The Campbell's Company (CPB)

The Campbell's Company, based in Camden, New Jersey, manufactures and markets consumer food products. Campbell operates two segments: Snacks and Meals & Beverages. The company's brands include Pepperidge Farm, Snyder's, Prego, Swanson, V8, Kettle, Cape Cod, and Spaghettios. The company has approximately 13,500 employees.

10 Questions About Artificial Intelligence: Part 1 Artificial intelligence (AI) has been and remains one of the fastest-growing technology developments in history. Just a few years after becoming broadly known, AI represents a significant portion of the global economy and is now the preeminent focus among investors. Developments in AI have progressed and continue to unfold rapidly, and new revelations on one day - say, the release of Meta Platforms Inc.'s Muse model - can make preconceptions about AI seem out of date. Based on input from clients, investors, industry experts, and our own analytic team, Argus has put together what we see as 10 current key and enduring questions about AI, along with our responses. We've broken these 10 questions into Parts 1, 2, and 3. Part 1, presented below, addresses three key questions about the beginnings of AI investing, the sustainability of the AI trade, and the massive investments being made by the biggest players in the industry. Following publication of Part 1, in coming weeks, we will be publishing Parts 2 and 3. We welcome feedback from clients and will continue to refine our 10 questions about AI in real time in response to this exciting and ever-changing opportunity. QUESTION 1: Why has the AI trade been so successful to date, while other investing phases based on technology breakthroughs have sometimes stumbled? ANSWER 1: The short answer as to why the AI trade has been so successful is decades of dedicated and now accelerating technology development, established mechanisms of delivery, profitable and fast-growing adjacencies, and receptive enterprise customers. The long answer touches on all those themes. AI technology development was far along when investors first became broadly aware of artificial intelligence. There was substance behind AI as an investment when OpenAI's ChatGPT first burst upon the scene in fall 2022, as well as many viable, investible entities. AI investment had been conducted for decades in university science departments and government-funded projects and agencies, providing building blocks for future development. Highly successful companies, including the major cloud service providers (CSPs), directed their huge cash flows to building robust AI science initiatives. Groundbreaking models include Word2Vec and Long Short-Term Memory, which 'vectored' (attached strings of numbers) to words based on syntactical and semantic relationships. This enabled development of large language models (LLMs) that could recognize and organize words in coherent patterns. The next generation of transformer models based on self-attention mechanisms led directly to ChatGPT. (GPT stands for 'generative pre-trained transformer.') CSPs determined that AI required massive infrastructure beyond the reach of most enterprises. These CSPs, or hyperscalers, recognized that for many companies, AI would need to be delivered as a service, which fit optimally with their cloud delivery models. Spurred by intense competition among financially strong hyperscalers seeking to develop the most cutting-edge models, AI technology development has been continuous and even shows signs of accelerating. The early focus on massive graphics processing unit (GPU) clusters in AI data centers used for training LLMs has been superseded by advances in inference and the unfolding age of agentic AI, which is highly useful in enterprise applications as companies adopt and deploy AI. On the hardware side, AI tentpole company Nvidia Corp. used its decades of GPU gaming technology development to invest in AI for data centers, with spectacular results. While AI implementations were initially centered in hyperscaler data centers, big swathes of the IT hardware industry - data storage, networking, standard servers, and more - are now being deployed to store, manage, and distribute AI capabilities broadly. On the customer side, enterprises were ready to embrace technologies that could simplify and strengthen business processes, improve return on research & development and sales & marketing initiatives, and bring customers closer, with more personalized interactions. By contrast, some past investment phases were based on promising technologies that either fizzled or were not ready for prime time. The first internet boom of 2000 was built more on promises (vaporware) than on substantive products. Ultimately, the dot-com boom imploded due to the limitations of then-prevailing dial-up internet. The AI trade, by contrast, has been profitable almost since day one because it was built on firm foundations across science, cloud, software, and hardware and because AI capabilities are ever advancing and improving, creating new monetization opportunities. And the leaders of the AI trade are financially successful, with strong and growing revenue, profits, and free cash flows. QUESTION 2: What are some of the key technology developments that are keeping investors excited about investing in AI three-plus years into the AI boom? ANSWER 2: Key developments in the proliferation and monetization of AI include both technology developments, spurred by the horse race among industry leaders, and market developments, as more and more companies embrace technology with demonstrable return on investment (ROI). At the time of the public launch of ChatGPT in November 2022, key elements of AI technology development included training of LLMs; deep learning, which enables image recognition and natural language processing; and generative AI, which can generate a response to a specific question. These technologies were well established when the public first encountered generative AI. In terms of technology development, inference and agentic AI are significantly more useful and applicable than generative AI solutions. In the early post-ChatGPT years, hyperscalers created large GPU clusters with sufficient size to train models, but in many cases, the output of these models - and thus their financial potential - remained isolated in AI data centers. Once a model is trained, including pre- and post-training, it can be assigned tasks from describing complex event sequences to modeling genome sequences, to generating a tailored investment strategy. Taking an input (your query) and producing output is called inference. LLMs can have hundreds of billions or trillions of parameters, generating multiple trillions of tokens. Inference can use many times the number of tokens generated for LLM training. Inference is vital to agentic AI, which differs from generative models by actively pursuing goals rather than passively generating outputs. In terms of market development, enterprises see agentic AI enabled by inference as much more useful and applicable than generative AI when it comes to improving business outcomes. Based on inputs, an AI agent can decide what actions to take, draw on appropriate models, and develop and deliver a response or strategy based on a continuous loop of observation, perception, reasoning, and action. Enterprises can incorporate AI agents across their organizations to drive efficiency and spur new and existing market development. This broadening out of AI users from hyperscalers to a wide range of enterprise and sovereign customers represents vast new markets for AI vendors. The horse race among the top AI companies and particularly the hyperscalers to develop new models results in a steady flow of new and ever more impressive LLMs coming before the public and before investors. A year ago, Anthropic PBC's Claude was having its close-up, followed by Claude Code. In fall 2026, Meta Platforms' Muse LLM triggered over 2.5 million downloads in less than two weeks, as business and consumers reacted favorably to its advanced agentic capabilities. Six months from now, users will be moving on to even more advanced models, in a steady flow that keeps the industry evergreen. QUESTION 3: A small number of hyperscalers are investing huge sums of money in AI infrastructure. Will these investments prove to be profitable? ANSWER 3: One of the earlier sell-off triggers for the AI trade was the concern that AI sales were circular. Critics alleged that initial AI investment were largely conducted among a very small set of CSPs or hyperscalers and a few hardware infrastructure providers, led by Nvidia; these companies would then invest in one another. Circular AI trade fears are for now receding, following an exceptionally strong earnings cycle in which AI participants revealed massive order backlogs and as the market opportunity broadens out to enterprises, sovereigns (nations), and neoclouds (which lease dedicated AI infrastructure to customers). In the race to develop leading AI models and deliver cutting-edge AI services, leading hyperscalers - Amazon.com Inc., Microsoft Corp., Alphabet Inc., Meta Platforms, CoreWeave Inc., and Oracle Corp. - are expected to allocate a combined $650 billion-$725 billion in AI-related capital spending in 2026 alone. That is about double the roughly $365 billion-$380 billion this group spent on AI in 2025. Notably for these companies, more than two-thirds of 2026 spending ($450 billion-$500 billion) is dedicated to AI infrastructure, including GPU and central processing unit servers, networking, storage, and data center construction. AI infrastructure spending by these hyperscalers is forecast to reach $1.0 trillion-$1.2 trillion in 2027. These companies are accessing credit markets and in some cases issuing equity to fund their enormous spending. And the 'circular spending' argument will never fully go away. Despite these legitimate concerns, the AI investing environment remains positive. Hyperscalers' massive spending is creating enormous AI resources, and that is driving unprecedented demand and market opportunities, particularly as agents make AI applicable to enterprise needs. While these companies expect these investments to pay off over time, they are realistically reckoning the time to monetization in multiple years as they simultaneously pay interest costs on new debt and manage long-term depreciation. Risks to earning ROI include the relatively short useful lives of technology investments, which are five to six years for servers and as little as three years for GPUs; new technology developments, such as distillation, that enable comparable AI output at much lower token use and thus could make massive infrastructure unnecessary; and the potential for a steep drop in demand for AI services based on economic downturn, market maturation, or even consumer aversion to AI. Hyperscalers believe these risks are meaningfully outweighed by potential upsides. Even as the AI industry gets larger, growth rates measured in dollars and even in percentage terms are continuing to accelerate. Technology segments immediately adjacent to GPU compute, including data storage, networking, and x86 server, are experiencing record orders and building record backlogs. Products enabling data center functionality, including power supplies, cooling, rack architecture, and related areas, are also thriving and driving growth for industrial companies. The hyperscalers appear to believe that establishing market-dominant positions in infrastructure creates a competitive moat and will ensure competitive advantage as AI demand continues to broaden and deepen. They are using their huge cash flows from legacy businesses and playing a long game to invest in AI.

Over the past 150-plus years, Campbell's has evolved into a leading domestic packaged food manufacturer, with a portfolio that extends beyond its iconic red-and-white labeled canned soup. In fiscal 2026 (July year-end), meals and beverages accounted for 61% of its revenue, while snacks accounted for 39%. Beyond its namesake, its brands include Pepperidge Farm, Goldfish, Snyder’s of Hanover, Swanson, Pacific Foods, Prego, Pace, V8, and, most recently, Rao’s (a deal that closed in 2024). Around 90% of its revenue results from the US and the remainder from Canada and Latin America.

Over the past 150-plus years, Campbell's has evolved into a leading domestic packaged food manufacturer, with a portfolio that extends beyond its iconic red-and-white labeled canned soup. In fiscal 2025 (July year-end), snacks accounted for 43% of its revenue, followed by soup (27%), other simple meals (23%), and beverages (7%). Beyond its namesake, its brands include Pepperidge Farm, Goldfish, Snyder’s of Hanover, Swanson, Pacific Foods, Prego, Pace, V8, and, most recently, Rao’s (a deal that closed in 2024). Around 90% of its revenue results from the US and the remainder from Canada and Latin America.