On Synthetic Biology Strategy

The therapeutics constitution opened by carving platforms into genera. This one cannot, because synthetic biology is not in the discovery business — it is in the manufacturing business, and manufacturing is not carved by what you discover but by what you can make, and at what cost. A drug company is paid for finding a molecule no one had. A synbio company is paid only if it makes a molecule everyone already has more cheaply than the incumbent who already makes it. That single difference — novelty versus cost, a binary readout versus a continuous slope, the lab versus the plant — inverts every law the therapeutics canon runs on. This essay is the constitution for the inverted world. Everything beneath it, the protein thesis included, is one province obeying these laws.

The category error

Synthetic biology has been valued, funded, and written about as though it were drug discovery with a different output, and almost every disappointment in the field traces back to that one mistake. The two could not be less alike in the only way that matters: how value is decided.

A drug company discovers a new molecule — one that did not exist, that no one can legally copy for twenty years — and is paid an enormous premium for the novelty, gated by a binary clinical readout that tells it, on a single day, whether it has succeeded. A synbio company does the opposite. It manufactures a known molecule — a protein, a chemical, a material, a fuel that already exists and is already made by someone, usually a petrochemical plant, an extraction industry, or a farm — and it is paid nothing for novelty, because there is none. It is paid only if it makes the identical thing more cheaply, or more cleanly at the same cost, than the incumbent who has been making it for fifty years on amortised capital. There is no binary readout. There is no day the science "passes." There is only a price, and a slow, expensive question of whether you can get under it at scale.

The people who lived through the field's reckoning say this in plainer language than any outside critic. The synthetic biologist Jay Keasling, who co-founded Amyris, calls the recent collapse a reckoning. And Zach Serber, who co-founded Zymergen, drew the contrast exactly: pharmaceutical-facing synbio is disciplined because regulation imposes milestones — preclinical models, then human safety and efficacy — and a candidate that misses a milestone washes out cleanly, while one that hits it can be sold to a deep-pocketed pharma buyer. Non-pharmaceutical synbio, the kind that makes materials and chemicals and food, has no such milestones. Nothing tells you you are winning. There is no FDA, no Phase 2, no clean wash-out — only the open-ended, capital-devouring task of competing with an established industry on volume and price, with no scoreboard until the market either buys your ton or does not. The category error is treating a business with no milestones as though it had them. The constitution begins by refusing it.

The master variable: the cost curve to parity

If there is no readout, what is there? One number, and the entire field reduces to it: the cost, in dollars per kilogram, of the molecule you make, measured against the cost of the incumbent you are replacing — and the rate at which your number is falling toward theirs. Call it the cost-to-parity curve. It is to synthetic biology what the clinical readout is to therapeutics: the thing on which everything is decided. But where the readout is a coin that flips once, the curve is a slope you descend for years, and you are a business the moment you cross parity and a science project every day before.

This holds across every output synbio produces, which is why it can serve as a constitution rather than a sector note. A protein company competes on the price of a kilogram of protein against dairy or soy. A materials company competes on the price of a kilogram of bioplastic or biological leather against the petrochemical or animal version. A chemicals company competes on a kilogram of flavour, fragrance, or specialty ingredient against the extracted or synthesised incumbent. A fuels company — the graveyard's oldest tenant — competes on a litre against crude oil, which is the cruellest cost curve on Earth. The molecule changes; the law does not. Can you reach cost-and-performance parity, at industrial scale, with the thing you are replacing? That is the whole question, and a synbio thesis that cannot answer it in dollars per kilogram is not a thesis. It is a hope dressed in a genome.

Two curves, and the lethal gap between them

Here is where the field's optimism comes from, and where it goes to die. There are two cost curves in synthetic biology, not one, and conflating them has bankrupted more companies than bad science ever did.

The first is the enabling curve — the cost of reading, writing, and engineering DNA itself. This is the famous one, the Carlson Curve, biology's answer to Moore's Law, and it has delivered spectacularly: sequencing and synthesis costs have fallen for two decades, in some stretches faster than computing, so that the act of engineering a microbe to make a target molecule has gone from a moonshot to a routine. If the enabling curve were the business, synbio would have won already.

But the enabling curve is not the business. The product curve is — the cost of making a ton of the actual output, purified, at industrial scale, ready to sell. And making DNA cheap does not make product cheap. The microbe still has to be grown in enormous steel vessels, fed, kept alive and productive at density, and the molecule still has to be pulled out of a messy broth and purified to spec, all at a yield and a throughput that survive the price of the incumbent. The enabling curve collapses the cost of the design; the product curve is governed by the brutal, unglamorous physics of fermentation, separation, and scale, which improve slowly and resist the exponential. The gap between the two — between "we can engineer the organism," which is solved, and "we can make the molecule at a competitive price at scale," which usually is not — is the valley every synbio company must cross, and the one most of them die in. The pitch decks plot the enabling curve. The bankruptcies happen on the product curve.

And even the enabling curve is not a law. Rob Carlson, who drew it, is its most disciplined critic: the curve is a function of demand and industrial investment, not a property of the universe, and there is nothing that forces prices to keep falling. The honest reading of both curves is the same — they are possibility schedules, not promises, and they bend down only where enough demand exists to fund the next step. A constitution that took either curve as destiny would be writing the next decade's bankruptcies.

The value-destruction rule: the graveyard

The therapeutics canon has its ninety-per-cent rule — most platform value is destroyed, and the destruction happens upstream, at the clinical readout. Synthetic biology has its own version, and the inversion is the whole point: in synbio, value is destroyed downstream, at the plant, on the product curve, after the biology has already worked and the capital has already been spent. The graveyard is specific, recent, and instructive.

Read the right-hand column and one cause repeats: not "the biology failed" but "could not make it at a competitive cost at scale." Zymergen's microbes worked; its plastic film could not be manufactured cheaply enough, and a company that raised one and a half billion dollars sold for three hundred million. Amyris built a real strain-engineering platform and a real plant and still could not out-cost its incumbents durably. Keasling's own post-mortem names the disease precisely: mountains of investor cash that produced an overabundance of ideas and a lack of focus. And the graveyard has an afterlife worth noting — the dead leave behind working IP, which the chemical and ingredient incumbents then buy at a steep discount, exactly as Corbion absorbed Solazyme's. The value the synbio company created does not vanish; it transfers, downstream and cheaply, to the established industry it failed to beat. That is the rule, stated: in synthetic biology, value is destroyed at the plant and captured by the incumbent, not destroyed at the lab and captured by the inventor. Build accordingly.

The taxonomy: foundry, product, input — and why the foundry is the trap

If you do not carve synbio by modality — and you should not, because the modality is not what decides the outcome — then carve it by the only thing that does: how the company makes money. There are three models, and their survival rates are wildly different.

The foundry, or platform, sells the engineering itself as a service — "bring us your molecule, we will design the organism that makes it." This is the most-hyped model, the one Ginkgo built and the one Zymergen pivoted into as a Ginkgo subsidiary, and it is structurally the weakest, for a reason the cost curve predicts. A foundry has no product of its own, so it has no cost curve of its own to win; it is hostage to its customers' cost curves, and if its customers cannot make money making their molecules at scale — and most cannot — then the foundry's revenue never materialises, no matter how good its engineering. The horizontal "program any organism for anyone" promise sounds like leverage and behaves like dependency. It is the trap, and the layoffs and collapsed stocks are where the trap closed.

The product company makes and sells the molecule itself — Amyris, the precision-fermentation protein companies, anyone who owns the thing in the vessel. This model can win, because it owns its cost curve and captures the full margin when it crosses parity, but it is capital-intensive and unforgiving: it lives or dies on the product curve, and it dies if it picks the wrong molecule, as the next section explains.

The input layer sells into everyone's cost curve — DNA, enzymes, strains, reagents, the picks and shovels — and is the quietly profitable corner of the field, because it does not have to win any single product; it earns a little from every company that tries, including the ones that fail. (It is no accident that the incumbents buying the graveyard's IP are input and ingredient giants.) The synbio companies that, in Science's telling, kept living up to the hype were the focused ones in pharmaceuticals and additives — narrow, high-value, vertically disciplined — not the horizontal foundries.

The constitutional lesson, parallel in function to the therapeutics canon's archetype claim but opposite in content: the model that promises to do everything for everyone is the one that does nothing profitably. Focus is not a virtue in synbio; it is a survival condition. Keasling's "too much cash, too many ideas, no focus" is the foundry's epitaph.

The molecule-selection axiom

This is the single most important strategic decision in synthetic biology, and it is made once, at the start, when you choose what to make — because that choice fixes which cost curve you have signed up to climb. The axiom: start where your cost already beats the incumbent, and climb down — never start at the bottom and hope the curve saves you.

The bottom is commodity: high-volume, low-margin molecules — bulk chemicals, commodity protein, fuels — where the incumbent price is two to six dollars a kilogram and a synbio company must achieve enormous, not-yet-real cost reductions just to break even. A company that picks a commodity at founding has bet its entire existence on the steepest part of a product curve it does not control, bending down years in the future. That is the bet Amyris made reaching for fuels and bulk squalene, the bet Zymergen made on bulk plastic film, the bet that fills the graveyard. The top is specialty: high-value molecules — pharmaceutical intermediates, specialty enzymes, lactoferrin, flavours, fragrances, colours — selling for tens to hundreds of dollars a kilogram, where a synbio process can already be cheaper today, and the company is profitable before any curve has to bend.

The discipline is to win small and high-value first, bank the margin, and earn the right to descend toward larger markets with someone else's profit funding the climb — and to treat the commodity dream as a destination, never a starting point. Every dollar of the protein thesis's specialty-first argument is this axiom applied to one province. Here it is the law: the molecule you pick is the bet, and picking commodity is betting against the only curve that can save you.

Where the geography tilts

One consequence falls out of all of this and points at the rest of the canon. If synthetic biology is decided on the product cost curve, and the product cost curve is governed by the physics and economics of manufacturing at scale, then the geography with the lowest cost of manufacturing at scale has a structural, durable edge — not in the biology, which is a global commodity, but in the one thing the business is actually decided on. The scale-up valley that bankrupts a company in California is an operating condition somewhere else. This is the thread every application piece in the synbio canon develops: "The Protein Thesis" works it for precision fermentation and India's fermentation base; the materials, chemicals, and agriculture provinces will work it for theirs. The constitution only plants it: in a business decided at the plant, the cheapest place to build the plant wins, and that is a fact about geography, not genomes.

The provinces under this constitution

This essay is the law; the applications are the territory. What sits beneath it:

Precision fermentation / protein — the first province built, The Protein Thesis: the cost curve applied to dairy and specialty proteins, the scale-up valley as the killer, India's fermentation base as the edge. Biomaterials — leather, silk, structural materials, packaging, where the incumbent is petrochemical or animal and the parity fight is on performance as much as price. Specialty chemicals and ingredients — flavours, fragrances, cosmetics actives, the high-value corner where the molecule-selection axiom says the survivors live. Agricultural biology — nitrogen fixation, biologicals, crop traits, where the cost curve competes against synthetic fertiliser and chemical inputs. Fuels and bulk chemicals — the oldest and hardest province, the one that competes against crude, included here mostly as the cautionary boundary of the field. Each is the same constitution in a different incumbent's market. None escapes the cost curve; all are won or lost on the product curve at scale.

Close

The constitution of synthetic biology fits in one sentence, and the rest is consequence: synbio is the business of making known molecules more cheaply than the incumbent who already makes them, decided on a continuous cost curve, killed downstream at the plant rather than upstream in the lab, won by focus and molecule selection and the cost of manufacturing at scale — and the horizontal foundry that promises to do it all for everyone is the trap, not the prize. It is the inverse of the therapeutics canon at every joint: manufacturing not discovery, a slope not a readout, parity not novelty, the plant not the clinic, the incumbent not the inventor capturing the value. Read the drug canon and the synbio canon side by side and you have the two halves of how engineered biology makes money — one paid for finding what is new, the other for making what exists cheaper. The protein thesis is one province under this law. So is everything else that programs a cell to make matter. The cost curve is the constitution; the molecule is the bet; the plant is the verdict.

The umbrella of the Atoms and Cells synthetic-biology canon, above "The Protein Thesis." It is the synbio counterpart to "On Biotech Platform Strategy" but inherits none of its machinery — no platform-modality taxonomy, no clinical binary — because synbio is decided on a cost curve at scale, not a readout. The boom-bust receipts and the "no milestones" diagnosis are from Science (Aug 2024, quoting Jay Keasling and Zach Serber), Wikipedia, and bankruptcy filings: Amyris (Ch. 11, 2023), Zymergen (~$1.5B raised / ~$530M 2021 IPO → Ch. 11 2023 → sold to Ginkgo ~$300M), Ginkgo Bioworks (SPAC 2021 → 2024 layoffs of >1/3 of staff), Solazyme/TerraVia (bankruptcy; IP to Corbion), with Intrexon, LS9, and BioAmber as further graves. The enabling-curve framing is Rob Carlson's (the Carlson Curves; synthesis.cc; "Biology Is Technology"), including his insistence that the curve is demand-driven, not inevitable. Techno-economic anchors (biomass protein ~$4–6/kg vs ~$6–15/kg beef/pork; the commodity-molecule trap) are from the Good Food Institute. The field's essayist canon — Elliot Hershberg's Century of Biology, Niko McCarty's Asimov Press, Carlson's synthesis — is the conversation this piece writes into. All figures are directional; the discipline is that no curve here is a law. Manufacturing not discovery; the slope not the readout; the plant not the lab — that is the whole inversion, and the whole constitution.