The market's knee-jerk reaction makes enough superficial sense. Anthropic's AI-powered computer-coding platform Claude can easily help modernize complicated programs written in COBOL (Common Business-Oriented Language), seemingly posing a threat to IBM's (IBM 0.48%) biggest business. The technology giant's mainframes still utilize the aging programming language, after all.
That's why IBM's share price tumbled more than 12% on Monday in response to Anthropic's blog post -- not only could this option take a bite out of IBM's breadwinning software and consulting business, but it also conceivably sets the stage for a more sweeping transition away from IBM's mainframes that have worked so well with COBOL for so long.
The market's response, however, seems to be ignoring a handful of critical details that might have prompted Monday's sellers to rethink their decision. Here are the three most important of these details.
Image source: Getty Images.
1. IBM already provides such COBOL-modernization solutions
Anthropic's idea is a brilliant (as well as increasingly necessary) one. That's why IBM has offered such modernization tools for some time now.
In fact, the company's toolkit even includes options for migrating existing COBOL systems from legacy mainframe platforms to more modern ones like Linux or Windows. That's a key part of IBM's business right now, in fact -- engineering the eventual wind-down of its past in preparation for a future that will look considerably different, but won't necessarily be less fruitful.
2. AI-generated code usually works, but when it doesn't...
The advent of artificial intelligence has been a game-changer for several industries, but particularly for highly technical ones like computer programming. Engineers can simply describe in plain English what they want an app to accomplish, and a platform like Anthropic's Claude spits out the code. And most of the time it works great.
But when it doesn't work, it really doesn't work. CodeRabbit reports that a recent comparison of AI-generated code to human-coded programming shows the artificial intelligence-written code had about 60% more errors, jibing with plenty of other observations. And worse, as the coding work that AI is doing becomes increasingly complex, it's becoming more difficult to find and correct these bugs.
In short, AI's auto-coding tools aren't living up to the hype... at least not yet. It's unlikely Anthropic's Claude is an exception.
Now juxtapose this less-than-perfect modernization tool to the work that's still being done with COBOL. Though ancient, this programming remains at the heart of most ATM transactions, retail purchases, air travel bookings, bank transfers, and more, including supporting a bunch of governmental agencies. These institutions aren't apt to risk an attempt at an upgrade to a more modern solution until they're absolutely certain it will work at least as well as their current platforms. Claude's modernization tools can't provide such a guarantee.
3. Mainframes are still better at certain types of work
Finally, while Anthropic's ultimate unvoiced argument for modernizing COBOL is allowing institutional customers to move away from IBM's mainframes and toward the use of more flexible third-party cloud computing platforms, this premise misses an important point.
But first things first... what's the difference?
It's an oversimplified explanation, but broadly speaking, cloud computing is space rented out on someone else's servers. What the customer does with this space is entirely up to them. It could be simple storage of digital files, or when there's enough computing power available, some cloud platforms are capable of tackling AI tasks. All of it is off-premise, though, and as such, tends to operate more slowly than mainframes.
Conversely, mainframes are on-premise platforms that offer more operational and computational speed; IBM says just one of its new "Z" systems can handle 25 billion encrypted transactions per day. That's not where their advantages end, however. Mainframes are complete stand-alone systems often with built-in security features -- including quantum encryption -- a great deal of flexibility, and extreme dependability with uptimes nearing 100%.
Perhaps most important right now, though, while Claude's AI-powered potential to modernize COBOL coding is being viewed as a threat to IBM's mainframe business, ironically, AI may actually bolster the case for mainframes. As it turns out, IBM's newest mainframes are self-contained soup-to-nuts "full stack" systems that are each capable of 450 billion AI inferences per day. That's huge.
NYSE: IBM
Key Data Points
It's big enough, in fact, that mainframes could conceivably become the AI industry's future workhorses. As technology consulting and services outfit Kyndryl highlighted in its 2025 assessment of the industry, more than half the organizations that use them at all are now increasing their usage of mainframes, with the returns on these modernization costs often in excess of 300%. Indeed, nearly 9 out of 10 of these organizations are specifically using their mainframes to handle generative AI duties due to their strong performance.
IBM stock is far from being doomed
The point is, while it's likely -- maybe even inevitable -- that AI-powered computer coding will eventually become good enough that it can be trusted not just to modernize current COBOL programs but then facilitate the move of these operations off of a mainframe, that's not the case yet. And it may never be the case if mainframes can continue to prove their superiority to cloud-provided alternatives in several ways that matter.
More to the point for rattled investors, its recent setback makes IBM stock an even more compelling prospect. The market just jumped to the wrong conclusion on Monday, ignoring that the company's mainframe (and ancillary) businesses are apt to remain rock-solid well into the future.
In this vein, Straits Research predicts the worldwide mainframe market is poised to grow at an average annualized pace of nearly 8% through 2033. That's not enormous growth, but it's respectable growth that IBM can easily lead. You can plug into this growth while its stock is priced attractively at less than 19 times this year's projected per-share earnings.
