US semiconductor firm Micron Technology, like its peers, is angling for a piece of the growing pie driven by soaring demand for AI applications. Specifically, it is ramping up its production of high-bandwidth memory (HBM) chips, a crucial component used with AI market leader Nvidia’s graphics processing units (GPUs).
According to Koen de Backer, corporate vice president of smart manufacturing and AI, Micron’s HBM complements GPUs through enhancing processing power and enabling the scalability of AI models — a synergy vital for tackling increasingly complex challenges.
“HBM is really a new class of memory chip. It’s really specifically for the large kind of AI language models, and they require a lot of bandwidth, which is what HBM provides. So while it’s not the biggest part of our business currently, and it will take a long while before it gets to that level, it’s definitely a very fast-moving part,” explains De Backer at a Micron facility site tour in Singapore.
With some US$15 billion ($20 billion) and counting, Micron is one of the single largest foreign investors in Singapore.
Micron’s latest product, the HBM3E, is touted as the fastest and highest-capacity HBM in the market, which will enable Nvidia to incorporate more memory per GPU, thus enhancing AI-training and high-performance computing tools.
De Backer explains that the latest chips are “complicated” and expensive to manufacture due to additional steps needed. “Typically, manufacturing chips is more of a front-end type of flow, where it goes from a wafer into a single package. But here, the package becomes so much more complicated, you’re combining or packing different features in with a high-bandwidth memory, and Micron has come up with new techniques to produce it. That’s what makes our HBM so unique.”
See also: Australia’s social media ban for under 16s to become law
Such is the AI wave’s demand for the chip that Micron, at its earnings update on March 20 for 2QFY2024 ended Feb 29, announced that it has sold out its entire supply of HBM3E chips for the whole the year, with a large majority of its 2025 output already
spoken for. “These are new products and they are high in demand, so they help drive a lot of investments across our business,” says De Backer.
Unsurprisingly, Micron has been applying AI to its daily operations. “It’s been a core part of what we do for quite some time now. Our operations are highly automated, generating massive amounts of data. This data is where AI shines, allowing us to gain insights and optimise our manufacturing processes, for example,” he says.
Coding is another area of business where De Backer suggests AI could greatly benefit company operations with productivity gains of up to 20%. “We have more than 5,000 people in our organisation that code on a day-to-day basis in various functions, and we know that coding solutions could really help in productivity.”
See also: UBS-Credit Suisse integration opens up new tech for bigger plans
Delving deeper into Micron’s AI initiatives, De Backer delineates between artificial narrow intelligence (ANI) — the most common type of AI available — and artificial general intelligence (AGI) — the most advanced form of AI capable of human-level intelligence and beyond.
Although ANI’s proficiency in identifying patterns within data and making forecasts has been a capability leveraged extensively by Micron, he envisions AGI as the next frontier: “The AI landscape is evolving rapidly. While ANI has been instrumental for us, AGI holds the promise of more human-like interactions with technology, paving the way for exciting advancements.”
Meanwhile, generative AI continues to be the technology’s latest reiteration.
Although not quite the groundbreaking revolution seemingly offered by AGI, generative AI already stands as a noteworthy advancement over traditional AI, with its ability to source existing data from multiple sources to create content from prompts, a marked improvement over traditional AI’s function of intelligent streamlining and executing tasks from historical data.
On March 20, Micron reported that for its 2QFY2024 ended Feb 29, its revenue surged by 57.7% y-o-y to US$5.8 billion, while its EPS of 71 US cents improved 66.5% from a loss per share of US$2.12 in the previous period.
The investment community was cheered too by its bullish guidance of 3QFY2024 revenue of between US$6.4 billion to US$6.8 billion, along with an anticipated EPS of 45 US cents. Analysts were expecting revenue of US$5.99 billion and were projecting EPS of 24 US cents.
Year to date, Micron shares have gained more than a third to trade at US$118.80 on March 26.
Of the 37 analysts covering this stock, the vast majority of 33 have “buy” calls, with the target prices as high as US$150. Only Zacks Investment Research and Baird have a “neutral” call, while Joseph Moore of Morgan Stanley and William Kerwin of Morningstar stand out with their “sell” calls.
