While Intel experiences its worst year ever, Broadcom achieves a record gain in Silicon Valley.

• In 2024, Intel experienced its largest decline in value as a public company, losing 60% of its worth.
• In December, Broadcom experienced its best month ever, with its stock price more than doubling following the release of its earnings report.
• The driving force behind the diverging narratives was artificial intelligence.

The year was significant for silicon in Silicon Valley, but particularly challenging for the company that gave the region its nickname.

The chipmaker, founded by industry pioneers Gordon Moore and Robert Noyce and legendary investor Arthur Rock, experienced its worst year since going public in 1971, with a 61% loss in value.

In contrast to the story that unfolded at the chip conglomerate run by CEO Hock Tan and headquartered in Palo Alto, California, about 15 miles from Intel's Santa Clara campus, the opposite occurred.

Avago's 2015 acquisition resulted in the formation of Broadcom, whose stock price increased by 111% in 2024, marking its best performance to date.

The driving force behind the diverging narratives was the impact of artificial intelligence. While Broadcom embraced the AI revolution, Intel largely missed it. The shifting fortunes of the two chipmakers highlight the transient nature of leadership in the tech industry and how a few critical decisions can result in significant market cap fluctuations.

Broadcom produces custom chips for major cloud companies and essential networking equipment for large server clusters. In the field of AI, Broadcom has been largely eclipsed by NVIDIA, whose GPUs are used in most large language models developed at OpenAI, Google, and Microsoft. NVIDIA's GPUs also power the most demanding AI workloads.

Although Broadcom's XPUs, despite their lower profile, have emerged as a vital component of the AI industry.

Eric Ross, chief investment strategist at Cascend, stated on CNBC's "Squawk Box" that the reason for the increase is the frequent discussion about AI.

For years, Intel has been excluded from the AI market in the United States. Its server chips are significantly behind Nvidia's, and the company has also lost market share to its longtime competitor while investing heavily in new factories.

On Dec. 1, Pat Gelsinger was removed as CEO of Intel after four tumultuous years in the role.

"Someone more innovative may have predicted the AI wave," said Paul Argenti, professor of management at Dartmouth's Tuck School of Business, in an interview on "Squawk Box" following the announcement.

An Intel spokesperson declined to comment.

Broadcom is now valued at approximately $1.1 trillion and is the eighth U.S. tech company to surpass the trillion-dollar mark. As the second most valuable chip company, behind Nvidia, Broadcom has contributed to the AI boom's $3.4 trillion valuation. Nvidia's stock price increased by 178% in 2023, but it actually outperformed that in 2024, with a gain of 239%.

Intel, once the world's most valuable chipmaker with a market cap of nearly $300 billion in early 2020, is now worth about $85 billion and has been removed from the Dow Jones Industrial Average. The company is currently in discussions to sell off parts of its business and now ranks 15th in market cap among global semiconductor companies.

'Not meant for everybody'

After the Avago-Broadcom merger in 2015, the largest business of the combined company was the production of chips for TV set-top boxes and broadband routers. Broadcom continues to manufacture Wi-Fi chips used in laptops, the iPhone, and other smartphones.

In 2022, Broadcom completed its acquisition of server virtualization software vendor VMware for $61 billion, which accounted for 41% of its $14 billion in revenue in the most recent quarter. After a failed bid to buy mobile chip giant in 2018, Broadcom shifted its focus to software companies.

Wall Street is buzzing about Broadcom's collaboration with cloud providers to create custom chips for AI. Unlike Nvidia's GPUs, Broadcom's XPUs are more affordable and easier to manage, and they're optimized for running specific AI applications.

Nvidia's GPUs are being used by cloud vendors and other large internet companies to build their own models and run AI workloads for customers, with billions of dollars being spent annually. However, Broadcom's success with custom chips is creating a competition in the AI market, as hyperscale cloud companies seek to differentiate their products and services from their rivals.

Only a select few companies can afford to design and build their own custom processors using Broadcom's chips.

"To use those chips, you must be a Google, Microsoft, or another company, according to Piper Sandler analyst Harsh Kumar, who made the statement on CNBC's "Squawk on the Street" on Dec. 13, one day after Broadcom's earnings. These chips are not intended for everyone."

Broadcom's AI revenue increased by 220% in 2024, and the month of December has pushed it to new heights. The stock has risen 45% for the month as of Monday's close, surpassing its previous best month by 16 percentage points.

During the company's earnings call on Dec. 12, Tan announced that Broadcom had doubled its shipments of XPUs to its three major hyperscale providers, including Google, which uses the technology for its Tensor Processing Units, TPUs, to train Apple's AI software released this year. The other two customers, according to analysts, are ByteDance and Meta.

In approximately two years, companies may allocate between $60 billion and $90 billion on XPUs, as stated by Tan.

Tan stated that in 2027, each of the three hyperscale customers plans to deploy 1 million XPU clusters on a single fabric.

To train the most advanced models, AI server clusters require powerful networking components. In the fourth quarter, networking chips for AI accounted for 76% of Broadcom's $4.5 billion in networking sales.

In 2024, Broadcom reported that approximately 40% of its $30.1 billion in semiconductor sales were related to AI, and that AI revenue would increase by 65% to $3.8 billion in the first quarter.

"The success of hyperscalers in their initiatives here is a topic of debate, according to Cantor analyst C.J. Muse, who recommends buying Broadcom shares. However, regardless of how you look at it, the focus on custom silicon will continue to benefit those who are leveraged to it."

Intel's very bad year

Intel's worst year on the market prior to 2024 was 1974, when the stock dropped 57%.

Years ago, Intel missed out on the mobile chip market to Qualcomm, ARM, and Apple, which ultimately led to the company's latest struggles.

AMD has been gaining market share in the PC and server CPU markets due to its successful manufacturing partnership. Intel's manufacturing process has lagged behind for years, resulting in slower and less energy-efficient CPUs.

Gelsinger's removal as CEO of Intel was largely due to the company's failed attempts in the field of AI.

Nvidia's GPUs, initially intended for video games, have become essential in developing power-intensive AI models. Intel's CPU, once the main component in a server, is now overshadowed by AI servers. In 2025, Nvidia will ship GPUs that don't require an Intel CPU; many of them will be paired with an Nvidia-designed ARM-based chip.

Intel has been forced into downsizing mode due to Nvidia's revenue growth of at least 94% for the past six quarters. Sales have declined in nine of the past 11 periods, prompting Intel to announce in August that it was cutting 15,000 jobs, or about 15% of its workforce.

Frank Yeary, Intel's board chair, announced in a Dec. 2 press release that the company is working to become leaner, simpler, and more agile.

Intel faces a significant challenge in developing a comprehensive AI strategy, despite its efforts to showcase its AI capabilities through its laptop chips and the release of its Nvidia competitor, Gaudi 3. Despite these efforts, Intel's AI PC initiative and Gaudi chips have not gained much market traction, with Gaudi 3 sales falling short of the company's $500 million target for this year.

Intel will release a new AI chip named Falcon Shores, which is a GPU and not built on Gaudi 3 architecture, late next year.

Michelle Holthaus, Intel's interim co-CEO, stated at a Barclays financial conference on Dec. 12 that it will not be wonderful, but it is a good first step towards completing the platform.

Intel's interim co-CEOs Holthaus and Zinsner have pledged to prioritize the development of Intel's products, leaving the future of the company's costly foundry division uncertain.

Gelsinger, before leaving, advocated for Intel to establish a presence in the semiconductor market and manufacture chips to compete with TSMC. In June, at a Taipei conference, he told CNBC that Intel aimed to produce "everybody's AI chips" and offer an alternative to TSMC for companies such as Nvidia and Broadcom.

Intel announced in September that it intends to transform its foundry business into a separate entity with its own board and the possibility of obtaining external funding. However, Intel's main customer at present is still Intel, and the company anticipates that it will not see significant sales from external clients until 2027.

Zinsner stated at the Barclays event this month that the separate board for the foundry business is being "stood up" and the company is aiming to simplify and reduce costs.

Zinsner stated that we will continually examine our spending to ensure we receive the expected return.

WATCH: Intel plans to take chip subsidiary Altera public