Brainchip Holdings (ASX: BRN) seems to be doing most things right. It’s right in the middle of the truly sexy business of our time, artificial intelligence. It’s doing what strikes near everyone in the business as the right thing, coding AI into silicon for use across networks. So, BRN shares should be motoring, right? Except they’re not - they’re down 77% year to date. True, they’re up 16% today on decent stock turnover but that’s small comfort for long term holders. What is it that’s going wrong here? The answer is simply that chip development is an incredibly risky business. Get it right, become the new standard, and it’s a veritable gold mine. But there are always a lot of also rans in any such race.
The specific move is that at the end of last week the new generation of chips was announced for availability: “In an era marked by an insatiable appetite for artificial intelligence (AI) capabilities, BrainChip, a pioneer in neural network processors, has taken a significant stride towards empowering edge devices with unprecedented processing power. The company’s latest unveiling, the second-generation Akida platform, represents a groundbreaking leap in the realm of Edge AI, delivering the potential to liberate devices from cloud dependency.” OK, that’s obviously a bit over-reliant on the press release but still informative.
Everyone knows that AI will end up being implanted into a chip. That’s just the way this all works. But whose chip, whose design and which AI? That’s the battle that Brainchip is fighting.
Brainchip Holdings share price from Google Finance
We’ve talked before about Brainchip Holdings: “As it happens the Brainchip Holdings implementation is interesting. By putting much of the AI onto a specific chip - doing things in hardware is efficient after all - it should be possible to gain a large reduction both in computing costs and also response times and latency. However, it's also true that inserting operations into hardware risks being left behind in any fast moving field. Hardware does take time to develop and running new code in the cloud is always going to be faster in the development cycle. Hardware implementations require, almost as a definition, a certain stability of what is to be encoded into the hardware.”
And something we said about Brainchip: “One answer - a possible answer - is that perhaps Brainchip just isn't very good at AI. That's a possibility, obviously. We could also muse on the idea that as ChatGPT moves to version 4,0, Google launches Bard and so on then there's less room in the marketplace for Btainchip's approach. Far from the current popularity validating their ideas and sector, the technological path taken seems a little different (ie, large models in the cloud, not small ones on chips) and thus not quite what Brainchip is offering to customers.”
It’s simply going to be true that at some point AI is embedded in a chip. But when, and will it be Braincloud’s when it is?