October was a month about the gap between owning a right and controlling an outcome.
A valid patent on a drug’s mechanism that left its owner with no market. A licence signed years ago that now decides which defence contracts a company may bid for. A US$4.7 trillion profit shift that goes to whoever holds something a competitor cannot copy, which is rarely the model and rarely the product.
Nine posts published this month fall into three groups: what the contract decides, why owning the invention is not the same as owning the market, and what is left to protect when the capability itself is cheap. Read together they make one point from three directions. The grant, the licence and the filing are instruments. Control is the result, and it is decided elsewhere.
The contract decides what you control
The licence, not the patent, sets the price. The UK dropped its plan to set royalties on standard essential patents, a Delaware jury found Arm in breach of its Qualcomm architecture licence and awarded two dollars, and Liquidia paid US$380.7 million to keep the larger market while surrendering one indication. In each, the value sat in terms settled long before anyone reached a court.
The same terms decide market access. In India’s draft defence procurement rules, IP ownership is a condition of winning the contract: a vendor that holds its design data under a foreign licence cannot bid. An inbound licence is a market-access document, and chain of title is a tender question.
Two posts applied the test to AI. When a company adds AI to a product, the useful question is whether the move changes what it controls – the interface, the data, the customer relationship – or just adds a feature on a platform someone else owns. And as frontier models get cheaper, the buyer should ask what it is actually buying, because the model layer confers no exclusivity. Keep the contracts flexible, plan for a model disappearing, and put the protection around your own data and workflows.
Owning the invention is not owning the market
A failed clinical trial made the sharpest version of the point. A patent covers a mechanism, not the benefit the mechanism delivers, and where the existing standard of care already captures that benefit, owning the mechanism is not a position. Value rights against the incumbent default, and secure the combination claims early.
China’s energy plans showed the industrial version. Control of manufacturing and deployment capacity can matter as much as the invention, so who gets to scale it is a filing question: file where the product is made and installed, protect the process know-how, and read the supply contracts for IP terms.
For challengers, the lesson is timing. A better business model buys a window, not a barrier, and counter-positioning only converts into a durable position if the core mechanism is filed before the incumbent studies it or a second startup arrives without the incumbent’s constraints.
What is left to protect when the capability is cheap
Bain’s estimate that AI will move US$4.7 trillion in profit by 2035 raises the question of who keeps it. Only a quarter is productivity, which every competitor gets too. The rest goes to whoever owns something others cannot copy: trade secrets over the data and operating know-how, patents over the methods a rival could reach independently.
And whatever the asset is, it is worth what can be proved. IBM’s quantum result was built so the answer could be verified while it ran, and was challenged within days anyway. A company trading on a technical edge should expect the same test from an examiner, a court or an acquirer, which is why a lead you cannot prove is a lead you do not own.
What to watch
The thread through October is that the decisive document is rarely the one with the seal on it. It is the licence, the supply contract, the partnership terms and the evidence record, and those are written while the business is still deciding what it wants to control.
Three things are worth watching as the year closes. The Arm and Qualcomm royalty question goes to the judge next, and will show what a single licence obligation is worth. The UK’s court-anchored FRAND process will set the template for how standards royalties get negotiated without a regulator. And the AI contracts being signed now, as model prices fall, will decide who owns the data and workflows that the Bain profit shift rewards.
The question for the next strategy review is not what the company owns. It is what, after each deal and each filing, it actually controls – and whether the documents say the same.