Imagine a seismic shift in the world of AI hardware – what if Nvidia's ironclad dominance starts to crumble, paving the way for Google's cutting-edge custom chips to steal the spotlight? That's the tantalizing possibility sparking debate among investors and analysts alike. Let's unpack this transformative trend together, step by step, so even if you're new to the tech scene, you'll grasp the big picture. And trust me, this is just the beginning – but here's where it gets controversial, challenging everything we thought we knew about AI supremacy.
By Jules Rimmer
Citrini Research has pinpointed a lineup of firms poised to capitalize on a potential pivot toward Google's ultra-specialized AI microchips, moving away from Nvidia's graphics processing units, or GPUs.
Google's AI microchips, particularly its tensor processing units (TPUs), are grabbing more headlines than ever. Think of TPUs as specialized chips designed specifically for handling the massive computations in artificial intelligence tasks, much like GPUs are versatile workhorses for graphics and general computing, but optimized differently. The buzz stems from the impressive rollout of Google's Gemini 3, an AI program that runs on these TPUs instead of Nvidia's GPUs. This success has analysts buzzing about who might profit if this shift gains traction across the industry.
Citrini Research is posing a provocative question: What if the AI hardware story evolves from Nvidia's (NVDA) near-monopoly to a landscape where challengers like Alphabet (GOOG) lead the charge, growing the TPU ecosystem? It's a bold pivot that could reshape supply chains and market dynamics.
Citrini has spotlighted a group of companies that stand to see surging demand from this expansion. These include publicly traded giants such as TSMC (TSM), ASE (TW:3711), Amkor (AMKR), Micron (MU), Lumentum (LITE), TTM Technologies (TTMI), SK Hynix (KR:000660), Unimicron (TW:3037), Apple (AAPL), SiTime (SITM), and MACOM (MTSI). Each of these could benefit in unique ways, from manufacturing to components.
Over the past few months, Alphabet's stock has actually outperformed Nvidia's, reflecting a major market rethink on their roles in the AI boom. Investors are reevaluating who the true frontrunners are.
In a recent Substack piece, Citrini highlights Google's emergence as a leader in frontier models – those advanced, general-purpose AI systems that can tackle a wide range of tasks. Crucially, Google is achieving this with its own TPUs, bypassing Nvidia's GPUs, which dominate about 90% of the market and power tools like OpenAI's ChatGPT, the popular large language model for generating human-like text.
Earlier this week, The Information reported that Meta Platforms (META) is negotiating with Alphabet to acquire or lease TPUs worth billions, potentially marking a significant deal. But Citrini Research notes this potential has been brewing for a while, with the analyst's view shifting dramatically: 'While the signs have been evident for months, investor sentiment toward Google has flipped from viewing it as an AI underdog losing ground in search to a stealth competitor ready to challenge the top AI players.'
Citrini believes Meta's interest arises because training these cutting-edge AI models might soon be feasible without relying on Nvidia's hardware. This opens the door to diversifying beyond Nvidia's CUDA ecosystem – a proprietary platform that locks in users with its unique architecture for computing tasks. And this is the part most people miss: the high cost of switching from one hardware setup to another has historically protected Nvidia's profits, but now, partnerships between Alphabet and firms like Anthropic, Fluidstack, and Meta suggest that barrier might be eroding. Citrini's analysis warns that Alphabet's TPUs could pose a long-term threat to Nvidia's lucrative margins.
To illustrate this, Citrini compares the performance of the 'Google ecosystem' (companies tied to Google's AI push) against the 'OpenAI ecosystem' over the past year, showing how perceptions have inverted.
Delving deeper, Citrini emphasizes that the expense of migrating to a new AI accelerator architecture – those specialized components that turbocharge AI and machine learning processes – is steep. Yet, these emerging deals hint that Nvidia's competitive edge might not be as impregnable as once thought.
Identifying suppliers set to thrive in this new era, Citrini points to TSMC, ASE, and Amkor for their role in advanced packaging techniques like CoWoS (chip-on-wafer-on-substrate), which stacks chips for better performance and efficiency in TPUs. This is like building a multi-layered sandwich of electronic components to handle intense AI workloads without overheating or slowing down.
Citrini also notes that high-bandwidth memory – the fast data storage needed for rapid AI computations – remains a choke point for both GPUs and TPUs, boosting demand for providers like SK Hynix, Samsung (KR:005930), and Micron. Imagine trying to process a flood of data in real-time; without speedy memory, everything grinds to a halt.
For the printed circuit boards in TPUs, which require complex, custom designs for optimal routing, companies such as Isu Petasys, TTM Technologies, and Unimicron step in. Meanwhile, optical circuit switches (OCS) – devices that direct light signals in data transmission – are crafted by Lumentum, ensuring smooth, high-speed data flow in these AI systems.
Apple (AAPL) could leverage TPUs for cloud-based inference, which is essentially running a pre-trained AI model to analyze new data and make predictions in the cloud. For instance, this might enhance Apple's Siri by making it smarter at understanding and responding to user queries through real-time AI processing hosted online.
Lastly, SiTime and MACOM might capitalize on a faster adoption of 1.6T optical modules – transceivers that blast data at 1.6 terabits per second, enabling lightning-fast internet and AI communications. Picture sending massive amounts of data across networks as quickly as flipping a switch.
- Jules Rimmer
This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.
(END) Dow Jones Newswires
11-28-25 0744ET
Copyright (c) 2025 Dow Jones & Company, Inc.
The articles, information, and content displayed on this webpage may include materials prepared and provided by third parties. Such third-party content is offered for informational purposes only and is not endorsed, reviewed, or verified by Morningstar.
Morningstar makes no representations or warranties regarding the accuracy, completeness, timeliness, or reliability of any third-party content displayed on this site. The views and opinions expressed in third-party content are those of the respective authors and do not necessarily reflect the views of Morningstar, its affiliates, or employees.
Morningstar is not responsible for any errors, omissions, or delays in this content, nor for any actions taken in reliance thereon. Users are advised to exercise their own judgment and seek independent financial advice before making any decisions based on such content. The third-party providers of this content are not affiliated with Morningstar, and their inclusion on this site does not imply any form of partnership, agency, or endorsement.
What do you think – is this TPU revolution a fair challenge to Nvidia's empire, or just hype that won't pan out? Could diversification in AI hardware lead to more innovation, or might it complicate things unnecessarily? Share your opinions in the comments; I'd love to hear if you agree, disagree, or have your own take on where the AI market is headed!
