Changes to the global funding environment are putting pressure on artificial intelligence (AI) chip startups, according to Omdia, a London-based tech business consultancy group.
Omdia’s assessment of the AI chip production landscape comes from its February 2023 report, Top AI Hardware Startups Market Radar. The report indicated that over 100 distinct venture capitalists have invested over $6 billion since 2018 into the top 25 AI chip startups.
Half of the money has been directed into only one technology: large-die, coarse-grained reconfigurable array (CGRA) accelerators, which are often designed to load entire AI models on-chip. Omdia’s report calls this approach into question.
“In 2018 and 2019, the idea of bringing the entire model into on-chip memory made sense, as this approach offers extremely low latency and answers the input/output problems of large AI models,” said Alexander Harrowell, principal analyst for advanced computing at Omdia, in a press release. “However, the models have continued to grow dramatically ever since, making scalability a critical issue. More structured and internally complex models mean AI processors must offer more general-purpose programmability. As such, the future of AI processors may lie in a different direction.”
The report adds the global chip shortage and rising interest rates were concerns in 2021 and 2022 for AI chip startups. Other issues could add to the increasingly unfriendly environment for startups, including inflation, supply chain disruptions, the scarcity of certain materials, geopolitical tension, and the persistence of the chip shortage into 2024.
Omdia predicts that in 2023, at least one major AI chip startup will exit the market, likely through a sale to a hyperscale cloud provider or a major chip maker.
“Apple has $23 billion in cash on its balance sheet and Amazon $35 billion, while Intel, NVIDIA, and AMD have some $10 billion between them. The hyperscalers have been very keen to adopt custom AI silicon, and they can afford to maintain the skills involved,” said Harrowell.
The change in the production sphere matters to engineers because it may result in layoffs at one or more startups. Startup sales could also lead to less diversity of thought and further domination of AI chip production by large companies.
Startups tend to be in a weaker position than large companies because they consume a great deal of capital. They can easily run out of funding if they do not promptly share a valuable product that promises a quick ROI. In contrast, industry giants are presently flush with money. They can outspend smaller firms on developing code and hiring new employees. They also do not have to showcase successful products to keep their doors open. Late-stage startups face the additional obstacles of lower valuations and fewer exit options because of a freeze in IPOs.
AI chip startups are fighting for recognition and promoting their products with a range of tactics. SambaNova Systems, a Palo Alto-based AI chip maker that is among the most-heavily funded firms, is driving users to its platform by offering up to a million dollars in free compute credits for generative AI applications like ChatGPT.
Axelera AI, a Netherlands-based chip maker that produces AI acceleration cards and systems, announced it will share a new platform to improve computer vision at the edge at the Embedded World exhibition in Nuremberg in mid-March.
Cerebras is a Sunnyvale-based AI chip maker known for working with pharmaceutical companies to shorten the time for drug discoveries. It has developed a way to connect hundreds of chips to reduce the amount of power needed to perform AI-related operations.
One promising note for smaller companies and startups is that market leader NVIDIA’s late February “raise-and-beat” report for its fourth quarter boosted the share prices of competitors in AI chip production. AMD saw its stock up three percent after NVIDIA posted its results. NVIDIA’s continued success could benefit AI chip startups that deliver a high ROI.