Benson Hill
CropOS predictive breeding platform for ultra-high-protein soybeans
Benson Hill said its CropOS platform helped lift soybean protein 2% in one breeding generation, then filed for bankruptcy
Problem
High-protein non-GMO soybeans usually yield less than commodity GMO varieties. Benson Hill had to pick among thousands of possible crosses and traits while keeping yield close to market levels.
Approach
Benson Hill used CropOS, its cloud-based machine learning platform, to predict which breeding options were worth pursuing. The system took in genomic and field-trial data and narrowed several thousand options to a handful, within limits such as budget and lab capacity. The company applied it to its ultra-high-protein, low-oligosaccharide soybean lines.
Results
Takeaway
A breeding platform can narrow the choices and still not rescue a company's finances. Benson Hill filed for Chapter 11 in March 2025 and said it planned to sell its business.
Categories
- Industry
- Agriculture and food production
- Business function
- IT and R&D
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
The trade outlet reports that the third generation of Benson Hill's ultra-high-protein soybeans gained 2% protein over the previous generation. The yield gap against commodity GMO soybeans was only 3 to 5 bushels per acre. The CTO said CropOS drove the predictive breeding.
Fortune describes CropOs as Benson Hill's cloud-based machine learning interface. It narrows several thousand breeding choices to a handful and factors in budget and lab capacity. The article says the company's soybeans had been used on more than 300,000 acres.
The St. Louis-based seed company and its subsidiaries filed for Chapter 11 protection in Delaware on March 20, 2025. It planned a Section 363 sale of all or part of the business and had a commitment for about $11 million in debtor-in-possession financing.
Full details are in the original sources.