Posts

AI chips: Explosive growth of deep learning is leading to rapid evolution of diverse, dedicated processors

Artificial intelligence (AI) utilization has been accelerating rapidly for more than 10 years, as decreases in memory, storage and computation cost have made an increasing number of applications cost-effective. The technique of deep learning has emerged as the most useful. Large public websites such as Facebook (Nasdaq: FB) and Amazon (Nasdaq: AMZN), with enormous stores of data on user behavior and a clear benefit from influencing user behavior, were among the earliest adopters and continue to expand such techniques. Publicly visible applications include speech recognition, natural language processing and image recognition. Other high-value applications include network threat detection, credit fraud detection and pharmaceutical research.

Deep learning techniques are based on neural networks, inspired by animal brain structure. Neural networks perform successive computations on large amounts of data. Each iteration operates on the results of the prior computation, which is why the process is called “deep.” Deep learning relies on large amounts computation. In fact, deep learning techniques are well known; the recent growth is driven by decreasing costs of data acquisition, data transmission, data storage and computation. The new processors all aim to lower the cost of computation.

The new chips are less costly than CPUs for running deep learning workloads

Each computation is limited and tends to require relatively low precision, necessitating fewer bits than found in typical CPU operations. Deep learning computations are mostly tensor operations — predominantly matrix multiplication — and parallel tensor processing is the heart of many specialized AI chips. Traditional CPUs are relatively inefficient in carrying out this kind of processing. They cannot process many operations at the same time, and they deliver precision and capacity for complex computations that are not needed.

Nvidia (Nasdaq: NVDA) GPUs led the wave of new processors. In 2012, Google announced that its Google Brain deep learning project to recognize images of cats was powered by Nvidia GPUs, resulting in a hundredfold improvement in performance over conventional CPUs. With this kind of endorsement and with the widespread acceptance of the importance of deep learning, many companies, large and small, are following the money and investing in new types of processors. It is not certain that the GPU will be a long-term winner; successful applications of FPGAs and TPUs are plentiful.

The Big Six, the 150, and the future of Accenture’s alliances

Like nearly every IT services vendor TBR covers, Accenture professes to follow a technology-vendor-agnostic approach to making client recommendations, but we have noticed how the company — in addition to managing over 150 tech vendor partnerships — forms strategic business groups with core partners, such as SAP, Oracle, Amazon Web Services (AWS), Cisco, Pivotal and Google, not to mention the joint venture with Microsoft (Avanade). Given the initial structure of the Accenture-Apple partnership, we expect this may formalize as a Business Solutions Group as well. Augmenting its large partner ecosystem capabilities often requires Accenture to team up with regional firms (e.g., Sompo Japan Nipponkoa Insurance and Daiichi Kotsu Sangyo) to demonstrate how Accenture partners at the local level to gather insights and test technologies in specific industries. The alliance strategy makes sense for such a large firm with a diverse set of offerings and capabilities, but in today’s market, which is flooded with emerging technologies, we have to ask what market and competitor changes will alter Accenture’s strategy.

What’s changing?

As the market evolves, Accenture’s do digital, be digital approach has impacted the way the company forges and manages its alliance relationships. The influx of startups presents a great opportunity for Accenture to expand its reach into new areas such as artificial intelligence (AI), blockchain and the rest of the alphabet of new technologies. Part of the Accenture Innovation Architecture, Accenture Ventures has played a critical role in this evolution as it has taken Accenture’s alliance strategy to the next level, creating a bridge between acquisitions and strategic partnerships mainly through minority investments in vendors such as Ripjar. The additional commitment and risk sharing demonstrated through minority investments is a step in the right direction as new buyers can be skeptical when nontraditional vendors pitch new capabilities. But Accenture is an almost $40 billion organization with a predominant focus on Global 2000 clients, leading the company to heavily rely on its Big Six partners — SAP, Oracle, Microsoft, AWS, Salesforce, Google, plus the emergence of Workday on Accenture’s radar. TBR notes that recently Accenture’s leadership has recognized the importance of these partners, as platform-based services enabled by the soon-to-be Big Seven by these Big Seven are now generating over 25% of Accenture’s sales. So, is anything really changing? At the macro level, maybe not, but the “as a Service” economy has enabled Accenture to pursue opportunities within the midsegment market, which in our view is an even bigger strategic shift than bringing in Accenture Ventures. This move into the midsegment market creates opportunities for the small technology players to play in the same sandbox as the 800-pound gorilla.

With Accenture augmenting its strategy, how do the small guys get Accenture’s attention? TBR asks this $100 million question of the startups, but more importantly of Accenture too. Many rival IT service vendors have ramped up similar strategies and will now drop prices to meet clients’ demands for nimble, off-the-shelf solutions that may not require premium consulting expertise as much as strong back-end support (hello, India-centric vendors). So is Accenture ready for the next chapter of its alliance strategy? Possibly. Will data interoperability reduce the need for multiple large platforms and be the panacea for the small tech guys to fill in the blanks? Can Agile methodologies infect Accenture’s alliance strategy? Maybe. With Accenture, let’s recall that it’s all about process. To paraphrase the late Johnnie Cochran, if it doesn’t fit, you must quit. And if you’re one of 150, you’d better be better rather than good.

What Diane Greene’s Departure Means for Google Cloud

“While the AI-centric strategy played to Google’s strengths, it didn’t help much with more boring workloads such as storage and website hosting that drive Amazon’s dominance and are the bulk of the cloud market. ‘They’ve been making the right moves and saying the right things, but it just hasn’t shown through in performance financially,’ says Meaghan McGrath, who tracks Google and other cloud providers at Technology Business Research. She says Google is still hamstrung by a perception that it doesn’t really know how to work with corporate IT departments—an area where Microsoft has a long track record.”

Voice assistant volume is increasing

A survey conducted by Adobe Analytics found that 32% of U.S. consumers owned a smart speaker in September 2018, compared to 28% in December 2017. The report also projected that near half of the U.S. consumer base could own one by the end of December 2018, supported by Adobe Analytics’ finding that nearly 80% of smart speaker sales occur during the holiday season. It is just one study, and there are more conservative studies out there ― but even if the data isn’t completely on the mark, it does uncover the trend of voice-controlled devices gaining ground inside consumers’ households despite use cases and monetization still being blurry.

I own four Amazon Alexa-enabled devices myself: two Echo Dot smart speakers and two Fire TVs. Of the Echo Dots, one was given to me by a colleague to play around with, and another I bought for about one-third the list price from acquaintances who had received it as a gift from their extended family and left it unopened because they felt it was “too creepy.” In our household, the Echo Dots have been used as glorified hands-free music players in our kitchen and one of our bathrooms. The Fire TVs are used as media players first and foremost. Sometimes, we try some of the new skills Amazon sends along in update emails as a fun diversion, but usually that is a one-off activity. I am deeply invested in the Amazon ecosystem, having been a Prime member since its debut and a fan of Prime Video, but it is still challenging to find ways to use Alexa smart-home devices to enhance my other Prime benefits or drive me to Amazon’s e-commerce business.

Adobe’s research seems to align with my anecdotal experience, noting that among the most common voice activities* are asking for music (70% of respondents) and asking fun questions (53% of respondents). The only other activity above 50% is asking about the weather (64% of respondents). So yes, people are using them, but these are not skills that require much depth or complexity or that drive additional revenue for Amazon.

Therein lies the problem for voice-platform providers such as Amazon and Google (Microsoft and Apple are also players, but I don’t believe they are as developed as Amazon and Google are in the smart speaker and voice assistant space). In an ideal world, voice assistants would provide platform companies with a wealth of consumer data as users query the devices about their everyday needs. Also, voice assistants can be a new conduit to monetization through new applications or — especially in Amazon’s case — to lowering barriers to the purchase of goods. However, most complex tasks, such as ordering a ticket for the movie you’d like to see tonight, finding out when the beach is open, or buying an outfit for an upcoming wedding, are still much easier via a smartphone or laptop interface. The Adobe study found that of the 32% of respondents with a smart speaker, only 35% and 30% used voice interfaces for basic research or shopping, respectively.

Improving the use cases, or “skills,” of voice assistants will be critical for platform vendors to increase the use of these devices for complex tasks and to elevate smart speakers from smart radios and novelties to gleaming data gems. TBR expects this to be the major battleground between voice assistant and smart speaker providers moving forward as the form factor has been relatively proved. TBR believes Google has a slight advantage due to its heritage in data mining behind the façade of services as well as its Android and Chrome cross-platform tie-ins (a lot of relevant user data is already in Google, such as contacts, schedules, and often email). Amazon is no slouch either due to its investment spend, growing media empire and robust e-commerce platform, which Google lacks. Apple could be a dark horse; however, its Siri is still weaker on an artificial intelligence (AI) basis and the HomePod’s pricing makes it an unlikely easy gift.

The next frontier for all of these platform providers is in the commercial space, an area we may see Microsoft put much of its effort into while leaving the consumer space for better-suited peers. In fact, collaboration between Microsoft and Amazon on voice and smart speakers may confirm this. Using voice assistants and smart speakers to query analytics or gain business insights or employing them as a “smart secretary” in conference rooms are areas TBR sees as avenues for commercial expansion. TBR has seen slightly different approaches from Amazon and Google in the commercial space. Amazon, likely with Microsoft support, focuses on the office with Alexa for Business, while Google seems to be positioning its voice AI and smart speaker technology to serve as an interface for a business’s customers.

However, as with the consumer space, the use case must be proved, the skills must be ironed out, and existing commercial infrastructure must be modified to support voice assistants and smart speakers. And despite furious investment in these possibilities by the major platform players, TBR doesn’t expect to see Alexa widely adopted in the boardroom for at least another two to three years. For now, I believe smart speakers will continue to find their way into homes as a novelty or curiosity for tech-excited people and early adopters, contributing to slow but steady growth, or as an easy, cost-effective tech-based gift, driving additional bursts of increased unit sales during the holidays.

*Voice activity data includes devices that are not smart speakers, such as smartphones.

Customer preferences are forming around hybrid and shifting around open source as vendors focus on acquisitions

Prebuilt devices are a ray of clarity amid the fogginess of hybrid

Hybrid can be a difficult thing to define in cloud computing. The term “hybrid” is overused by vendors but underused by customers, causing general confusion over its definition as well as solid examples of hybrid solutions. An area of the market that cuts through those areas of confusion is hybrid cloud integrated systems. These are physical devices (appliances) that are designed to integrate with public cloud services and can be used in customers’ own data centers. The idea that customers can physically touch the box and also integrate with external cloud services makes integrated systems one of the easiest and most obvious hybrid scenarios.

Examples of integrated systems solutions include Azure Stack from Microsoft and its hardware partners and Cloud at Customer from Oracle. While adoption and usage of these hybrid cloud solutions remain limited, the trend is picking up momentum and is prompting vendors such as Amazon and Google to move closer to competing in the space, particularly as customer demand from heavily regulated industries favors local versions of vendor-hosted cloud infrastructure. For example, Amazon Web Services (AWS) and Microsoft are the two front-runners in the race to win the U.S. Department of Defense’s Joint Enterprise Defense Infrastructure (JEDI) contract. While AWS has largely been seen as the overall favorite, its Snowball Edge offering does not meet the same bidirectional synchronization requirement of the tactical edge device that Azure Stack does.

Kubernetes season is in full swing as OpenStack falters

For large enterprise customers, open-source technologies have garnered much interest as part of their cloud strategies. The ability to utilize solutions that provide the same backbone as large cloud providers while maintaining the control associated with open source has been an attractive value proposition for those with the resources to implement and manage them. However, predicting which technologies will be the most commonly adopted has been more challenging, creating uncertainty around frameworks such as OpenStack, which has yet to garner significant momentum in the market.

Compounding the hurdles for OpenStack to overcome continues to be the ongoing explosion in growth among public cloud IaaS front-runners AWS, Google, Microsoft and Alibaba. OpenStack founders and former OpenStack pure plays are making notable shifts toward Kubernetes. The difference, though, is that Canonical and Red Hat are still holding onto OpenStack, while others, such as Rackspace, Hewlett Packard Enterprise, IBM and Mirantis, de-emphasize it.

Customers increasingly understand the benefits of containers and container orchestration platforms and embrace the portability and interoperability they provide. According to a recent interview done as part of TBR’s Cloud Customer Research Program, a retail SVP, CIO and CTO said, “You need to make sure there are escape clauses in your contracts in case you need to get out. Once you’re in it, you’re pretty much married, and that divorce is really bad. That’s the reason we have a container. … Because if it starts to get too expensive, we want to pull it off quickly.”

This is just one example of the immediate enterprise benefits of container and container orchestration platforms, which can change the game for enterprises in terms of their cloud adoption road maps and long-term cloud plans.

Hybridization is becoming even more widespread than customers realize

While pre-integrated devices are the most obvious examples of hybrid usage, the vast majority of activity is occurring in more subtle situations. This activity is driven by the desire among vendors to sell broader solutions and the desire among customers to implement services that integrate with existing and other new technologies. The good news for both sides of the market is that there are more capabilities than ever to put those more cohesive, integrated solutions in place.

Salesforce, whose solutions are commonly integrated into hybrid environments, has taken a notable step into the hybrid enablement space by acquiring MuleSoft. The acquisition, which closed on May 1 at the start of Salesforce’s FY2Q19, brings MuleSoft’s well-known integration Platform as a Service (iPaaS) solution and services into Salesforce’s arsenal. The implications for Salesforce, its customers and the market are vast, as the company can create connections between its applications and the variety of other cloud and legacy systems residing in customers’ environments. Salesforce quickly leveraged the iPaaS technology, bringing Salesforce Integration Cloud to market within the first few months of having MuleSoft on board, enabling customers to augment their Salesforce applications and derive greater insights from their non-Salesforce data.

Maturing offerings, vendors and customers prompt long-term IoT vendor growth

The continued interweaving of the technology component market with Internet of Things (IoT) techniques delivers a well-defined path to long-term sustained growth for many IT and operational technology (OT) vendors, especially those vendors that are best able to differentiate their portfolio and position themselves as critical partners for a wide set of IoT solutions.

The hype surrounding IoT has only served to confuse and overwhelm customers and vendors, but efforts by both parties to cut through the hype is driving the growth of installed IoT solutions. As the hype fades, vendors are better able to rationalize their go-to-market strategies and messaging, particularly around how to assemble IoT solutions, leading customers to better understand how to apply IoT.

However, while it is becoming easier to assemble an IoT solution, it is still challenging to design and implement the IoT technique. We don’t expect a huge explosion of revenue; IoT itself isn’t a “killer app,” but it will enable moderate and slowly accelerating revenue growth for the various components involved in an IoT solution.

In our 3Q18 reports and thought leadership, TBR will focus on three topics that we believe are currently the most impactful on the wider IoT ecosystem: the increasing maturity of the IoT technique, the growing consolidation of generic platforms, and how increasing commoditization around IoT is working in favor of economies of scale and enabling the growth of installed solutions.

IoT is growing up: Increased ecosystem maturity will lead to increased customer adoption

TBR, through discussions with vendors and customers as well as our use case databasing, is noticing growth in installed IoT solutions, whether from net-new deployments or expansions of existing IoT deployments, signaling improved maturity. IoT maturation is not so much about the components of IoT as it is about businesses developing their ability to leverage technologies and techniques that are increasingly applicable to a growing number of business problems.

A major driver of this maturity is greater clarity around IoT techniques, led largely by go-to-market realignment and improved messaging by vendors, organization around IoT by customers, shifts from competition to coopetition by vendors, and general improvements in the construction of the technology that facilitate advanced usage of the IoT technique.

Signals of consolidation appear in the cloud IoT platform space

Infographic discussing signals of consolidation appearing in the IoT cloud platform space

The cloud IoT platform landscape consolidates around largest vendors as customers seek continuity, consistency and the best tools

Cloud services revenue grew 48.2% year-to-year and increased as a percentage of total benchmarked Internet of Things (IoT) revenue from 12.4% to 15.8% year-to-year in 2Q18. Growth is driven by customers, especially those without deep legacy ties, moving their workloads to the cloud. The public cloud ecosystem is beginning to consolidate, with the top vendors competing on best-in-class tools, partnerships and business-problem-solving messaging.

Software, while still a sizable portion of benchmarked revenue, is experiencing slowing revenue growth, from 19% year-to-year in 2Q17 to 4.2% year-to-year in 2Q18. Software, along with ICT infrastructure, will continue to play a role in IoT solutions with the advent of edge computing, but as providers’ cloud platforms mature and tie-in deals with application partners are cemented, demand increases.

ICT infrastructure revenue grew 14.1% year-to-year in 2Q18 due to increased IoT deployments as well as hybrid IoT becoming an increasingly common IoT framework. ICT infrastructure gross margin rose 80 basis points year-to-year. TBR believes the increase stems from the need for more specialized or powerful hardware to handle the more advanced needs of IoT and its components, such as artificial intelligence (AI) and machine vision. Despite the increased utilization of ICT hardware due to hybrid IoT and the need for specialization, the long view for ICT infrastructure will be complicated by commoditization. TBR expects most ICT infrastructure companies to deeply invest in software and service components to buttress the profitability of customer engagements as the threat of commoditization looms.

Vendors across the technology spectrum are all fervently trying to crack the code for the “killer app” within specific verticals that can solve common business problems and be widely adopted by customers. The vendors that win with building the first widely accepted solutions will be set up for success, while others in the oversaturated market will at best become acquisition targets and at worst become history.

For more information, contact Analyst Daniel Callahan at [email protected].

Canonical’s growth play: Make customers’ and partners’ lives easier (and more economical)

TBR perspective

At Canonical’s 2018 Analyst Day, CEO Mark Shuttleworth laid out a very compelling construct for Canonical’s vision of being the link between the operating system (OS) layer and the cloud control planes. Canonical has Ubuntu OS versions to run from the largest high-performance computers with NVIDIA graphics processing units to the smallest device OSes at the heart of offers from niche vendors such as Rigado. Throughout the event, Canonical stressed multicloud interoperability through Kubernetes. The big unknown on the horizon is how to provision infrastructure for edge analytics, which sits at the heart of the strategic relationship Canonical has with Google Cloud as Google donates Borg to ensure Kubernetes does not challenge Borg the way Hadoop forked from MapReduce.

Existing virtualization economics has stalled, with premium pricing models emerging from the major and better-established competitors Red Hat (NYSE: RHT) and VMware (NYSE: VMW). The Canonical play further compresses the economics of the infrastructure abstraction and OS components, where parts will be provided for free and the services and update provisions will become the basis for the monetization model. Akin to how free Android disaggregated the device OS space and gained share against Microsoft, Canonical bets on market projections showing devices used/owned per person growing from two to three devices today to as many as 20 devices within the next five years.

It is from this vantage point that one open-source Linux distro, Canonical’s Ubuntu, was taking direct competitive aim at another (Red Hat), while likewise suggesting VMware’s time as the market maker would quickly start to fade as more and more app modernization efforts move code from virtual machines (VMs) into lightweight Kubernetes containers (clusters).

 

Canonical hosted its 2018 Analyst Day in New York City on Sept. 20, 2018. The event featured presentations from the top leadership at Canonical, including Shuttleworth, Finance Director Seb Butter, SVP of Global Data Centre Sales Jeff Lattomus, and VP of Global Sales, IoT & Devices Tom Canning. Canonical focused on business and go-to-market updates as well as key presentations by partners, such as Paul Nash from Google Cloud, outlining how Canonical has accelerated or added value to their businesses. At this year’s event, there was a noticeable blurring of the lines between cloud and IoT discussions in comparison to years past where there were more definitive tracks. Regarding both Canonical’s own strategy and its conversations with customers, it is exceedingly difficult to have a discussion about one and not the other, which is reflected in the broader IT landscape as of late.

Increased market clarity drives 16.1% year-to-year growth in commercial IoT revenue

Technology Business Research, Inc.’s (TBR) 2Q18 Commercial IoT Benchmark recorded revenue growth of 16.1% year-to-year, to $10.3 billion, in 2Q18, among the 28 IT and operational technology (OT) vendors we benchmark. The revenue growth is largely a result of continued implementation of Internet of Thing (IoT) and growth of installed IoT solutions.

The dousing of rampant IoT hype, which only served to confuse and overwhelm customers and vendors, is helping drive the growth of installed IoT solutions. As the hype dies out, a wave of increased clarity and maturation is forming with vendors rationalizing their go-to-market strategies and messaging, leading to customers better understanding how to apply IoT and vendors learning how to assemble solutions. Packaged solutions are emerging as vendors cooperate, focusing on their strengths, and assemble components sets that solve verticalwide challenges. TBR believes these factors are driving tactical business-focused IoT projects to supersede overambitious projects stuck in proof-of-concept limbo.

However, while easier than in the past, IoT design and implementation are still a challenge. TBR does not expect a huge explosion of revenue beyond midteen growth going forward.

Total 2Q18 commercial IoT benchmarked gross profit increased 16.6% year-to-year to $5.1 billion. Reduced complexity in IoT due to increased knowledge around building and applying IoT as well as the streamlining of portfolios as a result of increased partnering is improving vendor profitability. Also, vendors are leveraging specialized tools, such as artificial intelligence (AI), to justify higher pricing.

 

TBR’s Commercial IoT Benchmark highlights current commercial IoT revenue and gross profit for vendors. TBR leverages financial models and projections across a diverse set of IT and OT components. Additionally, the benchmark outlines the major vendor drivers and trends shaping the market.

Postcards from the edge: Complexity is here — wish it were not

“Analog dollars to digital pennies” is a phrase used to discuss the continued compression on technology price points as Moore’s Law economics, coupled with continued IP abstraction, creates economic trigger events aimed squarely at legacy business model best practices. Recently, I attended analyst events in New York City — one sponsored by Lenovo and one by Canonical — that outline these economic trigger events, albeit from different sides of the same coin.

Canonical CEO Mark Shuttleworth used the term “economic trigger events” often in his opening remarks. The idea is that technologies and new price points create trigger events that result in new economic fundamentals where some participants will be disruptors and some will be disrupted.

The rise in the hype cycle around edge computing as it joins forces with cloud, artificial intelligence (AI)/machine learning, and Internet of Things (IoT) creates a veritable Gatling gun of economic trigger events. These events accelerate business model disruption as we pivot to the Business of One era.

The disruption sits atop the continued economic pressure from commoditized hardware. What’s behind it all is that while infrastructure is valuable, it is not valued. In short, the margin moves out of infrastructure and into the business outcome. Technology enablement is less constrained by affordability and more by determining what business value can be derived from the application or use case.

Rather than being the lead decision in business investment decision making, infrastructure acquisition becomes the derived decision. The service attach rate or services drag becomes the fuzzy guide point for new inventions and new business models. Broad, ubiquitous ecosystems become imperative to generating sufficient margin in the digital penny world to justify the ongoing development, monitoring, and maintenance of secure flexible infrastructures that won’t break and will keep data secure and private.

For Canonical, this means focusing on the connection between operating system and cloud control planes to ensure a single code set operates silicon as large as high-performance computers (HPCs) and as small as single-purpose IoT devices. Compute infrastructure is assumed to work, until it breaks, and then users realize just how valuable that hidden infrastructure provisioning is.

As Canonical is hardening the abstraction layer to ensure seamless interoperability, Lenovo (and many other hardware manufacturers) create purpose-built appliances optimized for edge workloads. In some instances, these will be small appliances simply capturing data and routing it back to clouds for ingestion into massive analytics engines. In other situations, it will be very-high-performance compute engines with GPU accelerators in simple, easy-to-operate form factors where AI inference in real time has to be performed at the edge. Here again, the assumption will be that the edge appliance can operate (in a retail convenience store, for example) without the need for any technically savvy personnel to monitor, manage or provision the device on-site.

Look for more detailed special reports from TBR on the Lenovo and Canonical industry events in the next few weeks.