Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/america-needs-chips-champion.23346/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

America needs chips champion

Y.H

Active member
1754526694561.png


A little over five years ago, the Trump administration announced Operation Warp Speed to deliver a vaccine for COVID-19. It was one of the most stunning successes of Trump’s first term. Recognizing a crisis, the U.S. government facilitated a public-private partnership that likely saved millions of lives in record time. Now we must do it again. As a country, we have a strategic imperative to win in artificial intelligence and secure our supply chains for critical technologies, including communications, computing, and advanced military systems. Time is of the essence. Yet the Administration’s plans for AI and self-sufficiency are in serious jeopardy unless we have American-owned, leading-edge chip manufacturing plants on American soil.

US advanced semiconductor manufacturing has been withering for some time. The once-leading Intel appears to be dropping out of the race. Missed deadlines, poor execution, and a misguided strategy to retain manufacturing within Intel while also serving as a foundry for its fabless chip competitors resulted in a dearth of customers. Recommendations (including those from the four of us) to split off Intel’s foundry business and create a fully independent entity to supply its competitors, thereby giving itself a fighting chance, have never been adopted.

Intel appears to have only a few external customers for its current technology (called 18A), and the CEO recently said on July 24 that “Going forward, our investment in … [Intel’s most advanced process technology, 14A will] be based on confirmed customer commitments,” continuing a business model that has largely failed. Unsurprisingly, the CEO also announced the shuttering of its plans for German and Polish plants, further delaying its proposed Ohio plant, and a massive lay-off. More spending reductions will inevitably follow.

All of these announcements strongly imply a gradual exit from the chip manufacturing business, turning Intel into a fabless company over time. Given that Intel’s internal demand is no longer big enough to justify continued capital investment in leading-edge technology, this may be the right strategy for Intel.

Still, it is the wrong strategy for the United States. With Intel’s likely retreat from advanced chip manufacturing, America’s future and the future of its leadership in AI and all advanced electronics will be firmly in the hands of two firms: Taiwan Semiconductor Manufacturing Corporation (TSMC) and Samsung, two firms headquartered on the other side of the planet. TSMC is by far the dominant player, controlling over 90% of the world’s most advanced semiconductor manufacturing output. The Taiwanese chip manufacturer produces nearly 100% of Nvidia’s GPUs, which are the engines that enable AI. It also manufactures most of the chips for iPhones and 5G communications.

While TSMC and Samsung have committed to building more plants in the U.S., these will not solve the problem. Neither company will bring its latest technology here. The newest generation of chips must first be developed in a plant geographically close to its R&D teams. In the case of TSMC, those teams are in Taiwan; for Samsung, South Korea. The only R&D team that has been developing advanced generation technologies on US soil, fabricated in the latest generation U.S.-owned plants, is Intel. But as Intel retreats, America’s future in AI and other advanced technologies is increasingly reliant on a single firm, located a stone’s throw from mainland China.

To be sure, TSMC’s technological prowess is impressive. Moreover, its promise to invest $100 billion in Arizona is laudable. However, the fact is that we are giving TSMC too much power over the allocation of capacity, pricing, and human capital to drive AI into the future. In the case of Taiwan, business risk is compounded by the obvious geopolitical risks attendant to its status. These dependencies are intolerable if the U.S. is to protect its own economic and national security interests.

Fortunately, the Trump Administration has dealt itself enough cards to rectify this obvious vulnerability. The Administration recognizes that the United States needs advanced chip-making capabilities within our own country. To this end, by executive order March 31, the President created the United States Investment Accelerator at the Department of Commerce. It is responsible, among other things, for administering the CHIPS Program Office. Billions remain unspent from this Congressional program. Perhaps billions more can be retrieved from Intel, given its apparent surrender in the race with TSMC. In addition to these billions, the Trump Administration on July 22 also wrested from Japan a commitment to invest more than $550 billion in the United States.

With CHIPS money, Japan’s partnership, and government investment —either direct or through Trump’s recent executive order to create a sovereign wealth fund —the federal government has the opportunity to launch “Operation Warp Speed II” and put America back on the leading edge of chip manufacturing. Speed is essential: As Intel downsizes and lays off thousands of people, we are losing and will continue to lose the best people. Soon, we will be without a viable foundation on which to build a new, world-class American foundry, for which Intel’s assets are critical.

Here’s a plan:
First, similar to the first Operation Warp Speed, the Trump administration should build a public-private partnership, where future customers (e.g., Nvidia, Qualcomm, Broadcom, Google, Amazon, Apple and others), Japanese investors such as Softbank, and private equity, backed by government financing and/or investment, would buy Intel’s fabrication assets before the lack of investment and the rust of time makes them worthless and leave the United States dangerously dependent on a single manufacturing firm.

Second, the Trump administration has been very effective in persuading leading U.S. companies to invest in America’s future. They should be encouraged to partner and invest in a new American Foundry and to buy from it. Nvidia, Broadcom, Google, and others may have turned down Intel’s offering, but they cannot as easily turn down the opportunity to help create an independent, leading-edge domestic competitor to TSMC. American companies want (and need) alternative sources of supply, and this plan can provide them.

Building a new American Foundry for advanced semiconductors is the best strategy to keep the United States and American firms at the leading edge of AI and advanced electronics, and to ensure that critical supply chains are not disrupted by geopolitics, pandemics, or natural disasters. There is no time to waste.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

 
Personally, I believe the whole "no one will use Intel foundry because of IP risk" thing is overblown. The issue is far more fundamental. Intel needs to learn how to service foundry customers. They need to get the PDK's out on time. Those PDK's need to be industry standard, and they need to hit their commitments on the timeline they promise. Failure to do those things is the biggest reason that 18A has seen almost no outside adoption.

18A-P should be better as Intel seems to be learning these lessons, and 14A should have a decent chance to close these gaps as most of the new IP that made 18A challenging will be 2nd Gen. The resulting lift should be much easier.

All that said, anyone who invests in a "US foundry" should plan on a very long term investment to see those returns. I don't see this list of potential investors making such a long term play. There are faster ways to get a large return on their investments. The time value of money makes this type of investment unattractive.
 
While I think Intel advanced node development still has a chance, it is a small one. The US is not so dependent on Intel the way Korea is dependent on Samsung and Taiwan is dependent on TSMC. That means TSMC and Samsung will not stop advanced node development, because it would be existential for their respective nations.

The problem is, how do you build a political consensus in the US around the idea that Intel advanced node development must be saved by government means if necessary? I don’t think you can. There is a wisdom of the crowd to this. It’s a strength the US still possesses—the ability to let failure happen, because it leads to renewal. Because rewarding failure is bad, and counterproductive, in the long run.

I think you can build a consensus around doing science and technology development differently. China seems to be doing something differently, linking national objectives to university and private funding, in 5 or 10 year cycles. That process has socialist characteristics, but so does the old US system. Trump happens to be destroying the old system right now, and I think it’s fine. Something new is needed. Something built for purpose to compete with China, which seems to be winnning.
 
While I think Intel advanced node development still has a chance, it is a small one. The US is not so dependent on Intel the way Korea is dependent on Samsung and Taiwan is dependent on TSMC. That means TSMC and Samsung will not stop advanced node development, because it would be existential for their respective nations.

The problem is, how do you build a political consensus in the US around the idea that Intel advanced node development must be saved by government means if necessary? I don’t think you can. There is a wisdom of the crowd to this. It’s a strength the US still possesses—the ability to let failure happen, because it leads to renewal. Because rewarding failure is bad, and counterproductive, in the long run.

I think you can build a consensus around doing science and technology development differently. China seems to be doing something differently, linking national objectives to university and private funding, in 5 or 10 year cycles. That process has socialist characteristics, but so does the old US system. Trump happens to be destroying the old system right now, and I think it’s fine. Something new is needed. Something built for purpose to compete with China, which seems to be winnning.
I don’t think the plan proposed in fortune article will work. That one seems to require long term commitment from the government. I don’t think it’s wise to expect that from USG given their track record in recent few decades. Also, the article seems to be suggesting they are building things from scratch which seems to me like trying to reinvent the wheel.
If they want a US-headquartered leading edge node manufacturer, the best way is to provide significant incentives and support, economically and more importantly administratively. I mean that worked in many places. Not too sure they want to try something that had not yet prove to work with something they deem with critical importance.
 
I don’t think the plan proposed in fortune article will work. That one seems to require long term commitment from the government. I don’t think it’s wise to expect that from USG given their track record in recent few decades. Also, the article seems to be suggesting they are building things from scratch which seems to me like trying to reinvent the wheel.
If they want a US-headquartered leading edge node manufacturer, the best way is to provide significant incentives and support, economically and more importantly administratively. I mean that worked in many places. Not too sure they want to try something that had not yet prove to work with something they deem with critical importance.
What about a hybrid approach where Intel’s fabs are sold to GF for -$10B to -$20B (similar to how GF bought IBM’s fabs for negative $1.5B). Apple announced a new foundry deal with GF. They might be more willing to bring up a pure play foundry than Intel for advanced logic.
 
What about a hybrid approach where Intel’s fabs are sold to GF for -$10B to -$20B (similar to how GF bought IBM’s fabs for negative $1.5B). Apple announced a new foundry deal with GF. They might be more willing to bring up a pure play foundry than Intel for advanced logic.
Some form of joint venture approach with Intel foundry will work better than trying to build ASMC from scratch.
 
Personally, I believe the whole "no one will use Intel foundry because of IP risk" thing is overblown. The issue is far more fundamental. Intel needs to learn how to service foundry customers. They need to get the PDK's out on time. Those PDK's need to be industry standard, and they need to hit their commitments on the timeline they promise. Failure to do those things is the biggest reason that 18A has seen almost no outside adoption.

18A-P should be better as Intel seems to be learning these lessons, and 14A should have a decent chance to close these gaps as most of the new IP that made 18A challenging will be 2nd Gen. The resulting lift should be much easier.

All that said, anyone who invests in a "US foundry" should plan on a very long term investment to see those returns. I don't see this list of potential investors making such a long term play. There are faster ways to get a large return on their investments. The time value of money makes this type of investment unattractive.

The potential for IP theft or leaks is just one of several risks and conflicts of interest that Intel faces under the IDM business model. IMO Intel must split its product and foundry divisions into two independent companies; otherwise, it will have no meaningful path forward. Why on earth would Apple, Qualcomm, Nvidia, AMD, or anyone else want to help Intel, a direct competitor, survive?

When Apple, Nvidia, or AMD were on the brink of bankruptcy, did anyone step in to advocate for a rescue plan?

What puzzles me is that, after more than 30 years, the semiconductor industry has successfully transitioned to the foundry/fabless model. The only two remaining advanced logic IDMs are Intel and Samsung, and both are struggling to stay competitive. Why should we believe that Intel’s IDM model is somehow so exceptional and superior that it can succeed against the whole semiconductor industry?
 
The potential for IP theft or leaks is just one of several risks and conflicts of interest that Intel faces under the IDM business model. IMO Intel must split its product and foundry divisions into two independent companies; otherwise, it will have no meaningful path forward. Why on earth would Apple, Qualcomm, Nvidia, AMD, or anyone else want to help Intel, a direct competitor, survive?

When Apple, Nvidia, or AMD were on the brink of bankruptcy, did anyone step in to advocate for a rescue plan?

What puzzles me is that, after more than 30 years, the semiconductor industry has successfully transitioned to the foundry/fabless model. The only two remaining advanced logic IDMs are Intel and Samsung, and both are struggling to stay competitive. Why should we believe that Intel’s IDM model is somehow so exceptional and superior that it can succeed against the whole semiconductor industry?
Personally, I am guessing Broadcom has a good chance of using IFS
 
Back
Top