JP Data LLC
101 Jefferson Dr, 1st Floor
Menlo Park CA 94025
Phone
(408) 623 9165
Email
info at jpdata dot co
sales at jpdata dot co
JP Data LLC
101 Jefferson Dr, 1st Floor
Menlo Park CA 94025
Phone
(408) 623 9165
Email
info at jpdata dot co
sales at jpdata dot co
I was recently at the AI Hardware Summit in Santa Clara. The event made a come back after two year hiatus and was packed with 700+ attendees. The conference was co-located with the AI Edge Summit and there was one overlapping day. Like any other conference, there were tracks, keynotes and of course exhibits.
To my surprise, the event was packed – all exhibit booths were busy, people were shaking hands and eating lunch together on a table. Companies were handing out goodies as they would in the past without having to worry about ‘spreading germs.’ There were many familiar faces (and yes, I mean unmasked faces) and it was just good to see overall enthusiasm coming back after two long Covid years. This was third in person event attended by me in last year or so in Silicon Valley and the attendance by and far beat all others. A lot that can be written about the content, keynotes, speakers but overall below are some of my key takeaways from the conference.
New AI chip companies are still coming out of stealth
I have been following the AI chip market since 2016, and have a list of 150+ companies that I track. Despite this, I found few new companies that I had not heard of before at the conference. This just goes on to show that there are still undiscovered companies in the area and start-ups continue to up. Some of the companies had more than powerpoint and brochures to show and that is a right strategy at this point for a start-up to come out of stealth mode.
Lot more companies are demonstrating their first version of their chips but very few are on second
The AI chip industry did not take shape until 2017. This is when the number of companies I was tracking jumped suddenly from single digits to high double digits. Given that its been five year period since, you would assume that there would be lot more second generation products out by now. But majority of the chip demonstrations focused on first generation products at the show. This is just another indication of the fact that AI chip remains a difficult business to get to market and patience is indeed a virtue here! Of course, there were exceptions – some companies are onto their second or even third generation, but majority of companies are still on first. In general, no AI chip company has successfully monetized their first generation product and so these companies will have to get fast to their second generation products.
There was no major product announcement from a chip company.
There were a few announcements about different products but nothing that I would call a big step forward. With the exception of some demoes, majority of focused on same old facial and people recognition. These demonstrations nevertheless showed increase in accuracy, at low power consumption and better optimization. Unlike previous conferences, where the focus was on ‘whats new’ this time around the focus seemed to be on ‘how good’.
Market remains as fragmented (and confused to some extent)
As a part of my job, I am always bombarded with questions about my opinions about the market, companies and future. This time though, almost everyone asked me this question: ‘so what do you think is the future of all these AI chip companies’. The AI chip market has become overcrowded in the past few years and it seemed that everyone was wondering if there’s space for such a large number of vendors. During the conference, I had pleasure of talking to several chip vendors, and almost everyone said that they have secured one or more design win and are optimistic about the future. Of course, some of this can be attributed to what I call ‘entrepreneur’s optimism’ but companies seemed confidant about their market traction.
To get to a neutral answer to this question, we should be looking at the buyer side, rather than seller. During my briefings, several enterprise AI users have complained high cost per inference is the single most important reason for not taking their POCs (Proof Of Concepts) into production and they are eagerly waiting for new chip solutions to emerge. Edge device companies have opinionated that they are always interested in adding intelligent features but must get a chip solution that satisfies all their constraints (power, cost, performance, form factor….). So given that there’s still tremendous interest from buyers, it is now up to vendors to figure out how best to monetize. What amazed me about this question is that the sheer number of people who were doubting the future of AI chips. We are heading into uncertain economic times and some bumps and bruises are expected but we have seen very innovative technology come into existence driven by these chip companies. I am sure that in the long run , we’ll see intelligence everywhere driven by AI chips. I have some hypothetical scenarios about how this can evolve in terms of numbers and I’ll write about that some other time.
Until then, here’s to AI Hardware Summit 2023!