logo

Tag: Outlook

Blog Details

Webscale earnings and 2021 capex outlook

Posted on | by Arun

Amazon drives webscale capex jump in 4Q20, Alphabet lags; Oracle’s 63% YoY increase is sign of OCI ramp 

Webscale capex on track to close 2020 around $122 billion

Most webscale network operators (WNOs) have now reported 4Q20 earnings. For a large sample of reporting WNOs, 4Q20 capex jumped by an average of 52% YoY versus 4Q19. Amazon was by far the most significant company, spending over $14 billion on capex in 4Q20, from $5.3B the year prior. Remove Amazon, and the YoY increase was only 14%. This is a sign of both how influential Amazon has become in the cloud and how COVID-19 enticed the ecommerce side of its business to ramp up logistics and fulfillment capacity.

The figure below illustrates YoY changes in capex for key WNOs who have already reported earnings. As shown, all but Cognizant and Alphabet increased capex YoY on an absolute basis (in US$), but nobody comes close to the Amazon surge.

webscale prelim capex change 4Q20

In MTN Consulting’s recently published capex forecast, we projected that webscale capex would close out 2020 around $122 billion, up from $104 billion in 2019. Preliminary results from 4Q20 are consistent with this projection. Looking further out, we still expect webscale capex to continue solid growth, climbing to just over $200B by 2025. The network/IT/software portion of capex will come in at roughly 50% in 2020, as it did in 2019, decline to 42% by 2023 as a spate of new data centers come on line, but grow back up to 50% by 2025 as more of capex is for server/capacity expansion of existing infrastructure. Within the technology part of capex, data centers and their components (networking, compute, storage, power) will soak up the bulk of spending, but subsea cables and satellite networks will become increasingly important over the next 5 years.

Earnings calls: A rundown of webscale capex & plans for 2021

Oracle has big plans for expanding it data centers in 2021, according to its fiscal 2Q (calendar year 4Q20) earnings call on December 2020. The technology corporation, now headquartered in Houston, reported that it plans to continue adding new data centers through 2021, and expects to have 38 by YE-2021.

Oracle is a small cloud player relative to Amazon, especially considering Oracle’s heavy reliance on collocated facilities (mainly with Equinix). However, Oracle is clearly becoming more significant in the cloud, and its use of Ampere’s data center chips could also have an impact on the competitive landscape in the chip market.

Oracle’s frequent mention of its data center plans indicates a high commitment to its cloud rollout and a bullish outlook on its prospects in this business. During the first half of its earnings call, Chairman Larry Ellison noted the company’s “great quarter” would have been even better if it had not been capacity constrained in Oracle Cloud Infrastructure (OCI). He said to remedy this Oracle is adding OCI capacity and building OCI data centers as fast as it can. Ellison claimed that, as of December, Oracle had increased its footprint to 29 regional data centers worldwide, “more than AWS” (per Oracle), adding customers, and growing revenue well over 100% year-on-year.

Ellison explained that Oracle’s strategy recognizes that, because the company has a large existing business and installed base of software customers, “we just have to get into more countries than someone – than Amazon, let’s say, because we have to serve those customers where we have an installed base.”  He pointed to Indonesia and Israel as examples: Indonesia is a very big country with a lot of people where Oracle has a large installed base, “but a lot of people don’t have data centers there.” Ellison said it’s very important to get a good data center in Israel, but “some of the cloud companies have been late to get there.”

CEO and Director Safra Ada Katz noted Oracle has greatly increased its capital spending plans due to strong demand, and expected that it will increase 50% sequentially in 1Q21. Such an increase would imply around $850M of capex in 1Q21, a 144% increase relative to 1Q20. Oracle’s data center strategy has traditionally been capex-light due to the company’s partnership with Equinix, which provides many of the properties into which Oracle installs its equipment. That makes Oracle’s capex increases all the more significant.

Amazon’s earnings call revealed very little about its data center plans, in contrast to Oracle. That’s consistent with history but also may be due to management changes. Amazon promoted its AWS CEO, Andy Jassy, to replace CEO and founder Jeff Bezos just before its earnings call webcast on February 2, 2021. In response to a capex outlook question, CFO Brian Olsavsky said “We are working through our future plans.” Olsavsky did elaborate somewhat, suggesting that its 2020 capex reached $44B due to some unusual one-off investments.

Like Amazon, most other webscale providers avoided specifics on data center investment plans for 2021 on their 4Q20 earnings calls. Oracle was a standout. Clearly Oracle wants to make a splash in the market quickly, and assure its customers and prospects that its infrastructure is rapidly getting up to speed relative to the top 3 (AWS, Azure, and GCP).

Among the 17 WNO companies tracked by MTN Consulting, Alphabet had the second largest annualized capex through 3Q20, at $23B (behind Amazon). Looking forward to 2021, SVP Eric Schindler said he expected the pace of investment to return to normal levels, with spending on servers driving an increase: “Servers will continue to be the largest driver of spend on technical infrastructure.”

In addition to servers and new data centers, Alphabet continues to spend on subsea cables. For instance, Google reported on its blog on February 3, 2021 that its new Dunant subsea cable, which connects US and Europe, is now ready for service. The Dunant cable, which was made in partnership with SubCom, will have ultimate capacity of 250 Terabits per second, “enough to transmit the entire digitized Library of Congress every second.” Google notes that “Dunant is the first long-haul subsea cable to feature a 12 fiber pair space-division multiplexing (SDM) design.”

Facebook is the most bullish of the big webscale providers in terms of its 2021 investment outlook. During its 4Q20 Earnings call, Facebook CFO Dave Wehner said “We continue to expect 2021 capital expenditures to be in the range of $21 billion to $23 billion, driven by data centers, servers, network infrastructure and office facilities.” This compares to a 2020 capex total of just $15.1 billion. Wehner said several projects were delayed in 2020 due to the pandemic.

Microsoft CEO Satya Nadella noted that the company has announced 7 new data center regions recently, in Asia, Europe, and Latin America, which will contribute to capex over the next few quarters. CFO Amy Hood, speaking just after Nadella, said the company expects to see a “sequential increase on a dollar basis” in capital expenditures in 1Q21 as the company continues to meet growing global demand for its cloud services. Microsoft spending on finance leases, a.k.a. capital leases, is also significant and contributes to the company’s network infrastructure footprint.

IBM, which made no specific mention of data centers in its 4Q20 call, had much discussion of its focus on and efforts to bring AI-powered automation across its platform. IBM says it is building out its cloud integration offerings for this purpose. CEO Arvind Krishna also discussed quantum computing, and claimed that this “has the potential to unlock hundreds of billions of dollars of value” for its clients by the end of the decade. IBM says it has a roadmap to build a 1,000 qubit quantum computer by 2023.

Baidu is one of the smaller WNOs but, like Oracle, its 4Q20 capex grew significantly, up 162% from 4Q19 to reach $329M in 4Q20. Baidu did not speak specifically about data centers or capex projections in its earnings call. The company’s focus was on its self-driving and auto-driving projects, mentioning some combination of those terms 51 times. These projects require high levels of R&D spending. During its call, the CFO Cheng Chun Yu noted that it spent about 21% of its $12.1B USD core revenue in 2020 on research and development. During the call, a questioner asked how much time the company spends on different segments and what its priorities are. CEO Yanhong Ling answered that “half of our time and resources are invested in the mobile ecosystem part, and the other half in the AI cloud and intelligent driving and other growth areas…going forward, we will continue to do this”. Baidu has fallen in relative importance over the years, but it will remain an important player in the webscale market.

SAP made one minor reference to capex and data centers in its call. CFO Luka Mucic said there will be additional capex requirements this year and next for the purpose of harmonization and modernization of its cloud infrastructure. Apple and eBay made no mention of future capex or data center investments in their latest earnings calls.

Blog Details

India unlikely to rise above 4% share of telco NI in 2021

Posted on | by Arun

Prediction: India-based vendors almost certain to capture less than 4% of global telco network infrastructure market in 2021  

Prospects for 2021

The year 2021 is likely to see some significant shifts in vendor share in the global market for telecom network infrastructure products & services (“Telco NI”). Spending on 5G is picking up, interest is growing in open RAN architectures (and open networking in general), telcos are making more carefully thought out decisions without the chaos of COVID-19 dominating discussions, software’s share of capex is rising, and Chinese vendors continue to face supply chain and security constraints to their global position. Many telcos, and policymakers in the US, have pointed to India as a potential alternative player in the Telco NI market. India and the US are allies, after all, and India has a program aimed at developing local manufacturing – which calls to mind similar programs in China that helped Huawei and ZTE. The question arises, then: what are the prospects for India-based vendors globally in 2021? More specifically: what is the chance that India-based vendors can grow their share of Telco NI above 4% in 2021?

After reviewing the question, our conclusion is that there is less than 10% chance of this happening. Not impossible, but nearly so. Here is our assessment of a few key issues.

Current share below 3%

For the four quarters ended 3Q20, MTN Consulting estimates that vendors based in India captured roughly 2.6% of vendor revenues in the global telco network infrastructure (telco NI) market. That is based on sales to telcos for eight vendors we currently track (HCL, Infosys, Sterlite, Subex, TCS, Tech Mahindra, Tejas Networks, and Wipro), plus three vendors we are in the process of adding to our database: Cyient, HFCL, and ITI Limited. Annualized telco NI revenues for these 11 companies, by our estimation, amounted to $5.7 billion in 3Q20, out of a global market of $216.3 billion.  Nearly 70% of that $5.7 billion is recorded by three IT services players: Infosys, TCS, and Tech Mahindra, with the latter slightly ahead of the first two. (see chart, below)

India telcoNI vendors 3Q20-2

So, India’s starting point is low. Jumping from less than 3% to above 4% in the space of a few quarters is a challenge. Adding one percentage point in share is not the issue; larger vendors with the ability to ramp manufacturing to satisfy big one-time orders can do this easily. Most of India’s Telco NI strength lies in services, integration, and software development contracts, though, which tend to take more time to implement. Moreover, India is a relatively small market within global telecom. Local vendors do benefit from set-asides and easier access to local customers, but India accounted for less than 3% of global capex in the last four quarters. It has been higher in the past, in fact reaching 6.1% of global capex in the 1Q19 annualized period. But that was during Jio’s massive network buildout and aimed at helping the company jump ahead of the competition rapidly, which it succeeded in. 

India’s 5G buildout will be hobbled by pricey spectrum

Last month the FCC held an auction for 5G-suitable spectrum in the US. Winning bids totaled up to just under $81 billion, a bit more than 90% of the US telco market’s total annual capex of $88B (2019). Many observers have warned that this outcome could hobble the telcos financially, as they now also need to come up with capital to deploy the networks. These concerns are well-founded.

India could be in an even more precarious situation. Its primary 5G auction won’t be held until next month, but could potentially raise up to $50B or more for the Indian state treasury. This figure is several times India’s 2019 telco capex figure of $10.9B. 

Contrast this with China’s approach, where spectrum is generally given away for free (or nearly so), and the government uses its control of key operators to drive procurement practices and the pace of deployment. Because of this, 5G in China is seeing rapid adoption and local vendors have benefited massively. Huawei and ZTE gained share last year in the Telco NI market, in fact, despite all of their problems. 

A slow 5G rollout in India means fewer opportunities for local vendors to win new deals and gain crucial experience that they could leverage in global markets.

State-run operators are small and procurement is slow 

One area where India’s vendors clearly shine is with state-run operators BSNL and MTNL. Government procurement rules favor local suppliers strongly. However, the two together accounted for just $614M in 2019 capex, under 6% of India’s total. Further, the hoops they must jump through to build new network projects are cumbersome. Disputes about winners and losers can slow down final decisions and the actual spending connected to the decision.

BSNL’s long-planned 4G tender is a good example. It’s now more than 4 years since Jio debuted its 4G network in 2016, and BSNL is still at the early procurement stage with 4G. The project has been delayed several times. Most recently, BSNL did issue an expression of interest document covering a potential network with 57,000 sites. As promising as that sounds, it remains unclear what it means to be local. Partnering with a foreign vendor such as Samsung or Ericsson might qualify; having a large R&D presence in India (e.g. Mavenir) might also allow one to slip through. The challenge is that there does not exist a truly Indian company able to build a 4G network end-to-end, and it cannot be created overnight.

Hiring a truly Indian services firm like Tech Mahindra is one option for BSNL, as TM can get more experience putting together the piece parts for a network in way that will help with Open RAN opportunities. Tech Mahindra already works with Telefonica and Rakuten on Open RAN.   

Jio succeeded while avoiding local NI vendors

Indian telecom’s big success story in recent years is clearly that of Jio Platforms, which emerged from little in 2016 to become the market’s largest wireless provider in just over three years. Jio did many things differently than its rivals, one of which was to rely almost entirely on foreign vendors – but avoiding the Chinese. Jio has some gear from Tejas in its network and a few other small local vendors, but it’s minimal given the telco’s size. There is no current sign to suggest Jio will be more eager to adopt local tech as part of a future 5G rollout. If spectrum was to be issued on a cheaper (or even free) basis but come with some ‘buy local’ requirements, clearly the outcome would differ.  

Bharatnet is not that big a project in global terms

India has made impressive strides in expanding connectivity in rural regions over the last few years, due largely to the Bharat Broadband Network, or Bharatnet. This project has been especially helpful to local fiber manufacturers. It continues, but the scope of the project needs context. For the fiscal year 2020-21, approximately $800M was budgeted for Bharatnet. That is a sizable amount, and many local vendors (e.g. Tejas) cite national or state-level Bharatnet projects as contributing to recent revenue growth. But $800M works out to roughly 2% of telco industry capex in China alone for 2019. The project is just not big enough to help a local vendor go global. 

What could swing things the other way?

In my opinion, it would be great to see a globally competitive vendor emerge from India. At one point in the distant past, it seemed that Tejas had potential to be a real player in the optical market. That has not happened; lack of a local supply chain and manufacturing are among the many issues.

Currently Tech Mahindra is looking like it has real potential, with a growing focus on the blossoming Open RAN market, and helping telcos integrate these new networks. Tech Mahindra’s collaboration with Rakuten on this point is promising. TM will have competition from many other companies going after these projects, though – not just RAN vendors but other telcos, including NTT DoCoMo. It’s not just Tech Mahindra with potential, though; Infosys, TCS, and Wipro are already active in 5G RAN networks. If these IT services specialists can quickly help telcos review and implement new network architectures that have verifiable cost savings, then India’s vendor prospects may brighten. 

The biggest thing that could support the local industry, though, is a different government philosophy on spectrum auctions. Clearly the government needs to raise funds to operate, and auctions are one easy way. But if there is interest in developing local 5G ecosystems and helping local companies make it big globally, then maximizing auction proceeds cannot be the overriding goal.