Intel CEO Chris Clark: From Cadence's Turnaround to Rebuilding System-Level AI Infrastructure

Woofun AI reports that Intel CEO Chris Clark articulated a strategic pivot toward system-level AI infrastructure during a Tech Surge podcast hosted by Celesta Capital, emphasizing that the semiconductor industry's competitive frontier has expanded beyond individual chips to encompass packaging, storage, interconnection, cooling, and the entire software stack. Clark, who previously orchestrated the turnaround of Cadence, argues that Intel must leverage its vertical integration capabilities to address the complex bottlenecks of modern AI workloads, moving away from a purely product-centric view to a holistic platform approach.

This perspective was shaped by his long-standing investment philosophy and his observation that the value of hardware is being repriced as AI drives new demands for hash rate, power efficiency, and bandwidth. The conversation with host Michael Marks revealed Clark's intent to rebuild Intel's relevance by reconnecting with universities, venture capital firms, and startups, ensuring the company does not miss the next technological wave after previously falling behind in mobile internet, cloud computing, and AI.

The fundamental shift in semiconductor valuation stems from a broader repricing of hardware value, driven by the constraints of AI infrastructure. Over the past 20 years, software companies received higher valuations and more venture capital, while semiconductors were often viewed as unsuitable for venture investment due to long R&D cycles, heavy capital requirements, and limited exit pathways.

Clark recalled that 20 years ago, when visiting top venture capital firms, the entire partner team would usually be present at the start of meetings, but once he began discussing semiconductors, half of them would politely find excuses to leave, leaving only a few to continue listening "out of sympathy." At the end of these meetings, investors would often ask him, "Do you have any startups in software or services?"

This reflected a mainstream view that semiconductors were a declining industry, with capital flowing into software and internet services.

However, Clark maintained that chips remain the foundational element of the tech industry, arguing that without suitable performance, power consumption, and cost, many upper-layer applications cannot exist. Today, AI has turned hash rate, power consumption, and bandwidth into direct constraints on application expansion, making chips infrastructure that determines model costs and commercialization boundaries. This change is not merely a valuation shift but a repricing of capital driven by technical bottlenecks. Clark noted that some co-investors once asked him, "Can you name a semiconductor company with a market cap of over $1 trillion?" Today, this question no longer holds true, as NVIDIA and other semiconductor firms have achieved such valuations, signaling a return of hardware as the core of tech investment.

AI competition is evolving from focusing on individual accelerators to system-level engineering, with bottlenecks expanding beyond GPUs to CPUs, storage, high-speed interconnections, advanced packaging, and cooling solutions. The previous round of AI investment centered around GPUs and model training, but as workloads shift toward inference, agents, and physical AI, the limitations of individual chips are becoming apparent. Clark emphasized that cluster efficiency can no longer be improved solely by enhancing the performance of individual chips, leading to a distribution of hardware value across various weak links in the computing system.

This trend is evident in the shift from air cooling to liquid cooling and even microfluidic cooling, from electrical interconnections to photonics, and from traditional packaging to glass substrates and new materials. Clark believes that the next wave of hardware value may not be concentrated among top GPU companies but could be distributed across these peripheral infrastructure components. He pointed out that as AI clusters grow larger, data transmission between chips, servers, and cabinets becomes more critical than individual chip performance.

This insight led him to invest in companies like Credo Semiconductor and Astera Labs, which focus on high-speed interconnection, and to explore photonic interconnection technologies. Some of these companies were later acquired by firms such as Marvell and Credo, highlighting the strategic importance of interconnects in AI infrastructure. The shift also reflects a broader recognition that AI hardware competition is now determined by the synergy of computing, storage, interconnection, packaging, manufacturing, and organizational capabilities.

Intel's commitment to vertical integration is not about retaining as many businesses as possible but about reorganizing dispersed capabilities into cohesive platforms. In the past, the separation of design and manufacturing led to the rise of fabless models and specialized foundries, but Clark argues that Intel's simultaneous involvement in product development and chip manufacturing allows for cross-layer collaboration that is essential in the AI era.

He emphasizes the combination of CPUs, GPUs, software, advanced packaging, and chip manufacturing because product optimization increasingly relies on jointly optimizing architecture, packaging, and processes based on customers' specific workloads. This approach goes beyond mere in-house manufacturing; it involves creating greater value for customers through platform-based solutions. Clark acknowledges that vertical integration increases organizational and capital allocation complexity, and if product competitiveness and manufacturing execution cannot improve simultaneously, it might further increase costs.

However, he believes that future computing competition will not be about comparing individual chips but rather the synergy of entire systems. Intel could theoretically move further toward an outsourcing model and stop manufacturing chips in-house, but Clark still believes in vertical integration because it enables the company to address the complex requirements of AI workloads. This strategy requires attracting top CPU architects, GPU architects, system architects, and software talent to build full-stack capabilities from chips to systems to software. The goal is to create a larger platform that covers computing, interconnection, packaging, and manufacturing, rather than a set of separate products.

The relationship between CPUs and storage is being redefined by AI, with new architectural approaches emerging to address the demands of inference and agent operations. In the past, CPUs were Intel's most stable core business, while storage was often viewed as a commodity with volatile prices. As AI moves from training to inference, the demand for general-purpose computing remains, and CPUs continue to handle data processing, task scheduling, and agent operations. At the same time, memory bandwidth, capacity, and power consumption have become critical constraints on system performance.

Clark mentioned CPU-storage stacking and new storage architectures, but this does not necessarily mean Intel is returning to the traditional storage market. Instead, it reflects the need for better coordination between computing and memory at the architectural and packaging levels. Clark noted that Intel is researching CPU-storage stacking and new storage architectures, but the company is not yet ready to disclose specific plans. He has hired Lee Seok-hee, former CEO of SK Hynix, to lead these efforts, signaling a strategic focus on integrating storage into Intel's system architecture.

Clark explained that he was not keen on investing in memory chips in the past because traditional storage products have strong commodity characteristics, but as AI computing imposes new demands on bandwidth, capacity, power consumption, and packaging, storage is evolving from a standardized component to a key part of system performance. New technologies are transforming the storage industry, and exploring new storage architectures has become one of Clark's areas of focus. This approach aims to create a new system architecture where CPUs, storage, packaging, and manufacturing form a cohesive unit, rather than simply competing in the traditional storage market.

Woofun AI data shows that Clark's investment philosophy and early bets on alternative AI computing architectures reflect his belief that true investment opportunities often lie in underappreciated bottlenecks. He has been investing in chips since 1987, with a total of nearly 550 companies receiving his investment. For a long time, semiconductors were not a favored area for venture capital firms, but Clark remained a contrarian investor, believing that chips are the foundational element of the tech industry. He mentioned that in the past, some co-investors had asked him, "Can you name a semiconductor company with a market cap of over $1 trillion?" Today, this question no longer holds true, as some of the world's most valuable tech companies are semiconductor firms.

However, Clark's focus is not just on large companies like NVIDIA; he believes that semiconductors represent a vast technical ecosystem, with many key innovations coming from lesser-known small companies. These companies are working to reduce the power consumption of Chiplet designs, address high-speed interconnection issues, and bet on photonics, advanced packaging, and new cooling materials. Clark's perspective on AI hardware is a reflection of his investment approach. Having invested in graphics chip companies like S3 in the early days, he recognized early on the high power consumption problem of GPUs.

He also believed that as AI moved from model training to actual deployment, the market size for inference and agent AI could far exceed that of training. About nine to ten years ago, he backed two different computing architecture approaches: Cerebras' wafer-scale chip solution and SambaNova's RDU, or Reconfigurable Data Unit. Clark thought Cerebras' technology was very difficult to implement, but the problem it aimed to solve, as proposed by founder Andrew Feldman, was worth supporting, so he began investing from the seed stage. SambaNova's dataflow architecture sought to reduce power consumption while maintaining computing performance, offering an alternative path for AI computing beyond GPUs.

Specific investment details in SambaNova highlight the importance of adaptable teams in the semiconductor industry. In 2017, driven by Clark, Celesta made its first investment in SambaNova, pouring $2 million into the company, which was valued at around $12 million at the time. He subsequently participated in multiple rounds of financing for SambaNova and helped bring in new investors. Clark said that SambaNova is pursuing Series F financing, with funding targets ranging from $800 million to $1 billion.

This refers to the amount of funding raised, not the company's valuation. He emphasized that when investing in startups, one should not rely solely on a single founder but look for teams capable of continuously adjusting their strategies. Since markets change, about nine out of ten companies he invested in altered their initial business plans during development. Those truly worthy of long-term support are teams that can adapt to changes, establish a strong culture, and ultimately build world-class enterprises.

Clark's experience with SambaNova illustrates his belief that the ability to pivot and adapt is crucial for success in the fast-evolving semiconductor industry. This approach contrasts with the traditional view of semiconductor companies as rigid and slow-moving, highlighting the potential for agility and innovation within the sector. By supporting companies like SambaNova, Clark aims to foster a ecosystem of innovation that can address the complex challenges of AI infrastructure, from power consumption to interconnection and packaging.

Clark's turnaround of Cadence demonstrates the importance of humility, listening, and shifting from supplier to partner in rebuilding corporate culture and product strategy. When Clark took over at Cadence, the company's stock price had dropped to around $2.42. Initially, he agreed to serve as interim CEO for just three months while the company searched for a permanent leader, but those three months turned into 15 years. During this time, Cadence underwent a transformation in corporate culture and product strategy, and its stock price rose significantly from its lows.

Clark summarized the essence of this experience in three words: humility, listening, and responsiveness. Upon taking on the role of CEO, he told employees in a company-wide meeting, "This is my first time as CEO. If you have any good ideas, feel free to send them to me." After that, he received around 300 emails per day and responded to each one. For suggestions that warranted further exploration, he would go directly to the employees' desks to discuss them. This approach helped him identify internal information gaps and allowed management to hear genuine feedback from the front lines.

Customer relationships also needed to change. Clark recalled that some of Cadence's customers were extremely angry, demanding refunds and clearly stating they didn't want to use the company's products anymore. Others complained that their product issues were never addressed until contract renewals approached. As a result, Clark pushed Cadence to establish a rapid response mechanism. Later, one customer told him that less than 24 hours after filing a complaint, someone came to the office to resolve the issue.

One of Cadence's main competitors once told him, "The same customer sees me as a supplier, but you as a partner." In Clark's view, this is the key difference between the two types of relationships. Only when customers regard a company as a partner are they willing to share product roadmaps and genuine needs. Companies can also use feedback from other customers to provide more valuable advice. He is bringing this approach to Intel, though Intel's business is more complex—the company must not only rebuild its product competitiveness but also ensure success in its chip manufacturing business.

Intel's ecosystem strategy focuses on interconnects, cooling materials, and long-term platform building, leveraging both internal development and external partnerships. Clark believes that as AI computing scales up, bottlenecks are spreading from the chips themselves to peripheral infrastructure, including high-speed interconnection and cooling. Based on this insight, he invested in companies like Credo Semiconductor and Astera Labs and also explored the field of photonic interconnection.

Some of these companies were later acquired by firms such as Marvell and Credo, highlighting the strategic importance of interconnects in AI infrastructure. The next challenge is cooling. As the power consumption of CPUs and other AI chips continues to rise, cooling solutions are shifting from air cooling to liquid cooling and even further to microfluidic cooling. The same logic applies to advanced packaging.

Intel already has packaging technologies like EMIB-T, but Clark is also interested in new materials such as glass substrates and synthetic diamonds to improve the packaging, insulation, and cooling capabilities of high-performance chips. These technologies don't necessarily all have to become independent businesses within Intel. Clark's approach is to let Intel develop technologies internally where possible; those that aren't suitable for in-house development can first be supported by external startups, which could eventually be integrated into Intel's platforms through partnerships, acquisitions, or other means.

What Intel needs to build is not a set of separate products but a larger platform that covers computing, interconnection, packaging, and manufacturing. Intel was founded in 1968 and initially focused on memory chips, not microprocessors. When asked whether Intel might return to the storage market, Clark didn't reveal specific plans but sent a notable signal: Intel is researching CPU-storage stacking and new storage architectures. Clark has known industry figures such as Micron CEO Sanjay Mehrotra and NVIDIA CEO Jensen Huang for many years.

Today, there is both competition, investment, and cooperation among these companies. NVIDIA has become an investor in Intel, and the U.S. government and SoftBank are also among Intel's shareholders. But Clark says he doesn't simply view these people as competitors. The market is big enough, and the more important question is how to jointly create an even larger market. For Intel, the real challenge isn't to briefly lead in a particular product cycle but to re-enter areas where cutting-edge technologies keep emerging.

This means connecting with university professors, startups, venture capital firms, and AI labs to find opportunities in new materials, computing architectures, and system bottlenecks. Clark's time horizon is long-term; when joining Intel, he told the board that he isn't someone who focuses only on the short term. He is thinking about how Intel can build a larger platform in 10 or 15 years and how this platform can truly benefit the entire industry.

Intel's commitment to not missing the next technological wave is driven by the recognition that it has missed major waves such as mobile internet, cloud computing, and AI. Clark stated that his goal now is: "From now on, I won't miss any major wave again." This determination stems from his belief that Intel is a landmark company with significant importance to the semiconductor industry and the United States, and he wants to get involved personally and make a real impact. To achieve this, Intel must reconnect with universities, venture capital firms, startups, and AI labs to restore its sense of awareness and ability to respond quickly to external changes.

Clark argued that when a large tech company misses a technological wave, it's usually not because it completely fails to see new directions but because it fails to translate external changes into internal resource allocation and product decisions. By fostering a culture of humility, listening, and responsiveness, Intel can rebuild its relationships with customers and partners, transforming from a supplier to a partner who can share a roadmap. This approach will enable Intel to identify and address the complex bottlenecks of AI infrastructure, from computing and storage to interconnection, packaging, and cooling.

The focus is no longer merely on whether Intel can undergo corporate transformation but rather on whether a traditional chip giant can regain the ability to participate in the next generation of computing platforms. As Clark said, Intel has missed major waves such as mobile internet, cloud computing, and AI. Now, he wants to ensure one thing: when the next wave arrives, Intel must be there. This marks a critical juncture for Intel, as it seeks to leverage its vertical integration and ecosystem strategy to compete in the system-level AI infrastructure market, ensuring that it remains relevant in the rapidly evolving tech landscape.