Amazon unveils its 4th-gen Graviton4 chip amid rising competition

Amazon Unveils Fourth-Generation Graviton4 Chip

Amazon Web Services (AWS) has announced the release of its new Graviton4 processor, marking its fourth generation of the Graviton chip series. Unveiled exclusively to Yahoo Finance, the Graviton4 promises significant advancements in performance and efficiency. According to Rahul Kulkarni, AWS's lead product manager for AI infrastructure, the new chip offers three times the compute power and memory of its predecessor, has 75% more memory bandwidth, and delivers 30% better performance.

Boosting Price Performance

"Collectively it's delivering more price performance, which means for every dollar spent, you get a lot more performance," Kulkarni said. This launch underscores the escalating demand for semiconductors, which are crucial to the global economy. Currently valued at $544 billion, the semiconductor sector is anticipated to surpass $1 trillion by 2033, primarily driven by the growing need for AI technology. Major players like Amazon, Apple, Alphabet, and Microsoft continue to invest heavily in custom chip development to meet specific needs and reduce costs.

Nvidia's Market Dominance

Despite the burgeoning competition, Nvidia remains the dominant entity in AI chips, holding over 80% of the market share for GPUs. According to Patrick Moorhead, CEO and chief analyst at Moor Insights & Strategy, the market has enough demand to accommodate multiple competitors. Although the Graviton4 is not an AI chip, it supports AWS's Inferentia and Trainium chips tailored for AI, directly competing with Nvidia's market-leading offerings.

Focusing on Price-Performance

Rather than challenging Nvidia's supremacy, AWS aims to provide cost-effective alternatives. "Right now, if a customer is more focused on time to market, Nvidia-based products that we offer are a great option," Kulkarni explained. However, for customers focusing on cost-efficiency, AWS's Inferentia and Trainium products are positioned as strong alternatives for AI workloads like training and inference.

Cost Reduction Strategies

Amazon lowers development costs by not only purchasing chips but also designing its own. AWS designs two types of chips: those for AI and general-purpose chips like the Graviton4. During a visit to AWS’s lab, it was observed that engineers solder and test chip designs while optimizing performance using software, all aimed at cost reduction. "We can tune the product, tune this silicon, to just focus on things that really matter for customer workloads," Kulkarni stated.

Affordable Compute Power

AWS has not released specific pricing details for the Graviton4 but has revealed that it rents the processors at $0.02845 per second of compute power. This pricing strategy highlights AWS's focus on price-performance, aiming to save costs for its cloud infrastructure and server operations. "AWS customers can use Graviton4 to cut their bill for IT in half," commented Moorhead.

Long-Term Investment

The development of chips is costly, but it offers considerable savings eventually. Moorhead added that substantial savings could be achieved in the long term, which can be reinvested into further development. Investor expectations surrounding earnings from chip development play a crucial role in company profits, and AWS appears to be a growing influence in the semiconductor sector. Analysts have noted higher earnings expectations for Amazon compared to Microsoft or Alphabet, with AWS's profit margins reaching 38% in early 2024.

Gaining Credibility

"AWS has a lot of credibility in the semiconductor space," Moorhead stated, highlighting how AWS has effectively leveraged its capabilities to become a formidable player in the sector.