As Crypto and Risk Asset Worries Rise, Amazon Steps Into the AI Arena

Amazon Joins AI Arms Race as Crypto Miners Pivot to AI Infrastructure

Amazon is accelerating its AI ambitions with the launch of Trainium 3, a new chip engineered to compete with Nvidia’s dominant GPUs. The upgrade, available through Amazon Web Services (AWS), promises four times the training speed of the previous generation while keeping energy use constant.

The new chips are integrated into Amazon’s “UltraServers,” with each cluster capable of hosting up to 144 Trainium 3 units. The rollout strengthens Amazon’s position in large-scale language model training and other compute-heavy AI workloads, putting it in direct competition with Google and Nvidia as demand for AI infrastructure surges.

Google’s stronghold in AI model development—where it has an 87% chance of producing the leading model by year-end—has reportedly triggered a “code red” at OpenAI, highlighting the intensifying competition in the sector.

Scaling AI infrastructure, however, brings logistical challenges. High energy consumption and space requirements are hurdles few companies can tackle alone. Crypto miners, with their vast data centers, are stepping in to fill the gap, converting energy-intensive operations into AI-ready facilities.

Following the 2024 Bitcoin halving, firms such as Core Scientific, CleanSpark, and Bitfarms are rebranding as hybrid utility providers for AI infrastructure. IREN, a former bitcoin mining firm, surged after signing a $9.7 billion AI cloud contract with Microsoft, while TeraWulf launched a $9.5 billion AI joint venture with Fluidstack, backed by Google. These companies already control gigawatts of power capacity and data center infrastructure suitable for AI workloads.

Despite this momentum, risk remains. Bitcoin has dropped over 17% in the past 30 days, the CoinDesk 20 index fell 19.3%, and the NASDAQ 100 is down roughly 1.5% over the same period. Analysts warn that the AI infrastructure boom could mirror past tech bubbles. OpenAI has committed trillions to infrastructure spending, but much of this capital is recycled through the same players, creating vulnerabilities if AI demand slows. Bain & Co. predicts that a slowdown could leave companies with an $800 billion shortfall, requiring $2 trillion in combined annual revenue by 2030 to sustain projected compute needs.

If AI compute demand falters, hybrid operations could face liquidity pressures similar to the 2022 crypto crash, with potential knock-on effects for broader risk assets.

For now, crypto miners-turned-AI providers are betting on a new kind of digital gold rush—this time powered by GPUs instead of ASICs.