In the first quarter of 2026, Amazon delivered a report card that captured the market's attention. The company's total revenue reached $181.5 billion, a year-over-year increase of 17%. Revenue from its cloud computing division, AWS, was $37.6 billion, up 28% year-over-year, marking the fastest growth rate in 15 quarters. Within this earnings report, the most striking highlight was its rapidly emerging custom chip business.

Amazon CEO Andy Jassy, for the first time, systematically disclosed that the annualized revenue run rate of the company's chip business—encompassing Graviton processors, Trainium AI training chips, and Nitro security chips—has exceeded $20 billion and is growing at a triple-digit percentage year-over-year. Even more noteworthy, Jassy pointed out that if this business were treated as an independent company selling chips to both AWS and external customers like other chip giants, its annualized revenue run rate would approach $50 billion.

Three Major Chip Series Working in Concert

Amazon's rapidly expanding custom chip business consists of three core product lines: the ARM architecture-based Graviton server processors, the Trainium accelerator chips designed specifically for AI training, and the Nitro dedicated chips for virtualization and security.

Graviton processors have become the mainstay of AWS infrastructure, with 98% of the top 1,000 EC2 customers currently using them. Jassy noted that Graviton offers 40% better price-performance than any other x86 processor, giving it a significant advantage in general-purpose computing.

The Trainium series of AI chips represents Amazon's key strategic move in the artificial intelligence field. The Trainium2 chip offers approximately 30% better price-performance than comparable GPUs and is essentially sold out. Meanwhile, the Trainium3, manufactured using a 3-nanometer process, is set to begin shipping in early 2026. It offers a further 30% to 40% improvement in price-performance over Trainium2 and is already nearing sell-out status.

Explosive Growth in Customer Demand

The rapid growth of Amazon's chip business is fueled by strong market demand. Jassy revealed that over the past 12 months, more than 2.1 million AI chips have been deployed, with over half being Trainium chips. Revenue commitments for Trainium exceed $225 billion, demonstrating high customer recognition for Amazon's self-developed chips.

OpenAI has committed to using approximately 2 gigawatts of Trainium capacity via AWS to support its frontier models, with the agreement being phased in starting in 2027. Anthropic has committed to securing up to 5 gigawatts of current and future Trainium capacity for training and running its advanced AI models. Furthermore, Meta has signed an agreement to deploy tens of millions of AWS Graviton cores for its intelligent AI workloads. Uber is also partnering with Amazon to use Graviton4 and Trainium3 on its mobility and delivery platforms.

Future Outlook

Amazon is considering selling its self-developed chips directly to third-party customers for the first time, which would put it in direct competition with chip giants like NVIDIA and Broadcom in a broader market. This shift signifies that Amazon's chip business would transform from a cost center into a profit center, evolving from an internal tool into a core competency for external output.

CONEVO Professional Distributor

CONEVO Independent Distributor, specializing in providing emergency components and scarce electronic components for global electronic manufacturing enterprises with a one-stop procurement solution. It has a wide global supplier network and millions of scarce electronic component inventories, capable of quickly responding to customers' urgent needs and completing deliveries within a short period of time, effectively addressing the risk of production disruptions.

CONEVO not only provides IC chips, memory wafers, diodes, MCUs, etc., but also covers passive components such as capacitors, resistors, inductors, and structural components such as connectors and terminals, forming a complete material supply system. Here are several popular selected IC components.

● XCZU4EV-2SFVC784I: AMD Zynq UltraScale+ MPSoC EV series high-performance system-on-chip, integrating four-core ARM Cortex-A53, dual-core Cortex-R5 processors and 192K+ logic units of FPGA.

● TPS3850H01QDRCRQ1: TI automotive-grade high-precision voltage monitor, with 0.8% overvoltage/undervoltage threshold accuracy and programmable window watchdog timer.

● 5SGXMA3E3H29C2L: Altera Stratix V GX series high-end FPGA, containing 340,000 logic units and 19.5Mb embedded RAM, supporting multiple high-speed interface protocols.

● AT91SAM9260B-CU: Microchip 180MHz high-performance microprocessor, integrating rich peripherals including USB 2.0, 10/100Mbps Ethernet MAC and image sensor interface.

● LNK625DG-TL: Power Integrations LinkSwitch-CV series offline flyback switching power supply IC, simplifying low-power power supply design.

Website: www.conevoelec.com

Email: info@conevoelec.com