Icera sold to Nvidia

I suppose we’ve all known for some time that Icera was going nowhere, but it’s still very bad news for the UK industry as a whole that it’s been sold off to Nvidia for $367 million – barely $100 million more than the funding it raised.

The Icera story means that it will now be harder for UK start-ups companies to raise cash with VCs becoming more wary of giving backing to UK engineer-entrepreneurs.

It’s a shame that the $250 million of equity funding made to Icera was not given to ten companies with $25 million funding each.


In the VC world, as in other worlds, spreading your risks is a better bet than putting all your eggs in one basket.


For the UK VC scene. too much was riding on the success or failure of one company.


Icera had high hopes – its founding CEO and some of its team had come from Element 14 the Cambridge DSL start-up which was sold, pre-revenue, and even pre-silicon, to Broadcom for $600 million at the height of the tech boom in 2000.


The founders of Icera said that, this time, they would not sell out early but would build a significant European player in the consumer space.


It never happened.


Nine years on from its 2002 founding, Icera still hadn’t got a design win in a mobile phone handset – its original target market.


Instead it relied for its revenues on low value sales to dongle manufacturers.


It is thought that VCs would have pulled the plug on Icera some time ago, were it not for the sheer scale of the investment in the company. This way they will get their money back at least – but not much more.


To make 30% profit after nine years means the VCs would have done better to put their money into National Savings – not the sort of return VCs aspire to.


The problem for Icera was that, after it was founded, the wireless chip-set market transformed from a component business into a sub-system business with suppliers having to provide a complete module including software and RF.


This made the business extremely expensive for all but the largest participants. Even very big players like STMicroelectronics, Ericsson, Motorola and Infineon got out of the wireless chip-set business.


If Icera had got down and dirty and competed with the likes of MediaTek in the 2G market, it might have been able to generate enough revenues from Chinese handset OEMs to keep going.


But Icera chose to remain at the high end, and the consolidated handful of major phone makers weren’t going to put their trust in a start-up for a vital component.



  1. @Anonymous – one thing the standards bodies ARE being careful about after Qualcom’s domination of key patents and high licence fees over the last 15 years or so is to try to make sure this sort of thing doesn’t happen again.

  2. If Icera’s technology is so superior, with receive diversity and other features which benefit network efficiency and performance, why didn’t Icera sell these benefits to the operators rather than trying to convince the manufacturers? The operators have enormous power in the telecoms industry – they influence what hardware can connect to their networks and what is sold in their shops, and of course choose which manufacturers to use for their network infrastructure. If Icera convinced the operators that its technology would help maintain the performance of their networks, save them money and bandwidth and give more sustainable and reliable operation for their users I’m sure the manufacturers would soon have to incorporate it into their products – possibly through licensing or buying from Icera. Did Icera ever push these features in the various telecoms standards bodies (made up of technical representatives from the manufacturers and operators alike) working in these areas, or approach the operators directly?

  3. Anonymous: Well then you were right, analysts were wrong, and Icera was foolish to believe industry forecasts. But…
    1) Phone tethering is still a dubious solution to this day because 99% of cell phones still don’t have receive diversity. This can result in nearly a 2x real world performance difference (before we even consider their basebands are often slower than Qualcomm ones) and this also means a 2x higher cost per bit for the operator. So I think analysts expecting less tethering had a point even if they turned out wrong.
    2) As I said, I think Icera never got more than about 10% of the data-only market. They had hoped to get 33% and I’m sure they could have hit break-even and IPOed with that kind of volume despite the smaller total market. Competing with Qualcomm isn’t easy; Goliath usually wins.
    CustomGuy: ICE9000 is on 28HP, T40 is on 28HPM. I’m not sure the processes are fully compatible and if they can’t copy-paste their baseband synthesis, it wouldn’t be ready in time to make sense. Furthermore, NVIDIA still needs to support Qualcomm basebands for CDMA/Verizon, and integration does slow you down slightly because of certification time. And then there’s the question of whether you want 21Mbps HSPA+ or 150Mbps LTE.
    It just doesn’t make sense on 28nm IMO. It’s much more logical to wait for 20nm where all of these considerations go away. And integration won’t save you any power consumption or board space at this point – it’s all about saving a little bit of money. I don’t think that’s the right priority for NVIDIA/Icera today.

  4. I thought Sirifics was acquired to provide the RF skills and ChipIdea was tapped to provide the analogue baseband design, Robert, so surely Icera had a fair level of integration – even if not in a single SOC.

  5. David,
    as far as I know Icera never had the systems focus or the analog /RF skills to do really do an integrated SOC phone chip. The other point is MTK grew with the market opportunity that they created, they didn’t simply take over a TI socket or anything as straight forward as that. For this reason MTK needed a huge FAE base and a larger applications / software team. So the MTK path was never really open to Icera.
    I personally believe that Icera should have dropped cell phone in 2007 and focused on MoCA rather than messing around with 3G and LTE for so long. IMHO Icera is a better company than Entropic but they were not focus on the hot markets.

  6. Yes, Robert, I always assumed it was the Shenzhai/Shanzhai who gave MediaTek its initial thrust because, as you say, the MP3 guys wanted someone to show them how to bend handset plastic round wireless chip-sets. With the Western handset makers shrinking to a handful this, it seemed to me, was the big baseband opportunity on the planet. MediaTek took it and built a $250m revenue business. But there was nothing that MediaTek did which Icera couldn’t have done.

  7. There seems to be some misunderstandings about Mediatek,
    To start with Mediatek did not really go after Chinese OEM’s rather they constructed a complete cell phone “reference design” chips, protocol stack, analog PM, application software …..everything. With this package they went after Games & MP3 manufactures who viewed the cell phone area with envy. These guys created the so called “white box phones” often with names like Nokic or Nckia …Locally they are known as Shanzhai means (bandits). The point is MTK delivered a complete phone package, that was ready to ramp to production. From these humble beginnings they ramped to about 600M baseband chipsets / year. OEM’s like ZTE, Huawei came after the little guys, next came the bigger traditional phone makers. As a matter of fact Nokia refused to even talk to MTK until they did something to stop supporting all the white brand China rip-offs. When MTK stopped supporting these Nokia ripoffs Spreadtrum got it’s big opportunity….
    I know a lot is said about the technical inferiority of the MTK chipset, but from my experience, this is not true, it might have been in the very early days, but not now.

  8. David: yes you are right, you can do analog for less, but its more difficult to scale and you sell out for less (witness equally underwhelming exits Nanotech and Phyworks). Like many in the UK industry, I was hoping that Icera would turn into the BBC – the British Broadcom Corporation – but it didn’t work out that way and it looks like we never will have a really big fabless semi now.
    Arun: you are probably right as well – but I wouldn’t be surprised if they integrate sooner. Icera’s solution is so damn small – it just makes economic sense. As an aside – if nVidia want a full portfolio of RF to go up against Qualcom & Broadcom they could well go after CSR…

  9. Thanks for the info Arun Demeure – I thought most of the largest manufacturers either had their own silicon or were in long-term joint development alliances with Broadcom, Qualcomm, Infineon etc.
    As you indicate in your point 1), there certainly appears to have been opportunities for partnerships betweeen Icera and at least some of the newer manufacturers over the years, but I guess my main point was that this didn’t happen. Possibly due to the tighter RF/baseband integration requirement (predictable as designs evolved and got smaller), price (much more sensitive in the emerging markets where much of the action is today) or Icera simply failing to convince the market they were up to it. However, you said earlier that Icera gave up on trying to get phone design wins a long time ago and only started pursuing these again in the last 9 months or so. Fine if their other business was enjoying explosive growth, but this apparently was never the case.
    On point 2), despite Icera’s and the so-called analysts’ predictions of market share/size, I maintain that HSDPA data cards and dongles were never going to take off. In my view there are a myriad of dependencies and barriers to use of these items, such as availability of phone tethering, alternative connectivity e.g. wifi and SMS etc, need for SIM swapping or an additional subscription, and simply something to plug them into – they’re just not a mass-market product. Data cards date back to the early 90’s and GSM – nobody wanted them then either! With all the other options for data connectivity on the move they were never going to be a volume item. I’ve always thought Icera were ridiculous to think they could build a billion dollar company on the back of those things.
    As David said, MediaTek appear to have been far more savvy. I presume they went after the Chinese/Korean OEM box shifters, and tapped into the huge volumes this kind of company deal in. If you’re an ambitious and expensive start-up as Icera was, you have to think in terms of either completely dominating an emerging market with compelling price-performance advantages, or quickly gaining a reasonable share of a high-volume existing market which is strong and growing. By doing this, you can then grow the company on the back of hard sales rather than hopes and dreams. Perhaps Icera will now exploit some of their unique IP and patents, and licence to or partner with some of the leaders in the next-generation products rather than yesterday’s dodos!

  10. CustomGuy: I’d just like to point out that you can start semi companies on $25 million or less if you start an analogue semi company which is often 2 or 3 guys using mature processes. And analogue is not a bad market to be in. Of course you’re right that if you’re going to spin a digital CMOS chip on a 35nm process you do need $50 million plus.

  11. Thanks, Arun, that’s good analysis – perceptive and informative and thanks very much for it.

  12. Hi David, I don’t think any of your basic facts were completely wrong 🙂 I was responding to ‘Anonymous’ three posts above me. Also, from all I’ve heard on both sides, they definitely want the team as well as the technology.
    CustomGuy: I don’t think it makes any sense to do Bluetooth/GPS/NFC/etc. on Icera’s baseband. The majority of the cost is in the RF so there would be no real advantage. And it’d take a lot of R&D dollars to make it competitive with competitors that have 10+ years of experience. WiFi would make some sense, but since it’s getting integrated into Bluetooth/GPS chips there’s not really any market for it.
    If Icera had IPOed, I think 60GHz would have been attractive in the long-term (20nm in 2015+). But it’s not that great a match with NVIDIA so I think they should just keep focusing on cellular, including LTE-Advanced and single-chip integration. Given my understanding of NVIDIA’s roadmap, I think it’s most likely they start integrating on 20nm (but not before).

  13. Let’s hope they want the team as well as the technology, CustomGuy

  14. Thanks Arun that’s a good insight and very helpful. Still don’t know which of my ‘basic facts’ were ‘completely wrong’ but I appreciate what you’ve told us.

  15. Anonymous: While I don’t disagree with everything you wrote, several of your basic facts are completely wrong.
    1) There are only two OEMs in the entire industry which had their own 3G basebands in the 65/40nm generation: Nokia and Motorola. Apple didn’t, Samsung didn’t, LG didn’t, HTC didn’t, ZTE didn’t, Huawei didn’t, even Sony Ericsson is not exclusively EMP/ST-E. And both Nokia and Motorola are moving away from proprietary basebands (e.g. 2011 smartphones/tablets from Motorola use the Qualcomm MDM6200/6600, Nokia was moving to multiple merchant solutions even before WP7).
    2) The data market was more than big enough to reach break-even and IPO. The problem is they completely failed to reach their market share goals. In April 2008, Stan Boland said “We expect to have 33 per cent market share in the data card space by the end of next year.”
    But Strategy Analytics claims that Icera themselves said they only shipped 3-4M basebands in 2009 which represents only 5-10% share. And the market was slightly smaller than analysts expected, and because of tablets and smartphone tethering it didn’t grow that much in 2010.
    If Icera did manage to have just 25% of the 2010 data-only market, they could have had a successful IPO already. But competing against Qualcomm isn’t easy. They simply didn’t succeed despite their technical advantages.

  16. Hindsight is a fine thing, but back in 2002 you wouldn’t (couldn’t) have picked a much better team to throw $250m at to build a UK semi-startup.
    We all know 10X $25m companies wouldn’t have panned out – semis need $50m+ and the successful few would have been less ambitious and would have achieved less technically.
    Yes, things could have played out a bit better – perhaps marketing mistakes were made – but the market moved the goalposts and they were up against an entrenched player that is well known for, how shall I put it, Intel-esque marketing practices.
    Europe now has a fantastic team working for a good company – if nVidia has any sense it will put money in so the soft modem can support WiFi, Bluetooth, NFC and any other wireless standard…

  17. I do agree [Anonymous} that it’s the engineering team who were the heroes in the Icera story creating a fantastic technology platform which is admired by everyone in the industry. The sad thing about this deal is that it is the engineering team who are likely to get screwed by it, while all those with stock will get paid out pretty well. As you say the one great talent the management had was in extracting money from VCs – when it came to extracting money from the market they didn’t have a clue whereas, during the time Icera operated, MediaTek, with inferior technology, built a business with US $250 million of revenue a year.

  18. I hope Nvidia throw out Mr Boland, his senior management team and Icera’s entire sales and marketing department for their appalling lack of vision and marketing ability. Their main achievement to date was in convincing the VCs to keep throwing good money after bad. HSDPA dongles and data cards were always a dead duck – the market’s far too small and niche. Using your mobile as a data modem connected to your laptop or whatever via bluetooth etc makes these items redundant in any event. As for handset design wins, the huge phone companies of course had their own chipsets thanks very much.
    Nvidia/Icera should focus on exploiting the key features of their core technology to address a much wider market, both fixed and wireless, by leveraging the numerous technical advantages of their platform. They have a great engineering team, who have been focusing on a tiny subset of their technology’s potential for too long.

  19. Anonymous, I wasn’t wearing my uniform when I wrote this.

  20. That’s what I fear too, Arun, that the VCs will breathe a sigh of relief that they’ve got their money out and say – let’s find a better racket than semis. I see in the US there’s a lot of VC money going into start-ups which target consumers, but very little going into start-ups whose customers are business customers.

  21. Almost, Djonne, but there’s Sequans of France (LTE and Wimax). picoChip of the UK (Wimax and Femto) and, if you say trade-links make Israel EU-ish there’s Altair Semiconductor

  22. So now, ST-Ericsson is the only EU remaining player in the mobile field, struggling to sell anything? :/

  23. Gosh No, Rupert, I’m honoured to think you have the same view – great minds . . . .etc

  24. As long as you don’t think I’m ripping you off. If I’dve been smart enough to read this first, I might have got some of the facts a little more right…
    The FT, now, that I’m ripping off.

  25. Actually David, Icera did have phone design wins at this point, and they even demonstrated a real ODM Android phone at MWC that they expected to be available in Q3 (iirc). But I don’t know if they had any major OEM design wins (which is where the most money is) and even if they had I’m pretty sure it wouldn’t have come out this year.
    More importantly, Icera stopped even trying to get phone design wins a long time ago, and only started again in the last 9+ months. They never went through full voice certifications at large carriers for their 65nm baseband, and they only bothered to get their voice implementation to a sufficiently low number of dropped calls in late 2010.
    Icera made a conscious choice to focus exclusively on data-only products starting sometime in 2008. And the problem is, not only has Qualcomm managed to hold on to more share than they expected despite their cost advantage (see: their complaint to the european comission), but much more importantly both USB dongles and netbooks have significantly underperformed analyst expectations in the last 1.5+ years. You can probably blame the iPad for some of that. And it got especially bad this year:
    So basically Icera bet the farm on a niche, and the niche didn’t deliver. With 20/20 hindsight, they probably should have asked VCs for more money earlier, and attacked the smartphone market aggressively with their 65nm baseband in 2008/2009 instead of focusing only on data products. Then again maybe OEMs still wouldn’t have given them any major design wins out of fear they weren’t big enough to deliver.
    Either way it’s pretty sad that they only got $50M more than Beceem despite the latter simply licensing CEVA-XC for their LTE/WiMax chip. I think in retrospect it might have been better to be more aggressive and get more money earlier on rather than a steady trickle. But I doubt VCs would have approved and I don’t even know if it would have been enough.
    It could have turned out worse, but it’s still not the best outcome for the semiconductor industry. Yes, those VCs have got a lot of cash in hand now, but I bet they’re going to invest most of it in non-semiconductor startups.

  26. Have just read your piece, Rupert, excellent stuff. Here’s the link for anyone who wants to read something sensible about this episode:

  27. Thank Goodness you did, Rupert, I’ve been beginning to think I must be wrong-headed after reading multiple pieces about what a great deal it is.

  28. Hah. That’s embarrassing. I seem to have written much the same piece as you, only a few hours later…

  29. I suppose for us engineers in the Bristol area this means one more US employer to run to if things go awry… ahem, I mean we’re delighted that NVidia is expanding further the fabric of Bristol’s hi-tech eco-system… 😉

  30. You’re right, cheese, one start-up doesn’t matter one way or the other but a start-up with $250m of VC money in an economy which puts in barely $1bn of VC money a year, is going to get a lot of attention. VCs aren’t enamoured of semiconductor start-ups at the moment – they’re too expensive and the time to exit is too long. So this disproportionately-funded Icera start-up’s relative failure may have a disproportionately negative effect on investor sentiment. I very much hope not.

  31. I dont quite understand why a country (or a region) be judged on the basis of one start up. If I were a VC (how I wish!), I would look at not just Icera, but also ARM and Imagination. There are all kinds of companies with a “bell curve” of track record. So yes, if Icera is not the icing on the cake, so be it. The valley has more dead companies than Chinese takeaways. And man, they have a lot of the latter :-).

  32. Some did, Anonymous, or how else could Icera have got started nine years ago? But however late in the game investments were made, none of them would be regarded as successful by the crteria by which VCs judge these things. For them a success is a 10x return in five years.

  33. Well i doubt they put their $250M in nine years ago :), i would assume that was spread over the period, maybe not as bad as you suggest

  34. You’re right, Roger, it’s a better result than the VCs losing their money. However putting on a bet of $250 million to get back $360 million nine years later may not encourage the VCs to repeat the process.

  35. Roger Shepherd

    you say “The Icera story means that it will now be harder for UK start-ups companies to raise cash with VCs becoming more wary of giving backing to UK engineer-entrepreneurs.”
    I say “thank God the VCs have got some money back and are in a position to start funding again”

Leave a Reply

Your email address will not be published. Required fields are marked *