Intel's Gaudi 2 Surge: Pioneering AI's Future

The company's chip races ahead in the processing showdown

Summary
  • Intel's Gaudi2, developed by Habana Labs, is a state-of-the-art AI processing chip that outpaces many competitors.
  • Nvidia remains a dominant player in the AI processing market, but Intel's continual advancements, especially with the potential 5nm Gaudi3, show it's a significant contender.
  • The semiconductor sector experiences volatility due to U.S.-China tensions; however, Intel's strategic U.S.-based foundry plans and resilience position it as a robust alternative in the chip market.
  • Intel's initiative to construct a U.S.-based foundry ecosystem could offer an alternative to TSMC.
Article's Main Image

In the accelerating artificial intelligence space, there is a unique tug-of-war happening, with companies scrambling to offer the most powerful and efficient tools for processing. At the forefront of this is Intel Corp. (INTC, Financial) with its Gaudi 2 chip.

Engineered by the brilliant minds at Habana Labs, which the company acquired in 2019, Gaudi 2 is no ordinary chip—it is the embodiment of raw power and unmatched speed, promising to redefine the landscape of AI technology.

Introduction to Gaudi 2

Intel has introduced a new, supercharged computer chip designed to handle better tasks related to artificial intelligence. This new chip works twice as fast as some competitors. It is like upgrading from a regular car to a race car in terms of performance.

This speedy chip can help businesses get results faster, especially in areas like self-driving cars and medical scans. Companies can now buy this new chip, and a few products are already being developed using it. Simply put, this is Intel's way of making AI tasks faster and more efficient.

Understanding the AI processing market

Intel identifies two distinct markets for AI processing power:

  1. Dedicated AI infrastructure: This is the domain of Habana Labs' Gaudi. Here, the aim is to build robust systems dedicated exclusively to AI processes, whether for training or inference.
  2. Casual AI applications: Think of everyday tools and workloads, where AI is just one feature amongst many. Here, Intel believes that CPUs like its Xeon line are more capable of handling the demand.

The division of these markets stems from a belief that AI techniques will not remain relegated to niche applications, but will become pervasive in the mainstream IT realm. And while Intel strides forward with this vision, other CPU giants like AMD (AMD, Financial) seem to lag, especially in embedding advanced AI capabilities into their CPUs.

Performance distinctions of Gaudi 2

Notably, the HBM-equipped Xeon Max 9480 CPU's performance is approximately half that of the DRAM-based Xeon 8480+ CPU. But the significant revelation is the role of Intel's Habana Gaudi2 in the AI cloud space. In inference processing, an integral part of AI cloud, the Gaudi 2 chip has showcased impressive performance metrics, outperforming Nvidia Corp.'s (NVDA, Financial) A100 and coming close to H100 Hopper GPU's capabilities.

In the AI training segment, Intel unveiled the Gaudi 2, demonstrating its competitive edge even with software constraints. The upcoming Intel Innovation event is keenly awaited, with the company hinting at leveraging the FP8 precision quantization for the chip, which can substantially boost its performance.

Manufactured on the Taiwan Semiconductor Manufacturing Co. Ltd. (TSM, Financial) 7nm node, Gaudi 2 holds its ground against Nvidia's 5nm Hopper GPU. Rumblings of a 5nm Gaudi 3 on the horizon further amp up the anticipation.

Nvidia's dominance in AI processing

Nvidia, a juggernaut in this space, continues to hold its top position, clinching victories in every benchmark test. The Grace Hopper, Arm CPU and Hopper GPU Superchip are emblematic of the company's innovation in AI inferencing. Its recent announcement about a new TensorRT tailored for LLMs, which promises to double the H100's inference processing prowess, is a testament to the company's relentless push for supremacy.

However, this software update, arriving post-MLPerf submissions, highlights the software's pivotal role in optimizing hardware. Given that TensorRT-ML is open-source, other players like Intel and AMD could capitalize on these concepts, adapting them to boost performance metrics.

The dynamic landscape of AI processing

While Nvidia emerged as the victor in the recent MLPerf benchmarks, Intel's Gaudi 2 is a formidable AI training and inference contender. But the tech world is ever-evolving. Imminent software enhancements for Gaudi2 and the potential unveiling of the 5nm Gaudi 3 make the company a force to be reckoned with.

In the near future, with AMD's MI300 launch, Google's TPUv5 introduction and Intel 's software and hardware advancements, the AI processing battle is only intensifying.

In summary, while Nvidia currently reigns supreme, Intel's Gaudi2, with its promise and potential, underscores innovation as a continual journey, and the race for AI supremacy is far from over.

Impact of U.S.-China tensions on semiconductor stocks

Tensions between the U.S. and China have traditionally caused semiconductor stocks to dip, a trend that is evident once again. Reports suggested China might prohibit its government agencies from using iPhones. This news triggered a dip in Apple's (AAPL, Financial) stock and extended its impact on its suppliers and semiconductor stocks.

While many view this sell-off as an overreaction, it is interesting that Intel often emerges from such situations stronger than before. Despite the company's significant exposure to China, mainly through its PC and data center processors, it is making strategic moves that might shield it from such geopolitical tensions.

Intel's U.S.-based foundry ecosystem

Intel has started construction on a U.S.-based foundry ecosystem to challenge giants like Taiwan Semiconductor. Considering potential geopolitical challenges, such as the speculated invasion of Taiwan, which could cripple the company, Intel could become a primary alternative for chip designers.

Moreover, Intel recently garnered positive analyst remarks, further boosting its position. At an industry conference, CEO Pat Gelsinger unveiled that a substantial pre-payment had been made for its 1.8nm fab, which is set to be operational by 2024. This cutting-edge facility aims to produce chips that could compete directly with, or even outperform, Taiwan Semiconductor Manufacturing's offerings.

Resilience of Intel's stock

Despite U.S. export constraints, recent revelations about China's capability to produce advanced 7nm chips saw Intel's shares moving in a direction opposite to the broader tech sector. Speculations also abound that future restrictions on chipmaking equipment sales to China might favor the company, boosting its stock.

Takeaway

The semiconductor industry is navigating an intricate web of technological advancements, geopolitical tensions and fierce competition. At the intersection of these challenges, Intel stands tall with its innovative strides, particularly in AI processing with Gaudi 2 and its budding U.S.-based foundry ecosystem.

The company's remarkable resilience in the face of market fluctuations, underscored by its strategic shifts and bold ventures, speaks volumes about its adaptability and long-term vision.

Disclosures

I am/we currently own positions in the stocks mentioned, and have NO plans to sell some or all of the positions in the stocks mentioned over the next 72 hours. Click for the complete disclosure