
ARM hits the stock market with a bang

Shares in ARM jumped 25% on the opening day of its public offering after an enthusiastic response from the market.
Shares in ARM were priced at the top end of the valuation at $51, but rose to $66 by the end of trading on the Nasdaq exchange in New York. This boosts the valuation of the company to over $65bn, compared to the $32bn paid by Softbank of Japan to take the company private in 2016. Softbank has said it will not sell more than the 9.7% of shares already on the market.
“It’s a great milestone but it is just one milestone on our journey,” Paul Williamson senior vice president and general manager of the IoT business at ARM tells eeNews Europe.
“There are big changes since we were last public when we were just focussed on CPUs, now AI is going to be everywhere and even today the majority of AI is running on ARM on CPUs.”
- Customers agree to buy-in to IPO
- IPO filing reveals depth of Chinese risk
- IPO filing updated with reduced valuation
The share offering was underwritten by the majority of the largest chip makers in the world, including Apple, Nvidia, Intel, AMD and Samsung as well as Google-owner Alphabet and the world’s largest chip foundry TSMC.
“Having cornerstone investors is an exciting expression of the confidence they have in ARM, but ur overall business remains to grow across the markets, and grow across all the segments. The data centre requirements for Nvidia look very different to IoT with microcontrollers with more matrix capability into the deeply embedded CPUs as well as the Ethos AI accelerators.”
However the missing chip maker that is significant for smartphones and the IoT is Qualcomm, where ARM has on-going litigation. “I’m not going to comment on Qualcomm,” said Williamson.
To boost the adoption of ARM AI technologies in the Internet of Things (IoT) and embedded applications he points to the ARM flexible access programme. This allows companies to design with a range of ARM IP and only pay for it when the chip tapes out. This has been used by Raspberry Pi for its RP2040 microcontroller for the CPU cores, but Williamson points to the more integrated subsystems that can include an AI accelerator and cores such as the M23 and now the M55 and M85.
- ARM expands IoT solutions portfolio
- First Cortex-A pre-validated sub-system
- Compute subsystem extends to custom data centre chips
“In the power constrained IoT space we offer Corstone with the open source frameworks for AI and CMSIS for DSP and we provide Corstone in the flexible access programme,” he said. “With the two most recent cores, the M85 and M55, we are still at the point of introduction of new devices and these implement the Helium extensions with accelerated matrix and vector instructions. But its not just the hardware but the libraries tools with CMSIS and TensorFlow as well as our own Keil tools.”
More than 250 billion ARM-based chips have shipped to date, powering over 99 percent of the world’s smartphones but the design focus has shifted. “In 2017, we pivoted from designing general-purpose CPUs for smartphones and consumer devices to designing purpose-built CPUs for specific markets,” said Rene Haas, who took over as CEO of ARM following the failed $40bn acquisition bid by Nvidia last year.
- CEO replaced as ‘unwanted’ sale to Nvidia collapses
- German AI RISC-V joint venture poses challenge
- ARM battles RISC-V at Renesas
“ARM’s growth is no longer solely defined by the smartphone market. We have a much more diversified business with market-specific compute platforms deployed in more mobile devices, cloud infrastructure, automotive and IoT.
“But now our partners need even more from us. Chip manufacturing cycle times are increasing, and the pace at which they need to deliver their products is unrelenting, all while chips get more complex and are being broken up into smaller chips consisting of subsystems that are blocks of IP.”
This is driving the use of Corstone and the Compute SubSystem (CSS), combined with the software ecosystem, to fight off the rise of chips being developed using the RISC-V open instruction set architecture.
“To reduce design time, these subsystems require a fully verified, tested compute solution that works right out of the box. This is a new growth opportunity for ARM as we’re uniquely positioned to deliver these compute subsystems across our key markets,” said Haas.
