Part 1 argued that your IP portfolio is probably your most undervalued asset. Part 2 gave you a framework for choosing between two instruments: run a discoverability analysis, patent the perimeter, keep the internal parts secret.
That framework has a third variable. Regulation and caselaw decide whether the instrument you chose can actually be used, and they decide it differently in every sector.
The Legal Backdrop
Four things vary across biotech, energy and aerospace: the statutes that apply, the caselaw that has accumulated under them, the regulatory regime, and how the industry is commercially organized.
In biotech the structure works in your favor at nearly every layer. Composition of matter is the strongest type of patent claim and is usually the heart of biotech IP. Hatch-Waxman makes a generic entrant notify you before it can launch. Section 156 gives back years of patent protection lost to FDA review. The regulatory regime means a competitor who reads your patent still faces a decade of trials. The commercial structure cooperates too: acquirers are pharma companies buying discrete assets, and licensing is a normal transaction rather than an admission of weakness.
In energy the venue is the critical variable. Your infringer may be offshore and unreachable in district court, so enforcement collapses into a trade question governed by Section 337 of the Tariff Act rather than by patent law. Bayh-Dole may have already settled where you are permitted to manufacture if the IP came out of federal grant money. And a globalized industry with distributed production means the party taking your market is frequently one you have no practical jurisdiction over.
In space, there are unique limitations on the strength of patents. Section 1498 removes your injunction against the government and its contractors, a reading the Federal Circuit has extended through Astornet and Arlton. Section 105 lets a competitor register a spacecraft under a foreign flag and step outside your patent. Export control limits where you can file.
Four Practical Questions To Ask First
01How wide is the gap between knowing and doing?
02Who is the infringer, and can the law reach them?
03How will you even find out?
04Do the patent clock and the revenue clock line up?
Biotech: Where the System Cooperates
01 · Disclosure is nearly free
The knowing-doing gap in therapeutics is the widest in the economy. Publish the exact structure of your molecule and a competitor still faces eight to twelve years and several hundred million dollars of clinical work before they can sell anything.
Compare that to a battery founder publishing an electrode coating method, where a competent competitor with a pilot line closes most of the gap in two quarters.
This is why "file early and file broadly" is sound biotech advice and expensive advice almost everywhere else.
02 · Your infringer is legally required to announce itself
Under Hatch-Waxman, a generic manufacturer seeking to enter before your Orange Book patents expire must file an ANDA with a Paragraph IV certification and send notice to both the patent owner and the NDA holder. You get a letter. You get 45 days from receipt to sue, and suing triggers a 30-month stay on their approval.
Sit with how strange that is. In every other sector, detecting infringement is a real cost center: teardowns, market surveillance, watching customs data, hiring people to reverse-engineer competitor products. In small-molecule pharma, the statute makes your adversary mail you a warning.
Biologics are weaker here than founders assume. The BPCIA "patent dance" looks like a parallel process, but Sandoz v. Amgen (2017) held that the applicant's obligation to hand over its application and manufacturing information is not enforceable by injunction under federal law. The dance is effectively elective. What you are actually guaranteed is the 180-day notice of commercial marketing. Plan around the notice, not the disclosure.
03 · Biotech is the only sector with a patent term repair mechanism
Patent term extension under 35 U.S.C. § 156 restores half of the testing phase plus all of the approval phase, capped at five years of extension and fourteen years of remaining term post-approval, on one patent per product. Layer on regulatory exclusivities that run independent of patents: five years for a new chemical entity (four if the ANDA carries a Paragraph IV certification), seven for orphan designation, twelve for a reference biologic.
No such mechanism exists for a fusion company, an electrolyzer company, or a launch company. Nobody extends your term because your reactor took eleven years to build. Congress patched this problem for one industry and left the rest to absorb it.
Where biotech founders lose: § 112. Amgen v. Sanofi (No. 21-757, decided unanimously May 18, 2023) held that functional genus claims covering antibodies defined by what they bind and what they block were not enabled across their full scope. The claim you want is the claim you probably cannot keep. Acquiring pharma companies know this, and they price your portfolio on a small number of narrow, well-supported claims covering a specific molecule with a specific remaining term in specific geographies.
The diagnostics inversion. If you are building diagnostics rather than therapeutics, almost none of the above applies to you. Mayo (2012) and the Federal Circuit's Athena Diagnostics line (2019, rehearing en banc denied over four dissents, cert denied 2020) have left US method-of-diagnosis claims close to unenforceable. Europe is the inversion: the EPO excludes diagnostic methods practised on the body, but grants in vitro method claims on a removed sample, which is where most real diagnostic IP lives. So flip the usual geography. Build the patent moat in Europe, and in the US substitute a regulatory moat for the patent one you cannot get.
That substitute is CLIA, the Clinical Laboratory Improvement Amendments of 1988, which govern any lab running tests on human specimens for diagnosis or treatment. Certification is tiered by test complexity, and the tier that matters is high complexity, because any test a lab develops itself defaults to high complexity under the regulations. That tier opens the laboratory developed test path: an assay you design, validate, and run inside your own lab rather than sell as a kit. FDA asserted oversight of LDTs in a May 2024 final rule and lost. The Eastern District of Texas vacated it nationwide in American Clinical Laboratory Association v. FDA on March 31, 2025, FDA did not appeal, and in September 2025 FDA reverted the underlying regulation to its pre-2024 text. CLIA is the operative regime.
Here is why that functions as a moat. You cannot exclude a competitor with a patent, but they still have to stand up a certified high-complexity lab, validate their own version of the assay, pass proficiency testing, and accumulate enough specimen volume to match your performance data. Certification plus validation record plus reference dataset is a real barrier, and it happens to be the kind trade secret law protects and patent law does not. Slower to build than a patent. Considerably harder to design around once built.
Energy: Where the Venue Decides the Claim
Energy inverts nearly every biotech assumption.
The knowing-doing gap is narrow. Cell chemistries, stack architectures, coating processes and catalyst formulations are things a well-capitalized competitor can reproduce quickly once they know the answer. Part 2's LG Chem example was an energy story for a reason. The $1.8 billion settlement turned on manufacturing know-how, not on the patents covering the same battery chemistry.
Your infringer will not announce itself. It is frequently offshore, selling into your market, and effectively unreachable in district court.
So the venue changes, and the venue should change your claims.
The forum that works in energy is Section 337 at the International Trade Commission. The ITC cannot award damages. It can do something better: exclude infringing articles at the border, enforced by Customs, on a target schedule of roughly twelve to sixteen months to final determination, plus a sixty-day Presidential review.
First Solar's TOPCon complaint, ITC Inv. No. 337-TA-1494, instituted March 2026, seeking a general exclusion order on a single patent. First Solar moved to withdraw the complaint in September 2026 after a Section 232 proclamation on polysilicon and derivatives, and its parallel district court suits continue.
One article claim, asserted at the border, can reach an entire offshore supply chain at once, including companies you never named.
The precedent that the remedy is real: GE won a limited exclusion order and cease-and-desist orders against Siemens Gamesa on variable speed wind turbine generators in Inv. No. 337-TA-1218 (2022), later rescinded on GE's own petition once the conditions that produced them had passed. LG's Section 337 trade secret case against SK Innovation (Inv. No. 337-TA-1159) produced a ten-year limited exclusion order and a settlement on the last day of the Presidential review period.
Two drafting consequences follow directly, and most energy patents ignore both.
Claim the article that crosses the border. The ITC needs an imported article that infringes. A method claim covering a process practiced entirely inside a foreign factory gives you very little. 35 U.S.C. § 271(g) partially covers products made abroad by a patented process and then imported, but it excludes products materially changed by subsequent processes, a defense that is easy to run in multi-step chemical manufacturing. Claim the cell, the membrane, the module, the finished stack. Claim what shows up on the dock.
Your US pilot line is partly an enforcement decision. Section 337 requires a domestic industry that "exists or is in the process of being established," satisfied through significant investment in plant and equipment, significant employment of labor or capital, or substantial investment in exploitation including engineering, research and development, or licensing. The Federal Circuit's Lashify decision (2025) broadened what counts. Founders make the pilot-line decision on capital and talent grounds and then discover, years later, that it also determined whether they had access to the only enforcement venue that works. Decide it on purpose.
The second buyer nobody models. Energy IP has a customer beyond acquirers and licensees: the project lender and the technology risk insurer. At financial close on a $400 million facility, a clean freedom-to-operate opinion does more work than three additional patents. Underwriters price the possibility that a technology claim halts the plant.
The § 204 trap. If your foundational patents came out of a DOE-funded university lab or a national lab CRADA, Bayh-Dole's preference for US industry attaches. An exclusive license to use or sell in the United States generally requires that products be manufactured substantially in the United States, absent an agency waiver. Foreign-only exclusive rights sit outside the restriction. The waiver process is slow, discretionary, and not something you want on the critical path to a supply agreement.
Space: Where Your Dual-Use Customer Is Immune
Space has the strangest IP structure in deep tech, and the reason is a statute most founders in the sector have never read.
01 · 28 U.S.C. § 1498 removes your injunction
When a patented invention is used or manufactured by or for the United States, with the government's authorization or consent, the patent owner's remedy is a suit against the United States in the Court of Federal Claims for reasonable and entire compensation. That exclusive remedy is why no injunction is available.
The Federal Circuit read the shield broadly in Astornet Technologies v. BAE Systems, 802 F.3d 1271 (Fed. Cir. 2015), which barred claims against contractors accused of inducing the government's infringement. More recently, in a nonprecedential disposition, Arlton v. AeroVironment (Fed. Cir., February 2026) held that a contractor performing SBIR and STTR work is performing for the government and is shielded accordingly, rejecting the argument that the SBIR statute narrows § 1498.
Translate that into business terms. Your patent might only provide you with a compulsory license at a royalty a judge will set years later, with no leverage to stop the program in the meantime.
Now notice the sequencing problem. Most space startups take early revenue from NASA, Space Force and the NRO, and later expand into commercial markets. Their patent portfolio is worth least in the segment they enter first and most in the segment they reach last. That should change how much they spend on prosecution in year two, and it almost never does.
02 · § 105 lets a competitor register their way out of US patent law
35 U.S.C. § 105 treats an invention made, used or sold in outer space as occurring in the United States, but only for a space object under US jurisdiction or control, and it carves out objects carried on the registry of a foreign state under the Registration Convention.
Read that again if you are building in-space manufacturing, on-orbit servicing, or anything where the infringing act happens above the Kármán line. A competitor can arrange for a foreign launching state to register the spacecraft and operate outside the reach of your US patent while serving the same customers.
There is no legislative fix on the near horizon. The contractual workaround practitioners discuss is negotiating a registry commitment directly with the counterparty. It is worth raising in component supply and hosted payload negotiations, but state of registry is ultimately determined by the launching state's obligations under the Registration Convention, not by a private term sheet.
03 · SBIR data rights change the filing calculus
In space and defense-adjacent hardware, the patent-versus-trade-secret question has a companion that founders rarely ask: what does filing do to your government data rights?
Properly marked SBIR/STTR data carries a non-extendable twenty-year protection period running from the date of award, during which the government cannot disclose it outside the government. When that period ends, the government holds government purpose rights, meaning it can hand your data to a competing prime for government purposes. Get a marking wrong and the contracting officer can notify you; fail to correct within sixty days and the government may ignore, remove or correct the legend at your expense. DoD's rule codifying the twenty-year period took effect in January 2025, and the proposed changes to marking requirements were deferred to a separate rulemaking, so the marking rules are the unsettled piece.
The part that catches founders: publicly disclosing SBIR-derived technical content, including through a published patent application, risks moving that content into the category of data the government already holds unlimited rights in. No regulation addresses patent publication by name, which is precisely the problem. A filing decision made purely on patent logic can degrade a data rights position worth more than the patent, and nothing in the prosecution process will flag it for you.
04 · The moat might be the trade secrets that enable the qualification
Space has a wide knowing-doing gap, but for a different reason than biotech. The hard part is rarely the patentable architecture. It is the thermal-vacuum campaigns, the radiation tolerance characterization, the vibration qualification, and the flight heritage. Hardware in orbit is also hard for a competitor to get hold of and take apart, which further reduces the urgency of filing.
One more constraint that catches space founders late: 35 U.S.C. § 184 requires a foreign filing license before you file abroad, and ITAR defers to that process only for the patent filing itself. Technical data that goes beyond what the filing supports still needs its own State Department authorization.
Side by Side
Biotech leans towards patents. Disclosure is close to free, detection is done for you by statute, and term gets partially restored. Trade secrets are real but secondary, mostly formulation and process chemistry. The optimal portfolio is deep rather than broad: a small number of well-supported claims on the molecule, filed early, prosecuted hard, because § 112 is where these die and an acquirer is buying the remaining term on a specific asset. Diagnostics is the instructive exception. Strip out the patent layer via § 101 and the weight shifts immediately to CLIA certification and secrecy.
Energy leans towards trade secrets. The knowing-doing gap is narrow, which makes every filing expensive in the only currency that matters. LG Chem's $1.8 billion came from manufacturing know-how, not from the patents on the same chemistry. Patents still earn their place, but for a specific job: giving you an article claim assertable at the border, since that is the only venue where your actual infringer can be stopped. Narrow, article-focused, filed late enough to preserve term, sitting on a serious secrecy program.
Space splits by customer. Sell to the government and your patent degrades into a compulsory license under § 1498, while SBIR data rights do the protective work instead. Sell commercially and patents recover normal force, subject to the registry problem. Underneath both sits the qualification record. Protect the data rights position first, file for the commercial market you will reach in year five, and raise registry terms now while they are still cheap to negotiate.
What This Means for Founders Right Now
- 01Run a clock-alignment calculation this week. Take your honest time to meaningful commercial revenue, add it to your earliest filing date, and subtract from twenty years. If the remainder is under seven years and you are not in a sector with statutory term restoration, your filing calendar needs to move right. Provisionals and continuations exist for this.
- 02Map your funding history to your rights. Federal grants, university licenses, SBIR awards and CRADAs each attached conditions. Find out now which of them constrain manufacturing geography, which give the government a license, and which have data rights markings that were done wrong. This is a two-week exercise at seed and a deal-breaking diligence finding at Series B.
- 03Ask what your patent teaches, in your sector's units. Not "is this reverse-engineerable," which Part 2 covered, but "how many months closer to revenue does this document put a funded competitor?" Twelve years in therapeutics. Two quarters in cell manufacturing. The answer should determine how much you are willing to disclose.
- 04Get one freedom-to-operate opinion before you need it. In energy especially, it is a financing document, not a legal formality.
Castle Fund is a seed-stage venture fund backing deep tech companies with defensible IP moats. If you're building at the frontier of biotech, energy, aerospace, robotics, or semiconductors, we'd like to hear from you.
ZeroToIP helps startups build and execute IP strategies from day one. Learn more.