What’s The Deal With Atmel And Microchip?

It’s been nearly a year since Microchip acquired Atmel for $3.56 Billion. As with any merger, acquisition, or buyout, there has been concern and speculation over what will become of the Atmel catalog, the Microchip catalog, and Microchip’s strategy for the coming years.

For the Hackaday audience, this is a far more important issue than Intel’s acquisition of Altera, On Semi and Fairchild, and even Avago’s purchase of Broadcom in the largest semiconductor deal in history. The reason Microchip’s acquisition of Atmel is such an important issue is simply due to the fact the Hackaday community uses a lot of their parts. This was a holy war, and even changing the name of a line of chips to ‘MCMega’ would result in a consumer rebellion, or at least a lot of very annoying tweets.

For the record, I’ve tried my best to figure out what’s going on with Microchip’s acquisition of Atmel for the last few months. I’ve talked to a few Microchip reps, a few Atmel reps, and talked to a few ‘out of band’ connections – people who should know what’s going on but aren’t directly tied to either Atmel or Microchip. The best I’ve come up with is a strange silence. From my perspective, it seems like something is going on, but no one is saying anything.

Take the following with several grains of salt, but Microchip recently got in touch with me regarding their strategy following their Atmel acquisition. In a few thousand words, they outlined what’s going on in casa Microchip, and what will happen to the Atmel portfolio in the future.

Broad Strokes

In broad strokes, the Microchip PR team wanted to emphasize a few of the plans regarding their cores, software, and how Microchip parts are made obsolete. In simple, bullet point terms, this is what Microchip passed on to me, to pass on to you:

  • Microchip will continue their philosophy of customer-driven obsolescence. This has historically been true – Microchip does not EOL parts lightly, and the state of the art from 1995 is still, somewhere, in their catalog.
  • We plan to support both Atmel Studio 7 and MPLAB® X for the foreseeable future.
  • Microchip has never focused on “one core”, but rather on the whole solution providing “one platform.” This is also true. A year ago, Microchip had the MIPS-based PIC-32 cores, a few older PIC cores, and recently Microchip has released a few ARM cores. Atmel, likewise, has the family tree of 8 and 32-bit AVR cores and the ARM-based SAM cores.
  • We will continue to support and invest in growing our 8-bit PIC® and AVR MCU product families.


In addition to the broad strokes outlined above, Microchip also sent along a few questions and answers from Ganesh Moorthy, Microchip’s President and COO. These statements dig a little bit deeper into what’s in store for the Microchip and Atmel portfolios:

How will the 32-bit products complement each other? Atmel has a few 32-bit microcontrollers, like the SAM and AT32 series. Microchip has the PIC-32. The answer to this question is, “Many of the 32-bit MCU products are largely complementary because of their different strengths and focus.  For example, the SAM series has specific families targeting lower power consumption and 5 volts where PIC32 has families more optimally suited for audio and graphics solutions. We plan to continue investing in both SAM and PIC32 families of products.

Will Atmel’s START support 8-bit AVRs? “Yes, although it is too early to commit to any specific dates at this stage, we consider modern rapid prototyping tools, such as START and the MPLAB Code Configurator, strategic for the our customers to deliver innovative and competitive solutions in this fast-paced industry.”

Now that Microchip has a complete portfolio of low-power, inexpensive 32-bit microcontrollers, will the focus on 8-bit product be inevitably reduced? No, we see that in actual embedded control applications there is still a large demand for the type of qualities that are uniquely provided by an 8-bit product such as: ease-of-use, 5V operation, robustness, noise immunity, real-time performance, long endurance, integration of analog and digital peripherals, extremely low-static power consumption and more. We don’t think that the number of bits is an appropriate / sufficient way to classify a complex product such as the modern microcontroller. We believe that having the right peripherals is actually what matters most.”

Security, Memories, WiFi, and Analog products. For both Atmel and Microchip, the most visible products in each of their portfolios is the lineup of microcontrollers. This isn’t the limit of their portfolios, though: Atmel has space-grade memories, Microchip has some very useful networking chips, and both companies have a number of security and crypto chips. In the statements given by Moorthy, very little will change. The reason for this is the relative lack of overlap in these devices. Even in segments where there is significant overlap, no EOLs are planned, circling back to the, “philosophy of customer-driven obsolescence.” In other words, if people keep buying it, it’s not going away.

The Takeaway

What is the future of Microchip post-Atmel acquisition? From what I’m seeing, not much. Microchip is falling back on their philosophy of ‘customer-driven obsolescence’. What does that mean? Any non-biased assessment of Microchip’s EOL policy is extremely generous. The chip found in the Basic Stamp 1, from 1993, is still available. It’s not recommended for new designs, but you can still buy it. That’s impressive any way you look at it.

The one thing we’re not getting out of this pseudo press release is information about what Atmel will be called in a few years. Will the Atmel mark be subsumed by a gigantic letter ‘M’? Will the company retain two different trademarks? There is no public information about this.

Yes, I know this post is a nearly verbatim copy of a pseudo press release. I’m not particularly happy this information was presented to me this way, but then again, the Atmel/Microchip ecosystem has been impressively secretive. This is the only information that exists, though, and I’m glad to have it in any event.

That said, there are a lot of people in the Hackaday community that want to know what the deal is with Microchip and Atmel. Short of pulling Jerry Seinfeld out of retirement, this is the best we’re going to get for now. Of course, if you have any info or speculation, the comments below are wide open.

64 thoughts on “What’s The Deal With Atmel And Microchip?

    1. I did not try to compile the Duinomite BASIC specifically, but in my experience compiling on Linux is without problems. I’m on Linux exclusively, and MPLABX does run fine (even with PIC32 projects).
      (For Duinomite-BASIC: the project requires using the obsolete C32 v1.11b compiler. It also uses wrong upper-case for file names. So it seems whoever created the project was using a Windows system only. This is not Microchips fault…)

      1. I just hope they don’t pull some dick move like Microchip who also use GCC, but make you pay for their fork or compile it yourself (which they make as difficult as possible). The free version is crippled.

          1. GPL says they have to provide the code to anyone who gets it from them. If they only sell it, then the code does not need to be given away to anyone but customers. Now, those customers could just repost the GPL code for others.

  1. Microchip does not help open source projects at all. In fact, they sell old versions of GCC for $800 because they keep the build arguments secret. It’s time to abandon Atmel and AVR because they are going to same way.

      1. Prices of Atmel parts have rocketed already.

        I worry that they aren’t going to be developing some of Atmel’s best lines like XMEGA too. XMEGA is a great platform, I’ve used it in massive quantities in a few commercial projects over the past five years.

    1. Agreed. It’s only one data point but it has had a big effect for me–ATtiny88s nearly doubled in price over the last year. Consolidation inevitably increases prices and that’s not a good thing.

      1. I am not sure if they are supposed to apply in this case :-( There are several other MCU manufacturers. ST with STM8 and STM32, Infineon, Renesas and NXP. Of course some of them are not interested in the hobbyist market.

    2. Don’t know how it is possible, but one major EU supplier has ATmega328p approximately 40% cheaper than it was one or two years ago – and AFAIK cheaper than ever before. ATmega128a is cca 30% and ATmega2560 20% cheaper. I compared it against old invoices from the same supplier right now.

    1. I dunno, he may be, or may not be, thing is when you tell people the plain unvarnished facts of life, they might not want to hear it. 5 years from now, might be able to look back, pore over the accounts and figure if it was truth or bluff. However, in general, if there’s X dollars available total for something that means it only gives 0.5 the desired result, and you get lawyers and government involved, that means you can kiss goodbye to somewhere between 0.7 and 0.9 of X, and end up with a range of 0.15 to 0.05 of the desired result…. an even derpier move (Unions tend to like this one, I have no idea why..) is to force a company that’s adequately servicing it’s debt payments, and providing ongoing employment and economic activity, but has assets and reserves only about equal to debt, into bankruptcy, whereupon the lenders and lawyers get paid first, it’s shut down and there’s nothing left. Anyway, sounds to me like that’s what he was going on about with the burning house thing, facts of life.

      Now for example you have 2 types of people, get one of each and say, “Here’s a plate of good food, if you eat it right now it’s fresh and nourishing, if you leave it sitting there two weeks and then eat it, you may get sick and die from conditions including but not limited to botulism, listeria, salmonella, fungal toxins or other biological factors.” …. and the one guy will say “Heh, thanks.” and nosh down.. and the other guy will be all “OMG you totally tried to blackmail me into eating this food by threatening to murder me with poison!!!!! FFFFFFffffFFFFFF This isn’t okay!!! POLICE!!! LAWYERS!!!!!”

    2. The CEO of Microchip was simply put in a bad position in the terms of the benefits. What actually happened:

      1) Atmel promised benefits in case of layoffs after an acquisition (when Dialog merger was announced).
      2) Those benefits were unusually large.
      3) Those benefits EXPIRED in case the acquisition happened after a certain date.
      4) The Microchip acquisition happened AFTER that certain date.

      These facts put Sanghi in a bad position, and I think that he did the best that he could. I have a lot more trust in Microchip leadership than I ever did in Atmel.

      1. agreed. MC have a pretty good reputation (except here) if people continue using atmel products, they’re going to continue to supply them with better price and availability than existed before. if atmel staff do leave it will affect development of new products but not existing ones

    3. “No support sounds like a recipe for a slow death of products”. For me the only thing of value in atTiny,atMega and atXmega is the core. As CEO says: “We believe that having the right peripherals is actually what matters most.” If one compare the rich set of peripherals of 8 bits PIC to atTiny,atMega and atXmega, the good move should be to replace the PIC outdated core with the AVR core and keep the rich set of PIC peripherals.

  2. The one question this doesn’t answer is what will happen to the base tool chains.
    Will we see the Atmel parts move towards what the Microchip parts have (where you can either use the free gimped GCC compiler and libraries, the for-pay full-featured GCC compiler and libraries or a self-built GCC that gives you no license at all for the libraries) or will we see Microchip open up and stop charging extra for a full-featured GCC on these parts?

    Or will we see the new entity keep the same policies on both (gimped tools on the PIC parts and open tools on the Atmel parts)

    1. Microchip does not control the source code for gcc-avr, or avr-libc. Those will continue to be available for no cost. In the long term, I believe the 8 bit PIC architecture will be EOL’d or not recommended for new designs. The reason is lack of support for open tools (the gimped GCC compiler doesn’t help things), and the inefficiencies of the 8 bit PIC instruction set (Banks, lack of multiple and divide, C-compiler hostile instruction set, etc.). The 8 bit marketplace is still viable. We just need more powerful microcontrollers in that space. I’d like to see some 8 bit AVR’s refreshed to run at 3.3V and 2.5V and have clock speeds in the 10’s MHz.

      1. I for one don’t give a shit if the cpu is driven by a shit pic core, an uber efficient Atmel core or a fucking pixie riding a unicorn, that’s what my compiler is for, my product selection will be by peripheral set, price and availability. Bottom line I will EOL before the PIC 8 does and the ATMega is much more likely to be around for a long time now which is good for us as we have two products using them.

  3. I finally got with PICs (mostly 8-bit) about 9 years ago. I struggled with MPLAB and had some success… but things really took off for me when I tried sourceboost (C compilers, so-so IDE and some really nice emulation plugins). The pace of updating has been glacial, might almost be abandonware, but sourceboost’s free C compiler level plus the $20 plugins has met all my PIC 8-bit development needs so far.

    About a year ago, I finally tried out the Arduino ecosystem, as part of my playing with ESP8266. And wow… with all the libraries and their ‘harmonization’ of the build/deploy process to so many different boards… pretty addictive. Add to that the availability of useful form-factors at low low price (eg a Nano clone for $3). I’m now using the vMicro Arduino plugin for Visual Studio to get a better IDE. All-in-all, Arduino is a much friendlier and more active universe for embedded processer learning and development. By comparison, getting into PICs was a slog. I think I know where most hobbyists/hackers will be found.

    PIC now has MPLAB Xpress (online IDE) and a little board to go along with it… I’ve got one, but again, Arduino is still friendlier and cheaper.

    Now I’m not a power-user or a commercial developer. I suppose MicroChip is getting most of their sales from commercial users.

    1. “I suppose MicroChip is getting most of their sales from commercial users.”
      It is the case of all MCU manufacturers. Whatever the success of Arduino Atmel didn’t made their buisiness with it.

      How many electronic thermometers, remote control, etc are sold every year? a lot more than any Arduino board.
      Those who think that 8 bits MCU will be dead soon don’t understand the market. All those small electronic devices sold by million every year is where the 8 bits MCU market is, not the hobbyist/maker/hacker.

      quiz: how MCU there is in a car today?

      “It is estimated that today’s well-equipped automobile uses more than 50 microcontroller units (MCUs).”

      source: http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6004783

      1. All good points. And i still love my 8-bit PICs.

        I wouldn’t be surprised though, if designs that start out as Arduino prototypes start showing up in finished products as well. With proper testing for function and safety, of course… On that point, I have to say that some of the MCU designs in our appliances are downright f#@king embarrassing (google “dryer E68 error” Grrrrr.)

        I hope MicroChip continues in their efforts to make their entry-level stuff easier and more accessible.

        1. Stuck button error code?

          Check out the software requirements for making a dryer. EN/UL-60335, Annex R.

          Good luck with Arduino getting the necessary documentation, tool-chain relics and compliance with the product safety class required for the software “component.”

          As more and more products get swept into 60730 and 60335, the scope of what you can sell without at least Class B libraries will get smaller and smaller.

  4. I, for one, welcome our new microcontroller overlords. That being said, I’m a dyed-in-the-wool PIC user and have been since 1997 or so. What I expect to see more of (and which I am already seeing some of) is cross pollination of peripherals. What Microchip did right with the PIC32 line was introducing the concept of all peripheral registers having a normal r/w persona and three aliases for set, clear, and invert. That allows for bit-twiddling of peripherals on instruction sets that don’t provide those operations natively.
    This means you can harness any core to such peripheral devices and have a working microcontroller. I would like to see that model applied to the best peripherals from both IP portfolios and then see those available on a range of cores for the various niche needs as well as general purpose embedded smarts. This is what Microchip had excelled at over the years and I expect them to continue. They have already started this with the low cost PIC32MM motor control / power control MCUs which have the three-phase (six output) smart PWM in a $1.50 (quantity one) micro. If I am not mistaken that peripheral came from Atmel originally.

  5. While START is certainly a good idea, I found it unable to even load many examples using the SAMD21 There fail on missing dependencies. The USB examples are not functional at all. Look like there don’t have a script that automatically test all the example before releasing it on the web.

    1. >Meanwhile I can buy the entire Arduino Uno clone from aliexpress for about the same cost. Strange times
      But you don’t know what chip you get. Might be a bad clone, might be an original but off-spec recovered from the bin, …

  6. I’d like to point out that Microchip’s purchase of SST … those parts are still exclusively badged “SST”.
    Microchip’s purchase of HI-TECH … the compiler is still badged “HI-TECH C”

    I wouldn’t be surprised if “changing the branding” is simply not something they’re thinking about now, nor anywhere in the nearish future.

  7. 8-bit microcontrollers are old and antiquated. 32-bit microcontrollers are modern, offer more performance and ease of programming.

    With the exception of the AVR all other 8-bit architectures are not optimized for C programming. 2K of RAM even on AVRs is not enough to program in C comfortably. The Cortex-m0 core, costs less than 8-bit cores, exhibits similar or less power dissipation, improved performance and ease of programming in C/C++.

    Microchip 8-bit microcontrollers are exceptionally horrible on the tools, architecture and performance fronts. They have decent peripherals but nothing to write home about. The PIC32/dsPIC/PIC24 families are much better options in my mind and I’d use them if Microchip didn’t cripple the GCC toolchain optimization for these devices.

    The AVRs are still cool if you’re doing basic arduino stuff. For everything else I’d use Atmel cortex-m0 cores, TI’s TIVA C cortex m4s, or anything by STM32s. You can get STM32 Cortex-m0 in really tiny 20 pin SSOPS and 0.8mm 32-TQFPs packaging. They’re cheap, powerful easy to program in C and come with impressive peripherals.

    As far as I’m concerned Microchip and to a lesser extent Atmel are becoming increasingly less relevant in the Embedded industry, This becomes even more apparent when taking a good look at how many design wins the STM32 is taking as well as the rising popularity of the ESP32 and ESP8266 chips in the hobbyist/Maker arena

    1. You have never worked for a company period. That has turned a profit period. I have worked with companies that allow sloppy stupid engineers to use a 32 bit micro and we come in and put in a 16 bit, or you use a 16 bit and we put in in an eight. There are still products that use 555’s for things that weren’t supposed to be. Be real, get real and get a job.

      1. Since we’re going to start making assumptions about each other, most people who still advocate 8 bit/16-bit over 32-bit are outdated and keep clinging onto how things were done in the past because they’re unwilling/ unable or fear change. This is stupid and does not inspire. The 1990s called, they want their crappy 8/16 bit microcontrollers back (especially the ones from Microchhip).

        Even if i have to use an 8-bit microcontroller…I wouldn’t touch Microchip products PIC16/PIC18 with a 10-foot pole. Because let’s face it….most PIC16/PIC18 microcontrollers are GARBAGE. They’ve been garbage for at least 20 years now. They’re inferior to AVRs and always have been core wise. Some of their peripherals are better, bit so what. Most cortex-M0 have peripherals that blow the PIC16/PIC18 peripherals out of the water.

        The fact is that 32-bit micros come in so many varieties low pin count, low memory count, low power, low cost e.t.c32-bit can be lean, mean and low cost….much lower cost than the 30 year old crappy PIC18/PIC16. They’re also easier to program and have decent open source tools that can be acquired for free. Unlike Microchip’s cheap attempt at making its own 16/32 bit compilers by ripping off open source toolchains and crippling their optimizations.

        Oh well I guess we won’t be seeing eye to eye on this matter anytime soon.

        1. Most people working with 8-bit micro’s these days are just whipping up something modest, some digital dirt to replace a handful of CMOS, fart some serial, and put chars on a LCD display. Oh and blink a LED. 32 bit MCUs are lovely, but you don’t need a Cray to read a DS18B20 I must disagree on the C programmability of PIC16 stuff… they work for me, and maybe a few others. It would be interesting to compare the reliability of some venerable 8-bit parts vs some new low-pin-count 3.3v 32bit jewels in the rougher applications.

          I will continue to disagree until my last tube of PICs are used up… then I will try something newer, then jump onto HaD to tell you that THESE are the cat’s nuts :-)

    2. Nice opinion there.

      STM can’t seem to make a lot of money, on any of their products. That’s a big problem.

      “As far as I’m concerned Microchip and to a lesser extent Atmel are becoming increasingly less relevant in the Embedded industry, This becomes even more apparent when taking a good look at how many design wins the STM32 is taking as well as the rising popularity of the ESP32 and ESP8266 chips in the hobbyist/Maker arena”

      I don’t know of a serious business that is considering ESP32 or ES8266 in their products, only the maker / kickstarter crowd.

      STM gets design wins because their parts are the lowest cost with reasonable support, right now. I would know, I design them in over other vendors unless the customer requires me to use a particular microcontroller.

      I don’t know how long a company can have negative revenue year-over-year and still survive, but ST is running that experiment.

      Microchip’s issue now is that they are making and supporting a variety of heterogeneous architectures. When I asked the microchip reps this they don’t seem to know how it will all shake out. Microchip’s claim to fame was that they never obsoleted parts (except for failed PIC17.)

      The 32 bit micros may be easier to use but the lowest cost from ST seems to be STM8S003 series. It also has real EEPROM with a lot of write cycles. I2C EEPROM is cheap, but that’s an additional part and pins. STM32L series has real EEPROM but costs a whole lot more than the STM32F or STM8.

      The Flash rewrite life on the STM32F series is comically low. like 2K / 10K writes for many of the cheap parts.

      But whatever floats your boat.

      I could see like TI buying ST just for their MEMs stuff.

      1. You got to the point. At least all european union citizens are paying tge negative revenue of ST with their taxes and financing this system. So far Draghi is putting lot of printed money in this business but guess what will happen if this one day stops? For the non-european users here, buying st component also currently sponsors this kind of business.

    3. Then riddle me this, o wise one, why there are 4-bit microcontrollers still in use?
      Also how much computational power you need to run a fridge, washing machine or vending machine? How much MIPS you would need to make car windows that close automatically when you lock the car? 8-bit micros are here to stay, and you are wrong. 8-bit microcontrollers are in use for 44 years now, and I’m expecting them to stay for another twenty years at least. And as prices of more advanced microcontrollers drop, prices of 8-bit ones drop too, and they keep gaining better and more advanced peripherals…
      Read this:

      1. It’s not necessarily about processing speed it’s also about time to market. A 32-bit can comfortably be programmed in C with vendor provided peripheral libraries. The combination of both of these means than designs can get of the ground running quickly and effectively. The final program/firmware may not be as optimized as if it were written on a 4-bit microcontroller in assembly, but it works, its tested and is compliant with the necessary standards. It also took a significantly shorter time to develop than writing assembly for an 8-bit/4-bit microcontroller.

        Also please name me one 4 bit microcontroller. I’ve been working on and off with microcontrollers for 20 years never ever came across a 4-bit microcontroller. Besides I thought that many of the systems required to do simple things like control a vending machine or open a car window were made in China anyways nowadays.

        I never said that 8-bits are dead….but they’re definitely in decline and for good reason…..because 32-bit micros are so much better. It’s not about the MIPS, but rather overall convenience, cost and time to market. Anyone starting in Microcontrollers today should focus primarily on 32-bit micros.

        There’s a reason why no one uses the 6502 anymore….Heck when was the last time anyone here about a new product utilizing the 68HC11? or even the 32-bitter 68332? These microcontrollers/microoprocessors are all very old and not used in new designs because better options out there; primarily cortex-m0/m3/m4 micros. Despite their age (40 years+) all three actually have well designed cores that are superior to the PIC16/PIC18 core.

        1. The 68HC11 can be setup to use external data/program memory thru a bus, I miss that sort of funcionality.
          Does anybody knows if there are any newer parts (from any semiconductor manufacturer) with that functionality?

        2. “””There’s a reason why no one uses the 6502 anymore….Heck when was the last time anyone here about a new product utilizing the 68HC11?”””
          – extra support chips required, that’s why. Modern microcontrollers don’t need external components.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s