I. Introduction
The AI chip market is witnessing rapid growth, driven by the escalating demand for AI applications across various industries. This surge in demand is largely attributed to the rise of AI and its transformative potential across various sectors.
The market is projected to reach $53.5 billion in 2023 and is expected to increase by nearly 30% in 2024. This growth is causing a significant shift in the chip market, with governments, tech giants, and startups vying for a share of this expanding market.
The competition is driven by the pursuit of faster AI training, more efficient AI deployment, and control over this increasingly critical technology.
Jump to Sections
II. The Technical Side of AI Chips
- Semiconductors: Semiconductors form the base materials for chips. Semiconductors have properties that can be manipulated through manufacturing to control the flow of electricity. This makes them crucial for constructing the circuits found in chips. The most common semiconductor material is silicon, but other materials like germanium are also used. Over 90% of semiconductors utilize silicon.
- Chips: Chips, also known as integrated circuits (ICs), are manufactured from semiconductors. They contain millions, or even billions, of transistors and other components that perform electronic tasks. Chips can be designed for general or specific purposes.
- Processors: These are specific types of chips.
- CPUs (Central Processing Units): These act as the “brains” for most computers. CPUs are responsible for handling a wide range of instructions and general-purpose tasks.
- GPUs (Graphics Processing Units): GPUs are particularly adept at parallel processing, which makes them well-suited for complex AI calculations.
- ASICs (Application-Specific Integrated Circuits): ASICs are custom-designed for specific purposes, like processing AI algorithms, offering enhanced performance and efficiency for those tasks.
- Memory and Storage:
- RAM (Random Access Memory): RAM is a form of memory utilized by AI systems to store data for active processes.
- Cache: Cache is a specialized type of memory that is even faster than RAM. Cache is primarily used for the storage of frequently accessed data.
- SSDs (Solid-State Drives) and HDDs (Hard Disk Drives): SSDs and HDDs are examples of storage that can be used to store the large datasets required for AI model training.
Specialized Architectures: AI chips often incorporate features designed to accelerate AI algorithms.
These features include parallel processing units, high-bandwidth memory, and data paths optimized for AI operations.
AI chips commonly include high-bandwidth memory, such as HBM (High-Bandwidth Memory).
They often have multiple processing units, such as TPUs (tensor processing units) or GPUs.
AI chips are designed to efficiently perform the specific mathematical operations used by AI algorithms. These chips prioritize power efficiency to support deployment in environments with limited power resources, such as edge devices.
Open architectures like RISC-V are used for many AI chips to enable flexibility and customization.
AI chips are also scalable, allowing for the combination of multiple chips to manage larger AI workloads. High-speed interconnects and distributed processing frameworks contribute to the scalability of AI chips.
III. Demand for AI Chips Across Various Sectors
The rapid advancement and adoption of AI are fueling an unprecedented demand for chips. Training and running complex AI models require massive computational power, creating an insatiable appetite for high-performance chips like GPUs. This demand is further amplified by the proliferation of AI applications across diverse sectors, from self-driving cars to personalized medicine.
- The demand for AI chips extends beyond traditional tech giants to industries like fast fashion, lawn care, and agriculture, where AI models are being used for tasks like document summarization, customer service, and crop optimization.
- Companies providing cloud computing services, such as Amazon, Microsoft, Google, and Alibaba, are investing heavily in AI chip development to meet the growing demand for AI applications.
- The automotive, aerospace, and defense industries are significant consumers of AI chips, particularly for applications like autonomous vehicles, drones, and military systems, driving demand for chips with high processing power and efficiency.
- The rise of AI PCs, smartphones, and IoT devices, which incorporate specialized neural processing units (NPUs) for AI tasks, is expected to drive further demand for AI chips in the coming years..
- Emerging applications like quantum computing, neuromorphic computing, and brain-computer interfaces are also poised to create new demand for specialized AI chips in the future.
IV. Key Players and their Strategies
Nvidia
- Market Share: Nvidia controls a significant portion of the AI chip market, holding between 70% and 95% of the market share.
- Key Products: Nvidia’s dominance stems from its powerful GPUs, such as the H100 and A100, and its CUDA software.
- Factors Contributing to Success: Nvidia’s AI accelerators have been instrumental in AI training, making them essential for companies involved in machine learning and generative AI. The company recognized the AI trend early on and provided high-performance solutions for AI applications. Additionally, Nvidia offers a comprehensive package of AI solutions, including chips, software, and access to specialized computers.
- Challenges and Competition: Despite its dominance, Nvidia faces competition from several fronts:
- Alternative Software Solutions: Rivals and customers are collaborating on alternative software solutions to challenge Nvidia’s proprietary CUDA software. OpenAI’s Triton, an open-source software released in 2021, aims to break Nvidia’s near-monopoly by enabling AI applications to run on various chips, including those from AMD and Intel. Major tech companies like Meta, Microsoft, and Google, significant investors in Nvidia’s hardware, are contributing to Triton while developing their own AI chips.
- Tech Giants Building In-House AI Chips: Amazon, Microsoft, Google, Meta, and Apple are developing their own AI chips to increase profit margins and reduce reliance on Nvidia. Amazon started in 2015 by acquiring Annapurna Labs, followed by Google in 2018 with its TPUs, and Microsoft in November 2023 with its AI chips. Meta unveiled a new version of its AI training chips in April 2024, while Apple has been designing its own AI chips for its devices since 2017.
- Competition from Existing and Emerging Chip Makers: Established rivals like Intel and AMD are developing their own AI chips and software platforms. Intel, for instance, has released its Gaudi 3 AI accelerator, projected to be more affordable and power-efficient than Nvidia’s offerings. Startups and smaller companies, such as Graphcore, Cerebras, and SambaNova Systems, are also entering the market, focusing on specific aspects of AI chip design and architecture.
- Strategic Responses to Competition: Nvidia has taken steps to maintain its market position:
- Increased Investment in Software: Recognizing the importance of software, Nvidia has been investing heavily in software development, hiring more software engineers than hardware staff.
- Focus on High-Performance Solutions: Nvidia continues to focus on delivering high-performance solutions that meet the evolving needs of AI applications.
- Collaboration with Alternative Cloud Providers: Nvidia has reportedly provided preferential access to its GPUs to alternative cloud providers like CoreWeave, recognizing their growing role in the AI chip market.
- Future Outlook: While Nvidia is expected to maintain a substantial market share in the AI chip market for the foreseeable future, the emergence of competitive hardware and software ecosystems poses challenges. Factors such as maintaining a balance between performance, efficiency, and cost, along with ensuring compatibility and seamless integration with existing systems, will be crucial for Nvidia’s continued success. The evolving landscape of AI chip manufacturing, with increasing competition and government investments, suggests a potential shift towards a more balanced market in the future.
TSMC
- Market Share: TSMC holds a dominant position in the semiconductor industry with a global market share exceeding 55%. This makes TSMC the world’s largest contract chipmaker, responsible for supplying more than half of the world’s chips. In the first quarter of 2024, TSMC held a 62% share of the global foundry market.
- Production Locations: Currently, TSMC primarily manufactures its most advanced chips in Taiwan. However, TSMC is also building a facility in Arizona.
- Production Timeline: TSMC’s Arizona plant is expected to begin producing 3nm chips in 2025. As of May 2024, TSMC is mass-producing next-generation 5nm and 3nm chips in Taiwan.
- Customers: TSMC manufactures chips for various companies, including major players like Apple and Nvidia.
Intel
- Market Share: While historically dominant in the PC market, Intel has lost ground in recent years. In 2020, AMD held just above 10% of the server market share, while Intel held the remainder. By the first quarter of 2024, AMD’s server market share had increased to 23.6%.
- AI PC Market Projections: Intel estimates that AI PCs will constitute 60% of the AI market by 2027.
- Neural Processing Unit (NPU) Usage Projections: Intel projects that the use of NPUs will increase from approximately 25% in 2024 to 30% in 2025.
- Strategic Focus: Intel is actively working to regain its position in the AI market through several initiatives:
- Data Center Market: Intel has released its Gaudi 3 AI accelerator to compete with Nvidia and AMD in the data center market. The Gaudi 3 accelerator kits aim to deliver high performance at a lower cost compared to competitors.
- AI PC Market: Intel is developing Lunar Lake processors specifically designed for next-generation AI PCs. These processors are projected to offer up to 40% lower power consumption compared to the previous generation. Lunar Lake processors are expected to ship in the third quarter of 2024 and will power over 80 different AI PC designs from 20 original equipment manufacturers (OEMs). Intel aims to deploy over 40 million Core Ultra processors, including Lunar Lake, in 2024.
AMD
- Release Strategy: AMD is actively competing with Nvidia by introducing new AI chip models annually.
- Product Offerings: In direct competition with Intel, AMD released the Ryzen AI 300 series for AI laptops and the Ryzen 9000 series for desktops. Notably, the Ryzen AI 300 series will also power laptops featuring Microsoft’s Copilot chatbot.
- Performance Claims: AMD claims its Ryzen 9000 series will be the “world’s fastest consumer PC processors” for both gaming and content creation.
- Launch Timeline: Both the Ryzen AI 300 and Ryzen 9000 series are expected to become available in July 2024.
- Market Share: One source states that AMD held slightly above 10% of the server market share in 2020. By the first quarter of 2024, AMD’s server market share had risen to 23.6%. In the overall x86 CPU market, AMD achieved a record-high market share in the first quarter of 2024, reaching 23.9%, an increase from 19.8% in the fourth quarter of 2023. However, this source does not specify how much of this share can be attributed to AI-specific chips.
Big Tech
- Reliance on Nvidia: Big tech companies like Amazon, Microsoft, and Google rely significantly on Nvidia’s chips for their data centers. This is partly due to customer preference for Nvidia’s top-of-the-line chips.
- Market Share: Nvidia commands a near-monopoly in the advanced AI training chip market, with a company value larger than the GDP of 183 countries. Sources estimate that Nvidia holds between 70% and 95% of the AI chip market share.
- Efforts to Reduce Dependency: Despite their reliance on Nvidia, these tech giants are actively trying to decrease their dependency by developing their own AI chips to potentially reduce costs and increase their profit margins.
- In-House Chipmaking Efforts:
- Amazon: Began in-house chipmaking efforts in 2015 with the acquisition of Annapurna Labs.
- Google: Launched its own AI chips, known as TPUs, in 2018.
- Microsoft: Introduced its first AI chips in November 2023.
- Meta: Unveiled a new version of its AI training chips in April 2024.
- Partnerships: Alongside in-house chipmaking, these companies are investing in partnerships to offer customers more choices.
It is important to note that while trusted numerical data about the market share and production volumes of Big Tech companies’ in-house AI chips are not available, their efforts indicate a strategic move to decrease reliance on Nvidia and potentially reshape the AI chip market landscape.
V. Alternative Cloud Providers
- CoreWeave:
- Funding and Expansion: CoreWeave secured $2.16 billion in venture capital and $2.3 billion in debt financing, facilitating rapid expansion from 3 to 14 data centers. The company raised a total of $5 billion in debt and equity financing, with the most recent funding round in May 2024 valuing the company at $19 billion.
- GPU Capacity: Leveraging the substantial funding, CoreWeave amassed approximately 47,600 GPUs across its data centers. The company received an allocation of 40,000 Nvidia H100 GPUs in 2023, further enhancing its capacity for AI workloads.
- Pricing and Revenue: CoreWeave offers an Nvidia A100 40GB GPU for $2.39 per hour, significantly lower than Azure and Google Cloud. This equates to a monthly cost of $1,200 on CoreWeave versus $2,482 on Azure and $2,682 on Google Cloud. Based on an estimated rental income of $5,270 per GPU over four years with a blended average of on-demand, one-year, and three-year instances, CoreWeave is projected to generate $15.68 billion in rental income over four years from its current GPU fleet.
- Competition: CoreWeave, initially a cryptocurrency mining operation, has emerged as a significant competitor in the alternative cloud market, particularly for generative AI workloads. Its growth is attributed to offering competitive pricing and potentially better availability of GPUs compared to larger cloud providers. However, the sustainability of this growth depends on securing a consistent supply of high-volume GPUs and navigating potential competition from custom AI hardware developed by companies like Google, Microsoft, and AWS.
- Lambda Labs:
- Funding: Lambda Labs received $320 million in funding in December 2023 and has raised $932.2 million in total funding since 2017.
- GPU Capacity: Lambda Labs obtained an allocation of 20,000 Nvidia H100 GPUs in 2023, further enhancing its capacity for AI workloads.
- Market Position: Lambda Labs initially focused on AI cloud services, then shifted to AI system manufacturing before returning to its focus on the GPU cloud market.
- General Trends:
- Demand and Growth: The alternative cloud provider market is experiencing a surge in demand due to the increasing need for GPU capacity to train and run generative AI models. This demand has fueled rapid expansion and significant venture capital investments in companies like CoreWeave and Lambda Labs.
- Sustainability Challenges: The long-term sustainability of this growth hinges on the ability of alternative providers to secure a consistent high-volume supply of GPUs at competitive prices. They also face potential competition from established cloud providers developing custom AI hardware, such as Google’s TPUs, Microsoft’s Azure Maia and Azure Cobalt, and AWS’s Trainium, Inferentia, and Graviton. Additionally, the sustainability of this market relies on the continued growth of the generative AI sector, as a decline could lead to a surplus of GPUs and reduced customer demand.
Market Outlook:
The alternative cloud provider market is expected to continue growing, driven by the increasing demand for GPU capacity to train and run generative AI models. However, the sustainability of this growth depends on the ability of alternative providers to secure a consistent high-volume supply of GPUs at competitive prices and navigate potential competition from established cloud providers developing custom AI hardware.
VI. Global Competition and Geopolitics in AI Chip Manufacturing
Geopolitical Factors and Government Investments:
The AI chip industry is at the heart of global technological competition, particularly between the US and China. Governments worldwide are recognizing the strategic importance of semiconductors and are implementing “CHIPS Acts” to boost domestic production and research. The US CHIPS and Science Act, for example, allocates billions in subsidies and incentives to attract chipmakers like TSMC and Intel to build facilities on US soil.
The US CHIPS and Science Act, a $280 billion initiative, aims to boost domestic semiconductor manufacturing and research to reduce reliance on foreign supply chains and counter China’s technological rise. The US government is actively promoting domestic chip production and has allocated $39 billion for semiconductor R&D through the CHIPS and Science Act of 2022. The US government has implemented export controls on advanced AI chips to certain regions, such as the Middle East, due to national security concerns.
In response to US export restrictions, China has established a $47.5 billion semiconductor fund, known as Big Fund III, to bolster its domestic chip industry and strive for self-sufficiency. China, a major player in the technology sector, is actively investing in its domestic chip industry to reduce reliance on foreign suppliers. The US is not alone in its efforts to bolster domestic chip manufacturing; Japan is spending $13 billion, Europe is spending more than $47 billion, and India announced a $15 billion plan to build local chip plants. These global investments highlight the increasing recognition of semiconductors as strategically important assets. Other countries, including Japan, Europe, and India, are also investing heavily in their domestic chipmaking capabilities.
The US Commerce Department imposed restrictions on exporting advanced semiconductors and chip-making equipment to China, with the most recent restrictions occurring in October, significantly impacting AI chip shipments. These actions underscore the geopolitical tensions surrounding the industry. TSMC, the world’s largest contract chip maker with a market share exceeding 55%, plays a pivotal role in these geopolitical dynamics. TSMC’s dominance in semiconductor manufacturing is concentrated in Taiwan, creating geopolitical risks due to tensions between China, Taiwan, and the US. China claims sovereignty over Taiwan, while the US is committed to defending Taiwan. US government subsidies are being offered to companies like Samsung, TSMC, and Intel to establish chip manufacturing facilities in the US.
Disruptions to chip manufacturing in Taiwan could have global consequences, as TSMC is a critical part of the global semiconductor supply chain. To mitigate potential disruptions, the US and Europe are subsidizing new chip plants outside of Taiwan. In the US, these subsidies are being deployed as part of the CHIPS Act. For example, Samsung Electronics received $6.4 billion, and TSMC received $6.6 billion to establish US-based factories.
China’s efforts to bolster its domestic chip industry extend beyond the Big Fund III. The country has been actively investing in domestic supply chains and developing advanced chips. The Chinese government, as part of its Made in China 2025 program, has been using state capital to fund domestic chipmakers like SMIC since at least 2015. SEMI, a global industry body, estimates that Chinese companies will launch 18 chip manufacturing projects in 2024, increasing China’s chip-making capacity by 12%. In 2023, Chinese companies purchased over $30 billion worth of chip-making equipment, approximately one-third of the global total, according to SEMI. While most of these new factories are focused on older-generation chips not subject to current US restrictions, analysts believe these investments could help Chinese companies catch up in advanced chip-making capabilities and strengthen mature-node production. Despite facing US restrictions, China appears to be accelerating its efforts to build a domestic semiconductor ecosystem.
These developments highlight the complex interplay between global competition, geopolitics, and the AI chip industry.
VII. Software and Open Standards
Nvidia’s Dominance and Open-Source Alternatives:
Nvidia holds a dominant market share between 70% and 95% in the AI chip market, largely attributed to its CUDA software and powerful GPUs like the H100. This dominance is challenged by open-source solutions like OpenAI’s Triton, aiming to break Nvidia’s near-monopoly by allowing AI applications to run on various chips.
Triton, first released in 2021, is gaining traction for its efficiency and adaptability, with support extending beyond Nvidia’s GPUs to AMD and Intel chips.
The Ultra Accelerator Link (UALink) Consortium and Its Goals:
UALink, formed by AMD, Google, Intel, Meta, Microsoft, and others, aims to establish an open industry standard for connecting AI accelerators within data centers. Notably absent from this consortium is Nvidia.
UALink seeks to improve AI system flexibility, scalability, and interoperability, allowing for diverse chip components.
The consortium is developing a system to connect over 1,000 AI accelerators in a single computing pod for faster and more energy-efficient AI training.
The first UALink components are expected to be available within the next few years.
Market Dynamics and the “CUDA Tax”:
Companies are motivated to explore alternatives to Nvidia’s dominance due to the high cost of its products and supply shortages.
Some developers refer to the “CUDA tax,” highlighting the costs and limitations of relying solely on Nvidia’s proprietary software.
Despite the emergence of alternatives like Triton, Intel’s OneAPI, and Chris Lattner’s Mojo, analysts predict Nvidia will retain a significant market share for years to come.
Intel’s Gaudi AI Accelerators:
Intel offers its Gaudi AI accelerators, particularly the Gaudi 3, designed for AI training and inference, as a cost-effective alternative to Nvidia’s offerings.
An AI kit with eight Intel Gaudi 2 accelerators is estimated to cost $65,000, about one-third the cost of comparable platforms.
A kit with eight Intel Gaudi 3 accelerators will be priced at $125,000, about two-thirds the cost of comparable platforms.
Intel projects its Gaudi 3 in an 8,192-accelerator cluster to achieve up to 40% faster training compared to a similar Nvidia H100 cluster.
A 64-accelerator Gaudi 3 cluster is projected to offer up to 15% faster training throughput than a similar Nvidia H100 cluster on the Llama2-70B model.
Intel Gaudi 3 is projected to deliver, on average, up to 2x faster inferencing than Nvidia H100 on models like Llama-70B and Mistral-7B.
The Role of Open Standards in AI Chip Development:
The emergence of open-source software solutions and industry standards like UALink represents a crucial shift in the AI chip market.
These initiatives aim to foster a more diverse and competitive landscape by reducing reliance on proprietary systems and promoting interoperability.
VIII. Conclusion: The Future of AI Chips
Factors Shaping the Future of the AI Chip Market
The Future Landscape of AI Chip Manufacturing:
Nvidia currently dominates the AI chip market, but its reliance on TSMC for chip manufacturing creates potential vulnerabilities.
TSMC, the world’s largest contract chip maker, is facing pressure to diversify its manufacturing base due to geopolitical risks associated with its concentration in Taiwan.
The establishment of new chip manufacturing facilities in the US and other regions aims to address supply chain vulnerabilities and reduce reliance on a single geographic location.
However, building new chip plants is a complex and time-consuming process, and it remains to be seen how quickly these efforts will significantly impact the global chip supply chain.
The Future of AI Chips in the Broader Context of Chip Manufacturing:
While AI chips represent a growing segment of the semiconductor market, they currently account for a relatively small proportion of overall chip manufacturing.
The majority of chip manufacturing is dedicated to producing chips for smartphones, consumer electronics, data centers, and other established markets.
However, the rapid growth of the AI chip market is expected to continue, driven by the increasing adoption of AI across various industries and the need for more powerful and efficient AI hardware.
The future of AI chips will be intertwined with the broader trends in the semiconductor industry, including advancements in chip design, manufacturing processes, and the development of new materials.
Frequently Asked Questions
What is driving the demand for AI chips?
The rapid advancement and adoption of AI are fueling an unprecedented demand for chips. Training and running complex AI models require massive computational power, creating an insatiable appetite for high-performance chips like GPUs.
Who are the key players in the AI chip market?
Nvidia, Intel, AMD, and TSMC are the major players in the AI chip market. Nvidia holds a dominant market share, while Intel and AMD are actively competing with Nvidia.
What are the strategic responses of companies to the competition in the AI chip market?
Companies like Nvidia, Intel, and AMD are investing heavily in software development, focusing on high-performance solutions, and collaborating with alternative cloud providers to maintain their market position.
What are the market share projections for AI chips?
Nvidia commands a near-monopoly in the advanced AI training chip market, with a company value larger than the GDP of 183 countries. Sources estimate that Nvidia holds between 70% and 95% of the AI chip market share.health data protection regulations, but always read the privacy policy before using a new app.
What are the efforts of big tech companies to reduce their dependency on Nvidia?
Big tech companies like Amazon, Microsoft, and Google are actively trying to decrease their dependency by developing their own AI chips to potentially reduce costs and increase their profit margins.
What are the production locations and timelines for TSMC?
TSMC primarily manufactures its most advanced chips in Taiwan. However, TSMC is also building a facility in Arizona. TSMC’s Arizona plant is expected to begin producing 3nm chips in 2025.
What are the market share projections for AI PCs?
Intel estimates that AI PCs will constitute 60% of the AI market by 2027.
What are the neural processing unit (NPU) usage projections?
Intel projects that the use of NPUs will increase from approximately 25% in 2024 to 30% in 2025.
What are the strategic focuses of Intel?
Intel is actively working to regain its position in the AI market through several initiatives, including the release of its Gaudi 3 AI accelerator and the development of Lunar Lake processors.
What are the market share projections for AMD?
AMD held slightly above 10% of the server market share in 2020. By the first quarter of 2024, AMD’s server market share had risen to 23.6%.
Key Takeaways
- Demand for AI Chips Across Various Sectors:
- The rapid advancement and adoption of AI are fueling an unprecedented demand for chips. Training and running complex AI models require massive computational power, creating an insatiable appetite for high-performance chips like GPUs.
- The demand for AI chips extends beyond traditional tech giants to industries like fast fashion and lawn care, where AI models are being used for tasks like document summarization and customer service.
- Key Players and their Strategies:
- Nvidia holds a dominant market share, with a company value larger than the GDP of 183 countries. Sources estimate that Nvidia holds between 70% and 95% of the AI chip market share.
- Intel is actively working to regain its position in the AI market through several initiatives, including the release of its Gaudi 3 AI accelerator and the development of Lunar Lake processors.
- TSMC holds a dominant position in the semiconductor industry with a global market share exceeding 55%. This makes TSMC the world’s largest contract chipmaker, responsible for supplying more than half of the world’s chips.
- Geopolitical Factors and the Global Race for AI Chip Dominance:
- Nvidia faces competition from several fronts, including alternative software solutions, tech giants building in-house chipmaking efforts, and partnerships.
- Intel is actively working to regain its position in the AI market through several initiatives, including the release of its Gaudi 3 AI accelerator and the development of Lunar Lake processors.
- Future Outlook:
- Nvidia is expected to maintain a substantial market share in the AI chip market for the foreseeable future, the emergence of competitive hardware and software ecosystems poses challenges.
- Intel is actively working to regain its position in the AI market through several initiatives, including the release of its Gaudi 3 AI accelerator and the development of Lunar Lake processors.
Sources
- What’s next in chips | MIT Technology Review, May 13, 2024
- The Biden Administration opens $285 million funding for ‘digital twin’ chip research institute | Engadget, May 18, 2024
- China will focus on AI in its new chip fund amid U.S. export curbs: professor | CNBC, May 29, 2024
- China Asking Its Tech Giants to Ditch Nvidia Chips and Buy Local | Business Insider, May 2024
- US Is Slowing AI Chip Exports to Middle East by Nvidia, AMD | Yahoo Finance, May 2024
- CoreWeave’s $1.1B raise shows the market for alternative clouds is booming | TechCrunch, May 5, 2024
- Intel Announces Gaudi 3 Accelerator For Generative AI | Forbes, April 16, 2024
- With Apple entering the fight, the AI chip wars have gone nuclear | Yahoo Finance, May 2024
