Machine Learning as a Service — The Top Cloud Platform and AI Vendors
A Race for AI Leadership
Which platform is the best for starting off with implementing your AI solution?
Well, there is of course not one universal answer to that question. But in the long run it makes a great difference which vendor you choose as your platform provider — and you shouldn’t blindly go for the one that already is in your portfolio.
There are hundreds of AI solutions out there. And for most companies accessing them through cloud services is the easiest way of starting to get a feeling for AI projects.
And there are more benefits of going with a large vendor first: There is for example some stability regarding up-times, API accessibility, customer support and most importantly their unfair advantage in the forms of the vast amounts of data to train algorithms with.
Therefore, we narrowed down the solution providers to 6 major cloud platform providers:
Please note that this is by all means not a comprehensive listing!
But to make the field a bit more overseeable, we chose the top 4 providers of cloud services measured by revenue (Google, Microsoft, Amazon, IBM) and added Salesforce and SAP as a growing percentage of their revenue comes from cloud services and they both have a unique starting point as leading CRM (Salesforce) or ERP (SAP) solution providers.
But how to choose the right provider?
This is where a combination of the requirements of your use case, your budget allocation, your long-term strategy and the existing know-how of your company comes into play.
To give you some indications, we evaluated those 6 cloud providers on their solution maturity and strategic commitment in the fields of AI:
Google offers a very broad toolset regarding artificial intelligence and machine learning. Googles AI and machine learning products for example offer full machine learning automation with (hyper-) parameter tuning, container management and a dedicated API management.
Google Cloud fits you perfectly if you need flexibility in early trials but at the same time extreme scalability in the long run. If your developers already know Google Cloud Services, they will find their way around quite easily when starting off with machine learning. Downsides of using the Google platform are its business and real-time analytics tools that to this date are still immature in comparison to the other service providers.
Google’s AI strategy is to create a strong position in core data science with corresponding patents and related computer technology areas as its business relies heavily on machine learning. With TensorFlow, Google developers already delivered the leading open-source software library in the ML space. This, alongside with their staggering investments when acquiring AI start-ups, shows Googles dedication to be and stay the leader in AI development. It recently underlined its status as an ‘AI-first’ company by rebranding the research division to become ‘Google AI’.
Microsoft seemed to have missed the AI development for a while — just as it seemed to be the case for cloud services. But for both cases, Microsoft Azure has gained traction at an incredible speed.
Microsoft is adding to its Azure platform more and more services like Azure Data Lake, Azure Data Catalog and Azure Cloud Functions. It is close to Google and IBM in the race for AI dominance through Azure Machine Learning and manages to integrate other services like business intelligence solutions quite well. Therefore, it is no surprise that Azure gets closer to AWS regarding enterprise adoption every year.
Microsoft kept quiet concerning the acquisition of AI start-ups for a long time. But since 2016 it keeps the pace of Google with some very interesting moves especially in 2018. And still, the company’s entire AI efforts are well linked with the Azure cloud. It will be very interesting to see how this plays out in the future as more technology from start-ups is integrated.
Amazon leads the pack in regards of cloud services with AWS. At the same time, its voice assistant Alexa developed into the household name when talking about practically implemented AI from a consumer’s perspective.
As seen in the chart below, AWS currently has by far the highest adoption rate of all cloud incumbents. It therefore makes sense for Amazon to upsell AI and machine learning services to its existing cloud customers. Advantages of AWS can especially be seen in the performance when dealing with large data volumes, in the completeness and maturity of its services and in its openness regarding APIs to other services — even to competitors like Microsoft.
In 2017, Jeff Bezos wrote a letter to Amazon’s shareholders and focused noticeably on AI and machine learning when speaking about getting an edge over the competition. Following the success of Alexa, Amazon will continue to build around voice, virtual assistants and natural language processing. Also, AI-as-a-service is pushed more and more into the AWS environment to make Amazon a leader in the AI and machine learning fields as well.
If you’re interested in how Amazon executes its innovation strategy on an operational level, have a look at flywheel. This is how they managed to get a recommendation into the core of their business model and to ‘unsilo’ all further AI efforts.
Enterprise cloud adoption for those 4 main incumbents is displayed in the following:
IBM focuses heavily on Watson as it is trying to make it into a cloud-based data operating system and at the same time keeping its business clients happy in satisfying their need in the analytics field.
IBM supports AI-assisted experiences and combines cognitive services with data management, analytics, and the entire suite of Bluemix developer tools. What currently makes it unique are AI-guided ‘experiences’ for business analytics via a common application framework. On the other hand, Watson Analytics still requires manual data loading and is not yet a replacement for general-purpose business analytics.
IBM’s AI strategy is lazor-focused on its enterprise customers. In giving them the control of their data and insights, IBM tries to assist them increase efficiency, lower costs or augment human intelligence. Simultaneously, open source projects in the fields of AI are supported and APIs to solutions built by other vendors like Google’s TensorFlow are endorsed.
Salesforce is one of the leaders in cloud computing, offering applications around CRM, sales, ERP, customer service, marketing automation, business analytics, and so on. Having its CRM customer base as a backbone, Salesforce collaborates closely with Amazon with regards to its cloud services.
It offers a platform-as-a-service model that is mostly based on AWS. Its own solution portfolio is currently being expanded aggressively by some major acquisitions (e.g. CloudCraze and MuleSoft). One aspect that is missing — at least in visibility compared to the other vendors — is community and engagement around the platform. In the long run, this will most likely change as other big players have shown the vast potential that can be tapped in this area.
A major initiative of Salesforce is its Einstein artificial intelligence platform — since its launch in 2016 it has been a central focus for the development and marketing strategy. Salesforces CEO Benioff then described their AI strategy quite straight forward: ‘AI is the next platform — all future applications, all future capabilities for all companies will be built on AI.’
SAP is still a niche player in the fields of cloud computing in comparison to the big boys like AWS and Azure. We included it into our analysis though, because of its unique position as a leading ERP provider. SAP is already in touch with most operational company data, thereby at least potentially representing a trustworthy technology partner and being known to all the buying centers.
The SAP cloud platform also contains its analytics cloud, which is its biggest strength as it integrates BI, predictive analytics and planning in major enterprise applications. On the other hand, the cloud platform is still a bit immature as a fragmented toolchain and some gaps in technology like in deep learning and Python show.
Cloud revenue surpassed licensing revenue for the first time in 2018. And SAP will continue to follow that path in terms of earnings — but as the examples of acquiring Recast.AI and creating Europe’s first corporate AI ethics advisory panel, the strategy clearly includes artificial intelligence as a central building block.
M&A Activities in AI from 2014 to 2018
M&A cases in this area really took off in 2014, when Google invested heavily in AI start-ups. Since then, more than 30 AI companies were acquired by those 6 vendors alone:
As some rules of thumb, we suggest to follow these basic indicators:
- Do you need a flexible solution to start off with, have the opportunity to scale and use a vast number of highly mature machine learning tools? Try Google Cloud Services.
- Do you already use a Microsoft technology stack, want to integrate analytics solutions but still have extensive AI and machine learning possibilities? Check out Azure.
- Is voice integration or natural language processing your focus or are performance and API connectivity central issues? Go for AWS.
- Do you need a user-friendly interface with full control over the data flows and still cutting-edge capabilities in the fields of artificial intelligence? Give Watson a go.
- Do you use Salesforce already for enterprise apps or CRM and want to build applications around customer data? You can’t go wrong with trying Einstein and Salesforces’ AWS integrations.
- Is SAP your ERP system, you use business objects and you may even want to choose between a cloud and an on-premise solution? Have a look at the SAP Cloud Platform
But to take a bit of weight off the decision: In a UK-based survey, 77% of companies stated that they already or intend to use multiple cloud solutions. The same will most likely apply for AI and machine learning tools as well as there are just too many amazing solutions out there to miss out on some providers just because currently another vendor is used for a totally different use case.
The best take is to start off with some well-designed prototypical use cases!
There is a positive feedback loop between planning and implementation. You’ll gain valuable practical experience from AI pilots that can then be used to revise your AI strategy and to set more realistic expectations and viable objectives.