Building Rhea's Factory: How AI-Designed Enzymes Could Finally Solve Plastic Recycling

Key Takeaways

• AI platform designs enzymes that depolymerize PET at low temperatures
• Agentic AI scientist explores enzyme space beyond human intuition
• Selective enzymes target specific plastics, enabling mixed‑waste recycling
• Rhea's Factory plans 5,000‑ton per year demo plant in California
• Wet‑lab data of few hundred samples suffices for domain model

Pulse Analysis

Enzymatic recycling is emerging as a viable alternative to the mechanical and chemical processes that have dominated plastic waste management for decades. By using protein language models—similar to those that power AlphaFold—Rhea's Factory can predict protein structures and catalytic activity before ever stepping into a wet lab. This computational front‑loading slashes R&D cycles, allowing the team to iterate through millions of enzyme variants in silico and focus experimental resources on the most promising candidates. The result is a suite of biocatalysts that operate at modest temperatures, dramatically reducing the energy footprint compared with traditional pyrolysis or depolymerization methods.

The startup’s shift from a human‑orchestrated pipeline to an agentic AI scientist marks a broader trend in biotech: AI systems are now trusted to make speculative leaps, or "hallucinations," that human designers might overlook. By adjusting model temperature settings, Rhea's engineers intentionally broaden the search space, uncovering enzyme motifs that break polymer chains cleanly back to monomers. This level of precision enables selective targeting of specific plastic types, even within mixed waste streams, a capability that mechanical sorting struggles to achieve. The company’s early lab metrics show high conversion rates for PET and promising activity on polyamides, suggesting a pathway toward handling the diverse polymer mix that composes modern consumer waste.

Commercial viability hinges on cost parity with oil‑derived plastics, a hurdle Rhea's Factory addresses through scale and process integration. Their upcoming 5,000‑ton per year demonstration plant in California will test the economics of continuous enzymatic reactors, leveraging low‑energy heat exchange and enzyme recycling loops. If successful, the model could catalyze a paradigm shift, prompting legacy petrochemical firms to invest in bio‑based circularity solutions and prompting regulators to favor low‑carbon recycling pathways. The convergence of AI, protein engineering, and industrial chemistry thus positions enzymatic recycling as a cornerstone of the next generation of sustainable materials management.