Internet Giants and the Need for Faster AI: Part 1
The saving grace of the benighted US effort to isolate China from the global technology community and competition is its futility. We have just seen a great example of that, and by great, I mean good for some crucial American companies.
It’s a bit of a long story, so let me explain.
The Chinese reaction to the US restrictions on selling Chinese firms state-of-the-art components for AI, 5G, and other leading-edge technologies has been predictable. China’s tech sector — long a laggard in chip design and production — has decided it needs to start designing and manufacturing its chips.
For instance, Huawei has long been a multi-billion-dollar consumer of US chips from Qualcomm and others and is making chips for its communication networks. Thank you, President Trump, former speaker Gingrich, FBI Director Wray, and all you other China hawks.
But it’s not just Huawei — an obvious candidate to get into the business — that’s doing this and US interference is certainly not the only motive.
Another Powerhouse Enters the Ring
Alibaba (China’s version of Amazon) — that nation’s leading internet retailer and the world’s second-largest — is getting into the chip business too with a dramatic announcement last September about a new AI chip.
That’s a little more surprising. Why would an internet retailer and consumer services company make its own chips?
And Alibaba was not even the first Chinese consumer giant out of the silicon gate. Baidu and Tencent are also making custom chips for their own use. So are Google and Facebook.
There must be something else going on here in addition to trade friction.
My friend Clay Christensen, famous for his “disruptive innovation” paradigm explains it well via his even more powerful paradigm of “integration vs modularity.” All these companies, in their own competition to provide consumers even better service, are pushing the limits of technological performance, and wringing everything they can out of available technology.
And they all, in different ways, bump up against the same problem: off the shelf, or as Clay would call them “modular” solutions may not work when you are pushing the limits. General purpose technology, including in this case, generally available microprocessors cannot always handle the job.
The choice, even for a consumer product company like Alibaba, is either to build just the right technology or slip behind in a furious competition. And when you’re up against Amazon, or Tencent, or Google, there are no slips allowed.
Alibaba has a huge need for artificial intelligence (AI) solutions that are customized to its operations and faster, better, and cheaper than its rivals.
AI: The Core of the Competition
AI is not a generic technology. It is not plug-and-play and certainly not at the leading-edge of performance. Nor is it even remotely a “completed” technology, with its forward path foreordained.
Artificial intelligence is not something that these great companies — all to some degree competitors with each other — can just go to the store and pick up to support their competitive strategies.
You see, AI doesn’t just serve the competition. To a considerable extent today, and a greater extent tomorrow, it is the competition.
Who has the fastest, the most powerful, the most energy-efficient (a huge consideration), and the fastest to “learn” the most efficient pathway to a solution? Any of these factors can drive victory or defeat in the internet economy that is being born.
And it is not “all in the software.” If it were, then you could just go buy a box at the store and let the programmers — and Moore’s Law — handle the rest. But even Moore’s Law multiplication of circuit densities enabling ever more fantastic feats of programming won’t do it.
AI needs smarter chips. Artificial intelligence is a learning machine and it needs chips that learn fast.
That, just as much as the trade war was why Alibaba introduced its first AI inference chip, a neural processing unit called Hanguang 800. The new chip — far from generic — makes performing machine-learning tasks dramatically faster and more energy-efficient, vital to Alibaba’s operations.
Yes, the trade war has been a spur to all this. Huawei’s move into the chip business is certainly in part defensive, an attempt to ensure technology independence much as American politicians in the 1980s kept crowing (foolishly in my view) about the need for US “energy independence.”
However, for Alibaba and the other internet giants, it’s likely the primary motive is to reach performance levels they can’t get to with Harry’s Hardware.
Tomorrow, I’ll explain how all of this is good for certain American companies.
Editor, Gilder’s Daily Prophecy