February 21, 2023, London, UK – According to Omdia’s latest “Top AI Hardware Startups Market Radar” report, since 2018, more than 100 different venture capitalists (VCs) have invested more than $6 billion in the top 25 AI chip startups.
Although 2021 will be remembered as a special year, it is clear that the financing environment has changed. The shift from a global chip shortage to an inventory crisis, the turning point of monetary policy, and the 2022 recession mean that raising funds is now more challenging.https://www.stoneitech.com/
“The most financially abundant AI chip startups are facing pressure to provide developers with the software support they are accustomed to receiving from market leader Invista.” Alexander Harrowell, senior computing chief analyst at Omdia, said, “This is a key barrier to bringing new AI chip technology into the market.”
Omdia predicts that more than one large startup may exit this year, possibly through transactions sold to ultra large cloud suppliers or large chip manufacturers. “The most likely exit route may be through trade sales to major suppliers,” Harrowell said. “Apple has $23 billion in cash on its balance sheet, Amazon has $35 billion, and between Intel, Nvidia, and AMD there is about $10 billion. Very large companies have been very enthusiastic about adopting customized AI chips, and they have the ability to maintain relevant skills.”
Omdia also found that during this period, half of the $6 billion venture capital investment was invested in only one technology – Coarse Grained Reconfigurable Arrays Accelerator, which is typically aimed at loading the entire artificial intelligence model onto the chip. However, considering the continuous development of artificial intelligence models, this approach has some problems.
“In 2018 and 2019, the idea of embedding the entire model into chip memory made sense because this approach provided extremely low latency and solved the input/output problems of large AI models. However, since then, models have continued to evolve significantly, making scalability a key issue. More structured and internally more complex models mean that AI processors must provide more general-purpose capabilities Programmability. “Therefore, the future of artificial intelligence processors may be in another direction.” Harrowell concluded.
