Breaking the Bottleneck | Issue 44
[4/1/2024] Gen AI in Industrial Operations, Green Steel, Boeing QC & More!
Breaking the Bottleneck is a weekly manufacturing technology newsletter with perspectives, interviews, news, funding announcements, manufacturing market maps, and a startup database!
đşď¸Check out these market maps for new discrete/process manufacturing tech!
đĽ Have you seen any interesting startups recently? Iâd love to chat with them!
đ If you were forwarded this and found it interesting, please sign up!
Content I Enjoyed Last Week đđď¸đŹ đ
News:
Unlock the Potential of Generative AI in Industrial Operations [AWS]
There have been various challenges associated with implementing Industrial AI and ML technologies in industrial operations. The primary challenge lies in dealing with vast volumes of unstructured data from various sources, such as sensors and telemetry systems. The key to addressing this is using generative AI with large pre-trained foundation models (FMs) like Claude, which can generate content, including Python code, in response to text prompts. This approach streamlines operations by reducing the need for custom ML models for each specific use case, thereby democratizing AI access and enhancing productivity. An example would be using agents like PandasAI with FMs. PandasAI, a Python library, adds generative AI capabilities to pandas, a popular data analysis tool, thus enabling more effective processing of high-resolution time series data. AWS also proposes a solution to enhance code generation accuracy via multi-shot prompting detailing three distinct use cases within this framework:
Natural Language Query (NLQ) with time series data, involving a workflow that includes Amazon Monitron for monitoring equipment health, Amazon Kinesis and Amazon S3 for data processing, and PandasAI for generating Python code.
Summary generation of malfunctioning parts, where images of parts are processed to extract text data and generate summaries using the Amazon Bedrock Claude v2 model.
Root cause diagnosis involves analyzing various documents using Amazon Bedrockâs knowledge base and a Retrieval-Augmented Generation (RAG) approach.
US Pledges Up To $1B For Two ââGreen Steelâ Projects [Canary Media]
The Biden administration's recent announcement of up to $6 billion for projects aimed at reducing greenhouse gas emissions in heavy industries, including the iron and steel sector. This funding includes up to $500 million each for Cleveland-Cliffs and SSAB to develop "green steel" projects that would utilize clean hydrogen, rather than coal or fossil gas, in ironmaking. These initiatives are significant as currently, no large-scale, low-emissions ironmaking facilities exist in the country, with the only significant operation being the Hybrit project in Sweden. Cleveland-Cliffs' plan involves installing a hydrogen-ready ironmaking plant in Ohio, while SSAB intends to construct a new facility in Mississippi exclusively using hydrogen. The direct reduced iron (DRI) process, which both Cleveland-Cliffs and SSAB plan to employ, utilizes hydrogen gas to remove oxygen from iron ore, potentially reducing carbon emissions significantly. Cleveland-Cliffs expects to invest around $1.3 billion over five years in its Ohio project, which could reduce emissions by 1 million metric tons annually. SSAB's Mississippi project, employing Hybrit technology, aims to use only clean hydrogen, potentially reducing carbon intensity by over 90%.
How Boeing Favored Speed Over Quality [NY Times]
In the past year, several incidents involving Boeing 737 Max planes have raised concerns about the quality of Boeing's production. These incidents, including an automated system malfunction, a fire detection system problem, and an engine failure at high altitude, have not been widely reported but highlight ongoing issues with Boeingâs manufacturing. Key Factors in Boeing's struggles to maintain quality standards include reduced experience levels among Boeingâs workforce post-pandemic, weakened inspection processes, and supplier challenges in meeting quality standards under production pressure. A recent FAA audit revealed lapses in Boeingâs quality-control practices. The audit followed an incident where a panel blew off a 737 Max 9 midair, traced back to a Boeing factory in Renton, Washington. Current and former employees reported a culture shift starting around 2017, with increased production pressures leading to compromised quality and safety. The loss of experienced employees due to pandemic-related job cuts and retirements has exacerbated these challenges. The need for more training for new employees and the increased use of less experienced workers have been identified as contributing factors to the decline in manufacturing quality.
Inside Boeingâs Quality Control Process for 737 Max Planes [WSJ]
39,000 Lost Jobs Undercut Bidenâs Manufacturing Wins [Bloomberg]
Biden's focus on bringing back middle-class factory jobs and reshoring strategic industries has seen mixed success. Despite some economic gains, the closure of iconic factories like Master Lock raises questions about the long-term transformation of urban industrial areas and the political implications for the Biden administration. A decade ago, workers at the Master Lock factory could earn $100,000 annually, but the shutdown represents a shift in manufacturing jobs and the political implications of industrial policy. Some interesting takeaways from the analysis include:
Since the 1979 manufacturing peak, total private employment increased by 84.7%, while manufacturing employment decreased by 33.1%.
In the past five years, the US added 146,000 manufacturing jobs.
In Michigan, Pennsylvania, and Wisconsin, manufacturing employment decreased by 0.8%, 1.8%, and 3.9%, respectively, compared to 2019 levels.
In contrast, Southern and Western states like North Carolina, Georgia, Arizona, and Nevada saw manufacturing job increases ranging from 10.1% to 16.8%.
Manufacturing jobs in Milwaukee's surrounding counties increased by 6.9% from 2010 to 2021, while they decreased by 15.3% in Milwaukee itself.
Research:
State of Smart Manufacturing Report [Rockwell]
Some Key Takeaways:
94% of respondents expect to maintain or grow their workforce as a result of smart manufacturing technology adoption.
85% of manufacturers said they would use AI and ML in operations this year, and 83% said they would use generative AI.
95% are currently using or evaluating smart manufacturing technology, up from 84% in 2023.
98% have a sustainability or ESG policy in place, with almost half of those being formal, company-wide policies.
âI think the number one takeaway is manufacturers admitting and understanding that cybersecurity is a thing,â says Abbey, adding that for a very long time, manufacturers considered cybersecurity to be someone elseâs problem. âWe've had more and more published reports over the last couple of years, and they're starting to see them. And this report reflected that change, that mental awareness, that cybersecurity is a serious external risk,â
Generative AI: The Next S-curve For The Semiconductor Industry? [McKinsey]
As businesses increasingly adopt generative AI (gen AI), semiconductor leaders are investing heavily in data centers and semiconductor fabrication plants (fabs) to meet the soaring demand for advanced chips. The primary components of gen AI compute demand come from B2C and B2B applications, each involving two phases: training and inference. The semiconductor industry must prepare for increased wafer demand, especially for high-performance components like logic, memory, and data storage chips. By 2030, this could result in a significant supply gap, necessitating the construction of three to nine new logic fabs. DDR and HBM are two types of DRAM used in gen AI servers. The industry faces challenges in increasing memory capacity due to a "memory wall" problem. McKinsey presents two DRAM demand scenarios, "DRAM light" and "DRAM base," reflecting different growth rates in memory demand. NAND memory, used for data storage, is also expected to see increased demand, with up to eight million wafers required by 2030. This translates to the need for one to five additional fabs. Other components that will see increased demand include high-speed network and interconnect devices for server connectivity and power semiconductors for electricity consumption in AI servers.
Chart of the Week:
U.S. Semiconductor Ecosystem Map
Podcasts/Video:
The Future of Energy: ML in Production [Northvolt]
Manufacturing Dealsđđľ
Viam - A company building a singular software platform that streamlines interaction with devices across IoT, robotics, smart home, and industrial automation.
$45 million [Seed] - Led by Union Square Ventures and Battery Ventures
Gather AI - A company providing drone-powered warehouse inventory monitoring.
$17 million [Series A Extension] - Led by Bain Capital Ventures and joined by Tribeca Venture Partners, Dundee VC, Expa, and Bling Capital
Robovision - A Belgian company computer vision platform for machinery manufacturers and production lines.
$42 million [Series B] - Co-led by Target Global and Astanor Ventures and joined by Red River West
Vizcom - A company building a platform to bridge the gap between 2D sketches and immersive visuals.
$20 million [Series A] - Led by Index Ventures
Planned Downtime đđ§âđ§
Inside Samsungâs Futuristic Smartphone Factory