IBM has been pinching pennies for years, trying to squeeze fading businesses for cash even as it desperately scoured the horizons of artificial intelligence (AI) and cloud computing for new sources of sustenance. It largely hasn’t worked. Though mainframes have propped up IBM’s last three quarters, the company’s revenue has declined every year since 2012, in part because “IBM has a Watson dilemma,” as The Wall Street Journal opined.

One area that is working for IBM, and in a highly-profitable way, is collecting patents. IBM touts its “25 years of patent leadership,” but others rightly warn that “A lot of the research that IBM conducts may not translate to immediate business gains.” Amen to that. The immediate gains are patent royalties. In short, instead of selling technology based on patents, IBM increasingly resorts to selling access to the patents themselves. This is the wrong move.

Collecting cash for cobwebs

IBM has been accumulating patents for a long time. Over the last 25 years, the company has amassed over 100,000 patents. In 2017, the company collected 9,043.

While patent collectors will often claim that their portfolio is a good indicator of the deep research and development they do, rarely do we see patent heft translate directly into product success. Why? Because rarely do products succeed simply because of technical merit.

Instead, the most successful companies are those that can execute (sales, marketing, etc.) around a product, whatever its technical merits. In this area, IBM has largely failed over the last decade.

SEE: Software licensing policy (Tech Pro Research)

Nor is it alone. Of the top-10 companies awarded patents in 2017 (IBM, Samsung, Canon, Intel, LG, Qualcomm, Google, Microsoft, Taiwan Semiconductor, and Samsung Display–in that order), many have struggled against less patent-rich, but more execution-heavy rivals.

Samsung may dominate the smartphone market, but Apple cleans up the profits. Intel still makes a lot of money but has been nowhere in mobile, allowing ARM to win big (and long-time customers like Apple to finally abandon Intel to build their own chips). And so on.

Companies also seek patents to use as a defense shield against patent trolls or other patent-rich companies. In IBM’s case, its recent lawsuit against Groupon (ultimately worth $83 million) belies that apparent intention. Prodigy was buried decades ago, yet its patents (some of which found their way into IBM’s hands) continue to be tools for extorting cash from more modern applications. In the case of Groupon, it was forced to pay $83 million to IBM, while Priceline settled for $34 million. Well over $100 million for patents that are worthless in terms of fostering innovation, and instead are simply being used to prop up IBM’s fading profits.

Focus on the other kind of innovation

It doesn’t have to be this way. IBM has long been one of the pioneers in open source software, which is where most usable innovation seems to be happening today. From TensorFlow to Apache Kafka to Kubernetes, if IBM wants to compete with modern technology giants like Google and Microsoft, it needs to innovate in the same way they do, too. Yes, they still gather patents, but their more interesting work emerges as open source software.

SEE: How Mark Shuttleworth became the first African in space and launched a software revolution (cover story PDF) (TechRepublic)

With IBM digging deep into blockchain, AI, and cloud, it can’t afford to try to lock up innovations in patents, eventually to squeeze comparative pennies from royalties. While IBM collects cloud patents Amazon Web Services (AWS), for example, collects tens of billions of dollars in cloud revenue. Real businesses execute, while others…sue.

IBM could take a cue from Microsoft in this. Microsoft (which unfortunately still seeks royalties from Linux) has turned most of its attention to working with the technologies it once tried to sue into oblivion, with Linux front and center. Open source provides a key way to turn innovations into applications that companies actually use…and pay for.