Nokia’s CEO, Rajeev Suri, has acknowledged that the Finnish company’s 5G chip strategy is partly to blame for its falling margins, as highlighted in its recent quarterly results announcement.
Margins were hurt by the high costs of its Reefshark family of chipsets, which power equipment from high end routers to base stations to Massive MIMO antennas. Although Reefshark was launched to acclaim in the engineering community, its reliance on programmable chip technology and inhouse designs makes it costly compared to platforms based on microprocessor or system-on-chip (SoC) technology.
The big three network OEMS have taken differing approaches to silicon. Nokia and Huawei have remained in the tradition of designing their key chips inhouse, to give themselves control and differentiation, while Ericsson has been the most open to using merchant platforms, not just for cloud infrastructure to support virtualized cores, but for cutting edge designs such as 5G vRAN systems. For instance, Ericsson is working with Nvidia on a potential vRAN platform based on graphics processing units (GPUs).
Nokia and Ericsson both have close Intel partnerships, but Ericsson has pushed this further into its network, while Huawei has gone as far as designing its own server processor, based on ARM cores. The centrepiece of Nokia’s bid to differentiate itself through selected silicon products is Reefshark, which powers its petabit router – a bid to extend its sales to cloud giants as well as telcos – as well as 5G products.
But a decision to base this on externally sourced field programmable gate arrays (FPGAs) has proved costly, Suri admitted. FPGAs are programmable after deployment and so very flexible. This was relevant to Nokia on two counts when it started Reefshark development – it was in the midst of acquiring Alcatel-Lucent, and unsure exactly how the two companies’ products would be combined and migrated; and the 5G standards were immature, so it was highly likely that functionality and features of the chips would have to change to support the final specs.
FPGAs “give you flexibility, they give you time-to-market advantage, but then they’re expensive,” Suri said on the earnings call, which saw the company’s shares fall on a reduced forecast, lower margins, and the suspension of the dividend until that margin recovers.
He also indicated that an FPGA supplier (reported to be Intel) had fallen short. Suri told analysts: “Admittedly, one supplier let us down as well.” In a research note, Michael Genovese, an analyst with MKM Partners, said the supplier in question had failed to deliver a low cost 10-nanometer product. On the earnings call, Andrew Gardiner, an analyst with Barclays, said to Suri: “You had this issue with ReefShark with Intel’s 10-nanometer process”, and this was not disputed by the CEO (Intel has not commented).
Nokia hopes that, now the 5G specs have stabilized, it can move to SoC products to reduce costs and help restore its profitability and its cash position. Sandro Tavares, its global head of mobile networks marketing, said: “FPGAs do cost more in unit price, and they do use more power. Now that things are pretty much set with Release 15 and 16, it’s good for us to get our SoCs. We have been working on the development for quite a while right now. We are now starting to ship the first products with our SoCs replacing some of the FPGAs.”
In its third quarter, its operating margin fell by five percentage points in the Networks division, and it does not see any improvement in the near to medium term. For full year 2019, it downgraded its margin outlook by two percentage points, and for 2020, by 5.5 points compared to previous guidance. It is now looking for non-IFRS operating margin of around 8.5% in 2019, and 9.5% in 2020, but insists it can improve that figure to between 12% and 14% in the 3-5 year timeframe.
Of course, this is not just about chip costs, but that is an aspect over which Nokia has some control, unlike some of the other factors which Suri cited as impacting Q3 margins – as well as “product mix and a high cost level associated with our first generation 5G products” he also pointed to “profitability challenges in China; pricing pressure in early 5G deals; and uncertainty related to the announced operator merger in North America”.
He told the analysts that the plan was to move “to equity SoC-based products, which we’ll progressively start shipping during 2020. To ensure that we execute on this fast and effectively, we are increasing investment in system on chip capabilities and moving aggressively to strengthen and diversify our supplier base.”
It has hired about 350 new people in Finland, most of them reportedly working to accelerate the process of migrating to SoC-based 5G silicon.
Tavares said: “We are repositioning our workforce to work on topics that are more critical for 5G development. We are hiring people in Finland and in other parts of the world and moving people from different programs to our 5G radio products.”
There are many other factors at work, including the suspicion that Nokia was poorly prepared for the rapid deployment of first-phase 5G networks, as enabled once the Non-Standalone subset of specifications was fast-tracked, allowing for roll-out of a RAN with an existing LTE core. Nokia was the most outspoken vendor in criticizing that approach when it was accepted by the 3GPP in response to lobbying by AT&T and others. There have been reports that some Nokia equipment has performed sub-optimally in early contracts, especially in the USA, and of course, it is still consolidating multiple legacy platforms, and R&D programs, that were inherited with Alcatel-Lucent.
It seems highly unlikely that FPGA migration alone will account for a margin reduction of 4-5 percentage points in 2020, but it makes sense for the firm to highlight an issue it has already begun to solve, to boost confidence, even while it manages expectations about the speed of margin recovery. As well as working on its own SoC, it says it will extend its supply chain so it is not over-dependent on one vendor – but Suri stressed that all these actions will only affect product costs significantly in 2021.
Nokia cut overall R&D expenses by 6% last year, to about €4.62bn ($5.14bn), and has spent €3.34bn ($3.71bn) in the first nine months of 2019, about 3% less than in the same period in 2018. This was at a time when its main rivals were ringfencing or increasing R&D investment even while cutting back elsewhere.