How Much Opex, Capex Could Mobile Operators Cut?

Most connectivity service providers have programs in place to reduce capital expenditures and operating costs, where feasible. Mobile operators face continuing need to invest in spectrum resources as well as network platforms about every decade, so capex requirements will remain elevated.

Fixed network service providers will, in some cases, be able to better control network capex if they are far down the path of converting their access networks to optical fiber or other high-bandwidth access methods. 

Vodafone plans to cut €1 billion in costs by 2026 through various initiatives. For its part, América Móvil achieved $2.1 billion in cost savings in 2022.

The TM Forum predicts that telcos will achieve 20-percent to 30-percent cost savings by 2025 through intelligent automation and cloud adoption.

Arthur D. Little estimates that network sharing can generate cost reductions of up to 20 percent for the participating operators.

Company	Year	Opex/Rev (%)	Capex/Rev (%)
Verizon	2019	76.5	14.0
	2020	75.2	15.6
	2021	74.1	14.6
	2022	73.5	14.2
	2023 (est.)	72.8	13.8
AT&T	2019	74.8	15.6
	2020	78.0	14.9
	2021	75.4	15.2
	2022	74.0	14.4
	2023 (est.)	73.2	13.9
NTT	2019	78.2	10.3
	2020	77.8	10.8
	2021	77.0	11.2
	2022	76.5	10.9
	2023 (est.)	76.0	10.7
Vodafone	2019	72.3	14.7
	2020	73.2	13.8
	2021	72.8	14.2
	2022	72.0	13.5
	2023 (est.)	71.5	13.0

Many have significant hopes that artificial intelligence will help by supporting automated fault detection, self-healing networks, and dynamic power management, for example.

Migrating to cloud infrastructure and virtualization processes reduces some amount of physical infrastructure. Verizon, for instance, aims to have over 20,000 virtual RAN cell sites by 2025.

Infrastructure sharing likewise has been an effective tool for reducing radio access costs for mobile operators, in some markets. 

But opex remains the big potential cost target, as a percentage of revenue. 

Provider	2020 Opex/Rev	2021 Opex/Rev	2022 Opex/Rev	2023 YTD Opex/Rev
Verizon	71.8%	71.1%	70.0%	68.9%
AT&T	70.9%	70.3%	70.4%	69.0%
NTT	50.4%	49.6%	48.9%	48.6%
Vodafone	66.4%	65.2%	64.4%	63.5%

"Cloud Computing as a Service" Revenue Earned by Azure, Google Cloud, AWS s Hard to Compare

For many observers, it might not matter what components are counted as “cloud computing” revenues earned by Amazon Web Services and Azure Intelligent Cloud or Google Cloud. Revenue is revenue, some will maintain. 

For some, it might well matter what components are contained in each firm’s reported revenues, as Microsoft, for example, includes lots of end-user applications revenue within its reported Azure Cloud category. 

Keep in mind that reported Azure “cloud” revenue includes applications such as LinkedIn, Office, and other enterprise or consumer apps related to gaming, for example. The other competitors earn revenues mostly or exclusively from  "computing as a service" sales. 

Gartner, for example, estimates that in 2021, software-as-a-service (end user apps) represented 25-percent to 30-percent of global cloud infrastructure and platform services (IaaS and PaaS) spending. 

So some analysts speculate that the contribution of applications like Office and Dynamics 365 to Azure Intelligent Cloud revenue could be as high as 30-percent to 40-percent of total “cloud computing” revenues.

Provider	Market Share (%)	Cloud Computing Revenue $ Billion	Revenue Growth YoY (%)
Amazon Web Services (AWS)	33.1%	81.0	29%
Microsoft Azure	22.5%	50.1	35%
Google Cloud Platform (GCP)	10.1%	23.0	41%
Alibaba Cloud	7.5%	17.2	40%
Tencent Cloud	5.1%	11.6	48%
Other Providers	21.8%	49.9	38%

If so, then “apples to apples” comparisons of cloud computing revenues, excluding retail applications revenue, would have Azure remaining the second-biggest provider, but with a lead over Google that is not as large as generally believed. 

Azure core cloud computing (infrastructure and platform) might be in the range of $30 billion to $35 billion, for example.

OpenAI Sora as an "iPhone Moment"

Sora is OpenAI’s new cutting-edge and possibly disruptive AI model that can generate realistic videos based on textual descriptions. 

Perhaps it is not too soon to make an analogy to what we now call the “iPhone moment.” 

The phrase "iPhone moment" is used to describe a pivotal event, product launch, or technological advancement that significantly impacts a particular field or industry, as was the iPhone’s shift to touch screen interface in 2007, where prior smartphones had used keyboards. 

An "iPhone moment" might be characterized as one that:

• introduces a novel technology, feature, or design that significantly changes how things are done and therefore is disruptive

• has a profound impact on the way people use technology and becomes widely adopted and impactful

• creates a new paradigm or sets a new standard for the industry, influencing future developments.

Prior such moments might include:

• the invention of the internet

• the World Wide Web and multimodal media (video, graphics, image and sound plus text)

• social media platforms

• cloud computing

Sora and other text-to-image (TTI) and text-to-video (TTV) models are likely to be a significant turning point in how we interact and consume information, similar to the way the multimedia web expanded the internet beyond text-only interfaces. 

Some examples of other text-to-video platforms include:

• Runway Gen-2: This platform, available on web and mobile, allows users to create short videos from text descriptions. It offers various customization options and editing tools.

• Google AI's Lumiere: Released as an extension to the PyTorch framework, Lumiere focuses on generating high-quality 3D animations from textual prompts.

• Make-A-Scene: While not exclusively text-based, this AI tool allows users to create and manipulate scenes using natural language descriptions, offering a different approach to video generation.

• Imagen Video: This research project from Google AI demonstrates the ability to generate longer and more complex video sequences from text descriptions, showcasing potential future advancements.

Other examples of text-to-image platforms include:

• Midjourney: This platform offers stunningly realistic and detailed images generated from text prompts, with a strong focus on artistic expression.

• DALL-E 2: OpenAI's counterpart to Midjourney, DALL-E 2 is known for its creative and often surreal interpretations of text descriptions.

• Imagen: A Tencent project with the same name as Google’s Imagen

• VQGAN+CLIP: This open-source project allows users to experiment and create their own text-to-image models, fostering accessibility and exploration within the field.

As did prior “iPhone moments,” Sora and other text-to-video platforms will democratize content creation. As the multimedia web and broadband internet access enabled YouTube, realistic gaming, video streaming, video advertising; ridesharing; turn-by-turn navigation on a smartphone; video calling and all “rich media” including the future metaverse, so Sora and other text-to-video platforms could create new industries, firms and use cases. 

Sora is a major advance in user-generated and professionally-created content that might rival earlier changes wrought by the multimedia web; the internet; home broadband; smartphones and mobile broadband. 

As with those earlier changes, major changes could happen to legacy practices, industries and behaviors. As the smartphone replaced cameras, watches, GPS devices, video screens and home phones, so might TTI and TTV platforms reshape existing industries, firms, products and behaviors.