D2D: Adéu Barcelona
Our findings from this year's Mobile World Congress
Mobile World Congress (MWC) offered its usual assortment of large crowds, intense data-gathering, fading relevance to semiconductors, and lots of walking.
Just a reminder that we operate our newsletters on a paid model. Paid subscribers will get three newsletters a month including our China Deep Tech notes. Paid subscribers will also get early access to the newsletter as well other benefits coming soon. Please subscribe and support our work.
Mobile World Congress 2025
Adéu Barcelona
This was our 20th year attending Mobile World Congress (MWC) in Barcelona. The show remains very healthy, with reportedly 109,000 attendees, a record. The halls, especially on the first two days, were packed. That all being said, we have begun to question the relevance of the show, at least for our work.
MWC is a show about selling things to mobile operators. Ten years ago, those purchase decisions were critical to entire industries - semiconductors, networking equipment and more. Decisions that took place at the show settled wide ranging battles and determined the competitive dynamics across much of the semiconductor landscape. But those battles are all settled. MWC is no longer a show where amazing new things are launched. The tenor of conversations has moderated.
To be clear, MWC is still an important show for the telecom industry, but for semis that is just one more end market, and not a terribly important one. MWC remains relevant, just not so much for us.
That all being said, we had a good time and learned a lot.
Telecom Doldrums
Part of the problem facing the show is the industry is not growing. The 5G build is several years past its peak, with most of the major operators now fully deployed.
This is particularly difficult for the telecom equipment majors, or at least two of them. On one side, Huawei continues its relentless rise. As usual, they had a massive booth which covered a dizzying array of topics. By contrast, Ericsson seems to have reached some sort of steadiness, or lest stasis. For the most part they are marinating share globally, losing business in China but gaining in North America. That leaves Nokia squeezed between the other two.
Nokia is about to get a new CEO, and so we will be watching to see what direction he charts for the company, but a few things seem clear. Mobile equipment is becoming less important to Nokia. It is still their largest segment, but its other businesses are growing faster and look to have more long-term potential.
In particular, Nokia’s optical business seems to be doing well, as does its more general purpose networking equipment for the enterprise and data centers. It also now has $1 billion IP licensing business. As one person told us Nokia looks like it is transforming into a combination of Juniper and Interdigital. It is unclear to us how sustainable enterprise networking will be over the long haul, but we think it is important to recognize the changes underway at Nokia.
Qualcomm
We had some serious concerns about Qualcomm’s booth last year, so we were heartened to see a return to form this year. Qualcomm’s booth was split between a healthy amount of raw wireless technology - infrastructure, waveforms and 5G standards, and devices with drones, glasses, goggles and all kinds of IoT on display. They showed off their PCs, but we were happy this was no longer a focus as MWC is not the right audience for those.
One of the best aspects of the booth was this was the first time we saw the company’s new branding scheme. Snapdragon covers phones and Dragon Wing covers all industrial and enterprise parts - mostly home networking and industrial IoT, but also the company’s infrastructure products. It makes sense, and we wonder how it will work behind the scenes.
One concern we had was the degree of difficulty required to tease out all of these. The company issued a modest number of press releases, but to be blunt they looked a lot like last year’s. The company launched a new modem-to-RF platform - the X85, as opposed to last year’s X80. The new system has some fairly remarkable advances in it, and we would like to learn more about operators’ appetite for those. On the infrastructure side, the company’s networking processor family has reached commercialization with real world deployments, but we could not help but notice that the company did not announce any new operators.
The biggest question we had was 6G. As we discuss below, this is an important moment for the new standard, and Qualcomm made only brief mention in its booth.
For anyone reading at Qualcomm, we would love to get a briefing on 6G and all the rest.
Time for 6G
As of next week the work for the new wireless standard will have officially begun. This will take a while, but the coming working group is important symbolically for the industry. New G’s bring good times for the industry, so at least we have this to look forward to in 2029 or 2030….
Beyond the symbolism, the reality is less upbeat. The mobile operators are in no hurry to move to the next standard, they are still struggling to get a return on their 5G investments. Moreover, it is not clear what the new standard will bring. There are no easy ways to manipulate spectrum or boost spectral efficiency anymore. Instead, we have a long list of small tweaks, and hopes to tap into new spectrum blocks. Unfortunately, all the best spectrum is pretty fully utilized, and so the industry is looking to higher frequency bands which offer terrific data rates but limited range. (To say nothing of pending conflict with existing users of the frequency blocks in many markets.) So 6G is coming, but it is going to be a while and offer few easy wins.
Networking Processors - Sometimes Dinner Eats You
For the past few years we spent a lot of energy trying to analyze the market for mobile base station processors. Two years ago, we shuttled back and forth from the Intel booth to the Marvell booth to the Nokia booth to the Ericsson booth and then back to Intel, lapping Halls 2 & 3 comparing notes. Should accelerators be in line or not? Is general purpose compute sufficient? What are the true technical requirements?
It turns out that none of that matters. Sometimes, probably often, commercial technology comes down to customer mix, luck and inertia, not the actual technology. Marvell won Nokia. Intel went with Ericsson, and Ericsson navigated its end markets better than Nokia, so Intel ends up doing well here. That’s all there is to it.
On that topic, does anyone know whose processors Huawei is using? Bit of a mystery there.
A side effect of the stagnant growth outlook in telecom is that it makes it less important to semiconductor industry as a whole. If you have two equally demanding customers and one is offering 40% gross margins while the other is offering 70% gross margins, which one should get the attention? This is 2025, so the answer is of course AI, and not telecom.
One silver lining in all of this is that as the big companies focus more on the big data center cloud market, it opens up doors to smaller, newer companies. In telecom that means Ampere. There has been some concern around them in the industry. They have great products, but are stuck underneath some very demanding hypserscaler customers. Despite this, they continue to plug away, with serious business at Oracle (who owns a good chunk of them), and they had a lot to say at MWC, with a number of partnership announcements and a solidly packed meeting schedule. Telecom is a tough business to win, but Ampere has been plugging away at it for years. Maybe it’s a niche market, but it’s a sizable niche.
Comic Interlude - Aduna
Another notable launch at MWC was telecom API provider Aduna. This is a private company with venture backing from the telecom operators and Ericsson. They offer an API that lets software developers make API calls into operators’ data. This can be used as a security/anti-fraud protection, or maybe advertising, or some other kind of software need. We heard it referred to as the operators’ response to Twilio, which we see in theory, albeit in practice Twilio seems to have a much wider offering.
We try very hard to be polite and respectful of the work that anyone does, but we have to admit that we had a very hard time keeping a straight face with this one.
For decades, the operators have tried to come up with compelling software. Anyone remember LiMo? How about WAC? Or OMA? We remember all of those, and more. But you would be forgiven for not remembering them, because they all came and went very quickly.
Aduna has a great user interface, real API designers did good work on it. And we see the utility in its offering. It’s just that all of this is coming very late to the game. Twilio has been around for 15 years, and Aduna’s value proposition is hard to see. We spent 20 minutes in their booth learning about it and not once did we hear the word ‘customer’. Now we are fairly certain that the Aduna team is serious and they have thought through a lot more than they were able to share with us in a brief encounter. It’s just that the carriers are not good at this. The chief differentiation between Aduna and WA COMA LIMO seems to be that it is structured as a private company, all those others were consortia. Maybe that will help keep the team motivated, but the Board will look like all those past consortia’s boards, and be fraught with the same challenges.
ORAN - End Days?
While we are on the subject of Ericsson’s Series of Unfortunate Events, Open RAN (ORAN) deserves a mention. ORAN was intended as a software layer that runs on telecom network equipment. The idea was that operators would decouple the hardware and software that traditional equipment vendors sell to the operators. In its ideal form, ORAN would allow the operators to buy commodity servers and pick and choose which software they wanted to run on that gear.
That is not going to happen. We wrote a lot about this last year when Ericsson captured all of AT&T’s network business with an offering that promised to be ‘Open’.
ORAN is not dead, but it is clear that it will never achieve that Ideal Form. Instead, elements of it will be incorporated into the offerings of all the legacy vendors. There will be some room for new entrants to provide software, but the complexities of the market and the hard requirements around interoperability will confine the scope of that software.
Telco Cloud
Despite ORAN’s constriction, the hope of improving telecom software is moving forward albeit not in a way that the operators may have hoped. The public cloud providers continue to inch their way into the telecom business. AWS, Azure and GCP all had sizable booths at the show. The most interesting was AWS. Their suite was full of demos showing various 3rd party software running telecom workloads on AWS. There are now providers for all the pieces of critical software operator software - from call centers, to OSS/BSS management and billing systems, to every flavor of security.
Our favorite example of this is a company we met last year called Symphonica. They caught our eye because they had one of the more tangible AI workloads out there. They took operating manuals for all of the telecom gear and fed them into an LLM. They then tasked the LLM with providing configuration scripts for that gear. This year, they demo’ed their offering in the AWS booth where they showed what is essentially a drag and drop interface for telco billing systems. It is hard to overstate how revolutionary this is. Telecom billing software is a dark corner of custom programming work. Typically small changes can take months and millions of dollars to implement. Symphonica radically reduces that to a fraction of those amounts.
Another example comes from legacy provider Proximus. This company has been building telco software for a long time, but in recent years they have started changing. They have conducted a series of acquisitions to widen their offerings, most recently adding BICS, Route Mobile and Telesign which add call routing, security and identity to Proximus’ core offerings. The combined portfolio is now essentially a full-blown Telco API, which looks much more promising than Aduna.
The debate remains how much of all this the operators will adopt, and no one really knows. The operators move slowly. They are heavily regulated. When their systems break they will get called in front of congress, and are thus extremely conservative. On the other hand, they are all now operating in a no-growth business. In this kind of environment cost-saving imperatives will steadily crack open the door to new approaches and new suppliers.
Random Semiconductor Notes
A few other interesting things we saw. For the first time we can remember Unisoc had a booth at MWC. Unisoc is the third largest cellular modem vendor behind Mediatek and Qualcomm. They were formed by the merger of Spreadtrum and RDA Micro into what is essentially the flagship PRC chip design firm. Unisoc launched its first 5G modem a little less than two years ago. They target the bottom end of the handset market, and still sell a lot of 4G parts.
We also discovered a maker of mobile hot spots that uses Huawei merchant modems for 5G, because “Meditaek is too expensive”. This is the first time we have seen merchant Huawei modems in the wild. It will be interesting to see if they broaden that business.
Mediatek also put in a strong showing. Their booth focused mostly on their mobile business, but we are planning to do a lot more work on everything else they do. They have broadened considerably and we should pay more attention to all those other product areas as well.
MWC From a High Level
We always get asked what the theme of the show is, and while there is never a perfect answer to this question, but a few high-level themes stood out.
The software content of the show is becoming more pronounced. Our sense is that the operators are going to look much harder at their software requirements in coming years and so there seems to be a lot more vendors trying to tap into this. We doubt MWC will ever become a pure software show, there will always be dozens of remote radio unit vendors, cable suppliers and tower operators, but software will start to stand out more.
This was a big year for PRC companies. Five years ago almost none of China’s handset vendors had booths at MWC. This year, Honor occupied the most high profile spot at the entrance to Hall 3. For their part, Xiaomi had the largest booth in the hall. In fact, this was probably the most popular booth at MWC. In part because they made prominent display of their car, which is currently only available in China. The company has a lot going on and their whole portfolio was pretty interesting.
Photo Tour
If you like this content, you should check out our podcast The Circuit
Thank you for reading D2D Newsletter. This post is public so feel free to share it.