The Data on AI
By Christopher Steiner
IBM now runs commercials on podcasts that invite companies, presumably ones that have enough money to not worry about overpaying for an underdelivered project, to supercharge their AI efforts using IBM's workforce, which, the company says, includes 64,000 AI experts. Sixty-four thousand experts.
IBM has 160,000 consulting employees across the world. Some of those are off-shore code-writers. Some of those are operations staff that never really see client work. Knowing that, some back-of-the-envelope math yields that at least half of IBM's front-line consultants must qualify as AI experts. Impressive.
But let's give IBM a pass here. There was a time when it was the world's foremost authority on AI, and that old IP is still laying around, but much of it may not be worth much.
In either case, IBM has ramped up its chatter about AI, an not just in commercials. For the five years before ChatGPT's release, IBM mentioned AI and related terms an average of 15 times per quarterly earnings call. That's a high number compared with most companies' transcripts from that period. After ChatGPT's release, however, IBM poured AI gasoline on its earnings calls, averaging 51 AI and related mentions per call since the fourth quarter of 2022.
But back to that legacy stuff. IBM has been doing this longer than any of the Magnificent 7 (Microsoft, Google, Amazon, Nvidia, Meta, and Tesla). BTW: We're not sure why Tesla is thrown in with this bunch.
IBM's durability in the space shows, it has more AI-related patents registered with the U.S. Patent Office than any other company, a startling statistic considering that it's not amongst the real leaders in the space. The thing about IBM, however, is that it used to be a technology company. But it hasn't been for some time. It's a consulting company, primarily. It still has pockets of people working on things such as Watson and Quantum computing, but its core business is the same as that of Accenture or Deloitte.
For decades, however, IBM was a technology company that primarily made products for decades, everything from software (databases, computer languages, OS packages, and on) to hardware—the bombest of PCs back in the 1980s were IBMs, something that continued into the 1990s. Its laptops, aka ThinkPads, continued as a strong brand until the company carved that business off and sold it to a Chinese buyer in 2005, which rebranded it as Lenovo in 2007.
The point: IBM used to be the alpha of technology companies in the United States. Its research into AI, machine learning and expert systems back in the 1970s and 1980s far surpassed that of anybody else. So it does have a pile of patents, the largest of anybody, when it comes to these subjects, but they're legacy patents filed in a different era of compute power. Some of the computer science holds, of course, but the sheer volume of IBM's AI-related IP is somewhat misleading.
The capex of IBM does not reflect that of a company that's actually building IP and investing in product. It's a services company whose product is sending hordes of people to complete things such as SAP implementations. Companies with growing products spend more, over time, on capex, especially in this era of the AI arms race.
By our count, IBM does employ about 3,200 data scientists, which is impressive, especially considering that Microsoft, the king AI at the moment, employs about 5,700. But also consider that Accenture has 3,700 data scientists, 500 more than IBM. Accenture and IBM are essentially renting these data scientists out to others, which is fine, but it's not the same as Microsoft, whose data scientists are working on internal projects and external products and AI.
Also, any company with 64,000 'AI Experts' and only 3,200 data scientists is coming up short somewhere. Most likely: it's a lot of pontificating by 'experts' with a limited amount of 'doing' by those who actually understand data and AI. Even with all those old patents.