Breaking the Bottleneck | Issue 68
[1/6/2024] China's Influence in Mexico, 2025 Simulation Trends, Global Semi Outlook, & China's Approach to Building Nuclear Power
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 chat!
🏭 If you were forwarded this and found it interesting, please sign up!
Content I Enjoyed Last Week 🏭🗞️🔬 📚
News:
How China is Setting Up Shop in Mexico [FT]
Chinese firms are steadily boosting their presence in Mexico’s industrial landscape, primarily by investing in parks such as the Alianza Industrial Park in Coahuila. Although the Mexican government’s official data registered only $450 million in Chinese and Hong Kong investment in 2022, analyses by organizations like Rhodium Group indicate the actual figure may be up to six times higher. Over the past few years, Chinese occupancy in Mexican industrial real estate has doubled, now representing about 7% of new absorption—even if that share still lags behind U.S. and Mexican firms. This is becoming a concern. U.S. policymakers, including Secretary of State nominee Marco Rubio, worry that Chinese manufacturers may use Mexico as a backdoor to circumvent American duties, focusing particular concern on industries such as steel, auto parts, and high-tech. Faced with these challenges, Mexico’s new president, Claudia Sheinbaum, has maintained a carefully balanced approach. She aims to preserve strong U.S. ties while acknowledging the potential risks of unchecked Chinese investment, even signaling a willingness to regulate illicit or counterfeit flows. Although smaller Chinese suppliers in Tier 2 and Tier 3 automotive markets have been active, larger-scale EV investments have yet to materialize fully. Only about 1.2% of Mexico’s auto parts exports to the United States originate from Chinese companies, although that proportion is rising. These developments are poised to shape the following review of USMCA in 2026, with the role of China in North America’s supply chains likely to be a pivotal issue. Canada has also threatened to exclude Mexico from the agreement if transshipment of Chinese goods through Mexico becomes more prevalent. In the meantime, Mexico’s automotive exports to the United States have ballooned, representing over one-third of the country’s overall import-share gains in the American market. Meanwhile, critics on both sides of the border fear job losses and security concerns. Adding to the complexity, automakers such as General Motors continue to bring Chinese-made vehicles into Mexico, making it even harder to envision cutting out Chinese components or products entirely. Overall, the Trump administration’s tariff threats—and the fact that 80% of Mexico’s exports go to the U.S.—are creating significant uncertainty. Concerns about the actual scale of Chinese FDI, the potential for security issues, and the upcoming USMCA negotiations in 2026 cast a heavy shadow over Mexico’s industrial future.
Great Breakdown on Tariffs [FT]
The US Starts Manufacturing Advanced Chips [IEEE Spectrum]
TSMC’s advanced chip-manufacturing facility in Phoenix is scheduled to begin mass production in 2025, representing the first large-scale deployment of the company’s cutting-edge technology in the United States. Initially, this fab will produce 4-nanometer process nodes—the same technology powering Nvidia’s most advanced GPUs—and has already demonstrated a 4% higher yield than comparable fabs in Taiwan. TSMC plans to open a second fab in 2028 that will offer 2-nm or 3-nm processes, possibly a third U.S. fab using even more sophisticated technology. Amid this expansion, TSMC faces cultural and workforce challenges. American engineers working in Taiwan have cited top-down, “military-style” management, while some Taiwanese staff claim American personnel lack the meticulous approach required for semiconductor manufacturing. On the U.S. side, a patchwork of city-level permitting regulations adds complexity, and the industry grapples with a talent shortage of qualified engineers and semiconductor technicians. Samsung and Intel are similarly active. Samsung’s planned fab in Taylor, Texas—backed by around $6.4 billion in CHIPS Act support—will not begin production until 2026, prompting questions about whether its capacity will meet actual demand. Intel, a major force behind the CHIPS Act, seeks to reclaim manufacturing leadership by opening or expanding facilities in Arizona, Ohio, and New Mexico, potentially tapping $8.5 billion in CHIPS funding. Still, Intel struggles to attract enough external customers to sustain its foundry ambitions and to keep pace with advanced process technology.
The World’s First Industrial Plan for Green Steel [MIT Tech Review]
Steel production emits around 8% of the world’s carbon dioxide, with conventional processes generating roughly two tons of CO₂ per ton of steel produced. A Swedish company, now renamed Stegra, is building the world’s first industrial-scale green steel plant, slated to begin operations in 2026 in northern Sweden. Unlike the high-CO₂ blast furnace route, Stegra plans to use green hydrogen—derived from hydro and wind power—to remove oxygen from iron ore, producing low-emission “direct reduced iron.” This iron is then melted in an electric arc furnace to yield steel. Stegra expects to make 2.5 million metric tons of steel annually with only water as a by-product during the iron-reduction step. The startup has already raised close to $7 billion and secured contracts for 1.2 million metric tons of steel over the next five to seven years, with major automakers like Mercedes-Benz, BMW, and Volvo willing to pay a 30% premium for cleaner steel. While EU carbon regulations and subsidies help Stegra’s economic case, the big question is whether its scaling of hydrogen-based steelmaking can deliver on cost and reliability. Other companies and processes are exploring green steel (like Hybrit’s similarly hydrogen-based approach and Boston Metal’s electrochemical method). Still, Stegra will be the first large-scale commercial test of producing near-zero-emission steel. If successful, it will demonstrate that customers are willing to pay extra for cleaner steel.
How a $12.98 T-Shirt Is Made in America—at a Profit [WSJ]
Walmart recently introduced a 100% U.S.-made cotton T-shirt priced at $12.98, rolling out in 1,700 stores. A key factor in bringing the shirt to market was Walmart’s commitment to noncancelable purchase orders, which gave suppliers and the shirt’s manufacturer, American Giant, the confidence to invest in automation and scale production. American Giant, founded in 2011 by Bayard Winthrop, typically sells premium, domestically produced apparel at higher price points for T-shirts and sweatshirts. After Winthrop publicly praised Walmart’s American-made initiative, the retailer approached him to explore collaboration. The guaranteed volume from Walmart provided American Giant and its supply chain partners with the demand needed to justify equipment upgrades, thereby reducing unit costs. To deliver shirts at $12.98, American Giant sources and processes cotton entirely in the United States, predominantly in Southeastern states. The company partly owns a sewing facility in North Carolina and established a dedicated Los Angeles plant for the Walmart program, hiring 75 new employees. Automation also played a pivotal role: the garment design was modified to knit shirts in a tubular shape, eliminating side seams, and specific tasks such as sewing patches and printing were automated. These efficiencies have lowered labor requirements to a point where domestic production can compete with offshore options. Although more brands have inquired about domestic production amid fears of future tariffs, the broader U.S. textile sector continues to face difficulties, with 23 plants closing in the last 18 months. Price sensitivity remains a key consumer issue: while many Americans favor U.S.-made items, few are inclined to pay higher prices. Walmart’s $12.98 shirt thus tests whether midrange domestic apparel can succeed. Winthrop’s collaboration with Walmart suggests that more mass-market “Made in USA” apparel lines could follow if large retailers commit to stable, high-volume orders.
Simulation Trends for 2025 [Engineering.com]
AI is quickly emerging as a powerful aid for simulation engineers, according to Comsol’s Bjorn Sjodin. He points to chatbots that can tutor beginners on fundamental tasks—such as choosing boundary conditions in a heat transfer model—and help more advanced users streamline coding by automating or debugging scripts. While the technology isn’t perfect, Sjodin believes it can save time and reduce barriers to learning. He also sees an ongoing trend toward reduced order modeling and surrogate models—approaches that use techniques like eigenvalue analysis or neural networks to compress large, computationally heavy simulations into lightweight versions that run in seconds rather than hours. This shift is driven by demands for faster system simulations, digital twins, and production-floor tools that cannot afford long run times. Traditional model-reduction methods rely on capturing the essential physics in smaller matrices (for instance, going from a 10-million-by-10-million matrix down to 100-by-100), while neural-network-based models rely on a training phase in which parameters—such as geometry or material properties—are sampled within known ranges. Once trained, the neural network can deliver near–finite element accuracy within those parameter bounds, making it suitable for real-time or near-real-time decision-making. Despite the promise of AI, Sjodin notes that current chatbots still lack sophisticated spatial understanding—making them less adept at interpreting CAD geometry or 3D details—an area he expects will improve over time.
Apptronik Partners with Google Deepmind [Apptronik]
Apptronix partnered with the Google DeepMind robotics team to merge advanced AI with its cutting-edge hardware. Apptronik, which spun out of the Human Centered Robotics Lab at the University of Texas at Austin, boasts a flagship robot standing 5 feet 8 inches tall, weighing 160 pounds, and designed for physically demanding tasks in industrial spaces. Meanwhile, Google DeepMind’s robotics team has been pushing the envelope in AI and simulation, from foundational model advancements to the latest Gemini system. The collaboration aims to create versatile, intelligent, and safe robots capable of assisting humans in dynamic environments. The partnership follows a year of rapid progress for Apptronik, which recently announced partnerships with GXO and Mercedes-Benz, with further collaborations expected in the coming year.
A Generative Physics Engine for Robotics [Genesis]
Genesis [See Paper] is a physics platform designed for general purposes, such as robotics, embodied AI, and physical AI applications. It is simultaneously multiple things:
A universal physics engine rebuilt from the ground up, capable of simulating a wide range of materials and physical phenomena.
A lightweight, ultra-fast, pythonic, and user-friendly robotics simulation platform.
A powerful and fast photo-realistic rendering system.
A generative data engine that transforms user-prompted natural language description into various modalities of data.
Genesis's physics engine is developed in pure Python while being 10-80x faster than existing GPU-accelerated stacks like Isaac Gym and MJX. It delivers a simulation speed of ~430,000 faster than in real-time and takes only 26 seconds to train a robotic locomotion policy transferrable to the real world on a single RTX4090. The company aims to build a universal data engine that leverages an upper-level generative framework to autonomously create physical worlds, together with various modes of data, including environments, camera motions, robotic task proposals, reward functions, robot policies, character motions, fully interactive 3D scenes, open-world articulated assets, and more, aiming towards fully automated data generation for robotics, physical AI and other applications.
Research:
Toward Security in Battery Raw Material Supply [McKinsey]
The global push toward net zero hinges on meeting the volume of supplies of key battery materials such as lithium, nickel, cobalt, and manganese, each facing potential shortages, regional supply risks, and mounting ESG concerns. Many of these materials are concentrated in just a few countries, posing geopolitical and sustainability risks, while refining often occurs in regions like China or Indonesia. Producers must balance three core objectives—availability, affordability, and sustainability (the “materials trilemma”). They can do so by adopting new production processes, investing in technology for refining and direct extraction (particularly for lithium), forging better supplier partnerships, and focusing on design and manufacturing efficiencies to reduce both costs and carbon footprints. Approximately 40% of a typical battery’s emissions stem from raw material mining and refining. Different chemistries (for instance, Li-NMC versus LFP) have varying emissions profiles. Over time, decarbonization of the highest-emitting metals (nickel, cobalt, lithium) will shift attention to less-obvious contributors like manganese. By 2030, leading battery makers could cut total battery-material emissions by more than 70% and approach near 90% reductions by 2040 through measures such as sourcing from greener producers, deploying renewable energy, using “green” chemicals, and boosting recycling (which could supply up to 50% of key materials by 2040). In the short to midterm, price volatility, material shortages, and ESG pressures will likely persist, creating both risk and opportunity for automotive OEMs and battery manufacturers to reshape their supply chains for a more sustainable, net-zero future.
The Space Industry Is Booming. Manufacturers Must Catch Up [BCG]
As global demand for rockets, satellites, and other space-related hardware continues to surge, manufacturers face an unprecedented growth opportunity and several critical challenges in meeting new production targets. The growing space boom, spurred by the proliferation of low-Earth orbit satellites, lower launch costs, and expanding commercial activity, has led to a record 221 orbital launches in 2023, with tens of thousands more satellites anticipated before 2031. Against this backdrop, addressing the industry’s acute talent shortage is equally essential; companies should revamp recruitment, enhance compensation, and offer professional growth to retain engineers who may otherwise depart when production transitions from cutting-edge R&D to routine manufacturing. Next, manufacturers can simplify prototype designs and shift toward modular product platforms, reducing expensive testing and engineering costs. Close supplier relationships, meanwhile, help secure critical components from propulsion systems to avionics by sharing investment in necessary machinery and ensuring transparency around future demand. Finally, streamlining operations from leveraging automation to reorganizing workflow allows producers to transition from one-off fabrication toward serial manufacturing, substantially accelerating build times and throughput. With focused execution in these areas, space manufacturers can yield remarkable gains, including tripled production volume, double-digit cost savings, and significant reductions in hours per build. More aggressive measures could boost throughput eightfold, underscoring the potential impact of a healthy operational transformation.
Global Semiconductor Industry Outlook for 2025 [KPMG]
KPMG and the Global Semiconductor Alliance conducted the milestone 20th annual global semiconductor industry survey in the fourth quarter of 2024. Here are some of the key takeaways:
Last year, AI surged to rank as the second most crucial application driving semiconductor company revenue. This year, AI ascended to the top position for the first time, displacing automotive. Correspondingly, microprocessors (including graphics processing units used for AI) again ranked as the top product opportunity for industry growth over the next year, ahead of memory and sensors/MEMS.
Cloud/data centers tied for third last year but rose to the second most crucial application, driving semiconductor company revenue.
In last year’s survey, 30 percent of respondents believed there was already excess semiconductor inventory. That sentiment remained stable and did not increase (29 percent in this year’s survey). Also, in last year’s survey, 45 percent believed an inventory excess would become a reality in the next four years. In this year’s survey, that belief dropped to 37 percent.
With demand for semiconductor products expected to increase and government subsidies fueling a boom in manufacturing, 84 percent of executives expect their workforce to either expand or remain the same in the next year. Smaller companies (97 percent) are more bullish on this topic than large and mid-size companies (80 percent and 77 percent, respectively).
Podcasts/Video:
How China Builds So Much Nuclear Power [Odd Lots]
Finance & Transactions 🏭💵
Venture Capital:
KoBold Metals - A company levaraging artificial intelligence to find deposits of minerals such as copper, lithium and nickel.
$537 million [Series C] - Co-Led by Durable Capital Partners and insider T. Rowe Price
Planned Downtime 🏭🧑🔧
THR Actors Roundtable Adrien Brody, Daniel Craig, Paul Mescal & More