The sale of Habana Labs to Intel for $2 billion just three years after it was founded is an impressive feat. The deal represents yet another exit for serial entrepreneur Avigdor Willenz and yet another Israeli acquisition by Intel.
Habana's field of endeavor is the development of artificial intelligence (AI) chips. The ability to develop AI technology - to analyze data, to identify recurring patterns and to train algorithms - preoccupies just about every tech company today that wants to be innovative. Even those not already developing AI, are planning to develop it.
However, in order to undertake the very complex calculations that will let autonomous cars drive themselves or will identify diseases, it is not sufficient simply to have sophisticated algorithms. Also required is the hardware infrastructure that will enable these capabilities to become fully realized. Intel's decision to acquire Habana Labs for $2 billion is designed to provide such an infrastructure.
The most prominent personality behind Habana Labs is Willenz, a veteran entrepreneur. He was the founder and CEO of communications chip company Galileo Technology which was sold to Marvell for $2.7 billion in 2000. He also founded semiconductor platform Annapurna Labs, which was sold to Amazon for $370 million in 2015 and Leaba Semicnductor, which was sold to Cisco in 2016.
Habana Labs is not the only company developing AI chips and in particular the Israeli scene has in recent years been buzzing with such startups, so what about Habana justifies a $2 billion price tag?
Habana's chips kill two birds with one stone for Intel
Habana has developed chips for two main types of AI tasks - the training stage and the application stage. In other words, the stage in which the algorithm "learns" is a very demanding stage in terms of computer resources, and the application stage in which the system uses the insights learned in the training stage. This while some of the other companies developing AI chips do so only for one of the stages, or develop chips designed for end devices, and not for data centers.
This means that Habana's chips are not only precisely suited for Intel's needs but also kill two birds with one stone. The fact that the products are already commercially available, or in advanced stages with customers, is also a big help. Habana's chips are also ideally suited for the market that most interests Intel - the data centers market. Intel sees this as both an entrance to that market and a major future revenue source. Thus Habana's Gaudi and Goya chips are precisely what Intel needs.
While Intel is developing its own chips in this field, with the development also being carried out in Israel, Habana's chips are substantially different.
While Intel's AI chips are based on the architecture of regular CPU processors, Habana's chips are based on completely new architecture, especially designed for AI processor tasks. The fact that Intel will soon have AI chips will also give it a competitive advantage against Nvidia, which originally developed graphic processors and only then discovered that they were also suitable for the field of AI.
Intel does not plan missing the AI revolution
The acquisition is part of the global competition in the world of computer chips in which Intel's fiercest rival is Nvidia. The competition between the two companies could not be more open and they tend to snipe at each other in public forums. This will by no means be the last AI acquisition and a high stake race is likely to develop.
There is also competition from outside the field of traditional chips from Google and Amazon, which offer cloud services and are developing their own chips for data centers to try and gain a competitive advantage over each other. Intel, which in the past missed the mobile phone revolution, would be prepared to pay vast amounts of capital not to miss the AI revolution. Consequently, Intel paid a high premium for Israeli auto-tech company Mobileye, which it bought for $15.3 billion three years ago, and is now prepared to pay $2 billion for a three year old company, which has raised just $120 million.
In contrast to the acquisition of Mobileye, the current acquisition is part of Intel's core activities. Mobileye has taken Intel into a new area of endeavor but if it misses out on AI, then the next decade could be wretched for Intel. Moreover, Intel does not have a good track record with acquisitions and its performance over the years has been marred by the failure to integrate some of the companies it has bought. With Mobileye, Intel seems to be taking a different approach and letting it operate independently.
Will Intel repeat this model with Habana? It is by no means certain. Last year Intel appointed Bob Swan as CEO - he was previously CFO and he has tried to open up the company to external innovative processes and change its archaic DNA. The Habana deal is an opportunity to prove that he has successfully implemented change and prove that he can integrate acquired companies. Two of Willenz's previous exits - Annapurna to Amazon and Leaba to Cisco - were successfully integrated into those companies. Intel would be happy for that to happen with Habana.
Published by Globes, Israel business news - en.globes.co.il - on December 16, 2019
© Copyright of Globes Publisher Itonut (1983) Ltd. 2019