In June, Swiss investment bank UBS published a report that estimated artificial intelligence (AI) will add $1.2 trillion annually to the global economy by 2027. The analysts divided the AI value chain into three layers:
The enabling layer includes companies that provide the chips and data center infrastructure needed to train models and build applications. Annual revenue will reach $516 billion by 2027.
The intelligence layer includes companies that provide data management tools and AI models that support application development. Annual revenue will reach $255 billion by 2027.
The application layer includes companies that use large language models and other machine learning models to build AI products. Annual revenue will reach $395 billion by 2027.
The report also predicted that “AI will be the most profound innovation and one of the largest investment opportunities in human history.” The application layer probably offers the greatest monetization potential over time, but UBS believes companies in the enabling layer will be the biggest beneficiaries during the next three years.
Chipmakers Nvidia and Broadcom and server manufacturer Super Micro Computer fall into that category.
Nvidia’s graphics processing units (GPUs) are the industry standard in accelerating complex data center workloads like AI applications. Forrester Research recently wrote, “Nvidia sets the pace for AI infrastructure worldwide. Without Nvidia GPUs, modern AI wouldn’t be possible.”
Nvidia has cemented its leadership and enhanced its ability to monetize AI by branching into other product categories. Its portfolio includes adjacent hardware, like central processing units (CPUs) and networking equipment, as well as subscription software and cloud services that support AI application development.
Toshiya Hari at Goldman Sachs sees that as a key differentiatorKanpur Investment. “We believe Nvidia will remain the de facto industry standard for the foreseeable future given its competitive advantage that spans hardware and software capabilities,” he wrote in a note to clients. “Nvidia’s annual introduction of new products and platforms sets a pace of innovation that keeps it at the forefront of the industry.”Agra Investment
Wall Street expects Nvidia to grow non-GAAP (generally accepted accounting principles) earnings per share at 38% annually through fiscal 2027 (ends January 2027). That consensus estimate makes its current valuation of 66.7 times adjusted earnings look tolerable. Investors should feel comfortable buying a small position in this semiconductor stock today.
Broadcom provides a broad range of IT solutions, but the company is best known for its leadership in data center networking chips (for switches and routers) and application-specific integrated circuits (ASICs). Specifically, Broadcom holds 80% market share in data center networking chips and at least 55% market share in ASICs. Demand for AI should be a tailwind in both cases.
Switches and routers move information between data center servers and other networks, and their ability to do so depends on semiconductor throughputAhmedabad Stock. Due to the enormous amount of data involved, fast chips are essential where AI applications are concerned, and Broadcom has the fastest chips on the market. JPMorgan Chase says spending on data center networking chips will grow 20% to 30% annually over the next few years.
ASIC refers to bespoke silicon designed for specialized use cases like AI. For instance, Broadcom helps Alphabet’s Google and Meta Platforms build custom AI chips, and it recently announced a third (unnamed) customer that could be Amazon, Apple, or TikTok parent ByteDance, according to various analysts. ASICs represent less than 10% of AI chips today, but UBS says the market will outpace GPU sales in the coming years, such that ASICs account for 30% of AI chips by 2027.
Wall Street expects Broadcom to grow non-GAAP earnings per share at 21% annually through fiscal 2027 (ends October 2027). That estimate makes its current valuation of 36.7 times adjusted earnings look quite reasonable. Patient investors should feel comfortable buying a few shares of Broadcom stock today.Jaipur Wealth Management
Super Micro Computer builds high-performance computing platforms for data centers. Its portfolio includes servers and storage systems optimized for use cases like analytics and AI, as well as server management software and server subsystems (chassis, motherboards, power supplies) that can be assembled into complete solutions.
Supermicro is the leader in AI servers, and its market share is climbing due to its in-house manufacturing capabilities and modular approach to product design. To elaborate, the company can quickly build a broad range of servers featuring the latest chips by using common building blocks, and it assembles most of those servers internally in Silicon Valley. That enables rapid prototyping and product roll-out.
Indeed, Supermicro can usually bring new technologies to market before its competitors, often two to six months earlier. Hans Mosesmann at Rosenblatt Securities recently highlighted that advantage. “Super Micro has developed a model that is very, very quick to market. They usually have the widest portfolio of products when a new product comes out from Nvidia or AMD or Intel.”
Supermicro accounted for 10% of AI server sales in 2023, but Bank of America analysts expect that figure to reach 17% by 2026. Tom Blakely at KeyBanc is even more optimistic. He thinks Supermicro could capture a 23% market share in AI servers by 2025. Either way, those forecasts bode well for shareholders. JPMorgan Chase says the AI server market will grow 460% between 2023 and 2027.
Varanasi Investment