Nvidia, long the dominant force in AI chips, is facing increasing pressure from competitors as demand surges and the market evolves. While still leading the pack, the company’s once-unassailable position is being challenged by tech giants like Google and Amazon, as well as established players such as AMD and Broadcom-and even its own customers who are now designing custom silicon. This report examines the shifting dynamics of the AI chip market and nvidia’s strategies to maintain its edge amid growing competition and geopolitical constraints, including restrictions on sales to China.
شركة إنفيديا الأميركية في معرض بالصين، بكين في 17 يوليو 2025 (Getty)
For nearly a decade, Nvidia has maintained a dominant grip on the market for advanced computer chips powering artificial intelligence and machine learning. Armed with cutting-edge graphics processing unit (GPU) designs and benefiting from the rapid innovation pace of Taiwan Semiconductor Manufacturing, which produces 90% of the world’s advanced AI chips, Nvidia has become almost synonymous with AI processing.
Most industry observers now anticipate that Nvidia will lose its market dominance as demand increases, according to a report in the Wall Street Journal on Saturday. However, the company consistently maintains that its computing systems are more versatile and have broader applications than specialized chips. The increasing demand, however, means Nvidia is no longer the sole player in the field.
In late November, Dylan Patel, founder of the tech consultancy SemiAnalysis, suggested that the growing popularity of Google’s chips could signal “the end of Nvidia’s dominance.”
Erosion of Market Control
Table of Contents
Nvidia’s customers are already beginning to challenge its dominance by developing custom integrated circuits, a category of chips designed in collaboration between AI companies and major chipmakers to optimize performance for specific computing tasks.
OpenAI and Broadcom, for example, entered into a multi-billion dollar partnership in October 2025 to develop custom chips tailored to the computing needs of the ChatGPT developer. On September 30, Meta announced it would acquire the emerging chip design company Rivos to bolster its efforts in developing in-house AI model training chips.
Microsoft’s Chief Technology Officer, Kevin Scott, stated during a panel discussion at Italian Tech Week in Italy in October, that the company plans to increasingly rely on custom-designed accelerator chips within its data center operations.
On July 18, AI company xAI, founded by Elon Musk, posted job openings for chip engineers to help “design and refine new hardware architectures” to support AI model training on the company’s official careers page.
A Shifting Landscape
Nvidia’s once-unassailable position is now subject to change, with new entrants emerging in the AI chip design market, including Google and Amazon, both of which are now discussing selling their latest chips—rivaling Nvidia in power and efficiency—to a wider range of external customers. Smaller competitors, including Advanced Micro Devices (AMD), Qualcomm, and Broadcom, are also entering the market with new products focused on AI computing in data centers.
Even some of Nvidia’s largest customers, such as OpenAI and Meta, have begun designing their own custom chips, further challenging Nvidia’s widespread adoption.
While a mass exodus of Nvidia’s customers is unlikely, the push for diversification among AI companies could make it more difficult for the leading company to achieve the supercharged sales growth that investors have become accustomed to. The market map is changing rapidly, with nearly every week bringing a new large-scale technology infrastructure deal or the launch of a new generation of powerful AI chips.
The New Challengers
Competition has intensified in recent weeks, with Google’s Alphabet and Amazon’s cloud computing arm, Amazon Web Services, pouring significant investment into AI chips, fueled by the cash flow from other activities, and seeing increasing demand from external customers.
For over a decade, Google has designed and internally used chips known as Tensor Processing Units (TPUs), and began offering these chips for third-party use in 2018, but they weren’t widely available to large customers for many years. Today, giants like Meta, Anthropic, and Apple are buying or negotiating access to these TPUs for training and running their models.
Amazon is currently expanding a “cluster” of data centers for Anthropic that will eventually house more than one million of its internally designed “Trainium” chips. AWS also recently launched a broader sale of chips it claims are much faster and more energy-efficient than Nvidia’s.
Emerging Competitors
Three years ago, Advanced Micro Devices (AMD) made a pivotal shift, directly challenging Nvidia. Recognizing the impending explosion in demand for advanced AI processors, AMD CEO Lisa Su told the board she planned to redirect the entire company toward AI, anticipating “insatiable demand for compute.” This gamble has already paid off, with AMD’s market capitalization nearly quadrupling to exceed $350 billion, and the company recently secured major deals to supply chips to OpenAI and Oracle.
Another emerging chip designer, Broadcom—formerly a division of Hewlett-Packard—has become a formidable competitor, expanding to a $1.8 trillion behemoth through a series of massive mergers. Today, Broadcom produces custom chips known as XPUs, designed for specific computing tasks, as well as networking hardware that helps data centers connect massive server racks.
Intel, a long-standing giant in Silicon Valley, faced a challenging period, missing much of the AI wave due to a series of strategic missteps. However, it has recently invested heavily in its design and manufacturing operations and is now seeking to attract customers for its advanced data center processors.
Qualcomm, known primarily for designing chips for mobile devices and automobiles, saw its stock rise 20% after announcing the launch of two new AI accelerator chips in October. The company stated that the new chips, the AI200 and AI250, feature very high memory capacity and significant power efficiency.
Nvidia Remains the Leader
Nvidia’s dominance in AI computing power has made it the world’s most valuable company, and has elevated its CEO, Jensen Huang, to celebrity status. Investors closely scrutinize every word Huang utters and follow the company’s quarterly results as a barometer of the overall AI bubble.
Nvidia describes its business as going beyond simply selling chips, preferring to talk about “server-level solutions” and calling data centers that use these solutions “AI factories.” However, the core product Nvidia offers—accelerated computing—is precisely what every AI company is seeking.
From February to October, Nvidia sold chips, interconnects, and other hardware components supporting the explosive growth of AI worth $147.8 billion, up from $91 billion in the same period the previous year. In July, Nvidia’s market capitalization surpassed $4 trillion, becoming the first company on the planet to achieve this milestone, then briefly exceeded the $5 trillion mark after five months, before concerns about a potential bubble swept the AI sector, causing Nvidia’s stock price, like that of most of its competitors, to fall slightly to more realistic levels. Even after this correction, Nvidia’s value remains more than double that of its closest competitor, Broadcom, which is valued at $1.8 trillion.
Working to Stay Ahead
Nvidia’s rise has been remarkable. In what has become part of corporate lore, Huang, Curtis Priem, and Chris Malachowski—three friends (all electrical engineers)—founded the company in 1993 while having breakfast at a Denny’s restaurant in San Jose, California. Their initial goal was to develop chips capable of producing more realistic 3D graphics for personal computers.
Unlike central processing units (CPUs) that power most computers, GPUs have the ability to perform parallel computing—executing millions or billions of simple tasks simultaneously. Initially used by video game developers, Nvidia later realized that its units were ideal for deep learning and artificial intelligence.
In 2006, Nvidia launched CUDA, a proprietary software library that allows developers to build applications that run on the company’s chips and operate at higher speeds. As demand for AI technologies grew, thousands of developers became heavily reliant on Nvidia’s ecosystem, which combines hardware and software.
Nvidia has accelerated the pace of launching each new generation of advanced AI chips. Late last year, it began shipping the Grace Blackwell series, its most powerful AI processors to date, based on its latest chips, and nearly all units were sold out immediately.
At a conference in Washington, D.C., in October, Huang said the company had already sold six million Blackwell chips in 2025 and received orders for another 14 million, representing a total of $500 billion in sales.
However, challenges remain. Nvidia has been effectively prevented from selling its chips in China for the past three years, a problematic situation as Huang insists that the country accounts for half of the world’s AI developers. Without the billions in sales represented by the Chinese market, the company’s growth will be constrained, and China’s tech sector will likely turn to domestically produced chips instead of Nvidia’s products.
