Huge opportunity for chips: Cloud spending to top $1 trillion in four years

Broadly speaking, there is some uncertainty about global IT spending in 2023 and beyond, but Synergy Research, which follows the cloud space closely, is very bullish on cloud spending in all its forms.

Across cloud services and infrastructure, the company sees total spending revenue from cloud operators and hardware, software and service providers (what it calls the public cloud ecosystem) growing 21% in 2022 to $544 billion, and It further stated in its forecast that sales of the entire cloud ecosystem will double or even exceed $1 trillion in the next four years.

Doubling would mean sales reaching $1.09 trillion, a four-year compound annual growth rate of 26%.

As we reported in October, Gartner forecasts sales of what it calls "data center systems" at $189.5 billion in 2021, growing 10.4% to $209.2 billion in 2022, and expected to grow 3.4% to $216.3 billion in 2023 dollar, and growth has slowed. If software and IT services are added to get a proxy for core IT spending, Gartner believes that this core IT spending will grow 12.8% to $2.13 trillion by 2021. Perhaps more importantly, Gartner expects this core IT spending to grow only 6% in 2022, with IT spending reaching $2.26 billion, and further states that it will grow 8.7% to $2.45 trillion in 2023 - an almost complete increase Driven by growth in enterprise software spending.

What's clear by comparing these two data sets is that the cloud ecosystem is growing faster than IT overall, and it looks likely to even continue that momentum in the coming year.

Interestingly, Synergy Research says the number of "hyperscale data centers in operation" will only increase by 50% during this period. John Dinsdale, principal analyst and research director at Synergy Research, added that data center network capacity will increase by more than 65% in the next four years.

Computing "hyperscale" data centers - not cloud zones, but independent data centers with a single network, connecting all their devices into a large virtual machine, which in turn constitutes a cloud zone - is interesting. But you must always remember that some of these data centers are for applications developed internally by us at The Next Platform, which we call "hyperscale applications" running on their infrastructure, but the cloud hardware that underpins it is a sales commodity. the cost of. Some hyperscalers are not in the cloud business at all, such as America's Facebook and China's ByteDance (which has dozens of apps, including TikTok). In the US, Google, Microsoft and even AWS have application and storage services that are really best classified as SaaS, as do Alibaba, Baidu and Tencent. We bring this up not to be picky, but to point out that when we say "hyperscale" we mean a very precise thing - it refers to those free or cheap applications, and the hardware that makes it up and software. We don't mean the cloud, but rather the capacity that is leased in some form.

“We track this very closely,” Dinsdale said of hyperscale data centers when speaking to The Next Platform. “In December, the number of operational hyperscale data centers passed the 850 mark. This number is growing by about 100 annually. These are large data centers and do not include CDN nodes, small on-premises POPs, and relatively small edge deployments. It does not include all data centers in the pipeline (planning, development or about to launch), which adds an additional 420 data centers.”

Here’s the 2022 public cloud ecosystem chart from Synergy Research:

We no longer use the word "public" to talk about cloud because it has nothing to do with public. Clouds are indeed utilities, but they are certainly not regulated like other utilities that manage the distribution of electricity, gas, or water.

The chart above shows growth, not revenue, so be careful when looking at it. We looked at a few numbers from Synergy Research's executive summary and other comments Dinsdale made to us and created this table to give you the revenue levels and growth rates for the data Synergy Research released:

To get the growth rates for the chart, we printed the chart and measured the bars, and once we had that, we could calculate the revenues for certain parts of the Synergy Research data in 2021. Dinsdale tells us that data center hardware and software sold for cloud and hyperscale data centers will be worth $120 billion in 2022, of which 81% will be spent on hardware and software acquisitions and 19% will be spent on data center leasing (including equipment and facilities) and construction of physical data centers.

We don't have enough data to calculate the data center rental and construction costs, but if you make an assumption, you can calculate all the missing data under the data center. What we do know is that the data center construction bar missing from Synergy Research's chart must have been either very modest growth or decline, with the category as a whole growing only 13%, while two of the sub-segments grew 14% and 20%, calculated from the bar chart . The numbers shown in bold red italics exceed estimates based on a 40-60 split between data center leasing and data center construction in 2022.

We don't have enough confidence in the managed private cloud, enterprise SaaS, and CDN segments to make estimates, but we have shown the growth rates from the bar charts.

As you can see, this "Public Cloud Ecosystem" data set mixes data center hardware and software spending for cloud computing with end-user spending on the cloud, and we can debate whether this is a wise approach. But, as Synergy Research shows, breaking the data into pieces means we can pull it apart. The way we think about it, data center capacity expenditures are the cost of sales of the actual cloud IaaS and PaaS services sold. Likewise, SaaS vendors that run their applications on one or more clouds include IaaS or PaaS services as a cost of sales for their SaaS products.

Synergy provides a breakdown of cloud revenue and cloud capacity in the U.S. and China, and as you can see from the table, the U.S. completely dwarfs China in terms of revenue. We don't know the capacity of these data centers - expressed in megawatts of critical IT load - because Dinsdale is keeping it secret, but we do know that the US is 53% megawatts and China is 16%.