Breaking the Bottleneck | Issue 61
[10/7/2024] AlphaChip, MIT & Prologis Supply Chain Reports, & More!
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:
How AlphaChip Transformed Computer Chip Design [Deepmind]
Last week, Google released an addendum to its 2020 model AlphaChip, a novel reinforcement learning method for designing chip layouts that accelerates and optimizes chip design by generating chip layouts in hours. AlphaChip approaches chip floorplanning as a game, similar to AlphaGo and AlphaZero. Starting from a blank grid, it places one circuit component at a time, receiving rewards based on the quality of the final layout. A novel "edge-based" graph neural network enables AlphaChip to understand the relationships between interconnected chip components and generalize across different chips, improving with each layout it designs. This method addresses the complex challenge of chip layout design, which involves intricate, interconnected blocks and layers of circuit components connected by extremely thin wires. The AI has been crucial in designing Google's TPU AI accelerator chips. AlphaChip has generated superhuman layouts in every generation of Google's TPUs since 2020, including the latest Trillium (6th generation). These TPUs power Google's advanced AI systems, such as large language models like Gemini and image and video generators like Imagen and Veo. AlphaChip achieved an average wire length reduction across three TPU generations, leading to more efficient chip designs. Beyond Google, it has been adopted by external organizations like MediaTek, one of the world's leading chip design companies. Professor Siddharth Garg from NYU Tandon School of Engineering stated, "AlphaChip has inspired an entirely new line of research on reinforcement learning for chip design, cutting across the design flow from logic synthesis to floorplanning, timing optimization, and beyond." The method has sparked an explosion of work on AI for chip design, extending to other critical stages like logic synthesis and macro selection. Google is developing future versions of AlphaChip and aims to collaborate with the community to revolutionize chip design further, envisioning a future where chips are faster, cheaper, and more power-efficient.
AI-Generated Images Can Teach Robots How to Act [MIT Tech Review]
Researchers from Stephen James's Robot Learning Lab in London have developed a new system called Genima that fine-tunes the image-generating AI model Stable Diffusion to create training data for robots. Traditionally, teaching a robot to complete a task involves training a neural network on an image and producing an output in a different format, such as coordinates for movement. Genima's approach differs by using images for input and output, making it easier for machines to learn and more interpretable for users. The system overlays data from robot sensors onto images captured by cameras, rendering desired actions into a series of colored spheres that indicate where the robot's joints should move in the next second. The researchers used another neural network, ACT, to convert these spheres into actionable movements. In tests involving 25 simulations and nine real-world manipulation tasks with a robot arm, Genima achieved average success rates of 50% and 64%, respectively. While these rates are modest, the team is optimistic about improving speed and accuracy, especially by applying Genima to video-generation AI models to predict sequences of future actions.
How to Restart a Nuclear Reactor [IEEE Spectrum]
Data centers' growing demand for clean electricity leads to the potential revival of shuttered nuclear reactors in the United States, including the infamous Three Mile Island Unit 1 in Pennsylvania. Constellation Energy announced plans in September to restart Unit 1, closed in 2019 for economic reasons and sits adjacent to Unit 2. The revived reactor would supply 835 megawatts to grid operator PJM, with Microsoft entering a 20-year power purchase agreement to buy the generated power for its data centers in the PJM Interconnection. Restarting Unit 1 involves overcoming significant engineering and regulatory challenges, such as restoring the turbine, generator, main power transformer, and cooling and control systems. Experts like Kathryn Huff, former assistant secretary of nuclear energy at the U.S. Department of Energy, note that existing infrastructure like the Susquehanna River for cooling and established high-voltage transmission lines make the revival feasible. If successful, Three Mile Island Unit 1 would be one of the first reactors in the world to be resurrected, following the Palisades Nuclear Plant in Michigan, which is on track to resume operations in 2025 after securing a $1.5 billion loan from the Department of Energy. The trend of reviving nuclear plants is driven by the energy-intensive needs of data centers seeking constant, clean power to meet emissions reduction targets. Amazon Web Services also invested $650 million in a data center powered by a neighboring nuclear plant. Regulatory processes, community support, and the physical condition of the plants play crucial roles in these efforts. While challenges remain, including public perception and technical upgrades, the movement suggests a potential nuclear renaissance to support modern energy demands, particularly for Microsoft and Amazon.
Michigan Bets on EVs and Hopes Jobs Will Follow [Canary Media]
Detroit is experiencing a resurgence driven by clean energy initiatives. Governor Gretchen Whitmer launched an ambitious plan to reinvigorate Michigan's industrial base, and private companies have pledged over $11 billion to build clean energy projects, which is expected to create nearly 22,000 jobs. Notable investments and clean energy projects include:
Our Next Energy (ONE): A lithium ferrous phosphate (LFP) battery company offering safer and more cost-effective options for EVs plans to build a $1.6 billion factory to create 2,000 jobs.
Ford: Constructing the $2.5 billion BlueOval Battery Park, employing 1,700 workers. However, Ford's EV division lost $1.3 billion in Q1 2024, selling only 10,000 vehicles, and canceled a planned electric SUV, citing profitability concerns.
GM: Building a $2.6 billion battery plant in Lansing
Gotion: A subsidiary of a Chinese battery company approved to build a $2.4 billion facility in western Michigan, expected to employ over 2,300 people
Nxlite: A startup producing energy-efficient windows using physical vapor deposition, potentially reducing a building's energy use by 40%.
LuxWall: Opened a facility in Litchfield producing vacuum-insulating glass, with plans to employ 450 people, supported by over $31 million in federal funding.
Nel: A Norwegian company committed to a $400 million factory in Michigan to produce electrolyzers for clean hydrogen production.
While Democrats promote their climate and clean energy policies, challenges remain, including delayed factory constructions and financial hurdles startups like ONE face. The full benefits of the clean energy manufacturing boom are yet to materialize, and the continuity of supportive policies is crucial for long-term success.
Is Your Frontline Workforce Strategy Right For Where You Are? [McKinsey]
In 2023, federal spending on U.S. manufacturing construction reached $196 billion, marking a 143% increase since 2019, mainly due to investments like the $39 billion CHIPS Act to boost advanced manufacturing facilities. Despite this influx of capital intended to fuel the manufacturing sector, the industry faces significant workforce challenges, including severe labor shortages and stagnant growth. As of January 2024, there were 622,000 unfilled manufacturing job openings nationwide, with the overall labor force participation rate declining from 67% in the 1990s to below 63% in 2023. Skills gaps exacerbate the issue, making it seem like a surplus of workers isn't meeting demand. Executives report higher absenteeism, increased attrition within the first 90 days, and mismatched expectations with newer generations of talent. Manufacturers must address these challenges by understanding local labor market dynamics, which vary significantly across regions. McKinsey's approach yields four distinct labor market archetypes: Productivity Imperative, Distinctiveness Imperative, Transition Imperative, and Stability Imperative. For example, areas like Chicago, Cincinnati, and St. Louis fall under the Productivity Imperative, facing labor shortages and a shrinking labor pool. Manufacturers in these regions should invest in advanced technologies such as artificial intelligence, automation, and robotics to enhance productivity and develop a workforce with advanced skills. Conversely, areas like Dallas and Houston fall under the Distinctiveness Imperative, with increasing demand and a growing labor base. These areas should focus on creating a best-in-class employee value proposition, including modern facilities, flexible work arrangements, and partnerships for technical training programs to attract and retain top talent. An example is Flex's factory in Sorocaba, Brazil, which deployed digitized workflows and robotic process automation to reduce non-value-added work by nearly 40%, upskilling over 200 employees and boosting output by more than 20%. Companies can better navigate labor shortages and productivity challenges by prioritizing talent acquisition, development, and retention approaches based on local needs.
To Revive Manufacturing, How Much Can a President Do? [NYTimes]
Presidential candidates are focusing on revitalizing U.S. manufacturing ahead of the election. Donald Trump proposes tariffs on nearly all imports to force foreign companies. Kamala Harris builds on the Biden administration's approach by advocating tax credits and apprenticeships to strengthen factory towns and invest in advanced technologies. However, economists note that broader forces like recessions and exchange rates have a more significant impact than presidential policies. Despite flat job growth due to higher interest rates and a strong dollar, policy initiatives like the Inflation Reduction Act and the CHIPS and Science Act have spurred significant private investment—$89 billion in clean energy manufacturing and $400 billion in semiconductors—though not enough to offset job losses elsewhere. The Biden administration links $80 billion in incentives to investments in economically distressed areas, promoting well-paid, unionized jobs. Companies are nearshoring supply chains, seeking reliable power and ample land, often in conservative-leaning states like Nevada, which has seen significant manufacturing growth. Meanwhile, the broader manufacturing sector remains sluggish with contracting orders and emerging layoffs, partly due to election uncertainty affecting future tax, trade, and regulations policies. While political leaders propose strategies to boost manufacturing, the industry's trajectory depends on a complex interplay of policy initiatives, economic forces, and private-sector decisions, with ongoing challenges in generating widespread job growth and addressing geographic shifts in manufacturing jobs.
Research:
State of Supply Chain Sustainability Report [MIT]
The MIT Center for Transportation and Logistics (MIT CTL) and the Council of Supply Chain Management Professionals (CSCMP) have released the fifth edition of the State of Supply Chain Sustainability report. Some key findings include:
Increased pressure from investors: Over five years, pressure from investors to improve supply chain sustainability has grown by 25 percent, making it the fastest-growing driver of sustainability efforts.
Lack of readiness for net-zero goals: Although 67 percent of firms surveyed do not have a net-zero goal in place, those that do are often unprepared to meet them, especially when it comes to measuring and reducing Scope 3 emissions.
Company response to sustainability efforts in times of crisis: Companies react to different crises differently regarding staying on track with their sustainable goals, whether it is a network disruption like the COVID-19 pandemic or economic turbulence.
Challenges with Scope 3 emissions: Despite significant efforts, Scope 3 emissions — which can account for up to 75% of a company’s total emissions — continue to be the most difficult to track and manage due to the complexity of supplier networks and inconsistent data-sharing practices.
Supply Chain Outlook Report [Prologis]
Prologis conducted its exclusive annual survey of more than 1,000 business executives in the U.S., U.K., and Germany, and the report underscored the critical need for nimbleness and resilience as companies navigate mounting challenges. Some key findings include:
85% of executives now regard sustainability as a core business objective, recognizing that green initiatives represent a compliance necessity and a competitive edge in today's marketplace.
Executives are increasingly looking to AI for predictive insights, with 80% saying, "There is a lot of external pressure to adopt AI practices to stay competitive in the industry." However, 94% experience challenges preventing them from adopting AI, including security concerns, data quality, change management, and implementation complexity.
86% of executives: "Rising economic and geo-political pressures are driving decisions on where to manufacture and store goods."
71% of executives: "I am under pressure to transition energy sources."
70% of executives: "It is difficult to identify where to allocate supply chain investment for maximum return."
The integration of AI (81%), automation (81%), intelligent data analytics (80%), and sustainability (79%)—all deemed essential by executives—is set to revolutionize supply chain management.
Podcasts:
Boeing is a Cautionary Tale About Current American Manufacturing
Manufacturing Deals🏭💵
Dexory - A London-based warehouse inventory automation and data startup
$80 million [Series B] - Led by DTCP and joined by Latitude Ventures, Wave-X, Bootstrap Europe, and insiders Atomico, Lakestar, and Capnamic
Polarium - A battery developer for the residential and industrial markets
$49 million [Venture] - From Altor Equity CEO Harald Mix and pension funds AMF and Alecta
Mstack - A company helping manufacturers source specialty chemicals globally
$40 million [Series A] - Co-led by Lightspeed and Alpha Wave
Anaphite - A British company moving automotive battery industry to dry coating, to decarbonise and improve electrode production
$13.7 million [Series A] - Co-Led by Maniv & World Fund and joined by EEI, Nesta, and Elbow Beach Capital
Salient Motion - A next-generation supplier for aerospace and defense, specializing in advanced electric propulsion
$12 million [Series A] - Led by Cantos and joined by a16z, AE Ventures, Hummingbird Ventures, and BoxGroup.
Downtime 🏭🧑🔧
SNL 50 Starts w/ theMile High Burger Challenge