The hardest thing to fake in a slice of pizza is not the flavor. It is a ball of protein about a hundred nanometers across called a casein micelle, and for the better part of a decade it has been the reason your cheese still comes from a cow.

Whey was easy. Egg white was easy. Perfect Day had animal-free whey protein in ice cream around 2020, brewed by engineered microbes, because whey is a tidy single protein that folds up and behaves. Casein does not fold up and behave. Casein is a crowd. Four different proteins, held together by tiny bridges of calcium phosphate, packed into a squishy sphere with a fuzzy outer shell that keeps the spheres from clumping. That structure is what lets mozzarella melt into strands instead of a puddle, what lets cheddar age, what makes milk white. You cannot express it in a tank and pour it out. You have to build it.

A small number of US startups have spent years figuring out how to build it, and the patent record now shows how far they have gotten. Three of them, all American, hold 19 granted US patents on making real casein without a cow. The newest was issued at the end of May. Almost nobody is covering this as a race, because from the outside it looks like a slow-moving corner of the fake-cheese aisle. Read the actual claims and it looks like something else: two rival engineering religions converging on the same 100-nanometer problem.

You already eat the first version of this

The strange part is that the dairy industry crossed this bridge once already, and nobody noticed. In 1990 the FDA approved fermentation-produced chymosin, the enzyme that curdles milk into cheese. It was the first genetically engineered food the agency ever cleared. Cheesemakers used to get chymosin by scraping the stomach lining of slaughtered calves. Now they get the same enzyme from microbes fed sugar in a bioreactor, and by most industry estimates roughly 90 percent of hard cheese made in the United States is set with the vat-grown version. You have almost certainly eaten it this week.

So the template already exists: take a gene from a cow, put it in a microbe, let the microbe make the cow protein. Chymosin was the dress rehearsal. The main event is the milk protein itself. That is the adjacent-possible move here, and the reason it took thirty more years is that chymosin is one enzyme and casein is a self-assembling supramolecular machine.

Camp one: brew it

The company closest to a plate of food is New Culture, a San Francisco outfit making animal-free mozzarella from precision-fermented casein. Its cofounder Inja Radman put the thesis bluntly to FoodNavigator this month: “The real prize is casein, which is the only protein on earth that can unlock the cheesy properties we all love.” New Culture brews alpha-s1 casein, one of the four caseins in milk, in microbes, then assembles the cheese around it. The company self-affirmed its casein as safe in 2024, filed a product label with California regulators, and has been serving its cheese at Nancy Silverton’s Pizzeria Mozza in Los Angeles.

The catch is cost. Brewed casein is expensive, so New Culture engineered its way around using much of it, cutting the casein in its mozzarella to about 28 percent of what conventional mozzarella needs, less than half. Its newest patent, granted in late May and numbered 12,635,703, covers exactly that trade: a mozzarella analogue built from recombinant alpha-s1 casein plus plant oils, salts, and minerals. It builds on a broader 2023 patent that claims the functional attributes of animal-free cheese regardless of formulation. In plain terms, they are patenting not just the protein but the recipe that lets a little of it act like a lot.

New Culture is chasing a big target. Its executives peg US mozzarella at a $9 billion segment, the largest slice of the domestic cheese market, inside a global cheese business north of $150 billion. Mozzarella is the hardest cheese to fake, because it has to behave one way cold on a Caprese salad and another way blistering on a pizza, and it is also the one worth the most to get right.

Camp two: grow it in a bean

The other camp thinks brewing casein in a steel tank is the expensive way to do it, and that the cheaper factory is a plant already grown by the millions of acres. Their idea is molecular farming: engineer a soybean to produce cow milk proteins in its seeds, then crush the beans the way the food industry already crushes soy.

The most-patented player here is Alpine Bio, the company that spent years as Nobell Foods before rebranding in 2024. Its founder, Magi Richani, raised a $75 million round on the pitch and has drawn roughly $100 million in total, with backers reported to include Bill Gates and Robert Downey Jr. Alpine holds 14 of the 19 casein grants, the deepest US patent stack in the field. The interesting ones are not about the casein gene. They are about a second gene: a kinase. Casein only assembles into a micelle after specific spots on it are tagged with phosphate, and plants do not naturally do that tagging to a cow protein. Alpine’s patents claim host cells and transgenic plants that co-express casein alongside a kinase whose only job is to phosphorylate it. They are not just growing the brick. They are growing the tool that makes the brick sticky.

Then there is Mozza Foods, a Los Angeles company founded in 2018 that has taken the most audacious position on the board. Its four granted patents all carry the same title, “Recombinant micelle and method of in vivo assembly,” and the claims describe something close to science fiction: not just making the four caseins inside a soybean seed, but assembling the finished micelle in the living plant cell, with kappa-casein enriched on the outer layer and the alpha-s1, alpha-s2, and beta-caseins packed into the inner core. That is the actual architecture of the thing in real milk. Mozza calls its engineered beans Cheesebeans, and CEO Adam Tarshis has argued the capital cost is lower because the company can ride existing soy-crushing and dairy-processing infrastructure rather than building fermentation capacity. The company is targeting a launch around late 2028.

Who cares

For a dairy processor, this is a sourcing question with a fork in it. One path buys casein from a fermentation company that can ship sooner but has to fight the economics of an expensive protein. The other path buys it from a crop that promises commodity-scale cost but is still years from a harvest and carries the regulatory weight of a genetically engineered food plant. Whoever hits cost parity first on functional casein does not win a niche vegan product. They win an ingredient that could sit inside ordinary cheese, sold to people who will never read the label, exactly the way vat-grown chymosin quietly took over the curd thirty years ago.

The through-line across all 19 patents is not a marketing word. It is a single stubborn piece of biophysics: the casein micelle, and the phosphate chemistry and self-assembly that hold it together. That is why whey fell first and casein came last, and why the companies solving it are patenting kinases and micelle geometry rather than flavors. The calf-stomach enzyme was the first cow part to get disassembled and rebuilt in a bioreactor. The milk protein is the second. The animal is being taken apart one molecule at a time, and the patent office is keeping the receipts.

Method note. Patent counts come from US utility grants in a database of 9.3 million USPTO grant documents, filtered to assignees Nobell Foods / Alpine Roads (now Alpine Bio), Mozza Foods, and New Culture, deduplicated across variant spellings. The 19-grant figure is the combined distinct count across those three assignees; Alpine Bio accounts for 14. Patent titles, abstracts, and claim language (including the kinase co-expression and in-vivo micelle-assembly claims) are quoted from the grant text itself. The fermentation-produced chymosin figures (1990 FDA approval, roughly 90 percent of US hard cheese) and company details on funding, product status, and executive quotes are drawn from FoodNavigator, AgFunderNews, and other trade reporting cited inline. Grant dates run through late May 2026; filings from the last few weeks may not yet be indexed. Nothing here is a claim about a company’s private strategy beyond what it has patented or publicly stated.