Google (NASDAQ:GOOGL) (NASDAQ:GOOG) is making big splashes in the cloud computing community this week. Under the leadership of Diane Greene, former VMWare co-founder, the company is striking deals left and right for its long-lagging cloud hosting business, thus attacking Amazon (NASDAQ:AMZN), IBM (NYSE:IBM) and Microsoft (NASDAQ:MSFT). If you follow these companies, you already know that. However, there is something else going on which is an absolutely fundamental shift in Google's strategy and it has to do with today's announced additions to cloud platform.
If you are a regular reader of my articles, you might be aware of my under-the-hood series for the artificial intelligence market. In this article, I explain what just happened and how it shifts the tectonics of the analytics market.
Today, Google announced a new machine learning suite for its cloud platform. It contains functionality for image analysis, speech recognition, statistical machine translation for thousands of language pairs and custom deep learning models (e.g. neural networks). Now, this might not sound like a big deal to you, especially if you are not in the technology industry, but it is, for a number of reasons.
When I began writing my 'under-the-hood' series, I started with Google's advances in beating human Go players (article). In that article, I explained what was going on in that work and how it played into Google's long-term vision of creating a general artificial intelligence and eventually utilizing that for its products. In a later article, I went on to contrast that with IBM's Watson platform. The quintessence of the nascent machine-learning-as-a-service market was that Google had a leading edge in general AI research, but had shown little interest in building tangible products. Instead, management seemed to be content with letting its subsidiary DeepMind doing ground-breaking research that might now make it into a product for another decade.
In contrast, other players like IBM with its flagship Watson Cloud were getting busy creating software services based on their research and experience from building Watson to beat a human player at Jeopardy.
For the past decade, if you attend any top-tier computer science conference in the fields of systems, databases, machine learning or cloud computing, Google researchers and engineers will be presenting papers. Google is also probably a major sponsor of conferences. Its papers have been hugely influential in modern web application infrastructure. What usually happens is that Google writes a paper about some internal tool of its search/analytics infrastructure (e.g. MapReduce, Bigtable, Spanner, Megastore, Percolator, Borg...) and people then go on to build open-source software based on the papers and soon everyone is using them. For instance, Google has recently released its software deep learning execution platform TensorFlow so anyone could use deep neural networks without having to worry too much about the operational side of training them on different types of devices e.g., you can write the same code for deploying a machine learning model on a server cluster or a smartphone.
What Google did not do was to commercialize any of these projects. Of course, most of these tools were just a byproduct of the company's hugely profitable search business and management could afford not to care too much about making a profit from them. This had the weird effect that a number of software companies were built around open-source software that was essentially the brainchild of Google's infrastructure and search teams.
For a long time, this decision was somewhat understandable. Prior to roughly 2010-2012, cloud computing was not as omnipresent as it is today. The tools Google built were deeply entangled with its internal infrastructure and it would have essentially required whole new divisions to spin them off into separate own product lines with support and sales teams. This would have taken Google away from its core business and might have proved a distraction.
What is more, there was not exactly a huge market for customers with data on the same scale as Google for a long time. When cloud computing came along, the search giant began tentatively offering hosted database and app solutions without really pushing the platform in the way Amazon did with Amazon Web Services i.e., through offering a surrounding suite of convenience features.
The big shift
In the past few years, Google has been very active in exploring deep learning in its products and research teams e.g., with its large scale image recognition project on YouTube. Note that these projects were entirely separate from DeepMind, which was only acquired in 2014 and has a focus on learning tasks from sensory inputs. So what was essentially going on is that Google has always had the technologically leading infrastructure and database tools as well as sophisticated internal machine learning projects for search, spam and fraud detection, image recognition and, of course, its self-driving car. For instance, one of the most influential deep learning papers in 2015 (link for anyone interested) was written by Google computer vision engineers.
Now, Google has decided to commercialize pretty much all of this AND also to become a much bigger player in the cloud hosting business and software-as-a-service business. This is a truly massive shift. See, Amazon has the most mature virtual machine hosting platform with tons of services around it e.g., virtual private clouds, caches, proxies, DNS services, databases and so forth, but it does not have the machine learning know-how and services Google has. IBM has lots of natural language processing and computer vision services in its Watson Cloud product as well as hosting in its SoftLayer product, but it does not integrate them into one smooth platform like Google. This is because IBM has obtained much of its technology in that space from acquisitions.
Microsoft is making progress with its Azure Cloud, which features a machine learning suite for training and learning traditional classification and prediction models. It also provides third-party APIs for speech and image recognition on Azure. Microsoft definitely has significant expertise in natural language processing and image recognition but has not really turned them into mature software products yet. Long-term, they might be Google's biggest competition in this area.
Can you hear that? It's the 500-pound gorilla pounding his its on the table. Google has decided it is no longer letting other companies eat its lunch in the cloud. It has the superior expertise, the superior tools, and apparently the will to take on everyone else with them. This is a strong bullish signal for the long-term because it will help Google alleviate stagnating growth from search.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.