AI4S.fyi · Case studies
Case studies
Each Buyer Chains piece follows one chain, the AI company, its customer and sometimes the customer's customer, through five roles in the public record: who states the need, who sets the research direction, who runs the experiments, who validates the result, who decides to keep paying. It records only what the public record states and says who said each thing; a missing step is marked not yet public, and a company's own figure is written as the company's. The timeline at the end of each piece stays open.
Series · Buyer Chains · 8 cases
AI-made materials: who buys them, and what does a buyer need to see before paying?
Each piece opens with an "At a glance" line, the chain's one or two key numbers, and puts a "Decision point" before its timeline: who decides next, and what. Together the two lines say how far the chain has got and whom it is waiting on.
Three rules run through the series: a company's own figure is written as the company's; a customer not named is a buyer not yet public, never "no buyer"; each role is read in the most charitable way the record allows.
CASE STUDY 001 · Discovery stage
CuspAI × Kemira: 300 trillion structures down to 20 candidates, and not yet
an experiment From the buyer's brief to 20 AI-designed candidates: who decides which experiment gets paid for next?
At a glance · About 300 trillion structures, 5,000-plus designs, about 20 priority candidates; synthesised or tested: none.
CASE STUDY 002 · External test stage
Radical AI × U.S. Air Force (AFWERX) × Purdue: a $1.2M contract and one
torch test Between an Air Force contract and a Purdue torch test, who says the alloy is good enough?
At a glance · A $1,197,902 AFWERX contract; one public acceptance number, a 0.25% mass gain for RAI-939 against a 32% loss for C103 after a minute above 2,500 C.
CASE STUDY 003 · Commercial delivery stage
Scimplify × an unnamed US metals manufacturer: what a buyer's acceptance
looks like When a buyer's acceptance goes full circle, what does the public record show?
At a glance · A repeat order covering one year of supply; spec, tonnage and price disclosed: none.
CASE STUDY 004 · Milestone contract stage (control)
The control: Isomorphic Labs × Lilly and Novartis: the money is explicit, the acceptance thresholds
are not The contract states the money; why are the acceptance gates still invisible?
At a glance · $82.5M in upfronts and up to $2.9B in milestones across the two deals; scientific acceptance thresholds disclosed: none.
CASE STUDY 005 · Pre-market stage
Holyvolt × an unnamed cell buyer × a procurement rule: buy the lab first, promise the market
in 2027 The buyer is not yet named, but the procurement rule is. Who validates a cathode that is due on the market in 2027?
At a glance · $45M raised and a 2027 market date; DRX performance figures disclosed: none.
CASE STUDY 006 · Platform stage
Singapore × A*STAR's MARVEL-AI × five industry partners: all five roles designed into
one platform A government has written who states the need, who sets direction, who experiments, who validates and who pays into one platform. Who makes the first material an industry partner adopts?
At a glance · S$150 million of government money and ten named partners; materials accepted by a partner so far: none.
CASE STUDY 007 · Incumbent chain, no AI lab
Entegris × Element Six × an unnamed chipmaker: who runs the chip-heat chain that AI materials companies want
to enter Heat out of stacked AI chips is the problem AI materials companies have named as theirs. Who states the need, who does the work and who accepts it in that chain today?
At a glance · Two moves, an acquisition and a long-term partnership; price, capacity, customer and performance figures in the release: none.
CASE STUDY 008 · List stage, buyer not yet public
Google DeepMind's GNoME × Berkeley's A-Lab × a buyer not yet public: after 2.2 million predictions, who states
the need A materials list nobody ordered. Who has validated what in three years, and how is DeepMind now looking for a buyer?
At a glance · 2.2 million candidate structures over three years; made from the list by an outside lab: one compound; buyers in public: none.