SILVACO 073125 Webinar 800x100
Semiconductor Expert Forum

Posted on

Predicting Stochastic Defectivity from Intel’s EUV Resist Electron Scattering Model

Predicting Stochastic Defectivity from Intel’s EUV Resist Electron Scattering Model
by Fred Chen on 11-22-2023 at 6:00 am
Categories: Intel Foundry, Lithography

Predicting Stochastic Defectivity from Intel's EUV Resist Electron Scattering Model

The release and scattering of photoelectrons and secondary electrons in EUV resists has often been glossed over in most studies in EUV lithography, despite being a fundamental factor in the image formation. Fortunately, Intel has provided us with a laboriously simulated electron release and scattering model, using the GEANT4 Monte Carlo simulator toolkit [1]. The simulator is able to track and keep statistics on all photons and released photoelectrons and secondary electrons and their trajectories. Consequently, it is possible to obtain the probability density for a given scattering distance at a given electron energy. In the resists studied, there were certain electron energies within the covered 10-82eV range where it was especially easy to trace electron scattering distances. Figure 1 shows the double Gaussian fitted probability density function for a 60 eV electron in SnOx resist and a 75 eV electron in chemically amplified resist.

Figure 1. Probability density functions for electron scattering distance for a 75 eV electron in chemically amplified resist (CAR) and 60 eV electron in SnOx. 25 mJ/cm2 EUV dose to a 100 nm x 100 nm patch of resist 50 nm thick. Data points sourced from Figure 7 of Ref. [1].

It is possible to get the population fraction beyond a given electron scattering distance, using the weighted average cumulative distribution functions of the two Gaussians. For example, a population fraction of 1e-11 goes beyond 20 nm for the 75 eV electron in CAR or the 60 eV electron in SnOx. This large electron scattering distance is a source of blur. Blur aggravates the impact of stochastic dose fluctuations because the peaks and valleys of the image are partly flattened by blur. A blur of 20 nm substantially destroys a 50 nm pitch image (Figure 2), regardless of exposure threshold, i.e., exposed critical dimension (CD). Essentially, there is a defect floor. For the smaller exposed CD, there is mostly unexposed area within the exposed region. For the larger exposed CD, there are exposed spots in the unexposed region.

Figure 2. Stochastic defectivity with decreasing resist exposure threshold (wider exposed CD), going from left to right. Exposure conditions: 50 nm pitch, 20 nm blur, 30 mJ/cm2 absorbed, secondary electron quantum yield=8. Pixel size: 1 nm x nm. More details in Refs. [2].

A defect probability of 1e-11 corresponds to a 1000 nm2 defective area per cm2 area. Or it can also mean, for example, one missing contact out of 100 billion contacts. When the blur probability matches the defect probability as in this case, we get a measure of the stochastic defectivity. A higher probability for defect-inducing blur can occur for smaller blur values, and this would happen for a smaller pitch. For the blur to play much less of a role in aggravating stochastic defectivity, the pitch has to be substantially larger than the blur. This also has the simultaneous benefit of making the stochastic fluctuations less significant by having more absorbed photons (per nm) per pitch.

References

[1] P. Theofanis et al., “Modeling photo, electron, and chemical interactions in a model hafnium oxide nanocluster EUV photoresist,” Proc. SPIE 11323, 113230I (2020).

[2] F. Chen, “Modeling EUV Stochastic Defects With Secondary Electron Blur,” https://www.linkedin.com/pulse/modeling-euv-stochastic-defects-secondary-electron-blur-chen/ (2023); also F. Chen, “EUV Stochastic Defects from Secondary Electron Blur Increasing With Dose,” https://www.youtube.com/watch?v=Q169SHHRvXE&t=94s (2023).

This article first appeared in LinkedIn Pulse: Predicting Stochastic Density from Intel’s EUV Resist Electron Scatterning Model

Also Read:

The Significance of Point Spread Functions with Stochastic Behavior in Electron-Beam Lithography

Extension of DUV Multipatterning Toward 3nm

Stochastic Model for Acid Diffusion in DUV Chemically Amplified Resists

Posted on

Navigating Edge AI Architectures: Power Efficiency, Performance, and Future-Proofing

Navigating Edge AI Architectures: Power Efficiency, Performance, and Future-Proofing
by Kalar Rajendiran on 11-21-2023 at 10:00 am
Categories: AI, Ceva, IP

CEVA Comprehensive Edge AI Portfolio

The surge in Edge AI applications has propelled the need for architectures that balance performance, power efficiency, and flexibility. Architectural choices play a pivotal role in determining the success of AI processing at the edge, with trade-offs often necessary to meet the unique demands of diverse workloads. There are three pillars of common AI processing architectures and related hardware and software requirements.

The Three Pillars of Common AI Processing Architectures

Scalar processing architectures are specifically designed for tasks that involve user interface management and decision-making based on temporal data with non-intensive compute requirements. These architectures excel in quickly and sequentially processing tasks, making them ideal for applications where swift decision-making is crucial. However, the trade-offs come in the form of limited parallelism compared to other architectures. While they are efficient for certain types of tasks, the sequential nature may pose limitations in handling workloads that benefit from parallel processing capabilities.

Vector processing architectures center on performing operations simultaneously on multiple data elements, making them particularly suitable for signal processing followed by AI perception tasks. Their key characteristics include enhanced parallelism, enabling efficient handling of tasks involving vectors and arrays. This parallel processing capability makes them well-suited for tasks that benefit from simultaneous data manipulation. However, a trade-off exists as vector processing architectures may not be as suitable for tasks with irregular or unpredictable data patterns. Their strength lies in structured data processing, and they may face challenges in scenarios where data patterns are less predictable or follow irregular structures.

Tensor (Matrix) processing architectures are specifically tailored for the demands of deep learning tasks, excelling in handling complex matrix operations integral to applications such as image recognition, computer vision, and natural language processing. Their key characteristics lie in the efficient handling of tasks involving large matrices and neural networks, making them essential for processing intricate data structures common in advanced AI applications. However, the trade-offs come in the form of intensive computational requirements, posing challenges in terms of power consumption. While these architectures deliver unparalleled capabilities in processing sophisticated tasks, the computational intensity may require careful consideration, especially in edge computing scenarios with limited power resources.

Hardware and Software Aspects

In terms of hardware for AI processing, a diverse array of solutions emerges to cater to varying requirements. Processor units, notably specialized ones like Neural Processing Units (NPUs), take the spotlight for their ability in handling intricate tasks associated with deep learning. Digital Signal Processors (DSPs) play an indispensable role, particularly in tasks involving signal processing, ensuring efficient and precise manipulation of data. Hybrid solutions mark a convergence of strengths, combining scalar, vector, and tensor processing architectures to achieve unparalleled versatility. This amalgamation provides a comprehensive approach to handling a broad spectrum of AI workloads. Further enhancing adaptability, customizable configurations enable hardware to be tailored and optimized for specific applications, ensuring efficiency and performance in diverse computing environments.

In terms of software for AI processing, a robust ecosystem is indispensable, characterized by optimized libraries, flexible frameworks, and developer tools. Optimized libraries stand as the basis for efficiency, offering specialized software components meticulously crafted to enhance the performance of AI tasks. Flexible frameworks play a pivotal role by providing a supportive environment that accommodates a variety of AI models, ensuring adaptability to evolving requirements. Complementing these, developer tools serve as catalysts in the configuration, optimization, and overall development of AI applications. Together, these software components form a cohesive foundation, empowering developers to navigate the intricacies of AI processing with efficiency, adaptability, and streamlined development workflows.

CEVA’s Pioneering Solutions

CEVA’s offerings address the full spectrum of AI workloads and architectures and stand out in fulfilling the power, performance, latency, versatility and future-proofing requirements of various AI applications.

CEVA-BX

The flexibility of CEVA-BX processors is a distinctive feature, enabling them to be finely configured and optimized for specific applications, offering a tailored approach to diverse computational needs. Their versatility delivers a delicate equilibrium between performance and power efficiency. This balance positions CEVA-BX as a fitting solution across a broad spectrum of edge computing applications.

SensPro2

Its vector DSP architecture, characterized by high configurability and self-contained functionality, positions SensPro2 as a versatile solution for a variety of applications. Particularly notable is its proficiency in parallel processing, where it excels in high data bandwidth scenarios, making it adept at addressing communication and computer vision tasks. In processing AI workloads, SensPro2 showcases remarkable efficiency, seamlessly handling tasks with a throughput of up to 5 TOPS (Tera-Operations Per Second).

NeuPro-M

The versatility of NeuPro-M shines through as it adeptly handles a diverse spectrum of Advanced Neural Network models, showcasing its adaptability to the evolving landscape of AI applications. Its Neural Processing Unit (NPU) IP stands as a testament to CEVA’s commitment to low-power, high-efficiency processing, making NeuPro-M a frontrunner in energy-constrained scenarios. A noteworthy feature contributing to its future-proofing capabilities is the built-in VPU (Vector Processing Unit). This allows NeuPro-M to efficiently manage not only current but also emerging and more complex AI network layers, ensuring sustained relevance and optimal performance in the dynamic and rapidly advancing field of Edge AI.

Summary

In navigating the diverse landscape of Edge AI architectures, the choice between scalar, vector, and tensor processing architectures involves weighing trade-offs and aligning solutions with specific workload requirements.

CEVA’s pioneering solutions, including CEVA-BX, SensPro2, and NeuPro-M, stand out by providing a comprehensive suite that addresses the three pillars of common AI processing architectures. With a focus on flexibility, efficiency, and future-proofing, CEVA empowers developers and businesses to navigate the complexities of Edge AI, making intelligent architectural choices tailored to the evolving needs of the industry.

Also Read:

Fitting GPT into Edge Devices, Why and How

Bluetooth Based Positioning, More Accurate, More Reliable, Safer

Democratizing the Ultimate Audio Experience

Posted on

Cadence Integrates Power Integrity Analysis and Fix into Design

Cadence Integrates Power Integrity Analysis and Fix into Design
by Bernard Murphy on 11-21-2023 at 6:00 am
Categories: AI, Cadence, EDA

Voltus Insight AI min

As integration levels increase, clock frequencies rise, and feature sizes shrink it is not surprising that all or most aspects of semiconductor design become more complex and demand more from design technologies. One example where the traditional approach is breaking down is in optimizing power distribution networks (PDNs) for electromigration and IR drop (EMIR). Historically designer have first run detailed EMIR analysis, then have implemented fixes to fold into the next design update, repeating periodically as implementation evolves. That method is no longer scaling effectively for very large designs at 7nm and below. To address this scaling problem Cadence has just announced a new AI-centric EMIR optimization tool supporting a combined analysis and fix in-design flow, following the path timing analysis took in earlier generations. They see this new approach to EMIR as an IR 2.0.

The problem statement

The trick in designing a PDN for a design is to ensure that metal routes are sufficiently low resistance, from the top level down through multiple metal layers to a gate that is switching. If resistance through that path is sufficiently high, that will result in a voltage drop below the nominal operating voltage causing the gate to switch more slowly than intended, therefore triggering unanticipated timing failures. Naturally other gates switching at the same time (aggressors) add to current demand in the same vicinity, further amplifying the voltage drop. In FinFET designs aggressors account for as much of the IR drop experienced by any given node. In effect, managing noise on the power network has become just like a signal integrity problem.

The IR 1.0 approach to the problem has been focused on analysis to find the hotspots, providing guidance to designers on what they need to fix. That has worked very well, but below 7nm resistance in the lower-level metal networks has become so significant that a victim in a block instance may have as many as 8-10k aggressors contributing up to 80% of the drop on the gate of immediate interest.

These compounded contributions complicate the analysis problem, but more important they create a deluge of IR drop violations (hundreds of thousands), making it impossible for a designer to fix any but a few of the most critical issues. Worse yet, in fixing one issue the designer has little insight into how an individual fix will impact timing for the other thousands of connected gates. The real goal should be to eliminate the great majority of violations without messing up PPA or DRC.

IR 2.0 with Voltus InsightAI

The only way to address the problem is to intelligently automate EMIR analysis/fixing together with PDN implementation, in the same we were forced to couple timing analysis and optimization with physically aware synthesis years ago. This is a perfect application for AI, not because a smart designer can’t figure out how to fix an individual IR drop problem or a problem coupled with a bit of surrounding complexity but because the sheer scale of coupled devices and interconnect has grown beyond human abilities to manage, no matter how skilled the engineer.

Voltus InsightAI uses a generative AI approach to develop and refine a power grid to minimize EMIR issues, maintain timing and power goals, while avoiding area overhead for an over-designed power grid. At the outset when reading in a design, the Learn capability creates AI models for the power grid, suitable for fast incremental IR drop analysis based on dynamic vector data. Discover runs root-cause analysis to prioritize large drops by finding aggressors and looks for resistance bottlenecks in the grid and power density hotspots.

Improve is a recommender technology which looks at many methods through which IR drop problems could be fixed and suggests optimal choices to address as many as violations as possible while honoring PPA and DRC constraints. The solution integrates through Empower with Innovus, Tempus, Voltus Power Integrity, and Pegasus in support of on-the-fly fix and check iterations.

In PDN design with Voltus InsightAI, designers use the tool first before signal routing. After signal routing has been inserted, they use the tool again to further refine optimization. Throughout the tool is self-training, updating after every few rounds of changes. No specialized pre-training is required.

All sounds good on paper; how does it deliver in practice? Several customers have been running trials, MediaTek among them. They have already provided an endorsement with hard numbers. They say they have seen a 65-70% reduction in IR drop violations at the block level, optimizing using both vectorless and vector-based flows. Not bad.

Very interesting stuff. You can learn more HERE.

Posted on

Webinar: Silicon Catalyst & EE Times Examine the Bigger Picture Aspects of AI – Nov. 29, 2023 – 9am to 10am Pacific

Webinar: Silicon Catalyst & EE Times Examine the Bigger Picture Aspects of AI – Nov. 29, 2023 – 9am to 10am Pacific
by Mike Gianfagna on 11-20-2023 at 2:00 pm
Categories: AI, Events, Semiconductor Services, Silicon Catalyst

Webinar Silicon Catalyst & EE Times Examine the Bigger Picture Aspects of AI – Nov. 29, 2023 – 9am to 10am Pacific

Artificial Intelligence (AI) is dominating the news cycle these days. It used to be about the latest (and largest) chips to accelerate AI algorithms. While that’s still relevant and exciting, AI news is taking a much broader, socioeconomic character. What does AI mean for job security, the economy, or even life on Earth? These aren’t hyperbolic comments. Just browse some recent headlines – OpenAI shakeup has rocked Silicon Valley, leaving some techies concerned about the future of AI, Biden signs executive order to promote ‘responsible innovation’ of AI,  Global Leaders Warn A.I. Could Cause ‘Catastrophic’ Harm. AI has become an important topic on the world stage. Recently, Silicon Catalyst assembled a group of technology executives to discuss what our “AI Wonderland” will look like. As an important follow-on to the 6th Annual Semiconductor Industry Forum, up next is a webinar hosted by Silicon Catalyst and EE Times to take a broader look at AI and its impact. Read on to understand how Silicon Catalyst & EE Times examine the bigger picture aspects of AI, beyond the technology innovation cycle.

Previous Live Event Recap

 

 

 

 

 

Silicon Catalyst hosted a well-attended live event at the SRI Conference Center in Menlo Park, CA on Thursday, November 9, 2023. Questions discussed at the event included:

  • What are the AI technologies that will create new business models and industries?
  • What are the implications to semiconductor industry success for incumbents & startups?
  • How do we address the power-hungry AI hyper-scalers’ impact on our energy resources?
  • What impact will potential government and industry regulations have on innovation?

You can read the pre-event coverage on SemiWiki here.  There is also an event replay link on the Silicon Catalyst website, The event was well attended and widely covered. The prestigious IEEE Spectrum covered the event . Part of that coverage included some memorable quotes from the panelists. Some of those deserve repeating here:

  • “If you don’t think of [AI] as a paradigm break or something that can put you out of business, you’re in trouble.” —Deirdre Hanford, Chief Security Officer, Synopsys
  • “AI can take analog libraries and move them from generation to generation [of technology] automatically; this used to be incredibly time consuming, error prone, and difficult.” – Moshe Gavrielov, former Xilinx CEO and now a member of both TSMC‘s and NXP’s board of directors
  • “We are entering an era of (AI-assisted) electronic design creation.” – Ivo Bolsens, Senior VP, AMD

The panelists also exhibited a sense of humor regarding the Alice in Wonderland theme for the event, as shown below.

[left to right] Richard Curtin, David French, Pete Rodriguez, Moshe Gavriolev, Deirdre Hanford, and Ivo Bolsens.

What’s Coming Next

Continuing the theme of the coming AI Wonderland, Silicon Catalyst is teaming up with EE Times for a live broadcast webinar on November 29, 2023 from 9:00 – 10:00 AM Pacific time. The webinar will be moderated by Nitin Dahad, EE Times Correspondent and will expand the conversation on AI to include topics such as:

  • Social Impact
  • Regulatory oversight
  • Investment strategies

The panelists for this webinar are high-profile individuals that span many areas touched by AI. They are:

Jama Adams, COO of Responsible Innovation Labs, www.rilabs.org  

Jama Adams

Jama Adams is COO of a coalition of leading startup founders and investors creating the next generation of high growth technology companies with responsible innovation at their core. Jama was the first hire on Bill & Melinda Gates and Warren Buffett’s Giving Pledge and took the pledge from zero to $1 trillion in three years, growing the ranks to nearly 100 billionaires (today it stands at $3 trillion from 236 signers). She raised $1.8 billion in Series B funding in 12 weeks for Commonwealth Fusion Systems. In other roles – as founder of a consulting firms, on the Omidyar impact investing team, as Communications Director for the City of New York, on political campaigns, on Capitol Hill, and at Harvard Kennedy School, Jama has brought together elite influencers, frontline organizers, business executives, and created new organizations and industries to make change.

Navin Chaddha, Managing Partner, Mayfield, www.mayfield.com  

Navin Chaddha

Under Navin’s leadership, Mayfield has raised eight U.S. funds and guided over 80 companies to positive outcomes. He has been named a Young Global Leader by the World Economic Forum and has ranked on the Forbes Midas List of Top 100 Tech Investors fifteen times, including being named in the Top Five in 2020, 2022 and 2023. Navin’s investments have created over $120 billion in equity value and over 40,000 jobs. During his venture capital career, Navin has invested in over 60 companies, of which 18 have gone public and 27 have been acquired. Navin holds an MS degree in electrical engineering from Stanford University and a B.Tech degree in electrical engineering from IIT Delhi, where he was honored with the distinguished IIT Alumni Award. Navin is an active philanthropist who supports education, diversity, equity, inclusion, and food scarcity groups.

Rohit Israni, Chair AI Standards, US (INCITS/ANSI), www.incits.org

Rohit Israni

Rohit is the founder and CEO of CertientAI. He is also the Chair AI Standards for the US national body responsible for developing US positions and contributions to international AI standards being developed by ISO/IEC (ISO/IEC JTC 1/SC 42). He is also the liaison for SC 42 with Organization of Economic Cooperation and Development (OECD). Prior to founding CertientAI, he led Intel’s global scale ecosystem enabling programs in the Developer Relations Division for AI, Datacenter and Network, of which he co-founded the AI partnership program (Intel AI Builders). Rohit earned a master’s degree in engineering with a specialization in Robotics and AI from Tulane University. He also holds a master’s degree in management science & engineering from Stanford University where he was a part of the Stanford Technology Ventures Program.

To Learn More

This is an impressive group of panelists and a must-see event. Registration information can be found at https://www.eventbrite.com/e/navigating-the-ai-wonderland-tickets-760825638967?aff=oddtdtcreator.  And that’s how Silicon Catalyst & EE Times examine the bigger picture aspects of AI on technology innovation and ultimately, on our lives.

Posted on

China’s hoard of chip-making tools: national treasures or expensive spare parts?

China’s hoard of chip-making tools: national treasures or expensive spare parts?
by Craig Addison on 11-20-2023 at 10:00 am
Categories: China, Lithography
1 Comment

Raiders of the Lost Ark warehouse photo courtesy Paramount Pictures
The Raiders of the Lost Ark warehouse. Photo courtesy Paramount Pictures.

In the closing scene of Steven Spielberg’s Raiders of the Lost Ark (1981), the camera pulls back to reveal an enormous warehouse stacked with crates of artifacts hidden away by the US government. 

One could imagine a similar warehouse in China today. The crates would hold a different kind of treasure – semiconductor production equipment.

Literally tons of foreign chip-making gear has been stockpiled by China in recent years – exceeding what is needed for current or even near-future production levels. That’s why ASML and Lam Research are reporting China sales reaching levels of 46 per cent and 48 per cent of total revenues, respectively, in the third quarter.

The biggest winner from this buying binge is Netherlands-based ASML. Even before the US government tightened export controls on chip-making tools to China, to include the Twinscan NXT1980Di, Chinese buyers were snapping up everything they could lay their hands on.  What’s more, the spending spree is bankrolled by China’s government through massive state subsidies that support local chip champions such as SMIC and YMTC.

Between January and August this year, China imported $3.2 billion worth of semiconductor manufacturing gear from the Netherlands, a 96.1 percent increase from the same period in 2022, according to figures cited in the 2023 Annual Report to Congress.

That’s pretty much all ASML. Do the math and it probably adds up to over 100 machines, given that most would not be high-end DUV systems costing 60 million euros (US$65 million) apiece. ASML has indicated that Chinese customers are buying mostly older machines, and they cost between 10 million and 20 million euros each.

China’s imports of semiconductor production equipment from all countries totaled $13.8 billion over the first eight months of 2023, according to the same report. This is by no means a recent trend. In 2020, Chinese chip makers spent almost $32 billion buying tools from suppliers in Japan, South Korea, Taiwan and elsewhere, a 20 per cent jump from 2019.

“China has already made the decision that it is going to buy all the equipment that it can possibly buy for as long as it can buy it,” said Gregory Allen, an AI and chip expert at the Washington-based think tank Center for Strategic and International Studies, speaking on the Stratechery podcast on October 26.

The Twinscan NXT1980Di system, a workhorse for many Chinese fabs, was introduced by the Dutch firm in 2015. Last week, at a major China trade expo in Shanghai, ASML China president Shen Bo told local media that by the end of this year the company’s installed base of “photolithography machines plus measurement machines” in China would be “close to 1,400”.

A few years ago, an ASML lithography system was even used by a wafer fab start-up in Wuhan, central China, to wrangle more than $2 billion in funding from the provincial government. Two weeks later, the machine was put up as collateral for a $80 million bank loan. That turned out to be a ponzi scheme, with the fab in question, Hongxin Semiconductor Manufacturing Co (HSMC), collapsing at the end of 2020 without producing a single wafer.

So far, China has not been able to get its hands on one of ASML’s EUV machines, which have been strictly forbidden all along. However, Chinese fabs like SMIC are proving adept at stretching their DUV kit to produce 7-nm chips, like the one in Huawei’s Mate 60 5G smartphone.

SMIC is accelerating its acquisition of fab tools. When announcing third quarter results last week, co-CEO Zhao Haijun said the company would ask suppliers to speed-up deliveries to “ensure ramp-up of production at plants that have been initiated”, due to the impact of geopolitics on shipment schedules. SMIC raised its annual capex budget to $7.5 billion for 2023, 18 per cent higher than last year.

What is the risk of China reverse engineering one of ASML’s top-end DUV immersion systems, thus insulating itself completely from export bans? Given enough time, anything is possible. That’s certainly the overarching goal of the Chinese state, according to experts.

“I have not heard of a machine that is currently in China that there is not a Chinese company working on reverse engineering,” Allen said on the Stratechery podcast. “China’s ability to reverse engineer a Lam etching tool in 2010 was pretty crummy. China’s ability to reverse engineer in 2023 has improved considerably… every single component is the same. The spare parts for our machines work as the spare parts for their machines,” he said.

That begs the question: what will China do with its stockpile of foreign semiconductor production equipment if it successfully transitions to locally made gear copied from the West? To be sure, they’ll be stuck with a big warehouse of (very expensive) spare parts.

Also Read:

The Significance of Point Spread Functions with Stochastic Behavior in Electron-Beam Lithography

KLAC- OK quarter in ugly environment- Big China $ – Little Process $ – Legacy good

ASML- Longer Deeper Downcycle finally hits lithography – Flat 2024 – Weak Memory – Bottom?

Posted on

Semiconductors Headed Toward Strong 2024

Semiconductors Headed Toward Strong 2024
by Bill Jewell on 11-20-2023 at 6:00 am
Categories: Semiconductor Intelligence, Semiconductor Services

Semiconductor Market Change 4Q23

The global semiconductor market is now solidly in a turnaround. WSTS revised its data for 2Q 2023 growth over 1Q 2023 to 6.0% from 4.2% previously. 3Q 2023 was up 6.3% from 2Q 2023. With our Semiconductor Intelligence forecast of 3% growth in 4Q 2023, the year-to-year growth in 4Q 2023 will be a positive 6%. This will set the stage for double-digit year-to-year growth for each quarter of 2024.

The top 15 semiconductor companies all reported revenue growth in 3Q23 from 2Q23. The growth ranged from less than one percent at Texas Instruments and Analog Devices to double-digit at Nvidia, Samsung, SK Hynix, and Media Tek. As reported in our September newsletter, Nvidia has passed Intel to become the largest semiconductor company. The 3Q23 Nvidia results will not be official until it reports earnings next week.

The outlook for 4Q23 revenue change is mixed. Of the ten companies providing guidance, five expect revenue to increase from 3Q23 and five expect revenue to decline. The five companies expecting growth are driven by recoveries in PCs and smartphones.

The five companies expecting revenue declines are heavily tied to the automotive industry. TI, Infineon, STMicroelectronics, NXP, and Renesas generally have fared better in 2023 than the memory companies and companies more dependent on PCs and smartphones. These companies are not pointing to a decline in automotive business in 4Q23 but are citing other factors. TI foresees a weak general environment. Infineon stated auto would be up in 4Q23, but other segments are sluggish. ST pointed to weakness in industrial. NXP blamed the communications infrastructure segment. Renesas cited inventory adjustments.

The United Auto Workers (UAW) union went on strike against the big three U.S. automakers – General Motors, Ford and Stellantis – from mid-September until the end of October. The strike led to a 10% decline in U.S. production of motor vehicles and parts in October. Generally, the above companies did not comment on the strike in their earnings reports or conference calls. Renesas stated it expected the strike to have some impact on revenue in either 4Q23 or in 2024.

For 2024, two of the key drivers of the semiconductor market are expected to recover. IDC forecast smartphone units will increase 4% in 2024 after a 5% decline in 2023. IDC projects PC units will grow 4% in 2024 following a steep 14% drop in 2023. In contrast, production of light vehicles should moderate to 3.7% growth in 2024 from 4.7% in 2023, according to Statista.

Thus, we should expect the strongest revenue growth in 2024 from the memory companies (Samsung, SK Hynix, and Micron), the PC and computing focused companies (Nvidia, Intel, and AMD), and the smartphone focused companies (Qualcomm and MediaTek). The companies focused primarily on automotive, industrial and IoT (TI, Infineon, ST, NXP and Analog Devices) will show relatively less revenue growth.

Forecasts for the year 2023 semiconductor market change are homing in on the minus 9% to minus 12% range. Our Semiconductor Intelligence (SC-IQ) projection is minus 9.5%. The momentum headed into 2024 is clear. Forecasts for 2024 range from a healthy 9% from Future Horizons to a strong 20% from IDC. Our SC-IQ projection is 16% growth. A major factor determining whether 2024 will be closer to 9% or to 20% is the degree of price increases in memory.

The semiconductor market recovery is solidly underway. Despite the bright outlook, many companies are cautious as the global economy is still recovering from the COVID-19 pandemic and its aftereffects. Political tensions are high, particularly between the two largest economies – the U.S. and China. Wars are continuing in the Ukraine and Israel/Gaza. Regardless of these uncertainties, we at Semiconductor Intelligence believe we are poised for a healthy semiconductor market in 2024 and beyond.

Semiconductor Intelligence is a consulting firm providing market analysis, market insights and company analysis for anyone involved in the semiconductor industry – manufacturers, designers, foundries, suppliers, users or investors. Please contact me if you would like further information.

Bill Jewell
Semiconductor Intelligence, LLC
billjewell@sc-iq.com

Also Read:

Electronics Production Trending Up

Nvidia Number One in 2023

Turnaround in Semiconductor Market

Posted on

AMAT- Facing Criminal Charges for China Exports – Overshadows OK Quarter

AMAT- Facing Criminal Charges for China Exports – Overshadows OK Quarter
by Robert Maire on 11-19-2023 at 8:00 pm
Categories: Semiconductor Advisors, Semiconductor Services
1 Comment

Criminal Investigation

– News that AMAT facing criminal charges for violating China sanctions
– Sounds like Ion implant tools sent to SMIC in China via South Korea
– Criminal probe goes back to shipments starting in 2021 & 2022
– Overshadows an otherwise OK quarter and outlook

Reuters broke the news on the day of earnings results

Link to Reuters story on AMAT Criminal probe for export violations

It sounds from the Reuters article that the probe into AMAT has been going on for quite a while as it goes all the way back to shipments starting in 2021 after SMIC was put on the “bad list” in the end of 2020.

The probe is being handled out of Boston by the US attorney’s office and the National Security Unit, which points to Applied Ion Implant business in Gloucester Massachusetts.

The report seems to indicate that Applied circumvented the export restrictions by shipping the tools to their subsidiary in South Korea, then cross shipped to SMIC in China.

If proven true, this is a huge violation of the sanctions on China by the biggest US equipment maker.

As with any criminal investigation, Applied refused to comment on the ongoing matter and made no mention of it in prepared remarks on the quarter

Potential Huge penalties

Given that SMIC managed to essentially give a slap in the face to the US by manufacturing 7NM devices that they shouldn’t have been able to, we would imagine that anyone who violated the export sanctions will be harshly dealt with and made an example of.

We would imagine hundreds of millions in potential fines and/or restrictions on future sales to China and living under a microscope with any future export licenses likely subject to delays and more restrictions.

Its very, very bad when you take into consideration that China is Applied’s largest customer and Applied has a huge operation in China.

Likely to threaten Applied’s request for CHIPS act funding

As we had previously pointed out in prior reports we were very dubious of Applied Materials applying for CHIPS Act funding given that they are actively exporting jobs to Asia and specifically Singapore by building an even larger facility in Singapore than they plan for California.

If they are indicted, cut a deal or are found guilty we would imagine that the US government would/should not give them any funding out of the CHIPS Act budget. We can’t imagine the government funding a company that would circumvent the China sanctions.

We also pointed out in our most recent note that increased equipment sales to China was at the expense of other US semiconductor companies like Global Foundries and Micron. Applied sales to SMIC clearly would/will hurt Global Foundries directly.

Porosity of Sanctions continues to haunt the semiconductor industry

We have continued to point out in a number of our notes over the last year or more that the China sanctions were clearly both porous and ineffective, especially in light of SMIC and 7NM.

Equipment companies are well incentivized to circumvent sanctions especially since the industry is in a downturn and China is one of the few regions that continues to buy in an accelerating manner, while the rest of the industry is way down.

We can only imagine that US allies such as the Netherlands and Japan will look at this, if proven true, and wonder why they are going along with sanctions and hurting their own business.

Comes on the heels of a 741 page congressional China report

We put out our most recent note focusing on China semiconductors this past Monday.

Coincidentally on Tuesday (we did not have advance notice: The Coming China Chipocalypse – Trade Sanctions Backfire – Chips versus Equipment , Congress put out a 741 page report pointing out that China sanctions were ineffective.

Now on Thursday we hear about the criminal investigation of Applied Materials.

Link to Congressional report on China Sanctions

Doesn’t matter that AMAT had a good quarter

The criminal news story way overshadows any hint of positive quarterly results in our view.

Applied reported flat Y/Y revenues of $6.72B and non GAAP EPS of $2.12. Guidance was for $6.47B+-$400M and EPS of $1.72 to $2.08. Both results and guide were above expectations.

Applied talked about leading edge business improving while ICAPS (trailing edge) will likely be lower. NAND will be less than 10% of business but DRAM will be stronger.

Given that we are in a downcycle the results are good but not stellar and certainly does not nearly offset the bigger news.

The Stocks

A huge “Wet Blanket”

As the news of the criminal investigation was out along with earnings, the stock was down 7.5% in the after market.

We would imagine there is more downside potential as other investors may sell tomorrow as they don’t want to hang around and wait on the results of the investigation.

The potential negative impact on CHIPS Act money or future licenses and business with China throws a huge wet blanket over the companies outlook.

Unfortunately, there could also be collateral damage on other equipment stocks through guilt by association and exposure to China.

We can only imagine that investigators may want to look into Lam or KLAC given their exposure to China which we previously called out as a significant risk.

Even if this is a “witch hunt” by a government that is pissed about SMIC spitting in their face….witches still get burned or drowned….guilty or innocent…..

Semiconductors in general have been under pressure over China and this will add to uncertainty and concern and related discount in the stocks.

This could be the beginning of a long drip, drip, drip of an investigation and bad news.

With the risks in front of the stock we would be hard pressed to own it given the downside risk as well as the huge uncertainty associated with it.

We would be cognizant of the potential collateral fallout as well.

Not a Happy Thanksgiving……

About Semiconductor Advisors LLC
Semiconductor Advisors is an RIA (a Registered Investment Advisor), specializing in technology companies with particular emphasis on semiconductor and semiconductor equipment companies. We have been covering the space longer and been involved with more transactions than any other financial professional in the space. We provide research, consulting and advisory services on strategic and financial matters to both industry participants as well as investors. We offer expert, intelligent, balanced research and advice. Our opinions are very direct and honest and offer an unbiased view as compared to other sources.

Podcast EP194: The Impact of Joining TSMC’s OIP From the Perspective of Agile Analog

Podcast EP194: The Impact of Joining TSMC’s OIP From the Perspective of Agile Analog
by Daniel Nenni on 11-17-2023 at 10:00 am

Dan is joined by Chris Morrison. Chris has 15 years’ experience of delivering innovative analog, digital, power management and audio solutions for international electronics. Currently he is the director of product marketing at Agile Analog, the analog IP innovators. Previously he has held engineering positions, including 10 years at Dialog Semiconductor, now acquired by Renesas.

Chris discusses Agile Analog’s recent entry into the TSMC Open Innovation Platform ecosystem. He shares his experiences attending the event in both Santa Clara and Amsterdam. The impact of access to TSMC’s advanced technology nodes is discussed, along with the very positive customer interactions Agile Analog experienced at both locations for OIP.

Chris discusses the markets where Agile Analog is finding significant interest and plans for the coming year.

The views, thoughts, and opinions expressed in these podcasts belong solely to the speaker, and not to the speaker’s employer, organization, committee or any other group or individual.

Posted on

CEO Interview: Dr. J Provine of Aligned Carbon

CEO Interview: Dr. J Provine of Aligned Carbon
by Daniel Nenni on 11-17-2023 at 6:00 am
Categories: CEO Interviews
1 Comment

Cyclotron Road Cohort Five

J Provine, PhD, is the CEO and co-founder of Aligned Carbon, a startup that has developed a breakthrough synthesis and purification technology for aligned, semiconducting carbon nanotubes (CNTs) at wafer scale. His team believes that Aligned Carbon is the crucial missing piece in an industry that has long been hopeful of – yet historically disappointed by – the promise of CNTs.

Dr. Provine has over 20 years of experience in nanofabrication, beginning with his PhD studies at Cornell University, with a deep expertise in MEMS, atomic layer deposition, and a variety of semiconductor technologies. Continuing through the decade-plus he spent as a senior research scientist and Adjunct Professor at Stanford University, Dr. Provine consulted widely with established companies (Veeco, Bosch) and early-stage startups (Quantumscape, Spark Thermionics). He continues to be deeply involved in activities with the NSF and DOE. Dr. Provine also mentors other deep tech startup founders.

Tell us a little about your journey. How’d you end up in the driver seat of a semiconductor company?
I have worked in nanotechnology, particularly semiconductor-related research, since the late ‘90s. Before that, I was fixated on particle physics. I never planned on starting a company. I felt at home in the university setting. Over my career, I started spending more time consulting companies. I had great collaborators and PIs at Stanford, but I was searching for an opportunity to stretch myself in more of a leadership role, building a team with a focused technology that could greatly impact computing. I thought that would happen in the academic setting, but I saw so many revolutionary technologies get stuck in academia, so I looked outside the ivory tower. The technology behind what would become Aligned Carbon was the most interesting, high-value opportunity I had ever seen, and so I decided to take the leap and start the company. I saw the potential for my particular skills and background to improve material manufacturing, and that continues to drive me forward today.

Tell Us about Aligned Carbon. What have you been working on?
Aligned Carbon manufactures carbon nanotubes (CNTs) for use in integrated circuits. There has been a tremendous amount of research showcasing the promise of CNTs for use in nanoelectronics. Particularly, my Stanford colleagues Prof. Philip Wong and Prof. Subhasish Mitra led research programs in this area for decades. This research never made a commercial impact: issues with the CNT starting material played a big role in inhibiting broader applicability. We’re changing that at Aligned Carbon.

The semiconductor ecosystem is a huge, global endeavor. Aligned Carbon’s north star has been to revolutionize high performance computing, where gains in speed and energy efficiency can make an enormous impact. The impact CNTs can have for integrated circuits is even broader, including RF data transmission, power circuitry, satellite/space based circuits, and sensors. At Aligned Carbon, we are continuing to push the industry’s understanding of and capabilities in these areas.

What problems are you trying to address and why are CNTs uniquely suited to address them?
For 60 years, silicon has been able to solve most of our semiconductor needs for integrated circuits. Unique materials have emerged over the last few decades that have added special capability where silicon was not able to provide optimal performance. We have seen this in GaAs and other III-V semiconductors for photonics, and now GaN and SiC for power electronics. The unique material properties of CNTs, including high mobility, small body length, and enormous current density capacity, have such enormous potential to drive circuitry forward. Additionally, CNTs can deliver high performance transistors for logic with a low thermal budget (<300C), which enables monolithic integration on top of existing circuitry or temperature-limited substrates. This match of performance and integration is unique and marks CNTs as the best candidate for monolithically-integrated 3D logic.

Carbon Nanotubes went through a hype cycle before but did not pan out. Why, and what have you guys solved uniquely?
At IBM in the 2000’s, so much great work happened, but there were many issues with manufacturability and reliability of the material. At the time, there was an emphasis on directed synthesis/growth of CNTs, growing the specific kind of CNT you wanted (semiconducting vs metallic or even single chirality*) right where you wanted it. That proved to be a very challenging – perhaps impossible – problem, particularly when targeting a semiconducting-to-metallic CNT ratio of greater than 100,000:1. Sorting CNTs by polymer conjugation, commonly referred to as solution purification, has made great headway in solving the semiconductor-to-metallic ratio, but at the cost of CNT order and maintaining pristine CNT material.

Aligned Carbon has been focused on a route to having high semiconducting purity while maintaining arrays of CNTs with perfect alignment (no crossings or bundling), all without polymer contamination. This approach means you can have a “thin film” of aligned, purely-semiconducting CNTs that you can pattern as needed with traditional photolithography and etch to make devices.

*Chirality defines the structural and electronic properties of a single-walled carbon nanotube. A vector connecting the centers of the two graphitic hexagons is called the chiral vector, and it determines the structure of a single-walled carbon nanotube. A chiral vector C can be specified by a chiral index (n,m) using basis vectors a1 and a2 of a graphene sheet.

You mention a breakthrough. Tell us more about that.
Earlier this year, Aligned Carbon validated its purification process to greater than 99% semiconducting purity while maintaining fully-aligned material and no discernable polymer contamination on the CNTs. Further, there is a clear line of sight to scaling this to >99.9999% semiconducting purity, with our throughput capable of purifying a 300mm wafer in a few minutes. There is more work to be done, but we are very excited to put this material to work and find its limitations as we continue to refine and improve our unparalleled CNT synthesis and purification.

With this breakthrough, what application areas are you putting your focus on?
While we have always had digital logic as our target application, there are many other opportunities that we are actively exploring. The industry has just scratched the surface on the incredibly powerful new properties of CNTs.

We are digging into the use of CNFETs in RF data transmission, due to sub-THz signal linearity, and power electronics, due to high current-carrying potential. We have also previously looked at a few different types of CNT sensors. Another interesting opportunity is circuitry for use in space, as CNFETs are ~16 times less radiation- sensitive than silicon devices because of their reduced interaction cross-section. All of these are focused around making transistors out CNTs – which is where our material has been focused – and is a very fertile ground.

We are getting some interest across each of these spaces, which is really encouraging, and we do see our expertise in the material lending itself nicely to working closely with partners that like working on major innovations at the device layer.

What does the competitive landscape look like, both within carbon nanotubes but also when it comes to other technologies. Why do you win?
Carbon nanotubes are a remarkable material that have the potential to far outperform all bulk semiconductors out there. But, they need to meet the stringent requirements of a mature and highly-optimized industry. The great promise of CNTs and the hard limits that silicon has reached make that effort well worth it. One type of competition is other emergent materials, such as 2D materials, which possess intriguing material properties. But, they represent a less mature technology/material set that needs even more substantial manufacturing and reproducibility work than CNTs. So for me, CNTs really represent the most promising transistor channel material out there. The key to having reliable CNTs for the IC industry lies in the purity of electronic properties of order of the material. Aligned Carbon has a unique approach to delivering this material that maintains the quality of pristine CNTs and integrates cleanly and at volume with established IC manufacturing.

Solutions for other materials, integration schemes, and architectures are another form of competition for us. That said, the needs of the IC industry are complex, multi-faceted, and large. We at Aligned Carbon will be pushing as hard as possible to bring CHTs into the mix, knowing others are doing the same with their technologies.

What sort of partnerships or companies are you looking for right now?
We are excited to put our material to the test and iterate with collaborators/partners on a variety of transistor applications. We are confident in our ability to supply consistent material for R&D validation. Further, we have a lot of options for tuning our material, including pushing to much higher purity. Qualifying a new material for use in a foundry takes significant time, resources, and effort, but we have a unique material offering that innovative partners can utilize to meet the various challenges in the IC industry.

We as a society have a nearly insatiable desire for greater computing capability. The IC industry is rightfully concerned about that hunger causing significant issues in energy consumption and environmental impact. CNTs are uniquely poised to play a larger role in threading that needle, but we need to be vigilant about validating and verifying the technology to make it ready for mass market prime-time.

How can people get in touch?
We are really excited to talk with folks about our advanced technology. With the major material-level renaissance currently happening in the semiconductor space (as shown in power electronics), we are excited to get our products in front of people.

If you are interested in getting in touch, please feel free to reach out to info@alignedcarbon.com, as our team would be happy to chat. Also feel free to follow our company’s LinkedIn [Linkedin] for regular updates or to connect with our core team.

Also Read:

Executive Interview: Tony Casassa, General Manager of METTLER TOLEDO THORNTON

CEO Interview: Venkata Simhadri of MosChip

CEO Interview: Pat Brockett of Celera

Posted on

RISC-V Summit Buzz – Ron Black Unveils Codasip’s Paradigm Shift for Secured Innovation

RISC-V Summit Buzz – Ron Black Unveils Codasip’s Paradigm Shift for Secured Innovation
by Mike Gianfagna on 11-16-2023 at 10:00 am
Categories: Codasip, Events, IP

Ron Black
Ron Black

Codasip is a processor solutions company with an expanding footprint. It is Europe’s leading RISC-V organization with a global presence. Codasip reports billions of chips already use its technology.  You can learn more about Codasip here, The company has made some announcements recently that expand its offerings in terms of application areas and processor options. Dr. Ron Black has been the CEO at Codasip for almost three years. Previously, he was president and CEO at Imagination Technologies, CEO at Rambus, MobiWire, UPEK, and Wavecom. At his RISC-V Summit keynote, Ron discussed some major challenges for data security and an innovative approach to address those challenges. His presentation definitely got some attention. Read on to understand how Ron Black unveils Codasip’s paradigm shift for secured innovation. 

Leading Up to the Summit – Expanding Footprint

In mid-October, Codasip announced the 700 RISC-V processor family, expanding its offering beyond embedded processor IP to stand-alone application processors. In its own words, “bringing the world of Custom Compute to everyone.” The 700 family is a configurable and customizable set of RISC-V baseline processors. It is intended to complement Codasip’s embedded cores by offering a different starting point to accommodate the need for higher performance.

Codasip Studio delivers a streamlined design process that unleashes the potential of the 700 family. Examples of the power of Codasip Studio include:

Enabling different levels of processor optimization for each use case:

  • Through profiling 
  • Giving predictable results
  • Improving time-to-market with a proven, highly automated approach​

Automating processor design and custom SDK & HDK generation:

  • Better results achieved in a more efficient way
  • ​Accelerated exploration, investigation, and configuration
  • Allowing maximal optimization

Fostering collaboration between hardware and software teams:

  • To improve system efficiency
  • ​Allowing teams to analyze and suggest changes to the hardware team
  • Delivering differentiation and time-to-market in a highly competitive environment

The first core in the 700 family was also introduced. Codasip A730 is a 64-bit RISC-V application core that is being provided to early-access customers. It supports single-core or multi-core configurations and is Linux capable (MMU) with up to 2x the performance of the previous core. It employs a faster, tightly coupled cache coherent memory system as well. The core has broad market application from edge IoT to AI and sensor fusions. As you will see, the 700 family is also relevant to Codasip’s paradigm shift for secured innovation.

Ron’s Keynote at the Summit – Addressing the Elephant in the Room

Data security is a huge and growing problem. The specter of a data breach and the subsequent havoc associated with identity and monetary theft is on everyone’s mind. The headlines below illustrate the scope of the problem.

The problem at times can seem insurmountable. Huge amounts of money are lost each year from data breaches. There are so many “attack surfaces” and loopholes to worry about. Ron provided a rather startling insight about the problem at his keynote. It turns out a very large part of the problem is due to illegal memory access – 70 percent or more of the problem has this root cause. And more interesting, it’s been that way for a very long time. The graphic below drives home the point.

Ron then explored the dimensions of the problem – how big is it and what, if anything can be done to reduce the risk and financial loss. It turns out there is a substantial amount of legacy C/C++ code that is vulnerable. In-house, third-party, and open-source streams all contribute to the problem. And the incidence of data breaches is growing exponentially, as shown below.

So, what can be done to plug this rather large security hole? Ron outlined the options.

One could deploy tools to identify and fix all leaks and potential leaks in existing code. This would entail re-writing trillions of lines of code, perhaps with a memory safe approach like Rust. A direct, but intractable solution.

But what if you could use a specialized security processor to catch vulnerabilities natively? Something that could integrate with existing applications. This is a much better way, but does such an option actually exist?

Ron announced the news – both good and bad. The good news is that such a system does indeed exist. Capability Hardware Enhanced RISC Instructions (CHERI) is a specification developed by the University of Cambridge. It’s an architecture extension introducing hardware technology that mitigates software security vulnerabilities. It extends conventional hardware ISAs with memory protection and scalable compartmentalization. Something like this holds great promise since it can be integrated into existing systems with minor work vs. a major re-write.

There are several research projects underway for CHERI from the University of Cambridge and Microsoft, among others. And there is the bad news. It’s a research effort, not a production implementation. Until now.

Ron announced that Codasip is taking CHERI from a research topic to a commercial product thanks to RISC-V, CHERI, and the Codasip Custom Compute methodology. He explained that RISC-V is modular and allows customization. The company has taken its new Codasip 700 baseline processor design and enhanced it for security with CHERI technology using Codasip Studio.

This approach includes built-in, fine-grained memory protection by extending the RISC-V ISA with CHERI-based custom instructions. To enable the use of these instructions, Codasip is also delivering the software environment to take advantage of CHERI technology, bringing a full software development flow to add memory protection. Thanks to these developments, the world can now be a safer place with regard to data protection. Ron announced that Codasip was demonstrating this new core at the RISC-V Summit and lead customers would get early access to core with CHERI in the second half of 2024.

Comments From Ron

I got a chance to speak with Ron after his keynote. He likened data security to automobile insurance. No one really wants to pay for it, but you definitely want it just before you’re involved in an accident. Data breaches are like those automobile accidents. You need insurance in place at all times to truly be protected. With the introduction of a commercialized version of CHERI, that insurance is now available at a more reasonable cost, when compared with re-writing a massive code base.

Ron pointed out that CHERI is a cross-platform technology. Its benefits are not specific to RISC-V, but the open architecture of RISC-V make the implementation of CHERI more accessible. Will this new capability for improved data security find widespread use in the RISC-V community and will other architectures follow with commercial CHERI capabilities? Time will tell, but this innovation from Codasip could be the start of something big. And that’s how Ron Black unveils Codasip’s paradigm shift for secured innovation.