So, you know how AI is eating the world? Well, it's also eating data centers. All those large language models and image generators need to move absolutely staggering amounts of information between servers, and they generate enough heat to fry an egg. Broadcom Inc. (AVGO) is stepping up with a two-pronged solution: a new chip to move the data faster and a partnership to keep everything from overheating.
Think of it as building a wider, faster highway for AI's data traffic while also installing a top-tier air conditioning system for the engines powering it all.
The Data Highway: Introducing the Taurus Chip
Broadcom announced it has introduced a new chip designed to help move large amounts of data more efficiently inside data centers that support AI. The product, called the Taurus BCM83640, is essentially a key component that lets equipment makers build faster, more energy-efficient optical modules. These modules are the workhorses that transfer data between servers and networking gear.
Here's why it matters: this chip enables 1.6-terabit data modules. That means each connection can move twice as much data as previous designs. For data centers drowning in AI traffic, that's a huge capacity boost. Broadcom says the tech can also support future systems capable of handling even larger data loads, like 3.2-terabit modules. The company has started providing samples of the Taurus chip to selected customers and partners. It's not just a concept; it's heading for the real world.
The Heat Problem: A Liquid Cooling Partnership
Separately, there's the issue of all that computational power generating enough heat to melt your average server rack. To tackle this, Broadcom is working with JetCool, a Flex Ltd. (FLEX) company, on liquid cooling technology for next-generation AI chips.
The collaboration aims to manage the intense heat produced by powerful AI processors. JetCool's system sends liquid directly to the chip to remove heat, which is a more efficient method than traditional air cooling for these high-power components. Flex will handle large-scale manufacturing of the cooling equipment, while Broadcom provides the AI chips. It's a full-stack solution to keep the AI engines running cool and reliably as computing demands—and temperatures—continue to climb.
The Financial Firepower: AI Demand Is Boosting Everything
This tech push comes on the heels of some seriously strong financial results for Broadcom, all fueled by the AI boom. The chipmaker posted fiscal first-quarter revenue of $19.31 billion, up 29% year over year. The real star was AI-related revenue, which jumped 106% to $8.4 billion.
In a move that should ease investor concerns about supply bottlenecks, Broadcom also said it has secured the high-bandwidth memory (HBM) and advanced-node manufacturing capacity at Taiwan Semiconductor Manufacturing Co. (TSM) needed to support its AI chip production through 2028.
CEO Hock Tan laid out a staggering forecast: the company expects AI chip revenue to exceed $100 billion in 2027. That growth is being driven by demand from major customers including Alphabet Inc.'s Google (GOOGL), OpenAI, Anthropic, and Meta Platforms Inc (META).
Unsurprisingly, this bullish outlook prompted several analysts to raise their price targets. JPMorgan's Harlan Sur reiterated an Overweight rating and lifted his forecast to $500 from $475. Goldman Sachs' James Schneider maintained a Buy rating and increased his forecast to $480 from $450. Rosenblatt's Kevin Cassidy also raised his forecast to $500, and Benchmark's Cody Acree reiterated a Buy rating with a $485 price forecast, citing strong AI demand and Broadcom's secured supply chain.
Broadcom shares were up slightly on the day of the announcement. Over the past 12 months, the stock has gained over 78.50%, a run largely powered by its positioning in the AI infrastructure race. With a new chip for data, a solution for heat, and a CEO predicting $100 billion in AI revenue, Broadcom is making sure it's built for the AI inferno.