The Circuit

This episode of The Circuit discusses the recent trend of companies developing their own custom CPUs and accelerators for AI workloads. Google, Amazon, and Microsoft have all announced their own custom chips, while Facebook and Marvell are also entering the custom chip market. The conversation also touches on the challenges and trade-offs of custom silicon, the impact on NVIDIA, and the difficulty of accurately sizing the AI market. The hosts highlight the uncertainty surrounding the future of AI outsourcing and the potential shift in the fortunes of the semiconductor and cloud industries.

SummaryIn this conversation, Ben Bajarin and Jay Goldberg discuss the recent Intel webinar on the foundry business and the financial breakout for Intel Foundry Services (IFS). They analyze the financial details, including the higher costs and lower margins than expected, and the potential for improvement in the future. They also discuss the impact of the CHIPS Act and the subsidies received by TSMC for their Arizona foundries. The conversation concludes with a discussion on Intel's progress under CEO Pat Gelsinger and the opportunities and challenges they face in the data center and AI markets.

In this episode, Ben Bajarin and Jay Goldberg discuss the competition faced by Nvidia in the semiconductor industry. They explore various competitors, including AMD, Intel, and startups like Grok, Etched, and Cerebras. They also delve into the threat posed by custom silicon and the strategies of hyperscalers like Google, Microsoft, Facebook, and Amazon. Overall, the conversation highlights the challenges and opportunities for Nvidia in maintaining its position as a leader in the market. In this conversation, Jay Goldberg and Ben Bajarin discuss various themes related to Nvidia and the AI market. They explore the growing moat of Nvidia and the dominance of CUDA as a software platform. They also discuss the ease of use and stickiness of CUDA, as well as the uncertainty of Nvidia's software adoption. The conversation delves into the market potential and consumer applications of AI, as well as the slow progression of the AI market. They also touch on the risks of AI factories and inventory cycles, the potential slowdown of performance gains, and the regulatory concerns for Nvidia. The conversation concludes with a discussion on the impact of China's market and US sanctions.

Nvidia's recent keynote highlighted their dominance in the data center and their position as a platform company. They offer a full stack solution and are recognized for their sum of the parts story. While some customers may be concerned about getting locked into Nvidia's ecosystem, many appreciate the simplicity and turnkey nature of their offerings. The importance of inference and the transition to generative AI was also discussed, highlighting the complexity of scaling and the need for compute power. However, there may be room for competition in the inference market, particularly in relation to Nvidia's NVLink technology. The conversation covers various topics related to NVIDIA's keynote, including the need for more compute power, the potential of robotics and 6G, the challenges of implementing 6G, the future of software at NVIDIA, and the uncertainty of NVIDIA's software revenue. The conversation also touches on the future of AI software, the computing S-curve, and the longevity of performance gains. Overall, the conversation highlights the excitement and optimism surrounding NVIDIA's advancements in AI and computing.

In this conversation, Ben Bajarin, Jay Goldberg, and Austin Lyons discuss the semiconductor industry and the role of AI accelerators. They cover topics such as the future of Intel, the spectrum of AI accelerators, the dominance of Nvidia in training, the failure of accelerator startups, and the debate between general-purpose and specialized accelerators. They also explore the issue of GPU bloat and the need for alternative solutions in the accelerated computing space. The conversation explores the challenges of porting software to new architectures, the reluctance to switch from Nvidia, the potential for startups to disrupt the AI hardware market, Grok's approach to inference as a service, the need for purpose-built AI architectures, the struggle of custom chip design, and the sustainability of custom chip efforts.

In this episode, Ben Bajarin and Jay Goldberg discuss the thermal problem faced by data centers and other electronic devices. They are joined by Carl Schlachte, the CEO of Ventiva, a company that provides innovative cooling solutions. The conversation covers the origin story of Ventiva, the challenges faced by the company, and the unique technology they offer. They also discuss the traditional cooling technologies in the market and how Ventiva's approach differs. The episode highlights the importance of rethinking design and the potential benefits of Ventiva's solution in various electronic devices. The conversation explores the impact of video communication and the challenges of designing systems for cooling. It delves into the trade-offs between system functionality and thermal management, as well as the market opportunities for thermal management solutions. The discussion highlights the importance of supporting AI in laptops and the thermal challenges faced by laptop manufacturers. It also examines alternative approaches to thermal management and the potential applications in handsets and large format displays. The conversation concludes by emphasizing the advantages of Ventiva's approach and the significance of thermal management in the industry.

SummaryIn this episode of The Circuit, Ben Bajarin and Jay Goldberg discuss the infrastructure side of telecom at Mobile World Congress. They highlight the importance of the show for telecom operators and the focus on networking and telecom infrastructure. The conversation covers the current state of 5G deployment and the upcoming transition to 6G. They also discuss the challenges and concerns in upgrading core networks and the potential costs and timing of 6G. The decline of millimeter wave technology and the use of custom silicon in telecom infrastructure are also explored, along with the advantages and disadvantages of custom chipsets. The conversation covers topics such as AI chip infrastructure in the telecom industry, Qualcomm's underwhelming presence at Mobile World Congress, Qualcomm's AI hub and centralized repository for AI models, Qualcomm's initiative to bring 5G to low-cost handsets, the competition between Qualcomm and MediaTek in the AI space, and Apple's update to the MacBook Air and its positioning as an AI platform.TakeawaysMobile World Congress is an important show for telecom operators and focuses on networking and telecom infrastructure.The transition to 5G is currently underway, with 6G expected to be discussed and developed in the coming years.Upgrading core networks and improving network efficiency are ongoing challenges for telecom operators.The decline of millimeter wave technology and the use of custom silicon in telecom infrastructure are notable trends. AI chip infrastructure is not yet a priority in the telecom industry.Qualcomm's presence at Mobile World Congress was underwhelming.Qualcomm's AI hub provides a centralized repository for AI models.Qualcomm's initiative to bring 5G to low-cost handsets may face challenges in competing with MediaTek.The tension between premium and low-cost devices in the smartphone market continues.Apple is positioning the MacBook Air as an AI platform.

SummaryThis episode of The Circuit discusses the recent Intel Foundry event and Nvidia's earnings. The conversation explores Intel's focus on becoming a leading foundry and its bet on advanced packaging. The hosts also discuss concerns about Intel's culture and execution, as well as the response from TSMC. The episode concludes with predictions about the future of the market and the role of Intel in the industry. In this conversation, Jay Goldberg and Ben Bajarin discuss the future of TSMC and Apple, Intel's commitment to Apple and chiplets, the push towards chiplets in PCs, Intel's relationship with TSMC, NVIDIA's clean quarter, retail interest in NVIDIA stock, the investor relations dilemma for NVIDIA, NVIDIA's product cadence and long-term expectations, and the inference market and competition.TakeawaysIntel is positioning itself as a leading foundry for complex systems of chips in the AI era.The success of Intel's foundry strategy will depend on its ability to deliver on technical innovation and overcome cultural challenges.TSMC remains a strong competitor in the market and has credibility in technical innovation.The customer dynamics and relationships with major wafer scale customers will play a crucial role in Intel's success as a foundry.The future of the market will likely involve a shift towards advanced packaging and chiplet designs. TSMC's future success depends on targeting a big volume platform like IFS in 2027.Intel's commitment to Apple may hinder their adoption and advancement of chiplets.PC CPU makers are under pressure to move towards chiplets, which may influence Apple's architecture decisions.NVIDIA's clean quarter and moderate beat and raise had a fairly moderate reaction in the stock market.NVIDIA's stock is heavily influenced by retail investors, which can lead to unpredictable swings.NVIDIA's product cadence and performance improvements, as well as supply chain constraints, are concerns for meeting revenue expectations.NVIDIA's long-term expectations and communication about the total addressable market may create heightened investor expectations.The inference market is still in early days, and competition is increasing.NVIDIA's software ties and workload advantage may give them an edge in the inference market.

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss the puzzle of Intel foundries and the challenges of determining their revenue. They explore historical economics data points and the exit zone for leading-edge fabs. They analyze the revenue and financial projections for Intel Foundry Services (IFS) and the need for external customers. They also discuss the long-term outlook for IFS and potential scenarios for the future, including strategic partnerships and acquisitions. The episode concludes with the understanding that IFS will require a long-term timeframe and significant investment to become a competitive foundry.TakeawaysDetermining the revenue of Intel foundries is a challenging puzzle due to the lack of access to data and cost information.Historical economics data points provide insights into the exit zone for leading-edge fabs and the challenges faced by companies in staying competitive.IFS revenue projections indicate the need for external customers to fund the investment in leading-edge fabs.The long-term outlook for IFS suggests that it will take time and significant investment to become a competitive foundry.Strategic partnerships and acquisitions may play a role in the future of IFS and the semiconductor industry.

SummaryThe conversation discusses Sam Altman's plan to raise $7 trillion for semiconductors and the potential challenges and motivations behind it. It explores the idea of negotiation and leaked stories as part of the process. The conversation also delves into the simultaneous demand for compute in both hardware and software industries and the need for breakthroughs in technology. It highlights the importance of economic efficiency and competitiveness in the semiconductor industry. The conversation concludes by mentioning upcoming events and guests.TakeawaysSam Altman's plan to raise $7 trillion for semiconductors has sparked discussions about the challenges and motivations behind such a massive fundraising effort.Negotiation and leaked stories may be part of the process as companies navigate the complexities of the semiconductor industry.The simultaneous demand for compute in both hardware and software industries is creating constraints and slowing down innovation.Economic efficiency and competitiveness are crucial factors in the semiconductor industry, and companies need to balance supply and demand.The industry may benefit from breakthroughs in technology and the commercialization of new processes and products.

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss the outsized expectations for AI growth and the impact on stock market reactions. They explore the challenges of modeling AI growth and the difficulty of charging more for AI features. They also discuss the potential for AI to accelerate refreshment cycles and the importance of realistic expectations. The conversation highlights the small gains of AI in software and the experimentation stage of AI. They conclude by emphasizing the need to measure expectations and be reasonable in the AI industry.TakeawaysOutsized AI expectations have led to negative stock market reactions, as companies have not met ambitious growth models.Modeling AI growth is challenging due to variables such as product availability and demand.Charging more for AI features is difficult, as customers may not be willing to pay a premium.AI may not accelerate refreshment cycles, as the average consumer may not see significant improvements that warrant more frequent upgrades.The gains from AI are often small and incremental, but still important in improving efficiency and productivity.

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss Intel's recent earnings call and the challenges the company is facing. They analyze the guidance provided by Intel and the concerns raised by investors. The conversation also delves into Intel's core businesses, including the consumer and data center groups, and the impact of competition from companies like Nvidia. The hosts explore Intel's partnership with UMC to increase trailing edge capacity and the potential for specialized trailing edge processes. They conclude by highlighting the need for Intel to prove itself in the coming quarters and address ongoing challenges.TakeawaysIntel's Q4 earnings call and guidance were a source of concern for investors, leading to a sell-off.The company's core businesses, particularly the data center group, face challenges and increased competition.Intel's partnership with UMC to increase trailing edge capacity signals a focus on specialized processes.The company needs to prove itself in the coming quarters and address ongoing challenges to regain investor confidence.

SummaryIn this conversation, Ben Bajarin and Jay Goldberg discuss the latest news and trends in the semiconductor industry, with a focus on TSMC's strong earnings and growth forecast. They explore the importance of AI in TSMC's business and the strong position of NVIDIA in the AI market. They also discuss the challenges of designing AI-first silicon and the need for differentiation in AI accelerators. The conversation touches on the dynamics of leading-edge manufacturing, the transition from financial objectives to innovation in semiconductors, and the significance of Supermicro's pre-announcement. They conclude by emphasizing the importance of delivering on promises and providing predictable guidance in the industry.TakeawaysTSMC's strong earnings and growth forecast indicate a healthy year for the semiconductor industry, driven by demand for AI and leading-edge manufacturing.NVIDIA is well-positioned in the AI market, with strong engagement from AI leaders and a focus on AI-specific processors.The challenges of designing AI-first silicon and the lack of re-architecting for AI in most designs highlight the need for innovation in the industry.The transition from financial objectives to innovation and R&D is crucial for semiconductor companies to stay competitive.The temptation to talk about AI without substance and the importance of delivering on promises and providing predictable guidance are key considerations for companies in the industry.

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss their experience at CES and the presence of AI in various products. They explore the evolution of massage chairs and the AI wash in the industry. They also analyze the two tracks of AI at CES, with a more muted presence in big vendor booths and a greater focus in Eureka Park. The conversation delves into the role of AI in semiconductor vendors and the importance of on-device AI. They discuss the balance between cloud and edge AI and the outlook for semiconductors in 2024. They also touch on the revival of VR and AR and the potential for exciting developments in the future.TakeawaysCES featured a more muted presence of AI compared to previous years, with a focus on specific use cases rather than generative AI.The massage chair industry has seen advancements in AI sensing and 5D rollers, but there is still an AI wash in the labeling of products.Semiconductor vendors are re-examining their roadmaps to incorporate more on-device AI, recognizing the need for advanced sensor fusion and compute at the edge. The balance between cloud and edge AI is becoming more symbiotic, with a recognition that both are necessary for a comprehensive AI strategy. The outlook for semiconductors in 2024 is mixed, with a cyclically good year expected but uncertainty about significant advancements in consumer-facing AI.Chapters00:00 Introduction and CES Experience01:09 The Fascination with Massage Chairs05:02 The Muted Presence of AI at CES26:06 The Compute Happening at the Edge27:29 The Outlook for Semiconductors in 2024

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss their hopes and predictions for the semiconductor industry in 2024. They cover topics such as the need for clarity of direction in automotive semis, the potential for partnerships and spin-offs in the industry, the importance of venture funding for semiconductor startups, the need for a more balanced and realistic narrative around AI, and the challenges and tensions between the US and China in the chip industry. They also make some speculative predictions about potential acquisitions and consolidation in the industry.TakeawaysThe automotive semiconductor industry needs more clarity of direction to understand where it is headed.Partnerships and spin-offs could be a productive way for semiconductor companies to work together and fill gaps in their offerings.There is a need for more venture funding for semiconductor startups, as the industry is capital-intensive and often overlooked by investors.The narrative around AI needs to change to avoid overhyping and misunderstanding the technology.Reducing tensions between the US and China in the chip industry would benefit both countries and the global semiconductor ecosystem.

SummaryIn this episode, Ben Bajarin and Jay Goldberg discuss recent semiconductor events, including Marvell's analyst day, AMD's Advancing AI event, and Intel's AI Everywhere event. They analyze the key takeaways from each event, including Marvell's focus on data infrastructure, AMD's new chiplet SOC and AI capabilities, and Intel's architectural advancements in client and data center products. The hosts also discuss the challenges and opportunities in the AI PC market and share their holiday wishes.TakeawaysMarvell is a data infrastructure company with strong product roadmaps and a focus on the data center.AMD's new chiplet SOC, Meteor Lake, represents a significant architectural advancement and could be a game-changer for the company.Intel's AI Everywhere event showcased their advancements in client and data center products, including chiplet-based SOCs and AI integration.The AI PC market presents challenges in terms of consumer adoption and pricing, but has potential for growth in enterprise and cloud applications.

SummaryIn this conversation Ben Bajarin and Jay Goldberg have a discussion with Rene Haas, President of ARM, provides an overview of the company and its role as the brain that powers electronic devices. He explains ARM's business model of designing CPUs and licensing them to chip manufacturers. Haas discusses the value capture and increasing ARM's share of the value in the market. He highlights the growth drivers for ARM in various industries, including PCs, cloud data centers, networking, automotive, and IoT. Haas also addresses the impact of US sanctions on China and ARM's success in the Chinese market. He concludes by discussing ARM's partnership with Intel Foundry Services and the potential for chiplets.TakeawaysARM is the brain that powers electronic devices and is present in a wide range of products, including cell phones, laptops, automobiles, data centers, and IoT devices.ARM's business model involves designing CPUs and licensing them to chip manufacturers, allowing them to build chips with ARM's designs.To increase its share of the value in the market, ARM is focusing on delivering full solutions, accelerating chip design, and developing new products for growing markets.ARM sees opportunities for growth in industries such as PCs, cloud data centers, networking, automotive, and IoT, driven by trends like AI and the need for more compute power.ARM is navigating US sanctions on China by complying with export control restrictions and focusing on areas where its software ecosystem and partnerships are strong.

The conversation explores the current state of the Japanese semiconductor industry and its history of decline in the 1990s. It discusses the structure of the industry, key players, and recent developments such as the establishment of Rapidus, a consortium aimed at advancing semiconductor fabrication in Japan. The conversation also touches on the visit of Japan's Minister of Economy, Trade and Industry to UC Berkeley and the potential for collaboration between the US and Japan in the semiconductor industry. Overall, the discussion highlights the renewed interest and efforts in revitalizing the Japanese semiconductor industry. The Japanese semiconductor industry experienced a decline in the 1990s but is now showing signs of revitalization. Key players in the Japanese semiconductor industry include Tokyo Electron, Renesas, Sony, and Morata. The establishment of Rapidus, a consortium for semiconductor fabrication, demonstrates Japan's commitment to advancing its semiconductor capabilities. Collaboration between the US and Japan in the semiconductor industry is becoming increasingly important.

In this week's episode, Ben Bajarin and Jay Goldberg chatted about companies pushing the limits of what is possible with technology and breaking new ground with new products and experiences.  They discuss the impacts this has on the semiconductor landscape and why its a necessary, even if awkward, step to progress.

Ben Bajarin and Jay Goldberg dive into the recent announcements from Microsoft around their new custom silicon chips. They also discuss MediaTek's ambitions, shared from the MediaTek executive summit last week, where they are doubling down on helping people build custom silicon for the data center.