IBM CEO Arvind Krishna told investors that open-weight AI models now make strong business sense for companies that run AI on their own infrastructure. Speaking on the company\u2019s third-quarter earnings call, he said the performance gap between top closed models and open alternatives has narrowed sharply.
In 2024, Krishna said, closed frontier models were plainly superior. \u201cIf you ran some assessment, some benchmark, and it was broad-based, not narrow, the top model might score a 70, and the open model might score a 20 or a 30,\u201d he said. \u201cIn late 2025, we began to see that gap shrink, where maybe the gap was not 70 to 20 or 30, it was 70 to 50.\u201d
The cost difference is what Krishna finds most compelling. He said open-weight models currently deliver \u201canywhere from 25 percent to 10 percent of the token cost of the frontier models.\u201d He added, \u201cThat is very attractive given that they\u2019re now 90 percent as effective.\u201d
He framed the decision as a simple one for any enterprise. \u201cWould you like something that\u2019s 10 percent of the cost if it\u2019s 90 percent as good as the best, or do you want to pay that 10 times higher cost?\u201d
Krishna believes most businesses will not pay the premium. \u201cThen the question is: how many tasks require that? I would tell you some do, but the vast majority, summarizing an email, examining a contract, answering a customer call, composing an email, all of those can be done as well with the much cheaper answer.\u201d
Those economics, he said, are driving booming interest in on-prem AI infrastructure at IBM. He thinks most businesses can build useful AI infrastructure in just two or three racks. Many, however, will struggle to get it running.
\u201cThe question becomes: how do you do that?\u201d Krishna said. \u201cThe question that the enterprise struggles with is: \u2018Where are my kernel engineers? Where are the people who worry about the underlying network? Who can make sure that the model is clustered on GPUs that are close to each other in network proximity? Who can make sure that you\u2019re doing the right caching of results so you get an incredible response?\u2019\u201d
He did not offer detailed answers. He said only that IBM will produce software to ease AI deployments. He also predicted that demand for on-prem AI will spur sales of IBM\u2019s Spyre mainframe accelerator cards.
Mainframe, software and cloud results
Krishna said the current z17 mainframes have delivered $1 billion more revenue than any previous model. He called the company\u2019s new rackable models \u201cthe strongest midrange launch in IBM Z in nearly two decades.\u201d Mainframe revenue still fell 20 percent, which IBM expected, since it has been 18 months since the z17 launched.
Infrastructure revenue grew 5 percent year over year to $3.7 billion. That trails the improvement reported by rivals Dell, HPE and Lenovo.
Software performed better. It grew 9 percent to reach $7.8 billion for the quarter. Red Hat was the standout, growing 11 percent and lifting annual recurring revenue by 30 percent. Red Hat\u2019s virtualization products have won $800 million in deals, with a pipeline of another $750 million. IBM believes there is more to come, given that some VMware users may be looking to move.
Total quarterly revenue reached $17.1 billion, up 5 percent. IBM\u2019s share price wobbled in after-hours trading and ended down a point or two.
A separate IBM Cloud incident
On the same day, IBM reported a severity one incident in its cloud business. The incident was caused by \u201cemergency maintenance\u201d that left users unable to provision new resources or virtual private clouds for 90 minutes.
Users of IBM\u2019s Classic Cloud, the service IBM acquired from SoftLayer in 2013, were told the company will \u201crestore new provisioning and resource management in our Classic environment\u201d by 08:00 UTC Thursday. That means the incident will last at least 16 hours. IBM began that work a handful of hours after critical bugs in PHP were announced. The Register has asked whether fixing that flaw is the reason for the incident.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.