Wednesday, May 16, 2018

"Artificial intelligence chips are a hot market." (NVDA; GOOG)

In March 2017 our thinking on NVIDIA was:
"We focus on the stock, not the company. The company should be fine for at least the next couple years until the artificial intelligence biz catches up to NVIDIA and either takes a different approach or a really different approach and goes quantum computer."
Here's one of the competitors.
From Barron's Tech Trader Daily, May 14

Cerebras: The AI of Cheetahs and Hyenas
Artificial intelligence chips are a hot market.

There’s plenty of debate over who will make the dominant parts. Today, it’s definitelyNvidia (NVDA). Intel (INTC) believes it has enduring relevance, as explained last week by its AI chip architect, Gadi Singer. AMD (AMD) can also be a contender.

And then, there are several small, venture-backed startups. I met last week with a co-founder of one of them, Cerebras SystemsAndrew Feldman. Feldman’s previous startup, SeaMicro, was bought by Advanced Micro Devices (AMD) in 2012. Feldman left AMD in 2014 as CEO Lisa Su shook up the company’s lines of business. After some time on the beach, Feldman re-emerged at Cerebras.
I interviewed Feldman back in September about Cerebras. That was a conversation in very broad terms about what AI silicon has to achieve. Feldman is still not ready to disclose details yet of Cerebras's work.

However, during a lunch of tacos in Palo Alto, not far from Cerebras’s offices, Feldman offered some intriguing tidbits.

Asked if there’s enough AI work to keep a startup going amidst giants like Nvidia, Feldman launched into a most interesting treatise about cheetahs and hyenas as a way to think about tech companies:
My father is an evolutionary biologist, and I like to make analogies between evolution and market forces. You can think about it like this: there are two kinds of animals in technology. One is the cheetah. The cheetah’s only prey is the Thomson’s gazelle. Anywhere you find the Thomson’s gazelle in large enough numbers, the cheetah thrives. The cheetah is a gazelle-hunting specialist. It is perfectly biologically engineered to hunt the gazelle. It is the only one of the large cats whose claws do not retract. Its claws are always out, like cleats, so it’s ready at all times to accelerate. Its nose is smooshed up against its face to facilitate its breathing when running at high speed. Its tail is long and narrow to stretch out behind it and give it aerodynamic balance.
At the other end of the spectrum is the hyena. It is evolved in very different ways. It will eat all kinds of prey, not just one. It hunts in the day, it hunts at night. It hunts alone and it hunts in packs. It will eat things it kills, and it will eat things other animals kill. It will even chew on the bones of long dead animals.
The cheetah is a specialist. It will thrive as long as there are gazelle in abundance, as long as the one thing which it is engineered to pursue is plentiful. The hyena is a generalist. When resources are scarce, when there lots of different types of animals present, but none in large numbers, the hyena will thrive. The question, then, in technology, is whether you are a specialist or a generalist. And if you are a specialist, is the market large enough to sustain you? 
Cerebras, in other words, is a cheetah, a specialist in AI.

Nvidia and Intel are generalists, he asserts. In each case, whatever specialist ability they may once have had — graphics rendering in Nvidia’s case — it has been subsumed under a pile of generalist programming:...MORE
Chipmakers Battle To Power Artificial Intelligence In Cloud Data Centers" (AMD; NVDA; XLNX; INTC)
Ahead of Today's NVIDIA Earnings: A Look at One of the Competitors (GOOG; NVDA)