arizona. The company would construct additional fabs in Ohio, New Mexico, and Oregon, with total U.S. investments exceeding $100 billion over subsequent years. Intel would build fabs in Europe to serve European customers and reduce geopolitical supply chain concentration in Taiwan. The company planned to achieve five process nodes in four years—an extraordinarily aggressive roadmap designed to leapfrog TSMC and reclaim manufacturing leadership by 2025.
The strategy required massive capital deployment. Intel's historical capital expenditures had ranged from $10 billion to $15 billion annually. Under IDM 2.0, capital expenditures would increase to over $25 billion per year, straining Intel's balance sheet and requiring government subsidies to fund the fab construction. Intel would compete for CHIPS Act funding—the $52 billion U.S. government program designed to reshore semiconductor manufacturing—to offset the costs of building U.S.-based fabs.
In March 2024, Intel announced preliminary terms for up to $8.5 billion in direct funding under the CHIPS and Science Act. In November 2024, Intel and the U.S. Department of Commerce finalized the agreement at $7.86 billion—a reduction from the initial announcement but still the largest CHIPS Act award granted to any company. Intel could also receive up to $11 billion in loans tied to the legislation and a 25% investment tax credit. Additionally, Intel won a manufacturing contract for up to $3 billion for the Secure Enclave program, separate from the CHIPS Act funding.
The funding would support Intel's semiconductor manufacturing and advanced packaging projects in Arizona, New Mexico, Ohio, and Oregon. The Arizona fabs were being built at a combined cost of more than $50 billion. The Ohio site comprised two plants on a 1,000-acre "megasite" that could eventually host multiple additional factories. The New Mexico investment focused on Foveros chip packaging, comprising two fabs built at a cost of $3.5 billion. Oregon expansion included the D1X research factory in Hillsboro.
Intel's planned U.S. investments were projected to support more than 10,000 company jobs, nearly 20,000 construction jobs, and more than 50,000 indirect jobs with suppliers and supporting industries. The employment figures justified government subsidies on national security and industrial policy grounds—reshoring advanced semiconductor manufacturing to reduce dependence on Taiwan and create high-wage American jobs in critical technology sectors.
By late 2024, Intel Foundry Services had failed to achieve its strategic objectives. The business had attracted a handful of customers including Microsoft and Amazon for specialized chips, but neither would bring huge volumes needed to ensure the factories' profitability. Intel had failed to win any significant customers for its 18A node—the advanced process technology that Gelsinger positioned as Intel's leapfrog over TSMC's leadership.
The 18A process faced severe manufacturing problems. Industry reports indicated yields stood at approximately 10%, making mass production impossible. Sources inside Intel confirmed 10% yields, up from just 5% late in 2023 but still far below the 70-80% yields required for commercial viability. Other reports suggested yields of 20-30%, which remained insufficient for Intel's goal of hitting mass production in the second half of 2025.
Only 10% of Panther Lake wafers—Intel's next-generation laptop processor scheduled for late 2025 launch—met deliverable standards. Defect rates were reportedly three times higher than acceptable thresholds for high-volume manufacturing. The 18A process introduced RibbonFET transistors and PowerVia backside power delivery, advanced technologies that promised efficiency gains but also introduced unprecedented manufacturing complexity leading to low yield rates.
After receiving silicon wafers from Intel's 18A manufacturing process, Broadcom was unsatisfied with the results and concluded the process was not yet viable for high-volume production. Broadcom's rejection represented a devastating blow to Intel's foundry strategy. Broadcom manufactures massive volumes of networking and storage chips and would have provided the scale economics that Intel Foundry Services desperately needed. The lost customer validated skepticism that Intel could compete with TSMC's mature foundry operations.
TSMC maintained overwhelming dominance in the foundry market. In Q3 2024, TSMC held 64.9% market share. Samsung held 12%. Intel had less than 5%. The market share numbers reflected both technical performance and customer trust. External customers including Qualcomm, AMD, Apple, and NVIDIA favored TSMC due to its stable yield improvements and track record of delivering advanced process nodes on schedule.
TSMC's 2-nanometer process was on track for initial mass production in the second half of 2025, specifically targeting high-volume manufacturing starting in late 2025. TSMC maintained a substantial lead over Intel in securing 2nm chip orders, with major companies including AMD, NVIDIA, MediaTek, and Qualcomm signaling their intent to place orders. Samsung's 2nm process was expected to begin mass production in 2025, though Samsung's history of struggling with yield rates and efficiency in previous nodes created concerns about its 2nm competitiveness.
Intel claimed one critical advantage: backside power delivery. Intel would be about a year ahead of TSMC in implementing the technology, giving the Intel 18A process a critical advantage as it ramped up production. TSMC planned to introduce backside power delivery in its N2P process node, which would not arrive until sometime in 2026. But the advantage was theoretical—contingent on Intel solving its yield problems and achieving mass production before TSMC launched N2P. If Intel's 18A remained stuck at 10-20% yields through 2025, the backside power delivery advantage would prove meaningless.
Intel's foundry strategy also faced an identity crisis. Fortune described Intel's unresolved challenge: the company remained both a chip designer competing against AMD and NVIDIA and a chip manufacturer competing against TSMC and Samsung. This dual role created conflicts of interest that undermined customer trust in Intel Foundry Services. Potential foundry customers feared that Intel would prioritize its own chip designs for the best manufacturing capacity and process technology, leaving external customers with inferior access.
TSMC faced no such conflict. As a pure-play foundry, TSMC manufactured chips for AMD, NVIDIA, Apple, and hundreds of other customers without competing in chip design. Customers trusted TSMC to treat all customers equally and optimize manufacturing for whoever paid for capacity. Intel Foundry Services could not credibly offer the same assurance while Intel's product divisions competed directly with potential foundry customers.
While Intel struggled with foundry manufacturing, the company suffered an even more devastating failure: missing the AI revolution entirely. GPU maker NVIDIA outpaced Intel in the AI space, capturing an estimated 80% of the AI chip market. The explosion of artificial intelligence highlighted Intel's inability to pivot swiftly as NVIDIA's GPUs became the backbone of AI advancements.
Intel had been building Gaudi chips since 2019, when it acquired Habana Labs. Intel officially unveiled the Gaudi 3 GPU in April 2024 as a competitor to the NVIDIA H100 GPU designed for AI workloads. Intel claimed the Gaudi 3 delivered 50% better inference and 40% better power efficiency compared to the NVIDIA H100—"at a fraction of the cost." The Gaudi 3 was priced at about $125,000 and Gaudi 2 at $65,000, compared to NVIDIA's solutions that could cost over $300,000.
The pricing strategy reflected Intel's competitive positioning: attacking NVIDIA on cost-effectiveness rather than absolute performance. Intel focused on customers looking for cost-effective AI infrastructure solutions rather than enterprises willing to pay premium prices for maximum performance. Intel also emphasized open software ecosystems, teaming up with Google, Qualcomm, and ARM to build software that was not proprietary and could enable companies to easily switch chip providers—a direct attack on NVIDIA's CUDA moat.
But uptake of Gaudi was slower than Intel anticipated. The company would not reach its $500 million revenue target for 2024. In the broader AI chip market, NVIDIA was generating tens of billions of dollars in quarterly revenue from data center GPUs while Intel's AI chip business remained a rounding error. Intel's weak position in the AI chip market contributed directly to the mass layoffs announced in August 2024 and the financial crisis that led to Gelsinger's forced resignation.
The competitive gap extended beyond datacenter AI accelerators. Intel also lagged in AI capabilities integrated into PC processors. Intel's Lunar Lake mobile processors, released in September 2024, delivered up to 48 TOPS (tera operations per second) of AI performance through a dedicated neural processing unit (NPU). But AMD's competing Ryzen AI processors delivered similar NPU performance, and Qualcomm's Snapdragon X Elite processors brought ARM-based AI PC chips to market with strong battery life and AI performance.
More fundamentally, Intel missed the architectural shift toward GPUs and specialized accelerators for AI workloads. Intel's historic dominance derived from x86 CPUs optimized for general-purpose computing. AI training and inference workloads favored massively parallel processing architectures—GPUs with thousands of cores running matrix multiplication operations simultaneously. NVIDIA's decades of GPU development for graphics and gaming positioned NVIDIA perfectly when AI researchers discovered that GPUs dramatically accelerated neural network training.
Intel had no equivalent GPU architecture. The company's discrete GPU efforts—including the Arc GPUs launched in 2022—targeted gaming rather than AI and achieved minimal market share against NVIDIA and AMD. Intel's integrated GPUs in its CPUs provided basic graphics capabilities but lacked the parallel processing power required for AI model training. By the time Intel recognized the AI opportunity and acquired Habana Labs in 2019, NVIDIA had already established CUDA as the standard software platform for AI development and secured dominant market share in AI data center accelerators.
Intel's strategic failures manifested in catastrophic financial performance. The company posted nearly $19 billion in losses in 2024—Intel's first annual loss since 1986. In 2024, Intel generated $53.1 billion in revenue, down from $54.2 billion the previous year. The company reported a net loss of $821 million in Q1 2025—its fourth consecutive quarterly loss. Most dramatically, Intel posted a $16.6 billion net loss in Q3 2024, driven by massive restructuring charges and asset impairments.
Intel's market capitalization collapsed from $502.7 billion in 2000 to just $89.78 billion in April 2025—an 82% decline representing a loss of nearly $413 billion in shareholder value. Intel's stock price plummeted 60% over the course of 2024. In November 2024, Intel was removed from the Dow Jones Industrial Average and replaced with NVIDIA—a symbolic changing of the guard as NVIDIA's market capitalization soared past $3 trillion while Intel became a sub-$100 billion company.
The financial crisis triggered waves of layoffs. In August 2024, Intel announced it was slashing 15% of its staff as part of a $10 billion plan to reduce costs, affecting approximately 15,000 employees. In April 2025, Bloomberg reported that Intel was preparing to lay off up to 20% of its current workforce—roughly 22,000 people given Intel's 108,900 employees at the end of 2024. In July 2025, Intel announced it would shrink its "core employee number" to approximately 75,000, down from 99,500 at the end of 2024—a planned reduction of nearly 25,000 individuals.
The workforce reductions devastated Intel's culture and employee morale. Intel had historically prided itself on providing stable, long-term employment with opportunities for career advancement. Layoffs of 20-25% represented the deepest cuts in the company's history, creating widespread anxiety among remaining employees and damaging Intel's ability to recruit top talent in competition with NVIDIA, AMD, and other semiconductor companies experiencing rapid growth.
Intel's revenue declined 6% year-on-year to $13.28 billion in Q2 2024, and the company registered a net loss of $16.99 billion. The losses stemmed from multiple sources: declining PC processor revenue as the post-pandemic PC boom ended, continued server processor market share losses to AMD, near-zero revenue from the foundry business despite massive capital investments, and failed AI chip products generating minimal revenue compared to projections.
Intel also suspended its dividend—a symbolic break with decades of shareholder returns. The dividend suspension conserved cash for fab construction and operating expenses but signaled to investors that Intel's financial distress was severe. The combination of massive losses, dividend suspension, plummeting stock price, and mass layoffs created the financial context in which Intel's board concluded that Gelsinger's turnaround plan had failed.
Intel's financial crisis reflected catastrophic market share losses in its core data center business. Intel's data center market share by revenue declined from 61% in 2021—the year Gelsinger became CEO—to just 11% in 2024. The 50-percentage-point collapse in three years represented one of the most dramatic competitive reversals in semiconductor industry history.
AMD drove the market share gains. AMD grew its CPU market share from under 10% in 2016 to over 30% by 2024. AMD's EPYC server processors, manufactured by TSMC on advanced process nodes, delivered superior performance-per-watt and performance-per-dollar compared to Intel's Xeon processors. Cloud service providers including Amazon AWS, Microsoft Azure, and Google Cloud Platform deployed AMD EPYC processors at scale, validating AMD's technical performance and eroding Intel's server processor dominance.
The competitive reversal stemmed from Intel's manufacturing failures. Intel's inability to keep pace with advancements in chip manufacturing led the company to fall behind TSMC, which produced the most advanced chips for AMD, Apple, NVIDIA, and other fabless semiconductor companies. While TSMC advanced from 7nm to 5nm to 3nm process nodes on schedule, Intel struggled with its 10nm node for years before finally shipping volume products. By the time Intel launched its Intel 7 process (rebranded from 10nm), TSMC was already shipping 5nm products with superior transistor density and power efficiency.
AMD leveraged TSMC's process technology leadership to design server processors with more cores, higher performance, and lower power consumption than Intel's competing products. AMD's chiplet architecture—disaggregating monolithic processor dies into smaller chiplets connected via high-speed interconnects—provided manufacturing yield advantages and design flexibility that Intel's monolithic designs could not match. AMD could mix and match chiplets manufactured on different process nodes, using advanced TSMC nodes for compute dies and mature nodes for I/O dies, optimizing both performance and cost.
Intel's data center revenue declined accordingly. In fiscal year 2024, Intel's data center and AI revenue fell substantially compared to previous years, reflecting both unit volume declines as customers shifted to AMD EPYC and average selling price pressure as Intel cut prices to remain competitive. The data center business had historically generated Intel's highest profit margins, so revenue and margin compression in this segment created severe financial stress.
Intel also faced competition from ARM-based processors in the data center. Amazon AWS launched Graviton processors—custom ARM-based chips designed by Amazon and manufactured by TSMC—offering cloud customers lower costs and strong performance for web servers, application servers, and containerized workloads. Microsoft and Google similarly developed custom ARM-based processors for their cloud infrastructure. Apple's M-series processors demonstrated that ARM architectures could deliver exceptional performance-per-watt, undermining Intel's decades-old argument that x86 provided inherent advantages over ARM.
The rise of AI workloads further eroded Intel's data center positioning. AI training and inference increasingly ran on NVIDIA GPUs and specialized accelerators rather than general-purpose CPUs. Cloud providers provisioned servers with fewer CPUs and more GPUs, reducing Intel's revenue opportunity per server. Intel's Xeon processors included AI acceleration features, but these could not compete with dedicated AI accelerators from NVIDIA, AMD, and startups like Cerebras and Graphcore.
Intel's board of directors lost confidence in Gelsinger's turnaround plan during the final months of 2024. The 18A yield problems, foundry customer losses, AI market share failure, and continued financial losses convinced directors that Gelsinger's strategy would not succeed. After a contentious board meeting in late November 2024, the board delivered its ultimatum: retire or be removed. Gelsinger chose retirement, effective December 1, 2024.
The board's decision to force Gelsinger's resignation represented an extraordinary failure of both CEO and board. Gelsinger had been hired specifically because of his deep Intel knowledge and semiconductor manufacturing expertise. The board had endorsed his IDM 2.0 strategy and approved the massive capital investments required to execute it. Just three years and nine months later, the board concluded that the same CEO and strategy were fatally flawed. The reversal raised questions about the board's judgment in both hiring Gelsinger and greenlighting IDM 2.0 without sufficient scrutiny of execution risks.
Intel hired executive search firm Spencer Stuart to identify potential successors. The leading external candidates included Lip-Bu Tan, Matt Murphy, and potentially Johny Srouji. Lip-Bu Tan had spent two years on Intel's board of directors between 2022 and 2024 before departing. He previously served as CEO of Cadence, a major electronic design automation (EDA) software company, giving him deep knowledge of semiconductor design and manufacturing ecosystems.
Matt Murphy served as CEO of Marvell Technology since 2016 and worked for more than 20 years at Maxim Integrated before joining Marvell. Intel's board approached Murphy in early discussions, though talks remained in early stages. Murphy's Marvell experience included managing a fabless semiconductor company that designed chips manufactured by TSMC—potentially providing relevant experience if Intel's board decided to separate or spin off the foundry business from the product design divisions.
Johny Srouji, senior vice president of Apple's hardware technologies group, led Apple's custom silicon efforts including the M-series processors that displaced Intel CPUs from Mac computers. Srouji's team designed ARM-based processors manufactured by TSMC that delivered exceptional performance and power efficiency, demonstrating what was possible when world-class chip design combined with TSMC's manufacturing. Hiring Srouji would require convincing him to leave Apple—a difficult proposition given his senior role and Apple's success with custom silicon.
Internal candidates included CFO David Zinsner and Michelle Johnston Holthaus, the interim co-CEOs appointed after Gelsinger's departure. Zinsner brought financial expertise at a moment when Intel desperately needed to control costs and restore profitability. Holthaus led Intel's product divisions and understood the PC and data center businesses. But neither possessed Gelsinger's deep technical background in semiconductor manufacturing, and the board's decision to conduct an external search suggested skepticism that internal candidates could execute a successful turnaround.
The external search broke with Intel tradition. Intel had historically promoted CEOs from within—Bob Swan, Brian Krzanich, Paul Otellini, Craig Barrett, and Andy Grove all spent decades at Intel before becoming CEO. Only Gelsinger's 2021 appointment broke the pattern, and Gelsinger himself had worked at Intel for 30 years before his VMware stint. Hiring a CEO from outside the Intel family—particularly from a competitor like Marvell or Apple—would represent a fundamental break with Intel's culture and operating model.
Gelsinger's forced resignation opened the possibility of fundamental strategic changes that the board had previously been unwilling to consider. The most consequential question: should Intel separate or spin off the foundry business from the product divisions, abandoning the integrated device manufacturing model that had defined Intel since its founding?
The foundry breakup scenario argued that Intel's attempt to be both a leading chip designer and a leading chip manufacturer created irreconcilable conflicts. Chip designers like AMD and NVIDIA would never fully trust Intel Foundry Services while Intel competed against them with x86 CPUs and Gaudi AI accelerators. Investors struggled to value Intel because the foundry business required massive capital investments with uncertain returns while the product business faced intense competition and market share losses. Separating the businesses would create two focused companies: a product company that designed chips manufactured by TSMC and other foundries, and a foundry company that manufactured chips for external customers without competing in chip design.
The product-focused scenario argued that Intel should exit or drastically scale back the foundry ambitions, halting construction of new fabs and focusing capital on chip design and returning cash to shareholders. Intel could become a fabless semiconductor company like AMD, NVIDIA, and Qualcomm—designing chips manufactured by TSMC and Samsung. This would eliminate the capital intensity of fab construction, reduce execution risk from manufacturing yield problems, and allow Intel to leverage TSMC's process technology leadership rather than fighting to match it with internal manufacturing.
The acquisition scenario recognized that Intel's financial distress and strategic challenges might make the company an attractive acquisition target. Potential acquirers could include Qualcomm, Broadcom, or private equity firms willing to break up Intel's product and foundry divisions. Qualcomm reportedly explored acquiring Intel in late 2024, though the deal faced massive regulatory challenges given both companies' positions in the semiconductor industry. A private equity acquisition would allow financial engineers to restructure Intel's operations, divest unprofitable divisions, and extract value through cost-cutting and asset sales.
Each scenario involved wrenching changes to Intel's business model, culture, and strategic positioning. Intel had invested over $100 billion in fab construction based on Gelsinger's IDM 2.0 vision. Walking away from that investment would trigger massive asset write-downs and effectively concede that the strategy had failed. But continuing to pour capital into foundry fabs with 10% yields and minimal customer traction would compound the losses without improving Intel's competitive position.
Pat Gelsinger's career spanned the arc of the semiconductor industry from the PC revolution through cloud computing to artificial intelligence. His technical contributions were genuine and significant. He architected the 80486 processor at age 25, etched his initials into the 80386 microcode, led 14 chip projects at Intel, and served as CTO during Intel's dominance of PC and server processors. He tripled VMware's revenue as CEO and earned recognition as one of America's best CEOs based on employee satisfaction.
His return to Intel as CEO in 2021 represented a genuine attempt to restore American semiconductor manufacturing leadership and revitalize a company that had lost its way. The IDM 2.0 strategy was bold and ambitious. The decision to launch Intel Foundry Services challenged TSMC's dominance and offered American and European customers an alternative to concentrated manufacturing capacity in Taiwan. The investments in U.S.-based fabs created tens of thousands of high-wage manufacturing and construction jobs. The vision of five process nodes in four years set an aggressive roadmap designed to leapfrog competitors.
But vision and execution are different capabilities. Gelsinger's strategy required near-perfect execution across multiple dimensions: achieving yield and cost targets on advanced process nodes, winning major foundry customers, competing successfully in AI chips against NVIDIA, defending data center market share against AMD, and managing through a massive capital investment cycle while maintaining profitability. Intel succeeded at none of these objectives during Gelsinger's tenure.
The 18A process remained stuck at 10-20% yields years after initial development. Intel Foundry Services won no major customers and generated minimal revenue. Gaudi AI accelerators failed to gain meaningful market share against NVIDIA. AMD captured data center market share, growing from 39% in 2021 to 89% by 2024 in market share terms. Intel posted its first annual loss since 1986 and laid off 25,000 employees. The stock fell 60%, destroying tens of billions of dollars in shareholder value.
Gelsinger's defenders argue that three years and nine months was insufficient time to execute a turnaround of Intel's complexity. Semiconductor manufacturing development cycles span 3-5 years from initial research to volume production. Fab construction takes 2-3 years. Customer qualification and design wins require 1-2 years. Expecting Intel to reverse TSMC's decade-long process technology lead in under four years was unrealistic. The board's decision to force Gelsinger's resignation in December 2024 effectively abandoned IDM 2.0 before it had time to succeed or fail on its own merits.
Critics counter that Gelsinger's strategy was fundamentally flawed from inception. Intel could not credibly compete as both chip designer and chip manufacturer. The foundry business required different capabilities, culture, and customer relationships than the product business. Attempting to catch TSMC while simultaneously investing in AI chips, defending against AMD, and building fabs globally spread Intel's resources too thin. A more focused strategy—either committing fully to foundry and exiting product competition, or abandoning foundry and focusing on chip design—might have had better chances of success.
The counterfactual question remains unanswerable: what would have happened if Intel had given Gelsinger five or seven years instead of forcing him out after less than four? Would the 18A yields have improved to commercial viability? Would Intel Foundry Services have won major customers once 18A achieved production readiness? Would Gaudi AI accelerators have gained market share as customers sought alternatives to NVIDIA's premium pricing? Or would additional years have simply compounded the losses and deepened the strategic hole?
Intel's board concluded that the trajectory was not improving fast enough to justify patience. The Q3 2024 loss of $16.6 billion, the continued yield problems, the foundry customer failures, and the AI market share loss convinced directors that the strategy had failed. Whether that judgment proves correct depends on what Intel's next CEO does—and whether Intel as an independent company survives the next decade.
Pat Gelsinger's forced resignation marked more than the end of one CEO's tenure. It represented the potential end of Intel's historic business model—vertically integrated design and manufacturing of leading-edge semiconductors on American soil. For four decades, Intel embodied American semiconductor leadership. The company invented the microprocessor, dominated PC and server computing, and operated the world's most advanced chip manufacturing facilities in the United States.
That era is ending. TSMC manufactures the most advanced chips. NVIDIA dominates AI accelerators. AMD captures data center market share. Apple, Qualcomm, Amazon, Google, and Microsoft design custom chips that displace Intel's standard products. Intel's stock has collapsed, its foundry strategy has failed, and its new CEO—whether Lip-Bu Tan, Matt Murphy, Johny Srouji, or another candidate—will likely pursue fundamental strategic changes including possible foundry separation or fabless transformation.
The implications extend beyond Intel. The U.S. government invested $7.86 billion in CHIPS Act funding in Intel's fab construction based on the premise that Intel represented America's best chance to restore domestic leadership in advanced semiconductor manufacturing. If Intel's foundry efforts fail and the company becomes fabless, that investment will have failed to achieve its strategic objective. The United States will remain dependent on TSMC's Taiwan-based fabs for the most advanced chips, creating the geopolitical vulnerability that the CHIPS Act was designed to address.
Alternative scenarios exist. Perhaps Intel's next CEO will fix the 18A yield problems, win major foundry customers, and vindicate Gelsinger's IDM 2.0 vision with better execution. Perhaps breaking up Intel into separate product and foundry companies will unlock value and create focused competitors that can succeed where the integrated company failed. Perhaps Qualcomm or another acquirer will buy Intel and extract value through restructuring.
But the most likely scenario is continued decline. Intel faces the innovator's dilemma in its purest form: the capabilities that made Intel successful—vertically integrated manufacturing, x86 architecture dominance, PC and server focus—became liabilities when cloud computing, mobile devices, and AI transformed the semiconductor industry. Companies facing the innovator's dilemma rarely execute successful transformations. Kodak failed to survive the digital photography transition. Nokia failed to survive the smartphone transition. Blockbuster failed to survive the streaming transition.
Intel may follow the same path—slowly shrinking from dominant incumbent to niche player as TSMC, NVIDIA, AMD, and ARM-based chip designers capture the semiconductor industry's future. Pat Gelsinger's 40-year Intel career thus becomes a case study in both the rise and fall of American technological leadership: the 18-year-old technician who etched his initials into the 80386, architected the 80486, became CTO, tripled VMware's revenue, returned as CEO to save his company, bet $100 billion on IDM 2.0, and failed.
The story is not finished. Intel's next chapter will determine whether Gelsinger's vision was premature but fundamentally sound, or whether it represented the final gasp of a business model whose time had passed. The answer will shape not just Intel's future but America's position in the global semiconductor industry and the geopolitical balance of technological power in the 21st century.