@Convert Du kannst Dir gerne einreden, dass IBM ein armes Opfer war. IBM und Global Foundries haben sich aus gutem Grund gütlich geeinigt. Es wäre eine Schlammschlacht geworden, bei der sich beide Seiten beschädigt hätten. Und dieser Schaden hätte in keinem Verhältnis zu einem möglichen Gewinn aus dem Prozess gestanden.
- Wie das Missverständnis begann
- Enttäuschte Erwartungen
- Keine Kunden für 7 nm
- Fast fertig?
- AMD hatte nie vor 7 nm bei GF fertigen zu lassen
Wie das Missverständnis begann
https://investors.gf.com/news-relea...ndries-acquire-ibms-microelectronics-business
GLOBALFOUNDRIES TO ACQUIRE IBM'S MICROELECTRONICS BUSINESS
October 20, 2014
PDF Version
Acquisition Enables GLOBALFOUNDRIES to Become a World Leader in Semiconductor Foundry Technology;
IBM to Focus on Fundamental Semiconductor Research and Systems Innovation
ARMONK, N.Y., and SANTA CLARA, Calif., October 20, 2014: IBM (NYSE:
IBM) and GF today announced that they have signed a Definitive Agreement under which GF plans to acquire IBM's global commercial semiconductor technology business, including intellectual property, world-class technologists and technologies related to IBM Microelectronics, subject to completion of applicable regulatory reviews.
GF will also become IBM's exclusive server processor semiconductor technology provider for 22 nanometer (nm), 14nm and 10nm semiconductors for the next 10 years.
The Agreement, once closed, enables IBM to further focus on fundamental semiconductor research and the development of future cloud, mobile, big data analytics, and secure transaction-optimized systems. IBM continues its previously announced $3 billion investment over five years for semiconductor technology research to lead in the next generation of computing.
GF will have primary access to the research that results from this investment through joint collaboration at the Colleges of Nanoscale Science and Engineering (CNSE), SUNY Polytechnic Institute, in Albany, N.Y.
As part of this Agreement, GF will gain substantial intellectual property including thousands of patents, making GF the holder of one of the largest semiconductor patent portfolios in the world. GF also will benefit from an influx of one of the best technical teams in the semiconductor industry, which will solidify its path to advanced process geometries at 10nm and below. Additionally, the acquisition opens up business opportunities in industry-leading radio frequency (RF) and specialty technologies and ASIC design capabilities.
"This acquisition solidifies GF' leadership position in semiconductor technology development and manufacturing,” said Dr. Sanjay Jha, CEO, GF. “We can now offer our customers a broader range of differentiated leading-edge 3D transistor and RF technologies, and we will also improve our design ecosystem to accelerate time-to-revenue for our customers. This acquisition further strengthens advanced manufacturing in the United States, and builds on established relationships in New York and Vermont.”
“The Agreement expands our longstanding collaboration, which began when GF was created in 2009, and reflects our confidence in GF’ capability,” said IBM Senior Vice President & Director of Research Dr. John E. Kelly III. “This acquisition enables IBM to focus on fundamental semiconductor and material science research, development capabilities and expertise in high-value systems, with GF' leadership in advanced technology manufacturing at scale and commitment to delivering future semiconductor technologies. We are grateful for the leadership and investments by the states of New York and Vermont in supporting the semiconductor industry.”
This acquisition bolsters semiconductor manufacturing and technology development in the United States. GF has robust capital expenditure plans of approximately $10 billion in 2014-2015, with the majority being invested in New York. GF has created nearly 3,000 direct jobs in New York and thousands more indirect jobs in the United States since 2009.
GF will acquire and operate existing IBM semiconductor manufacturing operations and facilities in East Fishkill, New York and Essex Junction, Vermont, adding capacity to serve its customers and thousands of jobs to GF' workforce. GF plans to provide employment opportunities for substantially all IBM employees at the two facilities who are part of the transferred businesses, except for a team of semiconductor server group employees who will remain with IBM. After the close of this transaction, GF will be the largest semiconductor technology manufacturing employer in the Northeast.
GF will also acquire IBM's commercial microelectronics business, which includes ASIC and specialty foundry, manufacturing and related operations and sales. GF plans to invest to grow these businesses.
IBM will reflect a pre-tax charge of $4.7 billion in its financial results for the third quarter of 2014, which includes an asset impairment, estimated costs to sell the IBM microelectronics business, and cash consideration to GF. Cash consideration of $1.5 billion is expected to be paid to GF by IBM over the next three years. The cash consideration will be adjusted by the amount of working capital which is estimated to be $200 million.
The transaction is subject to the satisfaction of regulatory requirements and customary closing conditions.
IBM hat nur bis 22 nm Prozesse entwickelt. Power 8 kam 2014 im 22 nm Node. 14 nm und 10 nm waren zum Zeitpunkt des Vertrags noch Zukunftsmusik. IBM hat die Halbleiterfertigung, die an GF verkauft wurde, mit 4,7 Milliarden USD abgeschrieben. Das hat im betreffenden Quartal für einen heftigen Verlust gesorgt. Aber IBM war die Halbleiterfertigung los.
Das hatte ich ganz vergessen: Eigentlich wollte GF 10 nm entwickeln.
Aber dann haben sie beschlossen 10 nm aufzugeben und gleich 7 nm zu entwickeln.
Bei IBM zeigt auf der Website keine Pressemitteilungen an, die vor 2019 veröffentlicht wurden. Über Google habe ich ein
PDF mit der Pressemittelung auf der IBM Website gefunden. Der Inhalt entspricht dem von GF, aber das Dokument von IBM enthält zusätzlich ungeschminkt die Begründung warum IBM den Vertrag abgeschlossen hat: Die Halbleiterfertigung arbeitete mit Verlust und es war abzusehen, dass die Verluste größer werden.
Aus dem PDF:
IBM’s microelectronics OEM business has been reported in IBM's Systems and Technology segment. For the full year 2013 and first half of 2014, IBM’s OEM revenue will be reduced by approximately $1.4 billion and $0.6 billion, respectively. In the same periods, the divested business reflected pre-tax losses of $0.7 billionand $0.4 billion, respectively.
Im ersten Halbjahr haben die an GF verkauften Unternehmensteile einen Verlust von 400 Mio erwirtschaftet.
Die Zahlungen an GF wurden offensichtlich geleistet, um die erwarteten Verluste abzudecken. Sie erfolgten in den folgenden 3 Jahren (2015 ... 2017).
Im Januar 2014 hatte IBM angekündigt das X86-Servergeschäft für 2,3 Mrd USD an Lenovo zu verkaufen. Hier hatte IBM im Jahr 2013 mit der Hardware einen Umsatz von 4,2 Mrd USD gemacht. Zusätzlich 400 Mio USD mit Wartungsverträgen. Der Deal wurde zum 1. Oktober 2014 abgeschlossen.
Enttäuschte Erwartungen
Natürlich hat es die
Server Roadmap von IBM durcheinander gewürfelt, dass GF die Entwicklung an 10 nm und schließlich auch an 7 nm eingestellt hat. IBM konnte Power 10 im Jahr 2021 auf den Markt bringen.
2013 hat IBM die OpenPower Foundation initiiert. Dies ist aber wirkungslos verpufft und angesichts RISC-V ist die OpenPower Foundation so ziemlich in der Versenkung verschwunden.
Der Absatz von IBM Power Systemen ist in den 2010er Jahren stetig zurückgegangen. Erst zum Ende des Jahrzehnts hat sich der Umsatz mit den Power Systemen auf niedrigem Niveau stabilisiert.
In den 2000er Jahren hatte IBM noch bei den Gamekonsolen einigen Erfolg und konnte so den Verlust des PC-Markts zumindest teilweise auffangen. Aber Sony war 2013 mit der Playstation 4 zu AMD gewechselt und Nintendo ging 2017 zu Nvdia. Die Wii U, die noch eine IBM CPU hatte, war kein Erfolg. Das heißt nicht nur die Stückzahlen für die Power Chips gingen zurück, auch das OEM Geschäft von IBM hatte Schlagseite.
Hat IBM jemals die 2014 angekündigte Anzahl an Chips abgenommen?
Bei den Fabs musste GF wegen zu wenig Fertigungsaufträgen die Reißleine ziehen und die hat die
Fab in East Fishkill 2019 weiterverkauft. ON Semiconductor hat dort in 45 nm und 65 nm fertigen lassen. Bis Ende 2022 hat GF die Fab noch für ON Semiconductor betrieben, dann ging sie auf ON Semiconductor über. ON Semiconductor isz im Bereich der Leistungshalbleiter tätig und hat von GF eine Fab übernommen die für ON Semiconductor Gewinn erwirtschaftet hat.
Auch eine Fab in Singapur hat GF verkauft.
Und trotz dieser Verkäufe kam GF erst 2021 aus der Verlustzone als die Chipkrise GF neue Kunden beschert hat. Erst ab 2021 hat die Strategie als Anbieter für Sondertechnologien wirklich funktioniert.
Keine Kunden für 7 nm
GF selbst hat als Begründung, warum sie mit der Entwicklung an 7 nm aufhören, gesagt, dass sie für 7 nm zu wenig Kunden hätten:
https://investors.gf.com/news-relea...reshapes-technology-portfolio-intensify-focus
“Demand for semiconductors has never been higher, and clients are asking us to play an ever-increasing role in enabling tomorrow’s technology innovations,”
Caulfield said. “The vast majority of today’s fabless customers are looking to get more value out of each technology generation to leverage the substantial investments required to design into each technology node. Essentially, these nodes are transitioning to design platforms serving multiple waves of applications, giving each node greater longevity. This industry dynamic has resulted in
fewer fabless clients designing into the outer limits of Moore’s Law. We are shifting our resources and focus by doubling down on our investments in differentiated technologies across our entire portfolio that are most relevant to our clients in growing market segments.”
Ich wüsste nicht, dass sich TSMC im Jahr 2018 über einen Mangel an Kunden für 7 nm beklagt hat. Aber natürlich ist es besser darüber zu reden, dass die Kunden eigentlich gar kein 7 nm wollten, anstatt über die eigentlichen Gründe reden zu müssen, warum GF keine Kunden für 7 nm hat. Es hat in der Branche offensichtlich niemand GF wirklich zugetraut, dass GF den 7 nm Prozess hin bekommt. OK einen gab es wohl, IBM als einzigen Kunde.
Klar hätten einige Kunden sobald GF fertig gewesen wäre, Testwafer beauftragt. Wenn die Testwafer die Anforderungen der jeweiligen Kunden erfüllt hätten, hätten diese Kunden auch Aufträge an GF vergeben.
Wenn man annimmt dass GF tatsächlich den Prozess Ende 2018 fertig gehabt hätte, und dann sofort die ersten Testwafer beauftragt worden wären, hätte GF trotzdem vor 2021 nur wenig Umsatz mit 7 nm gemacht. Schneller ist nicht möglich angesichts des Zeitaufwands
- für den Testchip entwerfen, fertigen und evaluieren,
- für das Chipdesign
- für das Erstellen der Masken, erste Testchips fertigen
- für das Bring up
- für die eigentliche HVM. Hier kann man Pi Mal Daumen 2 Tage je notwendiger Maske ansetzen.
Wenn alles glatt gegangen wäre hätte IBM GF 7nm für den Power 10 verwendet. Aber auch IBM hätte erst ab 2020/21 7 nm Chips abgenommen.
Fast fertig?
https://spectrum.ieee.org/globalfoundries-halts-7nm-chip-development
Dieser Artikel gibt die Stellungsnahme von GF wieder, nennt allerdings nette Fakten:
During a tour of Fab 8 last October, executives told
IEEE Spectrum that 7-nm chips would be, at the very least, in limited production by the close of 2018. At the time, the company was completing installation of its first EUV lithography machine, a delicate and extremely expensive task.
Wenn der 7 nm Prozess von GF tatsächlich auf EUV basiert hat*), dann war GF nach bestenfalls einem 3/4 Jahr mit einer EUV Lithographie Maschine
nicht fast fertig**). Und nur mit einer EUV Lithographie Maschine gibt es keinen hohen Waferdurchsatz. IIRC hat TSMC je Fab/Phase 6 EUV Lithographie Maschinen im Einsatz.
*) Was erklären würde warum GF 10 nm aufgegeben hat. Diesen Prozess hätten sie ohne EUV machen müssen und da 7 nm später kommen sollte war wohl die Hoffnung es dank EUV ohne Multi Patterning zu schaffen. Aber EUV hat eigene Herausforderungen. Und TSMC hat 10 nm (2017) auch nur als Zwischenschritt verwendet. TSMC war im August 2018 schon mit 7 nm in HVM.
**)
Volker hat geschrieben:
Die Anzeichen für Probleme bei 7 nm waren bereits seit Monaten deutlich und wurden dennoch stetig klein geredet, doch der neue CEO hatte hinter den Kulissen die Reißleine bereits gezogen.
Der neuen CEO Tom Caulfield kam durch den Kauf der IBM Halbleiterfertigung zu GF.
In einem anderen Artikel von Volker:
In vielen Abteilungen wurde dann aber auch der Rotstift angesetzt, sehr deutlich wird das im Bereich Forschung & Entwicklung. Hier lagen die Ausgaben im Jahr 2018 noch bei 926 Millionen US-Dollar, 2019 waren es bereits nur noch 583 Millionen, um im Jahr darauf auf nur noch 478 Millionen zu sinken. Auch im ersten Halbjahr 2021 gingen die Ausgaben in dem Bereich noch weiter zurück.
Beim Rückgang 2018 auf 2019 hat sich das Einstellen der Entwicklung von 7 nm ausgewirkt. Und bitte beachten, dass GF die 7 nm Entwicklung gar nicht mehr das ganze Jahr 2018 betrieben hat.
Mubadala hat 2018 den Geldhahn zugedreht, nachdem Mubadala endlich erkannt hatte, dass GF mit der bisherigen Strategie ein Fass ohne Boden ist. Tom Caufield hatte gar keine andere Chance als 7 nm einzustellen. Wenn GF 7 nm fortgeführt hätte, wäre GF Konkurs gegangen.
AMD hatte nie vor 7 nm bei GF fertigen zu lassen
Convert schrieb:
Ja, klar, nach der Einstellung der Entwicklung war AMD raus.
Nein. AMD ist schon am 30. August 2016 zwei Jahre vor der Einstellung der Entwicklung am 27. August 2018 bei GF 7 nm ausgestiegen.
D. h. im Grunde war AMD bei GF 7 nm
nie im Boot. Da GF 7 nm erst am
15. September 2016 angekündigt hat. AMD hat allerdings zu diesem Zeitpunkt noch so getan als wären sie dabei.
Im August 2018 waren die Tape outs bei TSMC schon längst erfolgt. Anders wären die folgenden Releases gar nicht möglich gewesen
- Radeon VII: 7. Februar 2019
- RX 5700 XT: 7. Juli 2019
- Zen 2: 7. Juli 2019
Das hat Mark Papermaster am 27. August 2018 in einem Blog Post verkündet hat, den
Toms Hardware ganz am Ende ihres Artikels zitiert.
AMD announced earlier this year that its 7nm Navi GPUs would come with TSMC silicon, and Lisa Su recently confirmed that the second-gen EPYC processors would also come with TSMC's 7nm process. AMD's Mark Papermaster penned a statement on the move, assuring customers that the company's roadmap remains intact:
AMD’s next major milestone is the introduction of our upcoming 7nm product portfolio, including the initial products with our second generation “Zen 2” CPU core and our new “Navi” GPU architecture. We have already taped out multiple 7nm products at TSMC, including our first 7nm GPU planned to launch later this year and our first 7nm server CPU that we plan to launch in 2019. Our work with TSMC on their 7nm node has gone very well and we have seen excellent results from early silicon. To streamline our development and align our investments closely with each of our foundry partner’s investments, today we are announcing we intend to focus the breadth of our 7nm product portfolio on TSMC’s industry-leading 7nm process. We also continue to have a broad partnership with GLOBALFOUNDRIES spanning multiple process nodes and technologies. We will leverage the additional investments GLOBALFOUNDRIES is making in their robust 14nm and 12nm technologies at their New York fab to support the ongoing ramp of our AMD Ryzen, AMD Radeon and AMD EPYC processors. We do not expect any changes to our product roadmaps as a result of these changes.
Den Inhalt des Blog Posts hat Volker in einem Update zu seinem
Artikel zum Ausstieg zusammengefasst.
Auf gut Deutsch: AMD hatte keine Produkte für GF 7 nm geplant. Wird aber mit GF bei 12 und 14 nm weiter zusammenarbeiten. Da AMD nach August 2018 nur noch die beiden IODs für Zen 2 in 12/14 nm aufgelegt hat, ist Zen 2 das Ende der Zusammenarbeit. Vielleicht braucht AMD in den nächsten Jahren noch Zen, Zen+ oder Zen 2 Chips und verlänger das Wafer Supply Agreement noch Mal. Ansonsten ist Ende 2025 Schluss.
Die Geschichte dass AMD Zen 2 bei TSMC und GF fertigen lässt, wurde bis zum 27. August 2018 falsch verstanden. Viele glaubten, dass AMD den
Zen 2 Chip bei TSMC und GF jeweils in 7 nm fertigen würde. Erst zum Jahresanfang 2019 hat AMD offenbart, dass Zen 2 aus Chiplets in 7 nm und in 14 nm besteht.
AMD hielt bis zum 27. August 2018 die Illusion aufrecht, AMD würde mit GF bei 7 nm zusammenarbeiten. Aber seit dem ist bekannt, dass AMD nie vorhatte 7nm Chips bei GF zu fertigen. In allen späteren Interviews, die sich um 7 nm drehen, haben sich Lisa Su und Mark Papermaster so geäußert, dass AMD frühzeitig auf den TSMC 7 nm Prozess gewettet hat.
Im Januar 2019 wurde die 7. Änderung des Wafer Supply Agreements vereinbart. AMD erhielt das Recht sich für 7 nm und darunter die Foundry ohne irgendwelche Einschränkungen auszusuchen. Ohne weitere Zahlungen an GF.
Der Anfang vom Ende
AMD hatte sich im August 2016 mit der 6. Änderung des Wafer Supply Agreements für 14 nm und für 7 nm von GF freigekauft. Und offensichtlich in diesem Zeitraum auch mit dem Chipdesign für TSMC 7 nm begonnen.
Pressemitteilung zur 6. Änderung des Wafer Supply Agreements:
https://ir.amd.com/news-events/pres...nces-multi-year-amendment-to-the-wafer-supply
Ich musste die Pressemitteilung 3 Mal lesen bis ich verstanden hatte, dass sich AMD freigekauft hatte. Vordergründig steht da dass AMD und GF bei 7 nm zusammenarbeiten. Tatsächlich steht drin dass AMD das Recht bekommt bei 7 nm andere Foundries zu beauftragen und dafür sehr viel Geld in bar und in Optionsscheinen an GF bzw. Mubadala bezahlt.
In der zugehörigen Mitteilung an die SEC versucht AMD nicht das Freikaufen zu verschleiern:
https://ir.amd.com/sec-filings?year=2016&page=4#gallery-0000002488-451-1
On August 30, 2016, Advanced Micro Devices, Inc. (the “Company”) entered into a sixth amendment (the “Sixth Amendment”) to the Wafer Supply Agreement with GLOBALFOUNDRIES Inc. (“GF”). The Sixth Amendment modifies certain terms of the Wafer Supply Agreement applicable to wafers for the Company’s microprocessor, graphics processor and semi-custom products for a five-year period from January 1, 2016 to December 31, 2020. The Company and GF agreed to establish a comprehensive framework for technology collaboration for the 7nm technology node.
The Sixth Amendment also provides the Company a limited waiver with rights to contract with another wafer foundry with respect to certain products in the 14nm and 7nm technology nodes and gives the Company greater flexibility in sourcing foundry services across its product portfolio. In consideration for these rights, the Company will pay GF $100 million, which will be paid in installments starting in the fourth fiscal quarter of 2016 through the third fiscal quarter of 2017. Starting in 2017 and continuing through 2020, the Company also agreed to make quarterly payments to GF based on the volume of certain wafers purchased from another wafer foundry.
Further, for each calendar year during the term of the Sixth Amendment, the Company and GF agreed to annual wafer purchase targets that increase from 2016 through 2020. If the Company does not meet the annual wafer purchase target for any calendar year, the Company will be required to pay to GF a portion of the difference between the Company’s actual wafer purchases and the wafer purchase target for that year. The annual targets were established based on the Company’s current business and market expectations and take into account the limited waiver it has received for certain products.
The Company and GF also agreed on fixed pricing for wafers purchased during the 2016 year and established a framework to agree on annual wafer pricing for the years 2017 to 2020. The Company currently estimates that it will purchase approximately $650 million of wafers from GF in fiscal 2016 consisting of approximately $495 million of wafer purchases under the Sixth Amendment in 2016 and $155 million of wafer purchases previously taken in the first fiscal quarter of 2016 under the Fifth Amendment to the Wafer Supply Agreement. The Company expects that its future purchases from GF will continue to be material under the Wafer Supply Agreement, which is in place until 2024.
The foregoing description is not complete and is qualified in its entirety by reference to the Sixth Amendment, a copy of which will be filed with a Quarterly Report on Form 10-Q.
Item 3.02 Unregistered Sales of Equity Securities
Warrant Agreement
Also on August 30, 2016, in consideration for the limited waiver and rights under the Sixth Amendment, the Company entered into a warrant agreement (“Warrant Agreement”) with West Coast Hitech L.P. (“WCH”), a wholly-owned subsidiary of Mubadala Development Company PJSC (“Mubadala”). Under the Warrant Agreement, WCH and its permitted assigns are entitled to purchase 75 million shares of the Company’s common stock (the “Warrant Shares”) at a purchase price of $5.98 per share. The Warrant Agreement is exercisable in whole or in part after the date of issuance until 5:00 p.m. Eastern time on February 29, 2020; provided that the maximum amount of Warrant Shares that may be exercised under the one-year anniversary of the Warrant Agreement shall not exceed 50 million. Notwithstanding the foregoing, the Warrant Agreement shall only be exercisable to the extent that Mubadala does not beneficially own, either directly through any other entities directly and indirectly owned by Mubadala or its subsidiaries, an aggregate of more than 19.99% of the Company’s outstanding capital stock after any such exercise.
The Company expects to record a one-time accounting charge in the third fiscal quarter of 2016 of approximately $335 million comprised of the $100 million payment under the Sixth Amendment and the value of the warrant under the Warrant Agreement which is approximately $235 million.
The securities under the Warrant Agreement were issued pursuant to Section 4(a)(2) of the Securities Act of 1933, as amended. The foregoing description is not complete and is qualified in its entirety by reference to the text of the Warrant Agreement, filed as Exhibit 10.1 to this Current Report on Form 8-K and incorporated herein by reference.
Im Rahmen des 10-Q für das Quartal 3 2016 liegt eine zensierte Version des Sixth Amendment des Wafer Supply Aggrements vor.
$5.98 war wahrscheinlich der Aktienkurs am Tag an dem die Vereinbarung gültig wurde. Die Aktie ist in Folge immer über diesem Preis gehandelt worden. Gegen Ende des Zeitraums (Februar 2020) wurde die Aktie bei über 40 USD gehandelt. An diesen Optionsscheinen hat Mubadala ordentlich verdient. Und eventuell auch noch Mal bei den Aktien.