The Intel (INTC) Conundrum: When Will the Bleeding Stop?

Prominent chipmaker Intel Corporation’s (INTC) shares plunged more than 14% over the past five days. This downward trend follows the revelation that Intel incurred a significant operating loss of $7 billion last year for its chip-manufacturing unit, also called the foundry business, about 35% worse than in 2022. The unit reported revenue of $18.90 billion for 2023, down 31% year-over-year.

During an investor presentation, INTC’s CEO Patrick Gelsinger discussed the company's projections, stating that 2024 would likely mark the peak of operating losses for its chipmaking division. He mentioned that Intel anticipates reaching break-even on an operating basis by around 2027.

Pat Gelsinger further acknowledged challenges in the company’s foundry business, attributing to poor decisions, including one year ago against extreme ultraviolet (EUV) machines from the Dutch company ASML Holding N.V. (ASML). Although these machines can cost more than $150 million, they are considered more cost-effective compared to earlier chip-making tools.

Partially due to these missteps, Intel has outsourced approximately 30% of its total wafer production to external contract manufacturers like TSMC, Gelsinger added. The company’s goal is to lower this number to around 20%.

Additionally, the semiconductor giant has now transitioned to using EUV tools, which are expected to handle an increasing portion of production requirements as older machinery is phased out.

“In the post EUV era, we see that we're very competitive now on price, performance (and) back to leadership,” Gelsinger stated. “And in the pre-EUV era we carried a lot of costs and (were) uncompetitive.”

However, on a negative note, investment bank Bernstein analysts recently remarked that there is no compelling reason to hold Intel stock until 2030.

Bernstein recognizes the potential for improvement in Intel’s foundry business, given the significant loss incurred last year and the optimistic projection for achieving a 25-30% operating margin by 2030.

However, analysts cautioned, suggesting that the road ahead for INTC might be challenging, even with the seemingly ambitious targets. They noted that reaching break-even may not happen until after 2027, and the ambitious goals set for 2030 are speculative and dependent on achieving optimal progress, which remains a topic of debate.

In the last reported earnings, INTC surpassed analysts’ estimates on revenue and EPS. However, the chipmaker announced a weak forecast for the current quarter. For the quarter that ended December 31, 2023, INTC’s net revenue increased 10% year-over-year to $15.40 billion. This surpassed the consensus revenue estimate of $15.17 billion.

Also, net income attributable to Intel was $2.70 billion, compared to a net loss of $700 million in the previous year’s period. The company reported an EPS of $0.63, compared to analysts’ estimate of $0.22, and a loss per share of $0.16 in the same quarter of 2022.

However, as of September 30, 2023, the company’s cash and cash equivalents stood at $7.07 billion versus $11.14 billion as of December 31, 2022.

Intel’s fourth-quarter 2023 report marked a return to growth after eight consecutive quarters of decreasing earnings and seven straight quarters of declining sales on a year-over-year basis. But for the first quarter, the chip company projected adjusted EPS of just $13 on sales of $12.70 billion. Analysts expect earnings of $0.14 per share on revenue of $12.78 billion.

During an earnings call, Intel CEO Patrick Gelsinger stated that the company’s first-quarter sales would be impacted by difficulties at Mobileye, where Intel holds a majority stake, and in its programmable chip unit.

Gelsinger also mentioned that Intel’s core businesses, particularly PC and server chips, were performing strongly, with sales expected to fall within the lower end of the seasonal range.

On March 21, INTC announced plans to invest $100 billion in constructing and expanding chip factories across four states in the U.S., following securing $19.50 billion in federal grants and loans and wishes to secure another $25 billion in tax breaks.

Intel’s primary focus in its five-year spending plan is to convert undeveloped land near Columbus, Ohio, into what CEO Pat Gelsinger described as “the largest AI chip manufacturing site in the world,” with potential commencement in 2027.

In addition, the chip giant intends to revamp sites in New Mexico and Oregon while expanding its presence in Arizona. This initiative aligns with rival Taiwan Semiconductor Manufacturing Company Ltd. (TSM) construction of a massive factory in Arizona, leveraging President Joe Biden's efforts to bolster advanced semiconductor manufacturing in the U.S.

Intel was at the forefront of the semiconductor industry for decades and was known for producing the fastest and smallest chips. The company commanded premium prices for its products and reinvested its profits into continuous research and development (R&D), aiming to stay ahead of its competitors.

However, in the 2010s, INTC’s manufacturing superiority waned, particularly in comparison to TSM. This shift resulted in a significant drop in profit margins as Intel had to lower prices to maintain its market share, even though its products were perceived as less competitive than its rivals.

In 2021, Gelsinger unveiled a strategy to restore Intel to its former top position in the semiconductor market, acknowledging the necessity of government support to ensure the plan’s profitability. With the federal support secured, the chipmaker is now gearing up for substantial investments.

Gelsinger mentioned that approximately 30% of the $100 billion budget will be earmarked for construction expenses, covering labor, piping, and concrete. The remaining funds will be utilized to acquire chipmaking tools from firms like ASML, Tokyo Electron, Applied Materials, Inc. (AMAT), and KLA Corporation (KLAC), among others.

Moreover, Intel's strategy for business turnaround hinges on persuading external companies to use its manufacturing services. In February, INTC announced that Microsoft Corporation (MSFT) plans to use its services to manufacture a tailored computing chip. Moreover, the company expressed optimism about exceeding its internal target of surpassing TSM in advanced chip manufacturing before 2025.

As a part of this plan, INTC recently told investors it would start reporting the results of its manufacturing operations as a separate unit.

Intel’s new reporting structure, effective from the first quarter of 2024, includes operating segments such as Client Computing Group (CCG), Data Center and AI (DCAI), Network and Edge (NEX), Intel Foundry, Altera (now Intel’s Programmable Solutions Group), Mobileye, and Other. CCG, DCAI, and NEX will be collectively known as Intel Products, while Altera, Mobileye, and Other will be referred to as All Other.

The newly established Intel Foundry segment, including foundry technology development, foundry manufacturing and supply chain, and foundry services, will recognize revenues generated from external foundry customers and Intel Products, along with technology development and product manufacturing costs historically allocated to Intel Products.

Intel’s CFO, Dave Zinsner stated, “This model is designed to unlock significant cost savings, operational efficiencies and asset value. As it begins to take hold, we expect to accelerate on our path toward achieving our ambition of 60% non-GAAP gross margins and 40% non-GAAP operating margins in 2030. Ultimately, improved cost competitiveness will help us deliver process technology, product, and foundry leadership while driving significant financial upside for Intel and our owners.”

Bottom Line

Last week, INTC confirmed its intention to separate the financial results of its foundry business, providing investors with a closer look at its historical performance. However, the revealed figures were disappointing: the foundry business suffered losses of nearly $7 billion in 2023, a 35% increase in losses compared to 2022, alongside a 31% decrease in sales.

Along with these figures, the company stressed that the new financial reporting structure is designed to boost cost discipline and higher returns by offering enhanced transparency, accountability, and incentives across the business. Moreover, this transition is expected to unlock unrealized value across Intel’s about $100 billion in capital assets.

Last month, Intel unveiled plans to spend those $100 billion on building or expanding chip factories in four U.S. states. As part of its turnaround strategy, the chipmaker aims to convince external companies to utilize its manufacturing services. The company has been heavily investing to compete with its main chipmaking rivals, including TSM and Samsung Electronics Co Ltd.

Despite Intel’s optimism about turning the business around and achieving break-even by 2027, with a projected adjusted operating profit margin of 30% by 2030, analysts, including those at Bernstein, are cautious. They view Intel’s forecast as overly ambitious, suggesting that actual margins might only reach 25% by 2030.

Further, CNBC’s Jim Cramer advises investors to avoid investing in Intel despite the company’s turnaround plans.

While INTC is actively pursuing its turnaround initiatives, it currently encounters significant challenges, including underperformance within its foundry business, fierce competition, and cash burn. So, it could be wise to steer clear of this stock now.

Intel Backed Astera Labs Goes Public: What Investors Need to Know

In 2020, Intel Capital, the global investment arm of Intel Corporation (INTC), made a significant move by unveiling a momentous investment initiative of $132 million. This injection of funds fueled the growth of 11 cutting-edge technology startups, each driving innovation in Artificial Intelligence (AI), autonomous computing, and chip design.

Among these 11 disruptive startups, Astera Labs, Inc. emerged as a key player, ready to redefine connectivity solutions and spearhead transformative industry shifts. Established in 2017, Astera quickly made waves in the semiconductor realm with its state-of-the-art connectivity solutions tailored for AI and cloud applications.

Notably, last year, the chip startup made a significant announcement regarding its entire product portfolio. The company revealed that its products are now equipped to fast-track cloud-scale workloads with unprecedented scaling capabilities.

This development comes at a time when the global cloud AI market is projected to experience substantial growth. It is estimated to reach $647.61 billion by 2030, growing at an impressive CAGR of 39.6% from 2023 to 2030.

Astera’s innovative approach to data center connectivity, bridging accelerators, CPUs, GPUs, memory, and networking, is proving vital in supporting the expansion of cloud infrastructure for AI and data-driven applications at scale.

Thad Omura, SVP of Business and Corporate Development at Astera Labs, underscored the prowess of the company's hardware-optimized PCIe, CXL, and Ethernet connectivity portfolio. He emphasized that these solutions have been meticulously crafted with a software-defined approach, enabling them to deliver unprecedented scale for AI infrastructure and cloud connectivity.

Moreover, encouraged by the prevailing AI frenzy, the chip startup recently announced its plans to transition into a publicly traded entity, set to trade under the ticker symbol “ALAB” on NASDAQ.

The company is gearing up to offer 17.80 million shares priced between $27 and $30 each, aiming to raise to $534 million. Following the completion of the deal, there will be 150.50 million outstanding shares, potentially valuing the company at $4.50 billion at the upper end of that range.

Morgan Stanley and J.P. Morgan have taken on the role of lead joint book runners for this Initial Public Offering (IPO). Additionally, Barclays, Deutsche Bank Securities, Evercore ISI, and Jefferies are also serving as book runners. Meanwhile, Needham & Company, Stifel, Craig-Hallum Capital Group, Roth Capital Partners, Loop Capital Markets, and Siebert Williams Shank have stepped in as co-managers.

The funds raised from this IPO will be allocated towards working capital and general corporate purposes, with the potential for utilization in acquisitions if suitable opportunities emerge. In a letter included in the filing papers, the three founders of Astera, Jitendra Mohan, Sanjay Gajendra, and Casey Morrison, share their unwavering commitment to the transformative power of AI.

The three founders kickstarted their venture in a garage. Their brainchild, Astera Labs, birthed three product families, each tackling crucial bottlenecks in AI infrastructure.

Aries, designed to enhance CPUs and GPUs, facilitates the scaling of data input/output bandwidth. Taurus focuses on providing AI servers with accelerated network bandwidth. Meanwhile, Leo enables the seamless scaling of memory bandwidth and capacity for CPUs and GPUs.

Astera projects its total addressable market (TAM) to skyrocket to nearly $27.40 billion by 2027, up from approximately $17.20 billion in 2023. In addition, the company boasts an impressive clientele, counting chip industry giants like INTC, NVIDIA Corporation (NVDA), and Advanced Micro Devices, Inc. (AMD).

Riding the wave of strong chip demand, the company witnessed a solid year-over-year topline growth in fiscal year 2023. In its Form S-1 filing, Astera reported revenue of $115.79 million, marking a robust 44.9% year-over-year growth. The company also saw a notable uptick in its gross profit, climbing to $79.83 million from $58.68 million the previous year.

On the other hand, as a result of making substantial investments in the design and development of new products and platform enhancements, the company remains unprofitable in 2023, incurring a net loss of $26.26 million and $0.71 per share. Yet, this reflects an improvement compared to fiscal year 2022, where the net loss was $58.35 million and $1.71 per share.

Bottom Line

Astera Labs has positioned itself as a frontrunner in the semiconductor industry, driven by its innovative connectivity solutions tailored for AI and cloud applications.

Backed by a significant investment initiative from INTC and boasting impressive financial growth, the company’s transition into a publicly traded entity signifies its confidence in its products and growth potential.

Furthermore, Astera’s IPO announcement arrives amid a perfect storm of opportunity. With the demand for semiconductor chips, especially those powering AI applications, soaring to new heights, the stage is set for Astera's ascent. For instance, NVDA’s stock has surged dramatically, seemingly unstoppable in its ascent.

Meanwhile, the shares of AMD and Taiwan Semiconductor Manufacturing Company Ltd. (TSM) have also witnessed significant rises, reflecting the broader trend of robust demand for semiconductor chips.

That said, with a strategic focus on addressing critical bottlenecks in AI infrastructure and a solid client base, including industry giants, Astera is well-poised to capitalize on the booming demand for semiconductor chips. Despite remaining unprofitable in fiscal year 2023, the company’s continuous investments in product development reflect its commitment to unlocking AI's full potential.

Moreover, the funds infusion from the IPO will provide Astera with a substantial financial boost, empowering the company to enhance its operational capabilities and drive strategic growth initiatives.

Overall, Astera’s IPO represents a significant milestone in its journey toward becoming a major player in the semiconductor landscape, with promising prospects for future growth and expansion. Keeping all these factors in mind, Astera emerges as a strong investment candidate.

 

NVIDIA (NVDA) vs. Advanced Micro Devices (AMD): Which Stock Is Proving to Be the Better Long-Term AI Buy

After its earnings release on May 24, the Santa Clara-based graphics chip maker NVIDIA Corporation (NVDA) stole the thunder by becoming the first semiconductor company to hit a valuation of $1 trillion.

NVDA has also blown away Street expectations ahead of its quarterly earnings release on August 23, with profits for the current quarter expected to be at least 50% higher than analyst estimates and the momentum expected to continue in the foreseeable future.

On the other hand, since its humble beginnings as a supplier for Intel Corporation (INTC), Advanced Micro Devices, Inc. (AMD) has come a long way. During its earnings release for the second quarter, despite persistent weakness in the PC market, the company’s result topped analyst estimates.

While NVDA has carved its niche and cornered a significant share of the GPU domain through advancements in parallel (and consequently accelerated) computing which began back in 2006 with the release of a software toolkit called CUDA, Chair and CEO Dr. Lisa Su is widely credited with AMD’s turnaround and transition from being widely dismissed due to performance issues and delayed releases to being the only company in the world to design both CPUs and GPUs at scale.

The New (Perhaps Only) Game in Town

As a general-purpose technology, such as the steam engine and electricity, Artificial Intelligence (AI) that has already been touching and influencing all facets of our life, including how we shop, drive, date, entertain ourselves, manage our finances, take care of our health, and much more.

However, late in November of last year, when OpenAI opened its artificial intelligence chatbot, ChatGPT, to the general public, all hell broke loose. The application took the world by storm. It amassed 1 million users in five days and 100 million monthly active users only two months into its launch to become the fastest-growing application in history.

The generative AI-powered application’s capability to provide (surprisingly) human-like responses to user requests equally fascinated and concerned individuals, businesses, and institutions with the possibilities of the technology. A large language model or LLM powers ChatGPT. This gives the application the ability to understand human language and provide responses based on the large body of information on which the model has been trained.

NVDA is reaping the rewards for all that invisible work done in the field of parallel computing. Parallel computing was ideal for artificial neural networks' deep (machine) learning. As a result of that head start in the AI tech race, its A100 chips, which are powering LLMs like ChatGPT, have become indispensable for Silicon Valley tech giants.

To put things into context, the supercomputer behind OpenAI’s ChatGPT needed 10,000 of NVDA’s famous chips. With each chip costing $10,000, a single algorithm that’s fast becoming ubiquitous is powered by semiconductors worth $100 million.

However, AMD isn’t too far behind either. According to Dr. Su, Data Center is the most strategic piece of business as far as high-performance computing is concerned. AMD underscored this commitment with the recent acquisition of data center optimization startup Pensando for $1.9 billion.

At the premiere, AMD’s ambitions to capitalize on the AI boom were loud and clear, with the launch of MI300X (a GPU-only chip) as a direct competitor to NVDA’s H100. The chip includes 8 GPUs (5nm GPUs with 6nm I/O) with 192GB of HBM3 and 5.2TB/s of memory bandwidth.

AMD believes this will allow LLMs’ inference workloads that require substantial memory to be run using fewer GPUs, which could improve the TCO (Total Cost of Ownership) compared to the H100.

The Road Ahead

The optimism surrounding both companies is justified.

With NVDA’s presence in data centers, cloud computing, and AI, its chips are making their way into self-driving cars, engines that enable the creation of digital twins with omniverse that could be used to run simulations and train AI algorithms for various applications.

On the other hand, AMD has also been training its guns to exploit the burgeoning AI accelerator market, projected to be over $30 billion in 2023 and potentially exceed $150 billion in 2027.

AMD is one of the few companies making high-end GPUs needed for artificial intelligence. With AI being seen as a tailwind that could drive PC sales, the company announced plans to launch new Radeon 7000 desktop GPUs at its quarterly earnings release. It is being speculated that the GPU will come with two 8-pin PCIe power connectors and four video out ports, including three DisplayPort 2.1 and one HDMI 2.1.

Caveats

AMD existed as both a chip designer and manufacturer, at least until 2009. However,  significant capex requirements associated with manufacturing, amid financial troubles in the wake of the Great Recession, compelled the company to demerge and spin off its fab to form GlobalFoundries Inc. (GFS), which has been focused on manufacturing low-end chips ever since.

Today, both NVDA and AMD operate as fabless chip companies. Hence, both companies face risks of backward integration by companies such as Apple Inc. (AAPL), Amazon.com, Inc. (AMZN), and Tesla Inc. (TSLA) with the wherewithal to develop the intellectual capital to design their own chips.

Moreover, almost all of the manufacturing has been outsourced to Taiwan Semiconductor Manufacturing Company Ltd. (TSM), which has yet to diversify significantly outside Taiwan and has become the bone of contention between the two leading superpowers.

With geopolitical risk being the potential Achilles heel for both companies, their efforts toward geographical diversification also receive much-needed political encouragement through the Chips and Science Act.

Dr. Su, who also serves on President Biden’s council of advisors on science and technology, pushed hard for the passage of the Act. It is aimed at on-shoring and de-risking semiconductor manufacturing in the interest of national security by setting aside $52 billion to incentivize companies to manufacture semiconductors domestically.

Bottom Line

Given its massive importance and cornucopia of applications, it’s hardly surprising that Zion Market Research forecasts the global AI industry to grow to $422.37 billion by 2028. Hence, this field has understandably garnered massive attention from investors who are reluctant to miss the bus on such a watershed development in the history of humankind.

Hence, in view of product diversification, increasing traction in the GPU segment, and relatively higher valuation comfort, investors in AMD could benefit from more sustained upside potential compared to NVDA.

Intel Corporation (INTC) Races to Dominate the AI Market – Will It Succeed?

In our June 3 post, we concluded that the resurgent Intel Corporation (INTC), which had weathered back-to-back quarterly losses amid softening PC demand, consequent surplus inventory, and realignment toward GPU-heavy and AI-centered enterprise demand, could be worth more than what was being suggested by its market price at that time.

By announcing its return to profitability during last week’s earnings release, the pioneer of modern computing lived up to our expectations while exceeding the ones on the Street. INTC posted a net income of $1.5 billion, compared to a net loss of $454 million during the previous-year quarter. Its adjusted EPS came in at $0.13 compared to an adjusted loss of $0.3 per share expected by Wall Street.

While the Market greeted the news with a 7% surge in the stock price in extended trading and a further 5% gain the following morning, INTC’s revenue, despite exceeding low expectations, declined 15.7% year-over-year to $12.9 billion, marking the sixth consecutive quarter of sales decline.
In view of a subdued topline, much of the outperformance in the quarterly results can be attributed to the progress INTC had made in cutting $3 billion in costs this year.

In addition to exiting nine lines of business since CEO Pat Gelsinger rejoined INTC to achieve a combined annual savings of more than $1.7 billion, the company slashed its dividend and announced plans to save $10 billion per year by 2025, including through layoffs.

With INTC’s cloud computing group, which includes the company’s laptop and desktop processor shipments, and server chip division, which is reported as Data Center and AI, reporting year-over-year declines of 12% and 15%, respectively, Pat Gelsinger forecasted “persistent weakness” in all segments of its business through year-end, and that server chip sales won’t recover until the fourth quarter.

With upside through cost optimization capped and customers prioritizing GPUs over CPUs to handle ever-increasing AI/ML workloads, INTC is eager to join the race currently being led by NVIDIA Corporation (NVDA) and Advanced Micro Devices, Inc. (AMD). The company is working on the manufacturing front, for which it significantly depends on Taiwan Semiconductor Manufacturing Company Ltd. (TSM).

By doubling down on the fab business, INTC aims to match TSM’s chip-manufacturing capabilities by 2026, enabling it to bid to make the most advanced mobile processors for other companies, a strategy the company calls “five nodes in four years.”

To that end, INTC is pursuing an aggressive IDM 2.0 road map with new manufacturing facilities in Oregon, New Mexico, Arizona, Ireland, and Israel in the pipeline to augment the capabilities of 15 fabs worldwide and facilities to assemble and test the manufactured chips in Vietnam, Malaysia, Costa Rica, China, and the U.S.

Among those, Arizona's new facilities would be manufacturing chips for the company and customers such as Amazon, Qualcomm, and others as part of Intel Foundry Services. While the company still depends on TSMC for 5nm chips used for AI applications, it aims to take a quantum leap in that direction with even smaller 18 A chips.

While companies such as Amazon.com, Inc. (AMZN) are resorting to chips designed in-house to support their cloud infrastructure, INTC, in addition to being the manufacturer of both wafers and packaging of AI accelerators, is also present with its Gaudi chips.

The company’s efforts are also receiving much-needed political encouragement in the form of the Chips and Science Act, which is aimed at on-shoring and de-risking semiconductor manufacturing in the interest of national security.

Recently, Pat and his team at INTC upped the ante by unveiling its new ambitious plans to incorporate AI into every product it creates. This announcement comes as the company’s upcoming Meteor Lake chips are rumored to feature a built-in neural processor specifically designed for handling machine learning tasks.

With an objective to “democratize AI,” Pat was loud and clear about INTC’s plans to make it a ubiquitous and integral feature of its products designed to cater to all segments of the computing ecosystem, including “at the edge in the Client, in the enterprise, as well as in the cloud.”
The upbeat CEO forecasted that AI would permeate all business domains, including the client-facing consumer electronics market, enterprise data centers, and even manufacturing, and make its way into personal devices, such as hearing aids and personal computers. AI is already present as a co-pilot for Windows 11, allowing users to type questions and perform specific actions, and it could play a significant role in the next iteration of Windows.

Bottom Line

For the third quarter of the fiscal, INTC expects adjusted earnings of 20 cents per share on $13.4 billion revenue at the midpoint.
Whether or not INTC manages to meet or exceed the above target could go a long way in helping investors determine if its ambitious turnaround is on track to restore the company to its former glory.

Chips and AI Advanced Micro Devices Inc. (AMD)'s Next-Level Breakthroughs!

Last month, we gauged the prospects of two semiconductor giants, NVIDIA Corporation (NVDA) and Intel Corporation (INTC), which have carved out their niches and cornered a significant share of the GPU and CPU domains, respectively. In this article, we have talked about another chip company and its agile efforts to grab the best of both worlds while creating a widespread following of its own.

Founded in 1968 by a group of 8 men led by the larger-than-life Jerry Sanders, Advanced Micro Devices, Inc. (AMD) released its first product in 1970 and went public in 1972. Despite starting life as a supplier for INTC, AMD parted ways with its client in the mid-80s, and by the late 80s, it reverse-engineered INTC’s products to make its own chips that were compatible with INTC’s software.

AMD existed as both a chip designer and manufacturer, at least until 2009. However, significant capex requirements associated with manufacturing, amid financial troubles in the wake of the Great Recession, compelled the company to demerge and spin off its fab to form GlobalFoundries Inc. (GFS), which has been focused on manufacturing low-end chips ever since.

With the acquisition of ATI, a major fabless chip company, in 2006, AMD began shifting its focus toward chip designing and turned to Taiwan Semiconductor Manufacturing Company Ltd. (TSM) as its exclusive chip manufacturer.

With manufacturing no longer weighing it down, AMD started catching INTC with its Zen line of CPUs. Earlier this year, the former made history by surpassing the latter’s market cap for the first time ever. Chair and CEO Dr. Lisa Su is widely credited with the turnaround and transition from being widely dismissed due to performance issues and delayed releases to being the only company in the world to design both CPUs and GPUs at scale.

We look at how Dr. Su and her team’s unwavering focus on great products, customer relations, and simplifying the company’s structure to respond to the dynamic business with agility are shaping AMD’s offerings in each product category.

CPU Portfolio

Despite a conservative outlook, AMD believes its Genoa CPU processors are superior to competitive offerings in terms of performance and efficiency across diverse workloads, including AI. During the recent AMD Data Center & AI Tech Premiere, the company expanded its EPYC server CPU portfolio by launching the highly anticipated Bergamo EPYC CPUs optimized for cloud environments.

Given the focus on single-threaded performance and energy efficiency, Meta Platforms, Inc. (META), which has collaborated with AMD to customize the design of the Bergamo server, reported seeing 2.5 times greater performance than AMD's previous generation Milan CPUs and notable improvements in total cost of ownership (TCO).

In addition, AMD also introduced Genoa-X as another workload-optimized alternative to Genoa for faster general-purpose computing and optimal technical computing tasks. The company also updated that its upcoming server CPU product, Turin, has shown promising initial results and remains on schedule for a 2024 release.

Data Center Portfolio

According to Dr. Su, Data Center is the most strategic piece of business as far as high-performance computing is concerned. AMD underscored this commitment with the recent acquisition of data center optimization startup Pensando for $1.9 billion.

At the premiere, AMD’s ambitions to capitalize on the AI boom were loud and clear, with the launch of MI300X (a GPU-only chip) as a direct competitor to NVDA’s H100. The chip includes 8 GPUs (5nm GPUs with 6nm I/O) with 192GB of HBM3 and 5.2TB/s of memory bandwidth.

AMD believes this will allow LLMs’ inference workloads that require substantial memory to be run using fewer GPUs, which could improve the TCO compared to the H100.

Lastly, the company aims to address the growing AI accelerator market, projected to be over $30 billion in 2023 and potentially exceed $150 billion in 2027.
Gaming and Other Applications.

While INTC and NVDA control most of the CPU and GPU market, respectively, AMD dominates gaming by designing 83% of gaming console processors.

The recently launched AMD Ryzen 5 5600X3D is equipped with AMD’s revolutionary 3D V-Cache technology. Despite being close to both the Ryzen 7 5800X3D and the non-3D Ryzen 5 5600X in terms of specifications, it comes with a lot of L3 cache, giving it an edge over the latter, thereby improving gaming performance.

Moreover, with Moore’s Law, which is the core of computer chip advancement, showing visible signs of a slowdown and the 5-decade-old x86 architecture gradually but surely being replaced by ARM, general-purpose computing using CPUs is making way for more customized solutions.

That has prompted AMD to acquire Xilinx for $49 billion to close one of the biggest acquisitions in semiconductor history. The investee is known for its reprogrammable adaptive chips called Field-Programmable Gate Arrays or FPGAs, which have diverse applications, such as robotics, telecommunications, agriculture, and space exploration.

As a result, AMD is expanding its footprint from PCs and supercomputers to Teslas and Mars Land Rover.

Road Ahead

Despite its future readiness, geopolitical tensions between the U.S. and China could turn out to be the Achilles heel for AMD since all of its chips are made in China and Taiwan. Also, Mainland China accounts for roughly 30% of the company’s revenues.

Dr. Su also serves on President Biden’s council of advertisers on science and technology, which pushed hard for the recent passage of the Chips and Science Act, aimed at on-shoring and de-risking semiconductor manufacturing in the interest of national security by setting aside $52 billion to incentivize companies to manufacture semiconductors domestically.

Geographical diversification, as a result of this Act, could act as a hedge against geopolitical tensions for AMD by reducing reliance on Asian manufacturing.

Bottom Line

As AMD continues to advance its x86 core computing chips along with diversifying to accommodate high-performance and customized computing, its more than 70% increase in stock price since the beginning of the year (and coincidentally during the AI wave) could be indicative of a company that is poised to gain market share and capitalize on the expanding demand for AI technology in various industries.