Of all major semiconductor companies, Intel (INTC -1.10%) has perhaps been the most disappointing over the past decade. Even with the advent of cloud computing and the AI revolution, the stock has returned only 86% over the past 10 years, a little more than one-third of the return of the S&P 500 index and less than one-tenth the return of the semiconductor sector, as defined by the Van Eck Semiconductor ETF.
INTC 10 Year Total Returns (Daily) data by YCharts
Intel made numerous missteps under prior management, including falling behind in leading-edge manufacturing and perhaps not anticipating the extent of the GPU-based AI revolution.
But the company is in the midst of a course-correction under CEO Pat Gelsinger, who just introduced new AI accelerators Intel claims will challenge Nvidia's (NVDA -0.18%) current AI dominance.
Image source: Getty Images.
Can Intel's Gaudi hold its own against Nvidia?
At Tuesday's Intel Vision 2024 conference, Intel unveiled its new Gaudi 3 AI chips. Intel acquired the Israel-based Gaudi chip parent Habana Labs in 2019 for just $2 billion, which is actually looking like one of prior management's smartest moves.
The new Gaudi 3 AI accelerators have some impressive specs. According to Intel's presentation, the new chip will have 50% better inferencing performance with 40% better power efficiency than the Nvidia H100 and at a much lower cost. In addition, Intel claims the Gaudi 3 also achieves a 50% faster time-to-train versus the H100. Intel tested the two chips on the open-source Llama models developed by Meta Platforms, as well as the open-source Falcon LLM developed by the Abu Dhabi Technology Innovation Institute.
It's not surprising Intel's testing involved open-source models. Intel is aiming to break down all barriers in the AI industry, especially Nvidia's moat involving its CUDA programming software ecosystem.
In conjunction with the Gaudi announcement, Intel announced a new partnership to develop an open AI enterprise platform with with VMware, Red Hat, SAP, and a host of other artificial intelligence startups like Hugging Face and others. That new partnership follows last month's announcement of a collaboration with seven other large tech giants to develop an open-source software alternative to Nvidia's CUDA software stack.
Intel is also leading a consortium on the networking aspect of AI. Today, the preferred method for AI networking is Infiniband, which Nvidia acquired through its Mellanox acquisition in 2019. But Intel is now leading the Ultra Ethernet Consortium, a group of leading tech companies aiming to accelerate the more common ethernet networking fabric to handle AI workloads. Intel is contributing its AI networking interface chips and chiplets as part of the collaboration.
Should Nvidia investors worry?
At first glance, the new Gaudi chip may not seem like much to worry about for Nvidia investors. After all, while Intel's Gaudi may outdo the H100 in certain tests, Nvidia is already shipping the more powerful H200 starting this quarter.
Furthermore, Nvidia will release its new Blackwell architecture later this year. While the Gaudi 3 chip may achieve 50% performance improvement over the H100, Blackwell promises between 2.5 and five times the performance of the H100, handily outpacing the Gaudi 3 as well.
Still, Intel's leadership of large consortiums innovating open-source AI solutions represent a potential long-term threat. Nvidia will likely maintain a large market share in the AI space, but it likely won't be the dominant 90%-plus it commands today.
Is Intel a buy?
Intel is quite a complicated turnaround story, which will depend not only on the inroads Gaudi can make in the AI market, but also whether or not the company's foundry will ever become profitable.
Recently, the company broke out its foundry segment's independent financials, revealing its foundry cost the company $7 billion in losses last year, contributing to Intel's minuscule current profitability. This is because Intel's foundry must invest in capacity ahead of revenues.
However, management said it was "confident" margins would improve with the foundry achieving breakeven in the coming years. And by 2030, Intel sees foundry operating margins between 25% and 30%, helping the overall company achieve 40% operating margins, reassuming industry leadership.
If Intel can eventually achieve that level profitability on higher revenues in the future, the stock is cheap today. But whether or not Intel is a buy is totally dependent on its execution on these two challenging fronts.
As long as Intel continues rolling out improved products like Gaudi 3 and scaling its foundry with more third-party customers, investors may start pricing in Intel's 2030 target scenario. If that happens, the stock could very well rebound before that time.
