After the Bell: AI and 5G - the cellular business of slow and fast deaths

When cellular companies were contemplating a move from 3G to 4G, governments around the world were licking their lips. 

Cellular companies are specifically invited to participate in the industry and licensing regulations specify the “G” mode. Hence, cellular companies would have to bid for licences, and the result would be a huge transfer of boodle from the private to the public sector. 

Mobile Network Operators require something called a carrier licence to operate networks, but this is only one part of the entire, enormous regulatory context. Theoretically, the cellular industry is uniquely regulated because of its reliance on scarce spectrum resources, but of course governments could waive the auction income on the basis that cellphones are a boon to society - which, of course, they are, apart from texting while driving and the Candy Crush obsession. 

Anyway, perish the thought that governments would not require a tax if it’s available to them. Most spectrum is allocated by auction, and every new “G” equals new auction. The amount of cash raised by the government (let’s just call that government giving your cash a new home, between us friends) is eye-popping. In 2021, the 3.7–3.98 GHz spectrum (C-band) auctions raised a record $81.11-billion. In 2022, the 5G auction in SA raised R14.4-billion for the government from six operators. 

The cellular companies despise these auctions (in some ways), which are really pretty arbitrary if you think about it. You are not required to pay an additional tax to the roads department every time a road is built. Er .. actually sometimes you are. But in general, the mere allocation of road space is not auctioned because building roads is just one of the things for which you pay tax - technically, there should be no additional tax for the mere allocation of road space. 

Anyway, at the time of the move from 3G to 4G, one of the smaller cellular operators told me that participating in the auction was like the difference between slow death and fast death. Either you participated in the auction and died a fast death because the company would be saddled with mountainous debt that it would have to service for years while trying to offload the cost on to its customers. Or you didn’t participate in the auction, and you died a slow death as your customers gravitated towards the larger players who would offer a faster service than you could.

As it happens, that company opted for a slow death, participating but getting saddled with debt. The good news is that the company managed to avoid death and stay alive. It’s a great illustration of the resilience of the cellular industry, not to mention the hallowed business principle that you should deal with today’s mess first, and tomorrow’s chaos can wait its turn.

The same sort of process is now taking place with artificial intelligence (AI). The Chinese company Tencent, in which South Africans have a sizeable interest, announced at the weekend a big update to its model called Hunyuan Turbo S, which it claims is faster than DeepSeek - the Chinese newcomer which shot on to the scene recently.

This follows e-commerce giant Alibaba, which released its Qwen 2.5-Max model in January, which it claims outperforms DeepSeek-V3 across the board. You have to say, this is all getting pretty damned hectic. What was most interesting about the Tencent announcement was that it substantially reduced its prices: 0.8 yuan (about $0.11) per million tokens for input and 2 yuan ($0.28) per million tokens for output. These are significantly cheaper than previous Turbo models. (Tokens are fragments of words or characters that AI models process. “Hello, how are you?”, for example, is five tokens.)

The crazy land grab for space in this industry is now palpable. For investors, the chief fear will be whether this mad investment flurry will actually pay off in the long term, and this we just don’t know. But for the companies involved, it’s the choice between slow death and fast death, and almost uniformly, the companies are choosing slow death. Or to put it another way, they know they must enter the fray and are hoping beyond hope that tomorrow’s chaos will look after itself.

One way to perhaps assess this is to look at the growth rate of industry leader chip provider Nvidia. The company recently reported 78% sales growth in the most recent quarter, which seems fabulously impressive.

But as The Daily Upside pointed out, this is down from tripling during the same period last year, and its stock fell as enthusiasm for expensive chips cooled and macroeconomic concerns (aka tariffs) heated up.

From a consumer’s point of view, you have to say, this is all great. The intensity of the competition is doing exactly what you would hope it would do: bring down costs and improve performance. For investors, it’s more of a nuanced story. DM