With theory (1) (2) finished - it's time to move on to practice. This article is about what microelectronic factories in Russia (and Belarus) exist and what they can do, as well as about the largest developers of the microchips themselves, and what working results they managed to achieve.

Under the cut is a dramatic story about plastic electronics, the Russian “humanitarian aid” familiar to many AMD companies and Metro tickets.

Well and in addition - a graph showing how the state of Russian microelectronics has changed over the past 22 years compared with the United States and China.

Graph of the comparative state of microelectronics

The numbers of technological lag / advantages are built “optimistically”: how many years does the USA / China have the following technology. For example, we have 90nm since 2012 - the USA has the following technology (65nm) since 2006, respectively 2012-2006 + 1 = 7 years.

Blue graph - USA, red - China.

As I wrote in previous articles - in the race to satisfy the demand for iPhones, you can’t make money (if you are not a leader), so in the current situation there is nothing dramatic as long as the industry can produce critical products (and 90nm is more than enough for industrial and military / space electronics).

But it’s time for us to stop considering China a technologically backward country: in fact, they have been ahead of us in microelectronics for many years. At the moment, they have launched production of 45nm, and have a license for 32nm.

Microelectronics production

Sitronics - Micron

Until 2010, Mikron worked at about-micron standards, for the most part on equipment left over from Soviet times.

In 2006, technology and equipment for production with 180nm standards (CMOS + EEPROM) with aluminum metallization on 200mm diameter plates were purchased from STMicroelectronics. Then, on the basis of this equipment, SiGe technologies (needed for radio frequency applications - GLONASS, radars, etc.) and SOI (for radiation-resistant microcircuits) were developed (or purchased - there is no data). The volume of production is 1500 plates per month.

At the end of 2007, however, the ribbon was solemnly cut; the first test plates were in 2010. Such an early “ribbon cutting” is solely a matter of tax optimization.

There were reports that Micron was making RFID chips for metro tickets - however, having opened 30 metro tickets in the last half year, he found only NXP chips. A similar result was reached in August 2009 at Chipworks . However, I do not see any drama here, it was not for the metro that they bought this production. In Micron itself, my questions on this subject were left unanswered. However, one person seems lucky to see the Mikron chip 1.5 years ago.

Here’s what’s inside each metro ticket now: (size - 0.6x0.6 mm) Then, the same STMicroelectronics acquired 90nm copper-plated technology that requires additional equipment (mainly scanners) with an increase in production capacity of up to 3000 plates per month:




ASML PAS 5500 / 750F 248nm / 0.7 120 plates per hour
ASML PAS / 1150C 193nm / 0.75 135 plates per hour (Russia's first scanner at a wavelength of 193nm)

Rusnano was financing. In September 2012, they announced the beginning of sales of 90nm products. Thus, at the moment, Sitronics-Micron has the most modern and versatile semiconductor production in Russia, spending very modest money by world standards for it.

Mikron has a younger brother - VZPP-Mikron, in Voronezh - they make discrete elements there, and microcircuits based on micron standards.

NIISI RAS

Or as it is also called - "Kurchatnik". There is a super-small-scale production, the stepper - ASML PAS 5500 / 250C with a resolution of 350nm. Production volumes are a few plates per day. They are engaged only in military products, from the little known, developed and manufactured in Kurchatnik - Comdiv-64 .

Integral

For a long time, Belarusian Integral possessed 800nm ​​technology, but several years ago it was finally able to launch 350nm production on 200mm plates, with a production volume of 1000 plates per month. There is also a production of “clean” plates with a diameter of 200mm for the 350nm line.

On the new line - they seem to be engaged only in military microcircuits, but in the old giant production (150mm / 800nm ​​- 10 thousand square meters / month, 100mm / 1.5μm - 30 thousand square meters / month, 100mm / 2μm - 15 thousand square meters / month) they produce discrete components and simple chips for export.

Angstrom ("old")

On the old Angstrom - production with norms of 600 nm on plates with a diameter of 150 mm (8 thousand plates per month) and 100 mm, 1200 nm silicon-on-sapphire / silicon carbide (4 thousand plates per month).

Steppers - Cannon FPA-2000 i1 (wavelength 365nm, NA = 0.52, resolution ~ 0.5um).

Like other domestic factories, civilian products are exported (LED drivers, power management chips for cell phones, etc.).

Angstrom-T

Angstrom-T has already managed to generate a lot of drama - therefore, we dwell on it in more detail. The project is based on the building of an unfinished plant, which has remained since Soviet times. The case was built soundly, and met modern requirements of technology 130nm.

The equipment was bought from AMD Dresden plant (130nm, copper metallization in 9 layers, 14,000 plates per month), and the banquet was financed almost entirely from the VEB credit line with a limit of 815 million euros, which in principle is a unique case - microelectronic factories are rarely financed for credit score more than half. You can also read an article on the 2010 construction site - habrahabr.ru/post/80142

In 2008, AMD paid ~ 182.5 million euros for equipment, ~ 120 million euros - a technology license, ~ 10 million euros - to Meissner & Wurst for stage A of the plant design. However, the owner of the equipment was not Angstrom-T, but the Cypriot offshore company Runica. The equipment was also not taken to Russia, but left in warehouses in the Netherlands and Dresden, so VAT did not have to be paid either in Europe or in Russia. You can watch the video shot at that moment - tv.cnews.ru/?video_id=410

Then suddenly the financing stopped (the testimonies of the parties differ - the roofing felts are crisis, the roofing payments on the loan are overdue), the equipment remained abroad (now it is also necessary to pay for storage in special warehouses and insurance each year). In fact - the minimum possible amount spent to pay AMD was spent - and the project stopped there. By the way, as we recall, AMD just at that time separated its microelectronic production into a separate company (it ended in March 2009), and the company very successfully pushed out its outdated equipment at a very good price.

After the cessation of financing, there was no visible progress on the project, only mouse fuss (who insures whom, where, whom, change of owners and managers) and the accumulation of interest on a credit line. In 2012, funding was resumed (however, there were reports of resuming funding earlier), and at the end of October 2012, Angstrom-T was “officially” bought by Leonid Reiman. At the same time, IBM has licensed 90nm technology, which will require the purchase of additional equipment. Construction has been resumed, and production is now scheduled to begin in early 2014. Of course, the transition to the next technology will make it possible to somewhat compensate for the extra 5 years that the equipment lay in the warehouse.

It is worth noting that it is not possible to “upgrade” up to 65nm in this way - 65nm usually requires 300mm plates for economic reasons.

Counting the money (very approximately)
Let's “on fingers” roughly calculate how the financial future of the project might look in the most optimistic way: The

initial credit line for the construction and launch of the plant is ~ 815 million euros. Forget about already running interest since 2008.

Suppose IBM is not so greedy, and additional equipment with a 90nm license will cost only $ 100mln (for comparison, Sitronics paid EUR 27 million for a 90nm license, according to the press).
Then the amount to return is ~ 815 * 1.29 + 100 = $ 1,151 million, the rate is 8.5%

According to the initial business plan - the cost of materials, electricity, water and gases for the production of one plate, minus the cost of the masks (they are paid by the customer) are $ 408.75. These are our conditionally variable costs per plate, we only bear them if we have production.

The annual cost of labor with a traditionally low salary:
1098 (state size) * 12 * 30,000 = 395.2 million rubles per year.
We assume that we do not need to pay anything else (for example, taxes). These 395.2 million rubles are conditionally fixed expenses, we bear them regardless of how much production the plant has.

Graph of the average cost of finished plates with a diameter of 200 mm in the world market:


I think it will be fair to consider that the average cost of a finished plate 200mm 90nm in 2014 is about the same as 200mm 130nm in 2009. By the way, pay attention to the same price of plates 350-150nm - this is due to the fact that all market participants paid a loan for equipment and construction for a long time, and everyone works almost at cost (and labor and materials cost about the same).

Then, for the year, the plant, after deducting the most basic expenses, will receive:
14000 * 12 * 1001 (revenue at maximum load) -395.2 million rubles / 31 (conditionally fixed expenses) -14000 * 12 * 408.75 (conditionally variable expenses per plate) ~ 86.75 million $
And only interest on the loan at a rate of 8.5% will be 1151 * 0.085 ~ 97.8 million $
Thus, even in greenhouse conditions, and 100% load - the plant will not be able to pay even interest.

When writing off a 50% loan — a loan under the same greenhouse conditions can be settled for 13 years, but in reality — there will be enough complications (for example, further reduction in the cost of plates over time, incomplete loading of the plant, etc.)

How did the project initially receive financing with such prospects ? The business plan was based on an optimistic assessment of the cost of plates of the order of $ 2000-2500 + some administrative resource (since the project was still important for the development of the Russian electronic industry).

Thus, without writing off a loan, nationalization, annihilation of the eurozone, massive multi-year subsidies in any form (at the level of $ 100 million a year) or a nuclear war in Taiwan and South Korea, the plant will not be able to work. But surely Reiman knows more than we do, once officially got involved in this matter - it remains to wait for 2014-2015 and see what happens.

Other

In addition, there are smaller enterprises with production thicker than a micron: Exciton , NPO IT (for space), Voskhod-KRLZ , production of photomasks - TsKP PKIF and others.

Related industries caught in the news

Plastic logic

The company was engaged in e-ink displays built on the basis of "plastic" transistors. Of the advantages are their flexibility and potential ease of manufacture, but the electrical parameters of plastic transistors are orders of magnitude worse than silicon ones (therefore, no revolution and killing of "outdated" silicon electronics is not expected).

At the end of 2010, Rusnano invested in Plastic Logic, subject to the construction of a factory for the production of reading screens in Zelenograd.

October 11, 2011 Interfax: RUSNANO increased its stake in Plastic Logic to 43.8% from 33.7%
[...] Plastic Logic valuation has not changed, the invested funds will go completely to the construction of a plant in Russia, RUSNANO reported.

The construction of a plant for the production of non-silicon displays in Zelenograd is scheduled to begin in December this year. The company will produce about 100 thousand displays per month. [...]

At the beginning of 2012, the company suddenly decided that the moment was lost and it was no longer profitable to build a plant . The company will now only deal with the development and licensing of technology (however, who now needs this licensing if the market is littered).

Epic fail.

Crocus nano electronics

May 17, 2011 RUSNANO Press Center: RUSNANO and Crocus Technology set up MRAM production - for the first time in the world using 90 and 65 nm technology
RUSNANO and Crocus Technology, a leading developer of magnetoresistive memory, today announced the conclusion of an investment agreement on the establishment of the next generation MRAM memory in Russia . The total transaction volume is $ 300 million. Under the agreement, RUSNANO and Crocus create Crocus Nano Electronics (CNE), which will build a medium and high density MRAM memory plant in Russia with design standards of 90 and 65 nm using Thermally Assisted technology (Russia). Switching - TAS) developed by Crocus.

However, this is not a full-fledged 65nm production ($ 300m is a penny for a full-fledged 65nm production) - they will take ready-made CMOS plates (with transistors and metallization), and apply a layer for magnetoresistive memory. We will wait for 2013 and see what happens.

GS Nanotech

09/25/2012 RIA Novosti: The center for the production of microelectronics components will start operating in the fall near Kaliningrad
The first large-scale enterprise in the Russian Federation for the production of components of microelectronics with a topology of 45 nanometers or less, the GS Nanotech Center , located in the Kaliningrad Region, will enter service in late autumn, reporters, production center director Alexei Yartsev.

If you look at their site , you can immediately see that the company is only engaged in cutting plates, packaging and testing chips. Those. at the input - finished plates, at the output - microcircuits. This is undoubtedly also an important matter, but “45nm” - here, in general, does not belong to the enterprise in any way.

MAPPER Lithography

Rusnano invested € 40 million in maskless lithography technology. This, in my opinion, is Rusnano's most delicious investment. Maskless lithography is a future revolution in small-scale production of microelectronics (goodbye FPGA), and it is especially useful for the production of military microelectronics: now for the production of 10 microcircuits, you don’t have to order a set of masks for $ 150-500 thousand, or wait for a dozen “neighboring” projects to printed on one mask and one production technology.

The essence of MAPPER technology is exposure of an electron resist by an electron beam. This has been done before, but now - the rays are not 1, but 13'260. Thus, one installation will be able to exhibit from 1 to 10 plates per hour, which is quite enough for many small-scale products.

Microelectronics Development

I note the most interesting, in my opinion, domestic developments that are not covered by a veil of secrecy:

ICST

MTsST-R500S (500MHz SPARC 2 cores)
Elbrus-S (500MHz, its own architecture)
MTsST-R1000 (1GHz SPARC, 4 cores)
ELBRUS-2C + (500MHz 2 Elbrus cores, 4 DSP cores)

Elvis

1288HK1T (MF-01) - 4-channel SDR receiver
1892VM5YA - 1.2GFLOP DSP (32-bit)
1892VM10YA - 4 GFLOP DSP (32-bit) with GLONASS support

Milander

1986BE92U / MDR32F9Q2I ARM Cortex-M3 128kb flash / 32kb SRAM, 80Mhz
K1986VE92U / MDR32F2QI ARM Cortex-M3 128kb flash / 16kb SRAM, 36Mhz

STC Module

1879BM3 - 4 8-bit DACs, 300 MHz, 2 6-bit ADCs 600 MHz, 150 MHz processor core and external memory interface
NM6406 - 300 MHz DSP (2 32-bit MAC per clock)

KM211

K5512BP1F (Quark) - 150MHz 32-bit RISC processor, manufactured on a Micron 180nm

Summary

In recent years, Russian microelectronics has come to life significantly - both in terms of production (90 and 180 nm on Mikron, 350 nm on Integral), and development.

Now it is already possible to develop and produce absolutely any industrial, military and space products - processors, microcontrollers, power electronics, radiation-resistant microcircuits, microcircuits for radio communications, GLONASS and radar. But as we remember, development is an expensive pleasure, so you won’t be able to do everything right away.

There is no “obsolete” production in microelectronics; products by any standards have their own market and price. Even Soviet factories still work for export - releasing power microcircuits, discrete elements (power transistors and diodes) and other small things - and then we buy them in motherboards and cell phones under foreign trademarks.