Arm will sell its shares for between $47 and $51, aiming to raise up to $4.87 billion.
That would value the company at around $52 billion, below the reported initial target valuation of between $60 billion and $70 billion.
Arm opted to IPO on the Nasdaq after British government and stock market leaders failed to convince its parent company SoftBank to run the offering in London instead. Analysts have noted that public companies are typically able to achieve higher valuations in the US, especially in the cases of tech companies and the Nasdaq exchange.
The lower offering price, however, suggests that US investors may not value Arm quite as highly as Softbank hoped when it opted to reject London.
The $50 billion valuation is also significantly below the $64 billion worth of the business implied when SoftBank bought a 25% stake in Arm from its own flagship Vision Fund just weeks ago.
A number of tech giants, including Apple, Google, Intel and Nvdia, have already agreed to buy a combined $735 million worth of shares when Arm goes public.
Arm was founded in Cambridge in 1990, and listed its shares in London from 1998 until it was acquired by Japan-based SoftBank in 2016.
In the year to 31 March, Arm’s sales dipped by 1% to $2.68 billion. Profit was down by 5% to $524 million.
When it first filed for the offering last month, Arm talked up the importance of its chips to the growing AI sector.
It said: “As the world moves increasingly towards AI- and ML-enabled computing, Arm will be central to this transition.
“Arm CPUs already run AI and ML workloads in billions of devices, including smartphones, cameras, digital TVs, cars and cloud data centers. The CPU is vital in all AI systems, whether it is handling the AI workload entirely or in combination with a co-processor, such as a GPU or an NPU.
“In the emerging area of large language models, generative AI and autonomous driving, there will be a heightened emphasis on the low power acceleration of these algorithms. In our latest ISA, CPUs, and GPUs, we have added new functionality and instructions to accelerate future AI and ML algorithms.”