UK 5G Auction – Why wide radio channels send prices rocketing

UK 5G Auction – Why wide radio channels send prices rocketing

The media and general public will be hailing Ofcom’s spectrum auction as a huge success. It raised over £1.355 billion for the Treasury. This has now become the popular measure for the success of a spectrum auction. In 2003 spectrum in the 3.4 GHz band only fetched £175,000 per MHz. At this auction it fetched £7.83m per MHz. A stellar result from the Ofcom spectrum “money printing” press.

However, as Ofcom constantly reminds us, the purpose of a spectrum auction is not to raise money for the Treasure but to get the spectrum into the hands of those that will make the best use of it. Against this measure, how successful was the spectrum auction?

The auction saw Telefonica/O2 buy their way back into a spectrum parity with the other mobile players. They paid £5.1m per MHz for 40 MHz of 2.3 GHz spectrum which is the almost identical price Vodafone and BT paid for their 2.6 GHz spectrum five years ago. They also bought 40 MHz of 3.4 GHz 5G spectrum for £7.9m per MHz. This has got Telefonica/O2 out of its uncomfortable corner of trying to support the highest number of customers per unit of bandwidth of anyone in the market. It is a success for Ofcom’s spectrum cap policy.

The third positive outcome is that there is now 5G pioneer band spectrum in the hands of the mobile operator and this opens-up a clear road to the launch of 5G high capacity networks in the UK before 2020.

These successes need to be balanced against the auction down-sides and particular from an industry perspective:

The reserve for price for each 5 MHz lot of 3.4 GHz spectrum was £1m and there were thirty lots. If the spectrum had sold for the reserve price, the industry collectively would have paid £30m. Instead they paid £1.16 billion.  The 5G spectrum also fetched 50% more than the 4G spectrum at 2.3 GHz. What was going on?

Branding it “a 5G pioneer band” probably sprinkled a bit of star-dust on the spectrum but that was also true of the Irish 5G auction. Vodafone UK paid £7.56m per MHz for its 3.4 GHz spectrum whereas my calculations show that Vodafone Ireland paid £1.53 per MHz on a comparable basis. That is nearly one-fifth of the price. (Note: my calculation averaged the city and rural spectrum prices and adjusted for the two different population sizes).  We also see later that the new entrant had no effect on the price.

This only leaves the Ofcom auction design generating over-heating. Let’s look at some numbers:

If the four mobile operators had entered the auction to just to secure equal shares ie 37.5 MHz each, they would have only paid £7.5m each or £0.2m per MHz. Instead they paid:

Average price of 3.4 GHz spectrum for each operator

The incremental bandwidth over the “fair share” that Vodafone secured was 12.5 MHz and bought at a price of £371.45m over the reservice price or £29.7m per MHz. That seems a high price to pay for bandwidth over “fair share”. But it is a bargain when set against what BT and O2 paid for their 2.5 MHz of bandwidth over the “fair share”. It was a staggering £118m and £125m per MHz respectively just to finish-up with marginally more than a “fair share” size of radio channel bandwidth.

It is normal to blame the mobile network operators when auction prices overheat. But the auction framework creates a classical “prisoners dilemma” in game theory. The theory explains why two completely rational entities might not cooperate, even if it appears that it is in their best interests to do so. It arises from the prisoners not being able to talk to each other or in this case the bidders. In the case of prisoners they finish up with long jail sentences and in the case of spectrum bidders – a huge price tag. The penalty for a single party behaving rationally is to leave the auction with nothing and locked out of the new market.

Ofcom auction design has made “incremental bandwidth” disproportionately expensive in the UK. Since it affected all the operators, the consquence is likely to be less investment in the 5G networks.

That is not the only problem with the Ofcom 3.4 GHz auction outcome. The auction was in two stages. The first settles how much spectrum each bidder gets and the second determines exactly where in the band their spectrum will be assigned. This was made more complicated by an incumbent service, called UK broadband, that Ofcom allowed to stay in place. Subsequently, the company was bought by H3G, so even before auction began H3G had 40 MHz of spectrum in the band. The following band plan emerged from the auction assignment stage:

3.4-3.6 GHz bandplan.

The decision of H3G to leave UK Broadband 40 MHz of spectrum where it was has resulted in their 60 MHz being fragmented. But it also left no choice where Vodafone, O2 and BT/EE could put their assignments as only lower part had 50 MHz space to accommodate Vodafone’s 50 MHz of spectrum. However, it is not the only fragmentation issue – the spectrum of the two site sharing groups has been split up.  The band plan is a mess.

The mention of the two site sharing groups leads onto another revelation. If one looks at how much spectrum each site sharing group secured: the Vodafone plus O2 group got 90 MHz and the BT/EE plus H3G got 100 MHz of the band. That is an almost equal split. Had the two site sharing groups been the bidders rather than the four individual companies they would have paid a total of £35 million (the reserve price) instead of £1.16 billion. Now, it has been argued that different commercial policies between the two companies in each sharing groups would have precluded this joint approach, as different retail offers would have led to unequal demands on capacity. But “traffic prioritisation” below the norm or some sort of “wholesale model” could have solved that issue.

Instead, each bidder went for broke and ideally wanted to secure a 100 MHz wide radio channel to themselves but would have been well pleased to have got 80 MHz. Nobody got what they wanted.  The 5G auction of the 3.4-3.6 GHz band has been a battle for bandwidth for which everyone bid themselves into a stalemate. H3G’s 60 MHz will not make a material difference in the market against Vodafone’s 50 MHz and that 50 MHz will make little difference against O2 or BT/EE’s 40 MHz.

Spending over a billion pounds and nobody getting what they wanted should make the industry think long and hard about the next spectrum auction for the 3.6-3.8 GHz band.

A final observation on the auction outcome was the failure of a new entrant to emerge. Airspan were a serious player with a new “neutral host” business model. Ofcom made no provision for a new entrant, despite their strong commitment to increasing competition and improving coverage. Airspan might have done both. Their contribution to pushing the boundaries of 5G coverage would have been driven by their business model, as they would only get wholesale income by finding “not spots” left by incumbent mobile operators.

What conclusions do I draw from this analysis?

  • The industry need to think long and hard how they adjust their strategies to reduce the price they pay for incremental bandwidth. Were the second auction to also fetch £1 billion, then what would have been spent on 5G spectrum alone could have been used to roll-out 100,000 small 5G cells. That is not a route to profitability for mobile operators, nor is it a route to UK leadership in 5G.

A new strategy that could make a lot of sense is to introduce two new elements. The first is to think about 26 GHz for those areas where all mobile operators co-locate in dense urban centres. There is huge bandwidth available at 26 GHz. The second is for mobile operators to encourage Ofcom and the Government to over-license the 3.6-3.8 GHz band with secondary (“non-interfering”) dynamic spectrum expansion rights. This would allow a mobile operator, in areas where all mobile operators are not co-locating, to expand their radio channel “free of charge” up to the entire 200 MHz band. It is free as it is essential re-cycling local unused spectrum that would otherwise be going to waste. Together, these two elements would take the price pressure off the cost of incremental bandwidth.

  • The Government and Ofcom need to think how the release of the 3.6-3.8 GHz 5G pioneer band could open a route for new entrants.

Reserving a licence for a new entrant is not a serious option as a market price of £7.8m per MHz would not be affordable. (Airspan exited the 3.4 GHz auction when the price had reached £0.37m per MHz). The IET has been a long-time advocate of carving out 20 MHz channel in a prime cellular band for shared innovative use. Couple this with dynamic spectrum expansion rights would be a powerful spectrum package to take into niche locations (rural areas) that the incumbent mobile operators have little interest in covering. It would be a far better public policy approach than say the policy in Ireland of earmarking 5G spectrum specifically for use in rural areas.

Getting the 5G pioneer band 3.4-3.8 GHz into the market is only at the half-way stage. It is premature to judge the first stage as a success or failure in getting 5G spectrum powering an up-grade of the UK’s mobile infrastructure without knowing what will happen in the sccond stage. The good news is that there is time enough to address the issues coming out of the first stage so that, after the auction next year, the UK finishes in a good place in support of its ambition to be a leader in 5G.

Comments are closed.