Amazon readies new AI chip to compete with Nvidia

Nvidia might have some serious competition in the AI chip space coming. Amazon is now seeking to bring its own chips to the market, as it attempts to break away from the chipmaking giant.

This effort is being led by Amazon’s 2015 purchase, Annapurna Labs, which it acquired for $350 million. The company is best known for providing Amazon’s Web Services (AWS) with its hypervisor (virtual machine management) hardware, Nitro.

It also built Graviton, a lower-power ARM-based alternative to AMD and Intel’s traditional server chips.

Now, Annapurna is testing Trainium 2, the latest bit of kit set to cement Amazon’s footing in the AI sphere. Unveiled in 2023, the name “Trainium” relates to its purpose in the AWS pipeline. Amazon intends to train AI models, supposedly providing four times faster training than the previous version, which launched in 2020.

Anthropic, the firm behind the ChatGPT alternate AI bot Claude, is currently testing the Trainium 2. The OpenAI competitor has received a $4 billion investment from Amazon.

It is also now trying to work a deal with Anthropic, where it will have the company train its AI models on the new Trainium 2, even though it prefers Nvidia hardware.

Amazon trying to feed its investment dollars directly back into Amazon Web Services and force Anthropic to use their chips instead of NVIDIA. Just incredible levels of financial manipulation, 100% sure they book this as revenue https://t.co/S0K97g6YIe pic.twitter.com/r0N8x02hXi

— Ed Zitron (@edzitron) November 7, 2024

Amazon, much like OpenAI and Microsoft, is desperately trying to build its own set of chips. Nvidia is dominating the market right now, with a whopping $26.3 billion in sales in the second quarter of 2024.

By removing Nvidia from the equation, it aims to not only reduce spending on production but on running the chips too.

Speaking with The Financial Times, Dave Brown, vice-president of compute and networking services at AWS said, “When you save 40 per cent of $1000, it’s not really going to affect your choice. But when you are saving 40 per cent on tens of millions of dollars, it does.”

Another goal is to have more control over the chips themselves. We’ve already seen Apple take back a vast majority of control over its products, as it eliminated Intel from its Mac machines in 2020. It spent $1 billion on the M3 chip’s development alone.

Apple’s M-series chip has been hailed as one of the best on the market, with the latest M4 model specifically designed to increase performance when using artificial intelligence.

However, Amazon, much like its competitors in the AI space, has dramatically increased its spending. In 2023, the shopping giant spent $48 billion and is expected to increase this to a massive $75 billion in 2024.

Microsoft saw an even larger increase, as it went from $10.7 billion in 2023 to an incredible $56 billion by August 2024. The race isn’t just to produce the best AI offering, but also to fight against the ongoing issues with power infrastructure and data centers.

Amazon is expected to spend around $150 billion on data center infrastructure in the coming fifteen years. AWS powers a vast majority of the internet at this point and with the constant rise of AI needs, it is seeking to expand that exponentially.

The post Amazon readies new AI chip to compete with Nvidia appeared first on ReadWrite.