One of the ways in which 5G differs from its predecessors is that operators are, in general, taking a more staggered approach to capital investment. They are leveraging existing assets, such as 4G cores or dense fiber, to reduce the capex hike that was often associated with a network upgrade; they also plan to retain and enhance 4G RANs for a decade or more so that the addition of 5G build-outs can be targeted just where it is needed, and the cost spread over many years.
Such patterns are becoming visible in some operators’ financial results. For instance, while AT&T and Verizon have both increased their capex spending to support 5G, their bills have been lower than most analysts expected, through a mixture of cautious 5G investment, and reduced outlay in other areas.
One factor is likely to be the new approach to spectrum. AT&T and Verizon are making heavy use of their millimeter wave assets, which they gained through acquisition rather than auction, while some operators hope to reduce their costs by using CBRS spectrum, which even in its licensed form is unlikely to come with the high fees that accompanied some recent spectrum sales.
AT&T said at its most recently financial results briefing that it planned to spend about $20bn in capex in 2020, down sharply from this year’s $23bn and less than the $22bn Wall Street had expected in 2020.
Analysts at Wall Street research firm Cowen lowered their capex forecasts for AT&T to about $19bn in 2021 and 2022, a decrease from their previous estimate of around $21bn. Reduced 2020 capex was expected, because AT&T has virtually finished its fiber build-out, but not at the level the telco indicated.
Verizon said it spent $4.4bn in capex in the third quarter, about $100m below most Wall Street estimates. But it still expects to spend between $17bn and $18bn in capex for full year 2019, and analysts at Nomura expect that figure to remain constant in 2020 and 2021.
Both US operators have said they will reduce capex in the coming years, despite 5G and some fiber expansion, by increasing their deployment of software-driven networks, virtualized and cloud-based platforms, and through rising automation.
Cable operators are also looking thrifty on the capex side. Nomura wrote in a recent research note: “Comcast and Charter missed 3Q expectations for capex and guided 2019 lower than previously planned. We have lowered our combined 2019 capex forecast for Comcast and Charter from $14.6bn to $14.2bn.”
In 5G, opex will always be a greater challenge to control than capex, especially as upfront costs fall with the advent of networks run in software on commodity hardware. Conversely, a growing reliance on ‘as-a-service’ systems and partners’ fiber or sites will increase the opex bill.
One of the major aspects of 5G opex which MNOs are keen to bring under control is energy. Although 5G used to be billed as very energy-efficient, its greater density means that its total power consumption and cost will be higher than 4G, even if the radio itself is more efficient.
According to a study by ABI Research, a 5G network will consume, on average, 3.5 times more power than 4G because of densification, Massive MIMO antenna arrays and the need to support more spectrum bands as well as legacy networks. ABI says a typical LTE cell site draws about 6kW in power, rising as high as 9kW in peak usage times.
But a 3.5 GHz site with 4×4 MIMO could draw 14kW on average and up to 19kW under peak load, with the kind of traffic levels expected within 4-5 years. “In that scenario, the base station will be also supporting 2G, 3G and 4G as well, in as many as seven different bands from 700 MHz up to 3.5 GHz. That’s where you’re seeing the multiplication take place.” And there could be 2-3 times more cell sites. Once 64×64 Massive MIMO is introduced, that 3.5 GHz site could be consuming up to 350% more power than a 4G base station.
Another analysis, by Huawei, says that a 2G/3G/4G cell sites supporting five spectrum bands today would consume 5.9kW of power on average, with a peak of 7.3kW. In three years’ time, with the addition of 5G in 2.6GHz and 3.5GHz bands, this would almost double to an average of 10.4kW, with 13.7kW at peak; and by adding mmWave, the levels would rise to 13.4kW and 18.9kW.
Given the huge numbers of base stations which Chinese operators are rolling out, the energy issue is an important one. China Tower forecasts that the electricity cost of 5G networks will be about 10 times the site rentals per year, though some regional governments are offering electricity subsidies rather than risk a slowdown in 5G deployment. For instance, authorities in the Jiangxi province are offering a power subsidy of CNY8,000 ($1,134) this year, for every additional base station built beyond a baseline target, with further subsidies to come in 2020 and 2021.
Over time, improvements will come as operators turn off 2G and/or 3G, and as new chipsets are developed to make Massive MIMO, the main 5G power culprit, more efficient. But there will be new demands on the electricity supply, such as edge computing nodes deployed at cell sites. All this means intelligent power management will become an increasingly important element of operators’ strategies to keep 5G total cost of ownership down.