Electronics Weekly Electronics Design & Components Tech News
Intel would have to borrow to buy Altera. Its $14 billion in cash is almost matched by its $12 billion of debt. It would have to pay North of $13 billion for Altera which has $2.3 billion in cash.
The only logical reason for the move would be to boost Intel’s server and networking business.
Altera’s server strategy has been to position the FPGA between the flexible but power-hungry CPU and the efficient but inflexible ASSP/ASIC. The FPGA delivers performance, power efficiency and flexibility, and flexibility delivers differentiation.
Intel is packaging FPGAs with CPUs to deliver reconfigurability which allows customers to customise their server designs. And with big server buyers increasingly wanting to dictate the specifications of their servers, the CPU/FPGA mix will play to the market trend.
Intel’s dominance in servers is coming under threat from ARM which said recently that it expects 20% of the server chip market by 2020.
So buying Altera could be a move to defend its already dominant server position and help it expand into networking markets.
Having just announced that revenues could be flat this year, Intel may also be looking for growth, though paying $13 billion to acquire $2 billion in extra revenues is an expensive way to get it and, if Intel borrows the lot, the deal will leave Intel with total debt of $25 billion and net debt of $11 billion.
Ironically, Altera licensed its programmable logic technology to Intel in 1984. As with countless other Intel diversifications it was a failure because Intel positioned its programmable logic business as a computer business whereas Altera, Xilinx, Lattice et al positioned their programmable logic businesses as telecoms businesses.
In 1994 Intel sold the business to Altera.
What comes around goes around. In circles, it seems.
Actually one area this does make sense is in the basestations. If Intel can combine the knowledge from the Infineon team with that of Altera to offer a still programmable but lower power 4G pico-station solution they might even gain market share.
I concur, AnotherDavid, at the moment Altera is focussed on getting market share from Xilinx and Lattice. When/If Intel gets it the guys will be focussed on how to survive/prosper in the Intel firmament. Eye-off-the-ball syndrome big-time. Xilinx will do well.
I can’t imagine it would do much for the Altera market share in FPGAs – it will drive people to Xilinx, big time. But as it takes a long time to design these things in, it will take a couple of years to show.
David
Yes indeed John, but Intel has an annual capex budget of $10bn and an R&D budget of $11bn to fund. With existing debt of $12bn and another $12bn debt to take on to buy Altera and $14bn of cash, it might be able to ‘swing a larger deal’ but would it be wise?
Last year Intel’s operating income( profits before tax) was over $15B – they should be able to swing a much larger deal than both Altera and nVidia combined. Compare this to the NXP/FS deal and they have $40B in debt with a combined $2B in operating income last year. This deal is really about new markets and getting the ability to have programmable instructions on Intel processors to reduce power and improve performance in high performance computing applications. It increases Intels served available market by at least $50B as well (ASIC, FPGA) providing some growth story. It also gets Intel back in the ARM game as Altera has an ARM based product.
I couldn’t agree with you more Gary I don’t think anyone has figured out why Intel bought Mcafee or what good it has brought Intel in return for $7.6bn of Intel cash. Intel still has a very airy-fairy attitude towards money which is becoming inappropriate as its debts grow hugely and its main business, PC, shrinks.
It makes no sense (To me) for Intel to buy nVidia. The GPU business is stagnant/declining in lock-step with their CPU business. nVidia (like Intel) have been pretty unsuccessful with the apps processor. With Altera there are new markets that Intel can address, and it is a good fit with their technology. As David noted Intels track record on acquisitions has not been very good. They do have the benefit of some working relationship with the Altera-Intel foundry deal. Certainly a better business opportunity than the McAfee acquisition that still looks stupid to me.
I agree – I said they should buy Nvidia and not Altera.
Well buying Nvidia will be wishful thinking if they buy Altera, Mike, buying Altera for $12-13bn will push Intel’s debt up to $24-25bn and buying Nvidia for $12bn+ would push the debt up to $36-37bn and, even in an age when debt is taken on pretty lightly, I think Intel would balk at having $36bn of debt.
It’s a trifle premature to decide the outcome before the jury have pronounced, Mike, and everything has prior art but it’s the big leaps forward that are significant.
Regarding the GPUs, I would say they are a world-class process design company – by far the best in the business – but AMD proved years ago that their circuit design techniques were a little second rate. Or in the case of their GPUs – third rate !!
And I don’t have a PhD yet have told them how to fix all their problems in two words – BUY NVIDIA !!!
” I rest my case M’Lud.”
And hence lose the case. A leap forward inherently accepts that there was prior art to leap forward from 🙂
Also as I’ve said the Cypress device was actually on the market before the Altera one, but the Altera one had the better software and won in the end.
Well Keith I wonder if the reason why Intel GPUs are so awful is that no one among the top 15 execs has a PhD and so there’s no one at the top capable of figuring out how to fix the problem.
“Oh and I still think they should buy NVidia. They are slightly cheaper than Altera and would solve both Intel’s appalling graphics performance and lack of a world class CEO in one go.”
Agree with Mike here, Intel GPUs are awful, and their drivers even worse.. Whether they would get a better leader, I’m not qualified to comment on that.
Yes indeed Mike Intel needs a world-class GPU but why does it have to buy Nvidia to get one? Intel used to be an engineering company, why can’t it develop a world class GPU in-house?
Thanks, Stooriefit, it seems to me a big obligation to take on for a company with only $2bn net cash. If they screw this one up, as they have with previous acquisitions it will, as you say, be disastrous for Intel.
An interesting piece Mike, from which I quote: “The big leap forward occurred in 1984, when newly formed Altera Corporation (which was founded in 1983) introduced a CPLD based on a combination of CMOS and EPROM technologies.” I rest my case M’Lud.
Given this is TV debate season I’ll follow the agreed format and say “I agree with Gary”. I think Intel will have to change their spots somewhat and make sure Altera silicon doesn’t play 3rd fiddle to server and mobile chips on leading edge process or they are doomed from the outset. I suspect they will have over paid for it and it may hurt their capacity to get their older fabs up to date. If they make a mess of this one if could be a horror show, so I agree with David too!
Oh and I still think they should buy NVidia. They are slightly cheaper than Altera and would solve both Intel’s appalling graphics performance and lack of a world class CEO in one go.
Yes Cypress had PALs earlier, but I still had a Cypress CPLD on my desk before an Altera one. And the GE device was called a PLD and bore no relation to the PALs that came later. So sorry but I don’t agree with your ‘generally acknowledged to be first’ statement.
Interestingly I found this discussion by your competition Max at EETimes and most people there seem to side with Signetics as having the first PLD.
http://www.eetimes.com/author.asp?section_id=14&doc_id=1285398
It would make sense, Gary, if Intel had a record of successful management of companies it has acquired. Intel will likely pay $11-12 billion (having deducted Altera’s $2.3 billion cash) and that will most likely be borrowed giving Intel total debt of $24 billion (with cash of $14 billion). Altera has $2bn a year revenues with 60% margins. It might work but M&A typically doesn’t in semis and IMHO this is not the way most people run their financial affairs.
Well you can define PLD as a PAL, PROM and EPROM and say others invented it before Altera, Mike, but in the final form of a CMOS CPLD, Altera is generally acknowledged to be the first. Cypress called its early programmable logic devices PALs and they were based on flash erasable EPROM technology.
This makes sense to me for Intel. FPGAs represent the only significant high margin silicon beyond CPUs. With CPU volumes declining Intel needs additional (high margin) business. Foundry silicon is not going to move their needle, FPGA business will but I don’t know if the FPGA business is enough to offset the CPU decline. Would not want to be in Xilinx position if this deal is done though.
General Electric merged Intel’s floating gate UV erasable process to form a programmable sequential state machine in the 1970s long before Altera. Intel fabbed this for them for military use before selling it themselves.
And Altera weren’t even the first with a CMOS ePLD. I think Cypress were the first parts on my desk that actually worked.
But what Altera did do was realise you didn’t have to completely populate the logic array with programmable switches and with clever software most boolean trees could be created with a much sparser matrix, thereby saving a lot of transistors. This was the fundamental breakthrough that everyone else, including Intel, had to licence or break, and eventually led to the LUT that FPGAs are based upon.
I’m intrigued to hear you say that Mike because Altera has always been credited with inventing the PLD and it was founded in 83 and launched its first product in 84. Yet you’re saying that, in 84, Intel had an in-house developed PLD. I never heard Intel claim it invented the PLD and I rather think it would have done if it had.
That’s not completely accurate David. Intel had its own ePLDs in 1984 but Altera had patents which prevented them being evolved so they cross-licenced each other patents. I don’t believe any money went either way but Intel agreed to foundry Altera’s devices, as it (and TI) used to do long ago in the those days BT (Before TSMC 🙂
The Intel devices were indeed intended for computer use and found pretty high volume use on motherboards with the Altera and Xilinx devices mostly finding lower volume but profitable uses in military and IP routers respectively. It wasn’t until the explosion of mobile phone use that they both established themselves in basestations (Ericsson and Nokia – can’t remember who did which) that made them grow quickly into the companies we know today.