Microelectronics and Rare Earth Elements Sectors

[Bogeyman]

​

ASML to Install 600 DUV Chipmaking Tools in China by 2025​

As the U.S. government and its allies are tightening their wafer fab equipment export regulations for Chinese entities, ASML remains optimistic and expects the number of installed DUV tools in the People's Republic to increase in the coming years, reports DigiTimes. Meanwhile, the company admits that most of the deep ultraviolet (DUV) lithography machines it supplies to Chinese customers are designed for 'mature and mid-critical' nodes.

"The total number of ASML deep ultraviolet (DUV) lithography and metrology machines installed in China is around 1,400," said Shen Bo, ASML's China country manager, in an interview with TMTPost. "By the end of 2025, DUV equipment installment in China is projected to reach 600." ASML does not break down the number of metrology, litho, and inspection systems used by Chinese entities, but we presume that the number of lithography tools is lower than 600 right now.

ASML intensified its shipments to China in Q3 2023. As a result, Chinese companies accounted for 46% of the company's net system sales of €5.308 billion ($5.688 billion) during the quarter, up from 24% in the second quarter. In total, China-based customers procured $2.6 billion worth of wafer fab equipment to the country, including DUV litho tools and metrology systems. Most of them are now aimed at mature and trailing process technologies, such as 28nm, 45nm, 65nm, 90nm, and thicker.

"We are shipping lithography systems for mature and mid-critical nodes to China while, of course, complying with export regulations," said Peter Wennink, chief executive of ASML, at the company's earnings call in late October. The head of ASML also admitted that the company did not ship enough tools to its Chinese customers in recent quarters as it tried to fulfill demand from other clients.

"For systems shipping this year to Chinese customers, the majority of the orders were booked in 2022," said Wennink. "The demand fill rate for our Chinese customers over the last two years was significantly less than 50 percent. So, the Chinese customers were, in fact, receiving a much lower number of systems than they ordered. This was because demand for our systems worldwide significantly exceeded supply. With current shifts in demand timing from other customers, we now have the opportunity to fulfill these orders to our Chinese customers."

We are pretty much certain that the orders made by Chinese companies in 2022 included lithography equipment for making 14nm/16nm-class chips and perhaps even more advanced ones. Meanwhile, we do not know what portion of all the lithography tools these advanced systems made up and whether the orders were fulfilled before the Netherlands restricted shipments of such tools to China on September 1.

ASML to Install 600 DUV Chipmaking Tools in China by 2025 (Updated)

ASML expects to keep shipping DUV tools for 'mature and mid-critical nodes' to China.

www.tomshardware.com

 

[Bogeyman]

China's AI Analog Chip Claimed To Be 3.7X Faster Than Nvidia's A100 GPU in Computer Vision Tasks (Updated)​

Original Article:

A new paper from Tsinghua University, China, describes the development and operation of an ultra-fast and highly efficient AI processing chip specialized in computer vision tasks. The All-analog Chip Combining Electronic and Light Computing (ACCEL), as the chip is called, leverages photonic and analog computing in a specialized architecture that’s capable of delivering over 3.7 times the performance of an Nvidia A100 in an image classification workload. Yes, it’s a specialized chip for vision tasks – but instead of seeing it as market fragmentation, we can see it as another step towards the future of heterogeneous computing, where semiconductors are increasingly designed to fit a specific need rather than in a “catch-all” configuration.

As noted in the paper published in Nature, the simulated ACCEL processor hits 4,600 tera-operations per second (TOPS) in vision tasks. This works out to a 3.7X performance advantage over Nvidia’s A100 (Ampere) that's listed at a peak of 1,248 TOPS in INT8 workloads (with sparsity). According to the research paper, ACCEL can has a systemic energy efficiency of 74.8 peta-operations per second per watt. Nvidia’s A100 has since been superseded by Hopper and its 80-billion transistors H100 super-chip, but even that looks unimpressive against these results.

Of course, speed is essential in any processing system. However, accuracy is necessary for computer vision tasks. After all, the range of applications and ways these systems are used to govern our lives and civilization is wide: it stretches from the wearable devices market (perhaps in XR scenarios) through autonomous driving, industrial inspections, and other image detection and recognition systems in general, such as facial recognition. Tsinghua University’s paper says that ACCEL was experimentally tried against Fashion-MNIST, 3-class ImageNet classification, and time-lapse video recognition tasks with “competitively high” accuracy levels (at 85.5%, 82.0%, and 92.6%, respectively) while showing superior system robustness in low-light conditions (0.14 fJ μm−2 each frame).

https://go.redirectingat.com/?id=92...stry-best-in-ai-acceleration-for-vision-tasks

a, The workflow of traditional optoelectronic computing, including large-scale photodiode and ADC arrays. b, The workflow of ACCEL. A diffractive optical computing module processes the input image in the optical domain for feature extraction, and its output light field is used to generate photocurrents by the photodiode array for analog electronic computing directly. EAC outputs sequential pulses corresponding to multiple output nodes of the equivalent network. The binary weights in EAC are reconfigured during each pulse by SRAM, by switching the connection of the photodiodes to either V+ or V− lines. The comparator outputs the pulse with the maximum voltage as the predicted result of ACCEL. c, Schematic of ACCEL with an OAC integrated directly in front of an EAC circuit for high-speed, low-energy processing of vision tasks. MZI, Mach–Zehnder interferometer; D2NN, diffractive deep neural network .

In the case of ACCEL, Tsinghua’s architecture operates through diffractive optical analog computing (OAC) assisted by electronic analog computing (EAC) with scalability, nonlinearity, and flexibility in one chip – but 99% of its operation is implemented within the optical system. According to the paper, this helps in fighting constraints found in other vision architectures such as Mach–Zehnder interferometers and diffractive Deep Neural Networks (DNNs).

That 99% number is relevant to explaining at least the disparity in energy efficiency between ACCEL and other non-analog approaches: Nvidia’s GPU is 100% digital, meaning that its operation is based on the continuous flow of electrons (and produces waste heat as a result).

A photonic, optical system, however, leverages non-electrical ways of transferring, operating on, and encoding information. This can be done via laser pulses at specific wavelengths (we explored this in our recent article on China’s Quantum Key Distribution [QKD] satellite system, also photonic-based) being used to extract and communicate features of visual data (an image) and operating on that light (changing it) virtually on-transit. As a result of this optical processing system, there are fewer energy requirements and electrons wasted in thermal dissipation. Getting rid of the high energy and latency cost of ADCs (Analog-to-Digital Converters) goes a long way toward the performance improvements unlocked by photonics. It’s also why photonics systems are used across quantum computing and HPC (High-Performance Computing) installations.

Simultaneously, we reap speed benefits from moving away from the orderly but messy movement of electrons across semiconductors and unlock operating speeds limited only by light itself. As a result, the research paper claims in-house tests of the chip showcased a low computing latency of each frame at 72ns – resulting in a throughput of approximately 13,000 frames generated per second, more than enough to make any Doom player lose track of reality. It also seems like there would be enough frames for a co-processor to analyze a selection of those images in any computing-vision task. It hardly seems like the deep learning processing of these images through ACCEL would be the bottleneck.

https://go.redirectingat.com/?id=92...stry-best-in-ai-acceleration-for-vision-tasks

a, The principle of OAC for feature extraction of large-scale images. b, Simulated examples of OAC processing. OAC encodes the 28 × 28 original inputs into 4 × 4 features. A three-layer fully connected digital NN (Supplementary Table 1) reconstructs the image with the OAC output features. c, The SSIM (structural similarity index) of reconstruction results with OAC outputs under different compression ratios obtained by numerical simulations on the MNIST dataset. Examples of reconstruction images corresponding to different compression ratios are displayed in the corner. Compression ratio is the ratio of the dimensionality of OAC output to the dimensionality of original images. The example images for the original input are adapted from the MNIST dataset40 with permission. d, Classification accuracy by using OAC output as the input connected to a three-layer fully connected digital NN (Supplementary Table 1) under different compression ratios of OAC obtained by numerical simulations. The pixel size of the phase mask in OAC is 3 µm and the diffraction distance is 3 mm. The neuron number in OAC is 500 × 500. The red dashed line is the classification accuracy of the digital NN using the original images without OAC as the input. e, Photo of the EAC chip. Scale bar, 500 μm. The chip consists of a 32 × 32 photodiode array, two capacitance compensation modules P-CCM and N-CCM, voltage output module and peripheral SRAM I/O and controller. f, The structure of the capacitance compensation module. g, The structure of the EAC array. h, Magnified circuit structure of each pixel. a.u., arbitrary unit; Max., maximum; Min., minimum; Int., intensity; PD, photodiode.


ACCEL seems to be an analog rendition of an Application-Specific Integrated Circuit (ASIC) design. That’s precisely the role of the electronic analog computing (EAC) unit, as it can reconfigure the analog pathways within it to accelerate specific tasks. Think of these as pre-programmed algorithms within the chip, with the EAC coordinating which configuration should be applied to which task.

Dai Qionghai, one of the co-leaders of the research team, said, “Developing a new computing architecture for the AI era is a pinnacle achievement. However, the more important challenge is to bring this new architecture to practical applications, solving major national and public needs, which is our responsibility.”

The new ACCEL chip being photonic and analog may bring to mind the recent IBM announcement of another analog AI-acceleration chip (Hermes). It’s perhaps interesting to witness how even with all sanctions being applied to China, the country’s research and development is allowing it to catch up – and in some ways, apparently improve upon – whatever it was that they were being impeded of. Being able to go around limitations is undoubtedly the way China is thinking about sanctions.

It’s also important to understand that this generation of photonics-based analog chips is being worked on at extremely relaxed lithography levels. ACCEL, for instance, was manufactured on a standard 180-nm CMOS technology for the Electronic Analog Computing unit (EAC) – the brains of the operation. Naturally, further efficiency improvements could be gained from further miniaturizing the process towards lower CMOS nodes (Nvidia’s H100 is fabricated at a 4 nm process). It’s unclear what further work can be done to miniaturize the Optical Analog Computing (OAC) module.

It seems that implementing analog computing systems such as ACCEL at scale is more of a question of fabrication throughput and industry adaptation rather than of physical impossibility. But there’s a reason high-performance AI analog chips still haven’t been deployed at scale: their manufacturing is currently too low to serve anything other than research efforts and prototypical work. We now don’t have the throughput nor the available capacity to add these chips to the already-commited-through-2025 manufacturing commitments at companies such as TSMC – but these experimental results are always needed before committing to scale anything. And the markets meant for chips such as these would very much like to have them. Ultimately, it’s all a matter of planning, spending, and time.

China's AI Analog Chip Claimed To Be 3.7X Faster Than Nvidia's A100 GPU in Computer Vision Tasks (Updated)

China's ACCEL races into view.

www.tomshardware.com

 

[Bogeyman]

US-China tech war: Baidu buys AI chips from Huawei, as US restricts Nvidia chip exports: sources​

• Baidu ordered US$61 million worth of Huawei’s 910B Ascend AI chips, which are marketed as an alternative to Nvidia’s A100 chip, sources say
• Huawei’s Ascend chips, though inferior to Nvidia’s in performance, are the most sophisticated domestic option available in China, one person says

Baidu ordered artificial intelligence (AI) chips from Huawei Technologies this year, two people familiar with the matter said, adding to signs that US pressure is prompting Chinese acceptance of the firm’s products as an alternative to Nvidia’s.

One of the people said Baidu, one of China’s leading AI firms, which operates the Ernie large language model (LLM), placed the order in August, ahead of widely anticipated new rules by the US government that in October tightened restrictions on exports of chips and chip tools to China, including those of US chip giant Nvidia.
Baidu ordered 1,600 of Huawei’s 910B Ascend AI chips – which the Chinese firm developed as an alternative to Nvidia’s A100 chip – for 200 servers, the source said, adding that by October, Huawei had delivered more than 60 per cent of the order, or about 1,000 chips, to Baidu.
The second person said that the order’s total value was approximately 450 million yuan (US$61.83 million) and that Huawei was to deliver all of the chips by the end of this year. Both people declined to be named because the details of the deal were confidential.

Although the order is tiny relative to the thousands of chips top Chinese tech firms have historically ordered from Nvidia, the sources said it was significant, as it showed how some firms could shift away from the US company.

Baidu, alongside Chinese peers such as Tencent Holdings and Alibaba Group Holding, is known to be a long-time client of Nvidia. Baidu was not previously known to be a AI chip customer of Huawei.
Although Huawei’s Ascend chips are still seen as far inferior to Nvidia’s in terms of performance, the first source said they were the most sophisticated domestic option available in China.

“They were ordering 910B chips to prepare for a future where they may no longer be able to purchase from Nvidia,” the first source said.

Baidu and Huawei did not respond to requests for comment. Nvidia declined to comment.


Huawei’s website says it has since 2020 collaborated with Baidu to make its AI platform compatible with Huawei hardware. In August, the two companies said they would deepen compatibility between Baidu’s Ernie AI model and Huawei’s Ascend chips.


Baidu has developed its own line of Kunlun AI chips, which the company says supports large-scale AI computing, but the company has mainly relied on Nvidia’s A100 chip to train its LLM.
After the US last year imposed rules stopping Nvidia from selling its A100 and H100 chips to China, the company issued new A800 and H800 chips as alternatives for Chinese customers, including Baidu. Nvidia is no longer able to sell those chips to China because of the October rules.

Analysts predicted last month that the US curbs would create an opening for Huawei to expand in its US$7 billion home market. The company has been the subject of US export controls since 2019.
The order adds to signs of technological advances for Huawei, as Beijing pours investment into its domestic semiconductor industry to help it catch up with overseas peers and urges state-owned firms to replace foreign technology with domestic alternatives.
Huawei drew substantial global attention in August when it unexpectedly unveiled a new smartphone that analysts said uses internally developed processors featuring advanced semiconductor technology, highlighting the company’s progress in chip development despite sanctions.
In September, Reuters reported that Huawei’s in-house chip design unit, HiSilicon, had commenced shipments of newly developed Chinese-made processors for surveillance cameras to clients in 2023 in another comeback sign.

Baidu buys AI chips from Huawei, as US restricts Nvidia chip exports

Baidu ordered US$61 million worth of Huawei’s 910B Ascend AI chips, which are marketed as an alternative to Nvidia’s A100 chip, according to people familiar with the matter.

www.scmp.com

 

[Oublious]

​

• Huawei’s Ascend chips, though inferior to Nvidia’s in performance, are the most sophisticated domestic option available in China, one person says

Baidu buys AI chips from Huawei, as US restricts Nvidia chip exports

Baidu ordered US$61 million worth of Huawei’s 910B Ascend AI chips, which are marketed as an alternative to Nvidia’s A100 chip, according to people familiar with the matter.

www.scmp.com

lol

 

[Bogeyman]

https://twitter.com/i/web/status/1722254272832946271

https://twitter.com/i/web/status/1722254275328512173

https://twitter.com/i/web/status/1722254284652445879

https://twitter.com/i/web/status/1722254291950563418

https://twitter.com/i/web/status/1722254298430742798

 

[Gary]

Rare earth and critical minerals wonk @Bogeyman might find this interesting

News > Commodities > Copper >

Military rearmament is just getting started — without enough critical minerals​

• November 8, 2023
• By Anthony Milewski

• Russia’s 11 million shells fired against Ukraine in 2022 contain same amount of copper as 10% of UK’s total wind turbine capacity
• US Department of Defense used 750,000 tons of minerals in 2017, when it faced no major military conflict
• US is 100% net import reliant for 12 critical minerals, and had net import reliance greater than 50% of apparent consumption for 31 others

Over the summer last year, Ukraine fired — on a daily basis — up to 7,000 artillery rounds, much of it supplied by the US and Europe. Russia fired an estimated 11 million artillery shells in 2022.

Most of these are 155mm shells, with a varying amount of metals depending on design, manufacturer, etc — but can contain, at least, an estimated 0.5kg of copper, depending on the shell and manufacturing process.

To put in perspective, just for Russia’s 11 million shells in 2022, that’s 5500 tonnes of copper. Or, the same amount of copper in 1,170 wind turbines, roughly 10% of the same number of wind turbines in the UK. And that’s just Russian shells.

Copper demand is already forecast to increase by more than 100% by 2035, from current production of 25 million metric tons (MMt) to 50MMt. But, as we warn in our latest report, we expect a copper supply shortfall of at least 10 million mt after years of underinvestment in the industry amidst falling ore grades.

And many of the demand and supply forecasts are not taking into account any significant increases in military demand.

Because the market is so tight, any increase in demand — especially military demand, which often takes priority — has the potential to have significant impact on the market.

And we’re not just talking about copper.

Military demand ramping up as stocks depleted​

Exact figures on how much ammunition and equipment remains in military inventories, as well as costs and minerals needed, is highly classified. However, there are numerous reports that give some insight into the seriousness of the shortage unfolding.

The US Department of Defense (DoD) used 750,000 tons of minerals, for everything from ammunition to aircraft carriers, in 2017 when the US faced no major military conflict.

Now, the US military is preparing to support two wars with allies — supporting Ukraine and Israel — soon after ending two other wars in Iraq and Afghanistan, while also maintaining commitments to more than 750 bases in at least 80 countries, as well as a military deterrence against China.

In CSIS war games, a war in the Taiwan Strait between the US and China, the US used more than 5,000 long-range weapons in just three weeks of fighting.

But stockpiles of weapons and ammunition are running low, with the US supply of 155mm artillery rounds so low that Biden administration decided earlier this year to send controverisal clusterbombs instead.

“The bottom of the barrel is now visible… We give away weapons systems to Ukraine, which is great, and ammunition, but not from full warehouses. We started to give away from half-full or lower warehouses in Europe”
— Admiral Rob Bauer of the Netherlands, the chair of the NATO Military Committee and NATO’s most senior military official, said during a discussion at the Warsaw Security Forum

Analysis by the Centre for Strategic and International Studies (CSIS) highlights the problem: significant amount of weaponry has been sent to Ukraine already with, not only stocks running low, but production time to rebuild potentially taking from 3-18 years depending on the equipment.

The US aims to increase production of artillery shells to about 100,000 shells in 2025, from the current rate of 28,000 per month (up from 14,000 per month six months ago).

To help meet production targets, the US Army is reportedly planning to invest US$18billion over the next 15 years to modernize it’s industrial base, with US$2.5billion invested in 2023 alone.

The US has committed over $15.2 billion in security assistance to Ukraine since January 2021.

The latest US$2.1 billion Ukraine Security Assistance Initiative — separate from previous presidential drawdowns that is drawn down from DoD stocks — plans to procure munitions directly from industry rather than delivering equipment. This includes:

• anti-aircraft ammunition: 30mm and 23mm, artillery rounds: 130mm and 122mm, grad rockets: 122mm, rocket launchers and ammunition, mortar systems and rounds (120mm, 81mm, 60mm) tank ammunition: 120mm, javelin anti-armor systems, anti-armor rockets, precision aerial munitions, about 3,600 small arms and more than 23,000,000 rounds of small-arms ammunition, additional munitions for National Advanced Surface-to-Air Missile Systems
• 7 tactical vehicles to recover equipment, 8 heavy fuel tankers and 105 fuel trailers, armored bridging systems, 4 logistics support vehicles, trucks and 10 trailers to transport heavy equipment
• 9 counter-unmanned serial system 30mm gun trucks, mobile counter-unmanned aircraft system, or C-UAS, 10 laser-guided rocket systems, secure communications equipment, 3 air surveillance radars

We are focused on the US, as it is the world’s leading arms manufacturer, but this is a trend across the Western world and beyond.

• in Europe, a dozen countries including Germany and Poland, have delivered up to 223,000 artillery shells, 2,300 missiles, and promised to deliver up to 400,000 rounds of ammunition, with approximately 70,000 estimated to be 155mm. But with a 12 month deadline, only 30% has reportedly been delivered
• in a high-intensity conflict, the UK would run out of ammo in 8 days, Germany after two days
• Germany reportedly needs over US$20billion worth of orders to replenish military stocks, but only 10% of this demand will be contracted by the end of 2023
• France has ordered the country’s military contractors to plan a “war economy” strategy to increase production, including over US$2billion worth of munitions
• NATO is expected to ask its members to raise its ammunition stockpiles
• Russia has a reported capacity to produce up to two million artillery rounds a year and 200 tanks
• and China, according to the latest US DoD report, “continues to modernize equipment and focus on combined arms and joint training in effort to meet the goal of becoming a world class military”

Critical minerals needed​

Secure critical mineral supply chains are essential to sustain any significant military.

The challenge: the US identified 50 critical minerals in 2022 but was 100% net import reliant for 12, and had a net import reliance greater than 50% of apparent consumption for 31 others.

In other words — as highlighted by recent export ban on gallium and germanium by China, which supplies 54% of all US Germanium — the US military critical mineral supply chain is extremely vulnerable.

“These minerals are absolutely essential to our national security and the security of our allies”
— US Senator John Barrasso, ranking member of the Senate Committee on Energy and Natural Resources

Some examples to highlight the importance of critical minerals to any modern military:

• copper: used in wide-range of military applications, from wiring to guidance systems, ammunition to naval vessels
• nickel: essential to produce stainless steel, used across armour plating in tanks, aerospace alloys, ships, and military batteries
• uranium: used in the production of nuclear fuel for nuclear submarines and aircraft carriers
• lithium: lithium batteries are found in nearly a significant number of weapon systems used by the US military, particularly for portable equipment and satellites
• rare earths: used in a variety of military applications, from strong alloys in aircraft engines to sensors, magnets and lasers; for example, in Javelin missiles and F-35 fighter jets
• antimony: an essential mineral in the production of armour-piercing bullets, night-vision goggles, tracer ammunition, and much more

The list goes on, from Germanium in infrared devices, to aluminium for aircraft, silver in Apache helicopters.

As we highlighted in our analysis on the US Military and Critical Minerals: The New Arms Race, the US is investing hundreds of billions to secure supplies.

US stockpile​

The Defense National Stockpile Center (DNSC), founded after World War I, maintains an emergency stockpile of critical commodities, but the market value of the inventory has fallen 98% from US$9.6billion in 1989 (at the height of the Cold War) to US$888 million in 2021 — and could be insolvent by 2025.

Europe’s military stockpiles, judging by reports, are in an even worse state.

Critical mineral military demand​

So, what do the current levels of stockpiles, global conflict and geopolitical tensions mean for critical mineral demand?

Military mineral orders and projections are highly classified, but it’s possible to judge the scale of the demand from the military budgets ramping up to try and resolve the massive shortfall:

• the National Defense Authorization Act authorizes US$816.7 billion — in absolute terms, one of the highest since World War II — to the Defense Department for 2023, including US$1billion to boost critical mineral stockpiles (in particular, doubling the value of Rare Earths)
• about US$1.2 billion in contracts are underway to replenish US military stocks for weapons sent to Ukraine, including about US$352 million in funding for replacement Javelin missiles, US$624 million for replacement Stinger missiles, and US$33 million for replacement HIMARS systems
• US$1.2 billion in contracts are underway for equipment promised to Ukraine under USAI, including 155mm ammunition, Switchblade unmanned aerial systems, radar systems and tactical vehicles
• the Infrastructure Investment and Jobs Act appropriated US$7 billion for battery supply chains, including critical mineral production
• the Inflation Reduction Act in 2022, included a 30% tax credit for investments in battery material projects and a 10% tax credit for critical mineral production costs
• according to Tim Gorman, a spokesperson for the Department of Defense (DoD), the Pentagon is awarding contracts at an “unprecedented pace”

Some examples of the impact of the growing demand from the military for critical minerals include:

• a US$37.5 million agreement between the DoD and Graphite One (Alaska) to fast-track a domestic graphite mine
• two awards — US$24.8 million and US$15.5 million — by the DoD to Perpetua Resources to secure a domestic source of antimony
• a US$90million agreement to secure lithium production between the DoD and Abermarle
• a US $20.6 million agreement between the DoD and Talon Nickel to increase domestic nickel production

And there are many more.

Slowly, then all at once​

The US military has identified numerous active national security threats on a scale much larger than anything the USA has managed since the end of the Cold War.

However, the global economy and supply chains have shifted significantly in the meantime.

We see the US military shifting it’s position and capacity to secure it’s critical mineral supply gaining more momentum than it has for arguably the past 30 years.

However, the US military America’s largest government agency, and it will take time.

And, if there’s another geopolitical crisis, it will happen all at once. And, as the US ramps up, so will other countries.

Exposure for investors​

As we mentioned at the top, any increase in military demand for critical minerals has the potential to have an outsized impact on already tight and price — especially because it will need to be a secure supply of minerals.

Private investment increasingly looks to be a key driver to finance the US critical mineral strategy, with the US working with a network of mining and technology firms to help encourage investment in up to 15 global critical-minerals projects.

Security will arguably be the most important factor for any investor looking to gain exposure to military demand for critical minerals. For example, a secure location for the mineral deposit and mine, allied with the US, as well as ease and security for exports (see our analysis on mining and US supply chains in Australia and Canada).

Demand will particularly focus on the minerals that the US is 50-100% reliant on for imports, from gallium to manganese, titanium to cobalt.

And, particularly if another crises occurs sooner rather than later, there will be a demand to secure existing mines rather than developing new ones. See our analysis: The critical mineral wars are coming.

Military rearmament is just getting started — without enough critical minerals - The Oregon Group - Investment Insights

Over the summer last year, Ukraine fired — on a daily basis — up to 7,000 artillery rounds, much of it supplied by the US and Europe. Russia fired an

theoregongroup.com

 

[Bogeyman]

The top five semiconductor equipment manufacturers have released their latest financial reports amid a slowdown in the industry. Applied Materials, ASML, Tokyo Electron (TEL), Lam Research, and KLA have shown varying revenue performances. Despite current market uncertainties, most companies remain optimistic about overall demand for next year.

#AppliedMaterials reported a slight YoY decline of 0.4% in revenue to US$6.723B for the fourth quarter of the 2023 fiscal year. The company estimates revenue of $6.47B for 1Q24 and believes that the recent tightening of U.S. chip export controls won’t significantly impact its revenue, maintaining a positive outlook on market demand for its products.

#ASML's 3Q23 revenue increased by 15.5% YoY to €6.7B, with a net income of €1.9B. Q3 net sales from system equipment in China accounted for 46%, a significant jump from 24% in Q2. ASML expects robust growth in 2023, with sales net revenue growth approaching 30%, but anticipates 2024 revenue to be similar to 2023 due to customer uncertainty about the industry's recovery pace and strength.

#TokyoElectron released its 2Q24 financial report, showing a sharp 39.7% decrease in consolidated revenue YoY to ¥427.8B, but a 9% increase QoQ. TEL stated that revenue had bottomed out between April and June, and with a significant increase in new customers from China, the Chinese market accounted for over 40% of TEL's total revenue for the first time from July to September, prompting an upward revision of the 2023 estimate for WFE market size.

#LamResearch reported an 8.6% QoQ increase in revenue to $3.48B for 1Q24, with the 2023 WFE investment estimated to be around $80B. Lam Research noted that investments in foundry/logic WFE were slightly below previous baseline assumptions due to weakened spending in advanced processes and mature processes outside China.

#KLA announced its 1Q24 (Jul-Sept) financial report, with a 30.7% YoY increase in revenue to $2.7B. China was the largest contributor to revenue, accounting for 31%. KLA is still assessing the impact of the latest U.S. export bans, predicting that the ban could affect its 2023 revenue by $600M to $900M.

https://twitter.com/i/web/status/1725437425244332534

 

[Bogeyman]

https://twitter.com/i/web/status/1730296001691525468

 

[Bogeyman]

With AMD releasing their AI accelerators, I wanted to dive into TSMC's AI exposure: $TSM's primary AI exposure is through $NVDA where they manufacture most of Nvidia’s top-end GPUs.

They manufacture AI accelerators for $AMD $GOOGL $AMZN and $MSFT as well. TSMC management noted that 6% of revenue this year came from AI.

They expect the AI business to compound at a 50% CAGR until 2027 where the business is a “low-teens” % of overall revenue. Additionally, $QCOM and $AAPL are integrating AI processors into their chips for edge inference.

If we assume that's an additional 1-2% revenue, we can estimate that TSMC's AI exposure will be ~14% of revenue in 2027. It's important to study the numerous risks to TSMC before investing, but AI will be a strong growth driver over the next decade.

https://twitter.com/i/web/status/1732797998642377026