Breaking the Bottleneck | Issue 65
[11/18/2024] On Chip Test, Gen AI in Supply Chain & Manufacturing Security
Breaking the Bottleneck is a weekly manufacturing technology newsletter with perspectives, interviews, news, funding announcements, manufacturing market maps, and a startup database!
💥 If you are building, operating, or investing in manufacturing, hardware, or robots, please reach out. My email is aditya@schematicventures.com – I’d love to connect!
🏭 If you were forwarded this and found it interesting, please sign up!
Content I Enjoyed Last Week 🏭🗞️🔬 📚
News:
Machine Learning Might Save Time on Chip Testing [IEEE Spectrum]
Engineers at NXP have developed a machine-learning algorithm that significantly reduces the number of tests required for chips destined for critical systems in cars, potentially cutting 42% to 74% of tests without compromising quality. Traditionally, extensive testing—adding 5% to 10% to the cost of a chip—is conducted on finished chips from the foundry, especially for automotive applications where safety is paramount. The new algorithm learns patterns from test results to identify which tests are essential and which can be safely omitted. The team analyzed data from seven microcontrollers and application processors built using advanced chipmaking processes. They were subjected to between 41 and 164 tests, depending on the chip. The algorithm recommended removing 42% to 74% of these tests. The algorithm is currently a pilot project, with plans to expand it to a broader set of parts, reduce computational overhead, and enhance usability. Sriharsha Vinjamury, a principal engineer at Arm, remarked, "Any novel solution that helps in test-time savings without any quality is valuable. Reducing test time is essential, as it reduces costs." He suggested that the NXP algorithm could be integrated with systems that adjust the order of tests to spot failures earlier.
See Also: Is Virtualization Greener Than Lab Work for Chips?
‘Robot Revolution’ Forces China’s Human Workforce to Adapt [FT]
Chinese manufacturers rapidly adopt advanced automation and robotics to stay competitive amid rising labor costs and a shrinking working-age population. Beijing has embraced this "robot revolution" to tackle labor shortages due to an aging population, offering tax breaks and subsidies to encourage investment in automation. According to the International Federation of Robotics, China has become the world's largest market for industrial robots, installing over 276,000 units last year—more than half of the global total. However, the success of this automation push hinges on whether the workforce possesses the necessary skills to operate and maintain sophisticated machinery. China's manufacturing sector relies heavily on nearly 300 million migrant workers, with 52% having only a middle school education and 14% having just a primary school education as of last year. These workers are the most likely to be displaced by robots and are increasingly shifting to lower-paid service sector jobs like food delivery. Furthermore, training remains a critical challenge. Provinces like Guangdong have launched programs to educate a new generation of workers. Still, local universities and technical colleges often lack up-to-date equipment, relying instead on textbooks or outdated machinery. The most effective training is provided directly by robot suppliers. Workers respond by either retiring early or engaging in technical training to gain an edge over machines. Osea Giuntella, associate professor of economics at the University of Pittsburgh, observed that workers perceive the economy as changing and feel compelled "to undergo training or to go into retirement because the investment in their human capital is not worth it."
Scientists Make First Mechanical Qubit [IEEE Spectrum]
Scientists at the ETH Zurich have created the first mechanical qubit, marking a significant advancement in quantum computing. Traditional qubits rely on superpositions of electronic states—such as different levels of electric charge—but these electromagnetic qubits suffer from short coherence times before their quantum states decay, limiting their usefulness. The new mechanical qubit behaves like a microscopic drum skin that can exist in a superposition of both vibrating and not vibrating simultaneously, akin to Schrödinger’s cat. Mechanical qubits depend on superpositions of vibrational states, which theoretically can possess longer coherence times than electromagnetic qubits. The development opens the door to mechanical quantum computers capable of running long, complex programs and creating novel quantum sensors. These mechanical qubits could perform quantum gates—the quantum equivalent of logic gates in classical computing. The researchers highlighted potential applications in sensing, noting, "Our system lets us measure gigahertz-frequency mechanical forces caused by, for example, gravitational waves. There's not really a basis of comparison for this as far as we know." The researchers believe they can significantly improve coherence times with different designs and materials, potentially making quantum sensors far more accurate than conventional devices.
$6B Bid to Clean Up Heavy Industry At Risk [Canary Media]
The Biden administration's $6 billion Industrial Demonstrations Program, aimed at reducing carbon emissions from heavy industries, which accounts for nearly one-third of U.S. carbon dioxide emissions annually, is at risk under the second Trump administration. Announced last year, the program supports U.S. manufacturers in electrifying and decarbonizing industrial processes by providing significant cost-sharing grants. However, most of the funding has not yet been distributed or legally obligated. Only ten projects have entered early award negotiations and received initial planning funds, while 23 projects are still in pre-award negotiations without formal agreements. This uncommitted funding could potentially be rescinded when President-elect Donald Trump takes office in January. Ryan Fitzpatrick, senior director of domestic policy at Third Way, noted, "The program is vulnerable." However, the jury is still out as the program is set to support tens of thousands of jobs and stimulate billions in economic development, particularly in Republican-held states and traditional manufacturing regions. For instance, Cleveland-Cliffs could receive up to $500 million to build a green steel facility in Middletown, Ohio. Century Aluminum was selected for a $500 million award to construct a carbon-free, energy-powered smelter in Kentucky or the Ohio and Mississippi River basins. Industry stakeholders express cautious optimism about the program's longevity. Todd Tucker of the Roosevelt Institute added that industries recognize the need to adopt these technologies to remain viable in a low-carbon global economy, suggesting they have little interest in seeing the programs dismantled.
Research:
How GenAI Reimagines Supply Chain Management [BCG]
Companies have been slow to realize the vision of AI-driven supply chains due to an overemphasis on AI's analytical powers without adequate focus on adaptive learning—the ability to improve over time with use. "AI solutions, laden with complexity, often overwhelm supply chain personnel, leading to low adoption and diminished returns on investment (ROI)," the research notes. Challenges arise from outdated processes designed around legacy systems, disconnected data foundations with poor quality and multiple systems of record, and employees who struggle with complex, non-user-friendly AI applications. BCG believes Gen AI provides four primary benefits:
Enhancing the Data Backbone: GenAI strengthens data foundations, enabling broader and more accurate data usage across applications by continuously cleaning and augmenting master data sets.
Augmenting Supply Chain Analytics: By extracting meaningful attributes from unstructured data, GenAI enhances capabilities like product demand forecasting and predicts disruptions more accurately, allowing for more proactive decisions.
Overhauling the User Experience: GenAI simplifies the use of sophisticated tools through natural language interfaces, making supply chain professionals more open to adoption and increasing user satisfaction and system usage by >60%.
Deeply Automating Processes: By coordinating multiple tools and driving workflows toward desired outcomes, GenAI automates processes that previously required significant manual intervention, reducing administrative and data reconciliation tasks by over 50%.
These benefits result in substantial improvements: accelerating the development of complex applications and supply chain solutions by 30%, boosting decision-making speed by more than 30%, and freeing up personnel to focus on higher-value activities.
Local Policies Enable Zero-shot Long Horizon Manipulation [CMU]
ManipGen leverages a new class of policies for sim2real transfer: local policies. Locality enables a variety of appealing properties, including invariances to absolute robot and object pose, skill ordering, and global scene configuration. We combine these policies with foundation models for vision, language, and motion planning and demonstrate the SOTA zero-shot performance of our method to Robosuite benchmark tasks in simulation (97%)
Podcasts/Video:
Security Bookmarked, Manufacturing [Bloomberg]
Manufacturing Deals🏭💵
Orderful - A company building a modern EDI integration platform
$15 million [Venture] - Led by NewRoad Capital Partners with participation from Andreessen Horowitz, 9Yards Capital, Flume Ventures, and GLP Capital Partners
Vecna Robotics - A material handling robotics company that delivers rapid ROI in under one year
$14.5 million [Venture] - Led by Drive Capital and Tiger Global
Brightpick - A company building AI robots to enable warehouses of any size to fully automate order picking, consolidation and dispatch
$12 million [Venture] - Led by Photoneo
Downtime 🏭🧑🔧
The Amateur