In principle, unlimited data usage plans offered by mobile operators in many markets “should” entice consumers to buy them, which, in turn, should lead to higher average revenue per account.
And that indeed arguably is the case for 5G unlimited usage plans offered by some mobile operators in some markets.
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
“Killer apps” often are important in spurring adoption of new technologies. It was the spreadsheet which drove financial personnel to buy and use personal computers. It was the ability to watch over-the-air television which drove early cable TV adoption in rural areas, while it was commercial-free Home Box Office which arguably drove urban area customers to buy.
Mobile email access drove adoption of the BlackBerry smartphone. The touchscreen interface arguably drove adoption of the iPhone and all subsequent smartphones. Turn-by-turn directions supplied by navigation apps such as Google Maps drove some mobile network and device upgrades.
And that has been typical for general purpose technologies: there is typically a particular use case that drives adoption.
Lighting is generally considered the first important use case for electricity. Pumping water from coal mines was the important early use case for steam engines, followed by its use to power textile machinery. Autos were the killer app for internal combustion engines.
It is less clear which apps drove consumer interest in the internet, though music sites such as Napster played a role.
But there were also several other contenders that drove interest in using the internet, including email, the multimedia web and web browsers; search engines and instant messaging. Electronic file sharing and e-commerce also were lead use cases.
Social media platform adoption reached about 10 percent use by the internet-connected population in the United States between 2005 and 2015, for example.
We will have to wait a while to identify the AI “killer app,” either for large language models or broader AI adoption in other ways, though personal assistants, entertainment, health and wellness, education or creative tools have been suggested as promising early use cases.
And some might suggest large language models and generative AI could eventually be seen as the killer app that spurred broader consumer use of AI. But beyond that, the issue is which implementations and use cases for LLM will drive obvious and broad consumer or entity value?
And at what cost?
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
Generally speaking, edge computing facilities such as those envisioned by the multi-access edge computing model impose higher costs than do hyperscale data facilities, including capital investment efficiency and operating costs. That does not mean MEC is unfeasible, but that the business cases need to be worked out, as there are alternatives including on-board or on-device computing.
Assume 20 percent to 40 percent of edge computing requirements might be suited for on-device processing, especially for simple, real-time tasks and applications with tight latency constraints.
Assume the remaining 60 percent to 80 percent of processing tasks might use either remote edge computing or cloud processing for more complex analysis, data aggregation, or situations where device limitations are significant.
Even in those cases, it is presently unclear how much latency improvement might be needed, and therefore when edge facilities are required. The answer matters, since, generally speaking, MEC or other edge computing facilities will not be as capital-efficient as hyperscale data centers.
Add to that the operating cost profile, which likewise tends to be higher than for hyperscale sites.
All of that means that MEC and other edge computing facilities are likely to be relatively costly investments for a data center services provider, simply because of lower scale at each facility, as well as the need for many such distributed facilities.
That includes hardware costs; deployment costs; energy profiles; cooling requirements; monitoring and maintenance and well as security.
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
Generally speaking, edge computing facilities such as those envisioned by the multi-access edge computing model impose higher costs than do hyperscale data facilities, including capital investment efficiency and operating costs. That does not mean MEC is unfeasible, but that the business cases need to be worked out, as there are alternatives including on-board or on-device computing.
Assume 20 percent to 40 percent of edge computing requirements might be suited for on-device processing, especially for simple, real-time tasks and applications with tight latency constraints.
Assume the remaining 60 percent to 80 percent of processing tasks might use either remote edge computing or cloud processing for more complex analysis, data aggregation, or situations where device limitations are significant.
Even in those cases, it is presently unclear how much latency improvement might be needed, and therefore when edge facilities are required. The answer matters, since, generally speaking, MEC or other edge computing facilities will not be as capital-efficient as hyperscale data centers.
Add to that the operating cost profile, which likewise tends to be higher than for hyperscale sites.
All of that means that MEC and other edge computing facilities are likely to be relatively costly investments for a data center services provider, simply because of lower scale at each facility, as well as the need for many such distributed facilities.
That includes hardware costs; deployment costs; energy profiles; cooling requirements; monitoring and maintenance and well as security.
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
Gary Kim has been a digital infra analyst and journalist for more than 30 years, covering the business impact of technology, pre- and post-internet. He sees a similar evolution coming with AI. General-purpose technologies do not come along very often, but when they do, they change life, economies and industries.
