Overview of R&D at Dow

• Dow does all of their own R&D. We have centers in Shanghai, Dubai, and others. However, many competitors in the basics market (Westlake, Oxycam) have no R&D on their own. Just bulk manufacturers. Need to figure out where to point R&D, both to develop novel materials and drive down cost for existing materials
• A lot of R&D is customer interactions, where they bring customers in and talk to them about what they want to achieve. Bring the customer into the R&D process. Pack Studios is an example of this
• These are obvious opportunities for AI. Use it to pre-select methods of research. The days of scientists sitting at the bench doing random experiments are pretty much done. Even from 2003 - 2007, we were doing a ton of automation, miniaturizing a bunch of stuff, using robotics to run 100 experiments at a time
• Dow spends ~$800m on R&D per year
  • Had a GE executive come who shook up the Dow R&D world. Before then, R&D had complete autonomy on what they wanted to spend their time on. Now much more focused on targeting positive ROI investments, implemented a process between R&D and commercial teams
  • if you look at Dow;s Path to Zero, we would have to do a billion-dollar retrofit to our Canadian assets to have zero-emission polyethylene product line. There were tons of research was going into this area, figuring out how to adjust our manufacturing techniques to leverage circular hydrogen etc. They partnered with a commercial team, they think they can get margin + customer growth by showing sustainable environmentally friendly products in the plastics space. They partnered closely with folks to figure this out
    • Linde is the biggest partner
    • Need to figure out how to run our processes best here, which is the opportunity for AI. All of the ethylene crackers are the same. The IP is the process, we have a flexible process that can support a bunch of feedstock, we can run it 2-3 years longer than others
  • Split of focus on new vs. existing products
    • Polyethylene: 80% existing products. This is a cornerstone business
    • Polypropylene, specialty intermediate products: 80% new products
    • For commodity products, mostly on existing techniques
    • Semis was probably 99% new R&D, which we sold to DuPont
• We’re using AI to streamline the collection and interpretation of data to refine the experimentation process (C3 AI)

AI materials R&D

• Don’t think foundation models are valuable in the short-term. More immediate areas:
  • See a lot of automation going on. See video analytics
  • Using AI to query the results, semantic search to make it easier for researchers to analyze data
• I think the FMs are longer-term science; we still need to learn how to interpret them, how to build them as we need to build these in house
• Openness to partnering with a 3rd party AI vendor
  • Think this would be interesting for Dow, especially if it targets specific areas of need
  • On the innovation side, we found C3 AI, which really improved the reliability of our hydrocarbon crackers. When we first introduced it, the operations teams laughed at it. But they got more and more interested, but it wasn't easy. If the vendor had done it directly, It would have been hard. They also had to do it for free first, show where they could have avoided outage
  • Same thing with R&D. R&D guys think they know it better than everyone else. Need to find right use case and right value proposition
    • There’s Erwin Quimio, Jack Robinson (retired) who were thought leaders in using new tech in the R&D process at Dow
    • Look for areas with most potential impact margin quickly. Pack Studios is a good example (packaging R&D). New product development in blown film, packaging. Want to quickly embed yourself and get to a final result as fast as possible
  • DOW wants to spend money on materials scientists, not the AI
    • C3 had domain experts, folks who knew the cracker experience that directly applied it
  • DOW typically does not like to do a royalty model. Would typically lean towards consulting or a services model
    • We don't have people there to support AI models directly so more opportunity to sell the services
    • Would be shocked if they did a royalty model, they hate doing that
    • Would characterize most of Dow’s competitors like this. Margins are tighter, R&D is scrutinized a lot, very conservative — BASF, Dupont, Exxon Chemicals are all pretty similar