Artificial Intelligence in Telecom: How Big, How Soon?

Artificial intelligence now is a “thing” like “computing” was once a thing. But it might be quite reasonable to assume that, in the not distant future, AI will be embedded into products, processes and venue as routinely as computing now is part of the background fabric.

In the communications business, that should range from customer-facing service to the operation of core networks, enhancing pricing and packaging decisions, service configuration, resource allocation, network management, energy efficiency, use of assets to simple routing decisions.

Commercial AI revenues are expected, by some, to climb from less than $1 billion to perhaps $5 billion by 2020. Revenues might reach $31 billion to $38 billion, up to perhaps $41 billion, globally, different forecasts suggest, by 2024 to 2025.

Some forecasts are more aggressive than that, calling for a U.S. market of perhaps $67 billion for robotics applications by 2025.

Humana, for example, using AI to assist agents in its call centers, to help agents predict customer behavior while on calls. Right now, AI sometimes provides a digital coach. In other cases, AI might handle an inbound call until escalation to a human agent is required.  

In the future, AI, with big data sets to work with, might go even further. Some are working on ways of using machine learning to directly handle inbound calls without agent intervention. Indeed, one premise of AI is that it can be used to automate any process.

“Amelia can, after two months of learning from her human colleagues, handle over 60 percent of support tickets on her own,” argue researchers at  Kairos Future.

It would be fair to note that researchers have been working on AI processes for half a century, including research into deep learning, machine learning, natural language processing (NLP), and computer vision, machine reasoning as well as core computing.

source: Frost and Sullivan

Bank of America Merrill Lynch

It is anecdotally worth noting that among the firms pushing to create commercial AI platforms are firms with huge exposure to consumer apps and business models, including Amazon, Microsoft, Google, IBM, Salesforce and Baidu.

source: oreilly.com

Private Networks Deliver 39% of Global Traffic

The Internet has caused the biggest change in global telecom since the advent of mobility. To wit, former telcos now are participants in the broader Internet ecosystem, not the leaders of the “telecom” ecosystem.

One example of the trend: application providers already operate private wide area networks that

account for 39 percent of global traffic. In a few years, it is likely OTT app providers will carry a majority of global traffic, largely on their own networks.

In other words, the app providers no longer rely on “telcos” to carry their traffic, and increasingly operate their own networks, as functional substitutes for telco services.

Likewise, OTT providers now offer services and apps that effectively replace telco voice and messaging services. That is highly significant as voice and messaging have historically represented the vast bulk of telco revenues.

Facebook and Google also have been active reshaping access networks as well. Access historically has been the province of “telcos.” But Google is testing fleets of balloons, fixed wireless, unmanned aerial vehicles, Wi-Fi and fiber to the home networks.

Facebook is developing unmanned aerial vehicles, multiple-antenna array fixed wireless radios, and launched the Telecom Infra Project to create open-source telecom platforms spanning access, backhaul, core networks and network management. The Open Cellular effort likewise concerns ways villages and other entities can create and operate their own cellular networks.

Google has launched its own tablets and smartphones, increasingly, with its new Pixel devices, set to challenge the likes of Apple and Samsung at the high end of the market.

Facebook, meanwhile, has launched the Free Basics program to allow users to sample key Internet applications without having to buy mobile data plans.

One might well argue that innovation, to a large extent, has passed from the tier-one telcos to the few big app providers able to operate at “web scale.”

Few Actually Need a Gigabit: RCN

As it launches gigabit service in New York, RCN has provided one of the easiest, consumer-friendly ways an ISP can easily explain which speed tier should be purchased.

“Gigabit” mostly is a marketing platform, not a reflection of actual end user demand, in virtually all cases. Indeed, how much bandwidth any account requires is a function primarily of the number of users sharing an account, and secondarily of the types of applications used (especially streaming video).

Though RCN suggests accounts using streaming video choose a 50-Mbps plan, that assumption is based on four devices being used. In fact, for single-user accounts, even 10 Mbps will be enough to stream Netflix and other services.

The gigabit plan is said to be best for accounts supporting 10 devices or more. The 330 Mbps plan is best for eight devices; the 155 Mbps plan best for six devices. The 10-Mbps plan is said to be sufficient for one to two devices, perhaps typically a smartphone or two, perhaps sometimes a PC and a smartphone.

Faster speeds are a good thing. But few accounts actually “need” a gigabit service. The value of higher-speed services is mostly a function of the amount of bandwidth sharing on the account.

source: RCN

30 Billion to 50 Billion "Things" to be Connected by 2020?

There might be 30 billion to 50 billion Internet of Things (IoT) end-points by 2020, driving a total IoT market of up to $8.9 trillion, according to the GSA.

That represents huge numbers of new communications links, some of which might happen over non-paid connections such as Wi-Fi, but many of which will require mobile or other paid wireless connections. which explains the huge interest in IoT on the part of the mobile industry.

New developing markets, driving new communications standards and formats, typically begin with many different protocols contending, before markets pick commercial standards. That was true for videocassette recording formats, PC and smartphone operating systems, WiMAX and LTE mobile standards, for example.

GSA believes the same sort of winnowing process will happen for wireless IoT connections. Though it is conceivable that some new platforms will survive as niches, GSA believes mobile-based platforms ultimately will emerge as the mass market standards, for most applications.

With the caveat that one would expect GSA to say such things, mobile operators globally should be able to leverage their scale, over time. That is not to say mobile narrowband IoT standards will have the majority of sales in the early going. It is quite possible they will not.

Still, over time, scale should matter, as it typically does in the communications business.

The largest segments for IoT are consumer electronics, automotive and healthcare, GSA argues.

CenturyLink Moves Further in Direction of Business Services

A proposed CenturyLink acquisition of Level 3 Communications might come as a bit of a surprise. Many of us thought Comcast would be the more-likely buyer. But the proposed deal, if it goes through, would provide ample illustration of changing business strategies by former rural telcos.

Simply put: a few of the larger rural telcos have raced to recreate themselves as provider of services to business customers. That is true for Windstream, Frontier Communications and CenturyLink. In fact, CenturyLink, with a market value of about $15.2 billion, would be acquiring Level 3 at a market value of about $16.8 billion.

In its second quarter of 2016, CenturyLink earned about 34 percent of total revenue from consumers. If one assumes the transaction is consummated, then business revenues might reach as much as 88 percent of total revenues.

That is the sort of deal that might be called a "transforming" event. In other words, CenturyLink would almost immediately become a company that earns substantially much more revenue from business customers than consumers.

In fact, CenturyLink might then become a business services oriented company with some legacy rural telco operations, as well as some metro consumer operations.

CenturyLink would be a firm that "used to be" a rural telco, but has become a business-focused entity.

That further illustrates one enduring principle of the telecommunications business, in the United States and elsewhere. That principle is that tier-one service providers earn most of their profits from business customers, and use those profits to subsidize service to rural consumers.

The general principle is basically that a tier-one fixed network makes high profits in urban areas, is profitable, but less so in suburban areas and loses money in rural areas.

In the same way, tier-one service providers make money from enterprise customers, significant money from mid-market businesses and then make slimmer profits from small business.

Will Mobile Ultimately Become the Commercial Standard for IoT Connections?

There might be 30 billion to 50 billion Internet of Things (IoT) end-points by 2020, driving a total IoT market of up to $8.9 trillion, according to the GSA.

That represents huge numbers of new communications links, some of which might happen over non-paid connections such as Wi-Fi, but many of which will require mobile or other paid wireless connections. which explains the huge interest in IoT on the part of the mobile industry.

New developing markets, driving new communications standards and formats, typically begin with many different protocols contending, before markets pick commercial standards. That was true for videocassette recording formats, PC and smartphone operating systems, WiMAX and LTE mobile standards, for example.

GSA believes the same sort of winnowing process will happen for wireless IoT connections. Though it is conceivable that some new platforms will survive as niches, GSA believes mobile-based platforms ultimately will emerge as the mass market standards, for most applications.

With the caveat that one would expect GSA to say such things, mobile operators globally should be able to leverage their scale, over time. That is not to say mobile narrowband IoT standards will have the majority of sales in the early going. It is quite possible they will not.

Still, over time, scale should matter, as it typically does in the communications business.

The largest segments for IoT are consumer electronics, automotive and healthcare, GSA argues.

Telecom Infra Project Reaches 300 Members

The Telecom Infra Project, a wide-ranging effort to create open source telecom infrastructure, now has 300 members, according to Lance Condray, Facebook infrastructure strategist.

Those members include Axiata, EE, Deutsche Telekom, Globe Telecom, Indosat Ooredo, MTN, MyRepublic, SK Telecom, Tata Communications, Telefonica and Vodafone.

Acadia Networks, Accenture, Adva Optical Networking, Amdocs, Broadcom, Ciena, Equinix, Facebook, Gilat, Infinera, Intel, Juniper Networks and Nokia are some of the suppliers also working with TIP.

Facebook, Intel, and Nokia have pledged to contribute an initial suite of reference designs, while other members such as operators Deutsche Telekom and SK Telecom will help define and deploy the technology as it fits their needs

“A few years ago, Facebook was faced with a data center problem familiar to many scale companies: We depended on proprietary systems and hardware that were inflexible and expensive,” said Jay Parikh, Facebook Global Head of Engineering and Infrastructure. “We realized quickly that this approach would not be sustainable; we needed to find a new way.”

Note the language: traditional rack and stack approaches were “unsustainable.”

“We recognized that telecom infrastructure could benefit from the same innovations taking place in the data center,” Parikh said.

“It was clear that the raw building blocks of what we were developing for our own infrastructure could be applied to telecom networks with great benefit,” he said.

At first, “TIP will focus on disaggregating the components of network infrastructure that are traditionally bundled together and vendor-specific,” said Parikh.

As one early example, Facebook has been working in partnership with Globe, deploying a low-cost, solar-powered network-in-a-box solution, bringing mobile coverage to a village. “In the first week alone, we connected more than 60 percent of the community,” said Parikh.

Project groups also have been created to address “the most pressing industry needs including connecting the unconnected or underserved populations, and augmenting the development of powerful new technologies like 5G.”

The access system integration and site optimization group is chaired by SK Telecom.

The unbundled solutions group is co-chaired by SK Telecom and Nokia, and will seek cost-effective, low-power and low-maintenance solutions.

Media-friendly solutions, chaired by Intel, will focus on mobile experience, especially for close-to-edge solutions.

In the backhaul area, Facebook heads the effort to develop “thin and extensible software stack to autonomously coordinate routing, addressing and security related functions in packet-switched IPv6 networks.”

The open optical packet transport project is co-chaired by Facebook and Equinix, and is working on Dense Wavelength Division Multiplexing (DWDM) open packet transport architectures that avoid supplier lock-in.

The core network optimization project is chaired by Intel, and seeks to disaggregate

core network components.

The greenfield telecom networks group is co-chaired by Nokia, Facebook and Deutsche Telekom, and will work on IT-based network architecture.

Public Internet WANs for Branch Office Networking?

Can the public Internet support reliable enterprise wide area communications. Yes, it now seems. And data center networking seems to be a key driver.

International Data Corporation estimates global SD-WAN revenues will exceed $6 billion in 2020 with a compound annual growth rate (CAGR) of more than 90 percent between 2015 and 2020.

Gartner predicts 30 percent of enterprises will actually deploy SD-WAN technology in their branch offices by the end of 2019.

SD-WAN is a networking technique that uses the public Internet to connect dispersed sites within an enterprise network, including branch offices.

Unlike earlier WAN technology that typically involves fixed circuits and proprietary hardware, SD-WAN is a cloud-based service.

source: CenturyLink

source: CenturyLink

LS Networks to Build Rural Gigabit Networks in Pacific Northwest

LS Networks will deploy a high-density fiber-optic broadband network in 25 rural communities in Oregon and Washington over the next two years.

The program will offer simple broadband plans at 100 Mbps ($40 per month) or 1 Gbps speed costing $70 a month.

The $1.2 million “Connected Communities” project launched in Maupin, Oregon, in July, and the first services will begin in January 2017.

LS Networks is a competitive local exchange carrier that will leverage the backbone and access facilities it has built for business customers to be used for consumer Internet access. LS Networks has built more than 7,500 route miles of high-capacity fiber broadband in Oregon.

That approach might illustrate a continuing reality of the rural Internet access business, namely that it is difficult to supply high-quality services in isolated areas without revenues generated some other way.

Historically, service providers have used profits from business customer segments to support networks serving consumers; and profits from urban customers to support services for rural customers. LS Networks will use profits generated by its CLEC operations to build the rural consumer Internet access networks.

Peak-Hour Data Consumption Dominated by Video Entertainment, Globally

Peak hour data consumption in the Asia-Pacific region is driven by the use of real-time entertainment, representing 49.6 percent of total downstream traffic during peak period, and up from 47 percent in 2015.

Asia- Pacific’s traffic composition is for the most part similar to that of leading networks in Europe and North America. After streaming audio and video, web browsing and social networking round out the top-three traffic categories in the region.

In most regions around the world, real-time entertainment is the most dominant traffic category. That observation tends to correlate with use of 4G networks: as 4G usage rises, so does consumption of entertainment video.

Real-time entertainment in Africa now accounts for 18.1 percent of peak downstream traffic, an increase from 8.6 percent in 2015, according to Sandvine.

That is a result of wider availability of Long Term Evolution 4G networks, and use of smartphones, Sandvine suggests.

Web browsing accounts for 31.4 percent of downstream traffic in Africa.

WhatsApp now generates over seven percent of network traffic, while Viber represents 3.2 percent of traffic. Within 18 months, YouTube could become the lead application across Africa.

Across the Middle East, real-time entertainment is the leading source of traffic, accounting for over 37 percent of peak downstream traffic. Social networking accounts almost 20 percent of traffic in the region.

During evening hours. Facebook and web browsing are among the top three applications. an identical order to that observed on North American networks.

Internet Access: "We've Gotten the Easy Ones"

Lower-cost new networks, better business models and creating more usable spectrum to “connect everyone across South Asia and Southeast Asia to the Internet” were key themes at the Spectrum Futures conference held in Singapore, Oct 20 and 21, 2016.

Creating sustainable business models--as always--was a key focus.

“We have gotten the easy ones,” said Chris Weasler, Facebook director of global connectivity. So connecting the next couple few billion users will be challenging.

Speakers included communications regulators from Egypt on the west to Indonesia on the east; Internet service providers; app providers and enablers; venture capitalists; wholesale capacity partners; policy advocates and platform developers.

New ways of supplying and using spectrum; new models for deploying infrastructure and new ways to reduce the cost of Internet access facilities were discussed. New business models also were highlighted. “5G is completely different,” said Bob Horton, consultant.

Will huge new allocations of millimeter wave frequencies for 5G work? Simple answer, “yes,” said Reza Arefi, Intel director of spectrum strategy. Are infrastructure costs coming down? “By an order of magnitude,” said community networking specialist Jonathan Brewer.

Are regulator and industry expectations out of alignment? “Yes,” said Mohammed Shafi, Multinet Pakistan CEO. Is much more spectrum required? Without question, said Rajan Mathews, Cellular Operators Association of India director general.

Are there still big cultural gaps between service providers and app providers who can help access providers move up the value chain? Yes, said VC Dennis Wong, with Golden Gate Capital.

Telcos and app developers “speak a different (business) language,” said consultant Srinath V. So telcos should invest in app providers “as limited partners,” said Kenrick Drijkoningen, Golden Gate Ventures expert in residence.

Major work on business models also is necessary, argued James Sullivan, J.P. Morgan equities analyst.

“Even after accounting for Wi-Fi and new technologies and alternate business models, there will be still significant global wireless data demand that is not economically possible to serve,” said Sullivan. Simply, emerging markets “ don’t have nearly enough revenue opportunity to bridge the gap.”

Attendees learned about a number of initiatives, large and small, to bring lower cost Internet access efforts to bear, ranging from the Telecom Infra Project (TIP) to community-owned cellular.

TIP in an open source telecom platform effort presently supported by 300 organizations, ranging from tier-one mobile operators and equipment suppliers to Facebook itself. Few recognize that TIP wants to create open source access, backhaul, core network and management platforms.

Google’s Project Loon, using fleets of balloons to provide Internet access across rural areas, is working with all four Indonesian mobile operators and is awaiting approval from the finance ministry to proceed, said Leo Sugandi, Ministry of ICT, Indonesia.

Steve Song, Network Startup Resource Center associate, does not necessarily believe large national service providers are “required.” Villages and communities can be encouraged to create their own access networks, either mobile or Wi-Fi.

We“Solutions that blur the lines between licensed and unlicensed are possible and represent lower risk for regulators and operators,” Song said. “We need new models for spectrum access.”

Telekom Malaysia Ordered to Double Access Speeds, Cut Prices by 1/2 Over 2 Years

Telekom Malaysia Berhad (TMB) fixed network Internet access services--by order of the Malaysian government--must deliver a higher speed for the same price, starting in January 2017.

The objective is to double speeds and reduce prices  by half over the next two years.

"Lower prices and the potential for an increase in capital expenditure to support this plan are credit negative for TMB as they may pressure cash flows or raise debt levels in 2017," said Annalisa DiChiara, Moody VP. "At the same time, the financial implications for TMB in 2017 and beyond are unclear at this stage.”

TMB's adjusted debt/EBITDA was two times at June 2016 and the company expects capex/revenue of around 30 percent to 35 percent in 2016.

AT&T Might Already be Poised to Disrupt Video Markets

AT&T argues its acquisition of Time Warner will help it disrupt video markets. Even in advance of the proposed acquisition, AT&T appears to be doing so. DirecTV Now, the new steaming service, will offer 100 channels for $35 a month, a price point that is disruptive for the streaming video market.

In significant part, the price point seems to suggest one key advantage for over-the-top services, compared to traditional linear services. DirecTV Now will not require a technician visit, a dish installation or use of in-home customer premises equipment.

The truck roll along represents costs upwards of $100. The traditional install also requires the installation labor, hardware and then in-home decoders (one to several). It might not be unreasonable to suggest such costs represent $600 or more, in operations and capital investment, for each new install.

DirecTV Now will have none of those costs.

Google Fiber Head Departs, What Will Google Fiber Do?

There are at least a couple conclusions one might draw from new developments at Google Fiber. Some suggest Alphabet has given up on Google Fiber. Others argue, based on Google Fiber’s own statements and actions, that it instead is looking at business models that include  fixed wireless.

As Google Fiber has been saying, as it “paused” the extension of Google Fiber to new cities, the organization is looking at “new technology and deployment methods.”

It seems clear enough that actual uptake of Google Fiber was less than Google Fiber, and many other outside observers, expected to see. So given the cost of fiber to the home, a look at alternate access platforms would make sense, at a time when a number other leading entities, including AT&T and Verizon, plus Facebook, have signaled they are actively researching the use of such technologies.

It remains possible that even the new fixed wireless business models and deployment methods will not work to Google Fiber's satisfaction.

But it might be fair to suggest Google Fiber has learned one important lesson. In competitive local access markets, when market share can top out below 20 percent to 40 percent, the business model is extremely difficult, and might not often support three sustainable, facilities-based providers.

In the Cloud Era, All Plumbing is Commoditized

The idea of what we now call cloud computing is not new, with the concepts dating back more than a half century. But it was largely an idea, back then. Over the past 20 years we have moved faster, though.

Remember “application service providers?” Those turn-of-the-century efforts to supply remote versions of enterprise shrink-wrapped software mostly failed. Salesforce was the successful model. Amazon Web Services was the modern precursor, launched in 2002. So the cloud era--as a commercial reality--is less than two decades old.

Most consumers use cloud computing without knowing it, every time they use an Internet-delivered app. That is a key point: use of software and “applications” now is something users invoke from remote servers, not a bit of software locally resident on their appliances.

In parallel fashion, the move of computing resources “into the cloud” has had key implications for use of computing infrastructure by enterprises, government, schools, hospitals, smaller business and and consumers.

Less hardware is needed “on the premises.” More money is spent on services and access. Local area networks no longer are primarily about structured cabling and local servers, but Wi-Fi, for example.

To a large extent, that has meant hardware is commoditized. Local area networks now simply involve ensuring that Wi-Fi is available. “Computing” or “application access” now increasingly means enough bandwidth to reach the cloud-based app sources. All that means less capital is spent on local hardware and software; more on cloud-based replacements purchased as services.

In the telecommunications business, the implications have been vast, as well. Applications once created and sold by service providers have become--in large part--applications consumers or businesses can consume as cloud apps (Skype displaces voice; OTT messaging displaces SMS; Netflix replaces HBO or increasingly, linear video).

Cloud computing means more demand for bigger pipes. But the retail cost declines, either on a cost-per-bit basis or in absolute terms.

In other words, most computing infrastructure and communications has been, and will keep being, commoditized.

The fortunes of firms and the fate of industries will be shaped by that fundamental reality.