Request for Startups: LatAM
For the past 3 years, our SF-based team at Baidu Ventures has focused primarily on AI startups operating in the US and Canada. In the past 12 months, we’ve begun to see a new generation of AI founder, like Plotlogic’s Andrew Job, who leverage technology to solve truly global problems. We believe that North America, with its world class research institutions and large industrial economy, will remain a leading market for AI. However, we’re developing an emerging thesis that as advanced technology becomes more democratized, and global trade reconfigures in response to COVID-19, many emerging economies will become prime markets for applications of AI. As the technology gaps between developed and developing economies become more apparent, savvy founders will increasingly seek to invent solutions that can be deployed around the world.
Brazil and Mexico
Brazil and Mexico are the largest economies in the Americas after the US and Canada, and have massive agriculture, mining, metallurgy and finance industries. Decades of questionable monetary policies have resulted in multiple episodes of large-scale currency devaluation in both countries. As a result of this poor fiscal policy, many of the factories, mines and farms that supply these industries in Brazil and Mexico have become outdated. We believe that with the right investment and the right technology, it is possible to revitalize and automate these industries. We envision a world where talented AI founders in Silicon Valley can partner with factory owners in Latin America to build and deploy industrial automation solutions at scale.
We’ve spent months conducting interviews and research in these countries, identifying the most promising industries for applications of AI. As part of our thesis-building, we’ve identified five opportunities for startup founders to develop solutions that could grow to billion dollar companies. We’ve specifically focused on opportunities that can be realized with largely off the shelf technologies. Our hope is that by publishing this ‘request for startups’ we can identify and fund solutions in these industry verticals. If you’re working in these areas (or want to be) please don’t hesitate to reach out!
Smart Greenhouses
Both Brazil and Mexico are major agricultural economies, producing millions of pounds of fruits and vegetables for domestic and export markets each year. These countries each have long, hot growing seasons, abundant water resources, and face daily food insecurity challenges. With its proximity to the US, Mexico in particular has the potential to generate billions by producing and selling high-value, organic produce. Currently, the US imports over $15B of fruits and vegetables from countries around the world each year. But the outdoor agricultural sectors in Brazil and Mexico also face some systemic challenges that limit further growth. Soil degradation and erosion is rampant, reducing the productivity of new and existing outdoor farms. In addition, in Brazil and Mexico over 50% of arable land is used for livestock cultivation, meaning that any expansion of outdoor farming would be at the expense of the massive (and politically connected) cattle industry.
We believe that greenhouse agriculture can address the endemic challenges of land use and soil degradation in these countries while generating a billion dollar opportunity in food production. A network of sensor-enabled greenhouses could be built to track and optimize environmental conditions for the needs of the desired products. These systems can be almost completely automated and can be built on degraded or abandoned lands. Smart greenhouses eliminate the need for human operators to make decisions regarding inputs (water, fertilizer etc) and harvesting, reducing labor cost and increasing productivity. In Mexico today, there exists over 1m acres of abandoned land. Since smart greenhouses rely on sunlight, no expensive LED lighting or power generation systems are needed.
In terms of unit economics, a smart greenhouse can be built at a cost of $65/square foot, and can produce up to 7 lbs of high value produce per square foot per year. By growing organically, each greenhouse can generate $14/square foot in revenue. We estimate the operating expense at $4.20/square foot/year, leaving $10/square foot in profit. With a $20m investment, it would be possible to build 7 acres of new, smart greenhouses in Mexico or Brazil that would generate over $4.2m in annual revenue (5 year breakeven).
Smart Mills
Brazil and Mexico are the largest steel producers in Latin America, and each sells billions of dollars in finished steel products to the United States. With abundant, unexplored ore deposits across both countries, the metallurgy sector will continue to be a major industry. But many of the regions’ mills are several decades old, and badly in need of improved process control, safety systems and emissions controls. It is increasingly difficult for these aging mills to produce the higher grade steel products that command the highest price premium.
Increasingly, mills in developed countries are adopting automation technologies to reduce downtime, spot defects and increase the efficiency of the melt shop. Running the electric arc furnace (EAF) that melts the metal represents approximately 2/3 of the operating expense of a mill. Any technology that can reduce the time required to run the EAF is a major source of savings.
We’re interested in startups leveraging AI to develop the Smart Mill of the Future. Advanced technologies can be deployed at multiple points in the metallurgy value chain, including advanced process control and automation in the ore processing/ball mill phase. Similarly, with real time characterization technologies based on hyperspectral imagery and laser induced breakdown spectroscopy (LIBS), smart mill operators can reduce the amount of time and energy required to run the EAF, generating massive savings. Finally, with advanced computer vision mill operators can spot defects in the final products before they leave the mill, and adjust production on the fly to correct them. These eliminate bottlenecks and complex decision trees that are common in metals production. Streamlining of decision-making ultimately leads to better resource utilization, lower working capital requirements and increased production.
A typical mill in Brazil or Mexico produces about 3.6m tonnes of steel annually, generating $2.5B in revenue. We estimate that by adopting automation across multiple points in the production process, a mill can reduce its operating expense by 5%. While the cost of implementing these solutions is high — approximately $40m — the potential annual savings could be north of $125m. Furthermore, with advanced technologies these mills can produce higher grade products that can traditionally only be produced in developed economies, commanding a price premium over lower grade outputs.
The Smart Mine
Mining represents another multi-billion dollar industry in Brazil and Mexico. Over 25% of the land area in these countries remains unexplored, and is believed to have significant ore and mineral deposits. With abundant labor, there are literally tens of billions of dollars underground, waiting to be extracted. However, like the agriculture sector, challenges exist that prevent growth in mining in Brazil and Mexico. In both countries, virtually all land is privately owned, and held by a small landowning elite that has not proven receptive to mining. With poor record-keeping and concentrated ownership, it is difficult for mine operators to cost-effectively identify and explore new lands. New technologies to identify and explore land at low cost are critical to expanding the mining sector.
Recent advances in remote sensing, aerial survey and hyperspectral imaging have dramatically reduced the cost of mine exploration. It is now possible for mine operators to remotely identify potential mineral deposits with high confidence. Whereas they used to have to first negotiate with landowners to conduct a field survey, operators can now generate a list of qualified exploration targets in a few days, without permission from landowners.
An average-sized gold mine in Mexico would typically yield around 31 tonnes of ore, worth about $1.6B USD. Of this potential revenue, about 5% covers the cost of exploration and prospecting. Existing remote sensing solutions for mining exploration can reduce the cost of exploration by 30%, generating $2.1m savings for an average mine. Across a single commodity in a single country, there is the potential for mine operators to save over $6m annually through improved exploration techniques.
Smart Manufacturing
Brazil and Mexico are the largest manufacturing economies in Latin America. They represent the largest exporters of electrical equipment and white goods (appliances) in the world, and are top 10 producers of medical equipment and automotive components. These countries, with their proximity and friendly relations with the US, are also major beneficiaries of the US-China Trade War. Billions of dollars of finished goods can pass freely from Brazil and Mexico into the US with little or no tariffs. However, like metallurgy, the manufacturing sectors in these regions still lag behind those of developed economies when producing the highest-value components, such as aerospace parts.
We believe that automation and AI-driven simulation can dramatically grow and improve the efficiency of the aerospace sectors in both regions. New simulation software can streamline the process of designing and testing new parts. In addition, it can allow domestic engineers in Brazil and Mexico to design advanced components that were previously only produced in developed countries. Furthermore, AI and computer vision-driven defect detection can reduce errors in component manufacture and increase yield. Likewise, anomaly detection can reduce downtime across a factory. All these technologies are currently in use in developed world aerospace manufacture, and could be readily ported over to Brazilian and Mexican factories.
Across the two countries, aerospace represents a more than $12B industry. Automation systems for aerospace leverage many of the same systems as automation solutions for metallurgy, with a similar cost basis. For a typical plant, the cost of advanced sensors and AI is likely between $20 and 40m. However, these solutions have the potential to increase savings by 5% annually, representing tens or hundreds of millions of dollars in value for a typical facility.
Smart Currency
Mexico City and Sao Paolo remain the financial centers of Central and South America respectively. In each country, financial services is by far the largest industry. Despite the size of these sectors, over 50% of the population in Brazil and Mexico is unbanked. In addition, both countries have experienced multiple rounds of hyperinflation and currency devaluation over the past 50 years. Poor or informal workers are hurt the most by these currency fluctuations, as they typically cannot access credit or banking facilities.
We believe that cryptocurrency has the potential to address many of the monetary challenges in these regions. In particular, stablecoins, or cryptocurrency tokens that are pegged to a particular value (typically the dollar), could help to provide informal workers with stable banking and credit facilities. With over 80% mobile phone penetration, Brazil and Mexico are some of the most connected countries in the developing world. We’re interested in startups developing mobile wallets for stablecoins, and low cost solutions to enable crypto-fiat transfers. Similar to how the M-Pesa system brought banking to millions of informal workers in East Africa, a stablecoin platform could quickly bring much needed financial services to the masses in these regions.
The informal sector represents 20% of GDP across Brazil and Mexico, or approximately $440B in value. If we assume a round trip fee of .1% to bring fiat currency in and out of a stablecoin system, the potential exists to generate over $440m in transaction fees annually. In addition, by offering credit products to informal workers using the platform, startups can provide low-cost loans where they are needed most.
These are just some of the areas where we feel advanced technology can make a big impact in Brazil and Mexico. Over the coming weeks, we will publish more of our insights on other developing regions where AI and other systems can be applied. If you’re working on these solutions (or want to be) please don’t hesitate to reach out. We can’t promise to fund everything, but will definitely review anything we receive. We look forward to the day when advanced AI helps to level the playing field between developed and emerging markets.
