Nvidia, no surprise, is very interested in AI adoption by telcos, and they’ve just released a 2026 survey on it. Let’s take a look, exploring the conclusions and also how the results compare with what I hear from the 88 telcos I’ve chatted with in the last 6 months.
Most of those who follow my blog know that I am uncomfortable with surveys, given that experience has showed me that people often don’t respond properly, that the survey design often dictates the responses, and that many who are asked to participate have no qualifications to answer the questions. However, I admit that the only reliable way to get data from users of technology would be to investigate in person, which is hardly practical. The Andover Intel approach is to glean information from users’ questions and spontaneous remarks rather than to ask what will almost surely be leading questions. Even this is biased, since we can’t assume that a representative sample of buyers will ask or remark. Keep that in mind here, please.
The Nvidia survey covers global responses from over a thousand people. I rely on somewhere between three dozen telco responses and almost 300, representing 88 telcos, and I consolidate multiple responses by organization. My fear on the Nvidia approach is that large a group is unlikely to be a representative sample, or contain only people qualified to answer. Keep this in mind too.
The keynote comment in the report illustrates the tension here. It says that 99% of respondents say that AI has helped enhance employee productivity, and 26 say the impact has been significant. Of the 88 telcos who commented on AI from January 2025 to the present, only 17 said they had seen any productivity benefit to AI, and none said the benefits were significant. The issue here, I think, is that AI might “improve productivity” in that it might make some tasks easier, but whether this is a benefit depends on whether that improvement translates to a bottom-line impact. So far, none of my 88 telco contacts claim any substantial cost reductions attributable to AI.
This is an issue I’ve pointed out with enterprise AI too. In fact, most people say that there are situations where they find AI helpful, but most won’t pay for it. Same among workers; you can say it helps you without proving there’s a benefit to offset your company’s paying for it. We’ve proved, with AI, that if you give something away, people will take it. That doesn’t seem to me to be either surprising, or helpful in making a business case to justify major AI infrastructure investment.
The next point is that 65% of Nvidia’s responders said that network automation was being driven by AI. I hear that AI is being assessed for network automation missions by over three-quarters of telcos. I’ve heard that most operators think network automation is a top use case, more than the 59% who say that in the Nvidia survey, but all of this comes under a subhead that “Autonomous Networks Take Priority”, and telcos tell me that they are not yet ready to allow AI to autonomously manage networks. Would they like to? Sure, but none thought they’d take the leap in 2026, and only 32 said they believed they would rely on autonomous netops in three years.
The next subhead says that “Distributed AI Computing is on the Rise, Fueling the Path to AI-RAN”. Three telcos tell me that they are doing anything, even seriously testing distributed AI for AI-RAN. The detailed comments in this section do admit that a lot of this is in the area of wireless R&D, but while the report says that telcos are “stepping up investments in AI-native RAN”, telcos are not telling me that, and the comment flies in the face of reports that more and more players are leaving open-RAN models, which is where the only three telcos with any AI-RAN activity are playing with it.
I think you’re getting the point here, so let me just highlight some key points:
- Survey says almost half report AI has helped open new revenue opportunities; no telco told me that.
- Survey says that 89% believe that open-source is important to their AI strategy, that 60% say their company uses generative AI, and 48% that they use or are evaluating AI agents. All my telco contacts say open-source AI is important, all say their company uses generative AI somewhere, and all say they are evaluating AI agents.
- The top use case for AI in the survey is network automation, but the top use case I hear is customer service chatbots, which top the list for 62 of 88 telcos. I note that the North America data from the survey more closely matches my responses, but my telco comments are worldwide. The survey said that half of respondents said that network automation topped their use-case list in ROI, none tell me that.
- In the survey, 54% said data-related issues were their top AI challenge; my comments put expertise at the top (70 of 88) with data a fairly distant second (43 of 88). However, that figure is close to the survey comments on data issues. The fact that data issues include privacy and sovereignty mean to me that those responding to Nvidia were focusing primarily on public-cloud generative AI.
- While I’ve noted that the attitudes expressed on the acceptance of autonomous network operation via AI are highly optimistic, the report later admits that only 4% report high or full autonomy and 25% report little just basic autonomy.
- The report shows that AI-native plans seem to focus on wireless networks, which my own telco comments validate. It also shows that the interest is greatest for small telcos, which I can also validate; they have less access to skilled netops personnel. Most of the plans are safely in the future (2027-2030), which is why those who comment to me don’t really see anything happening; something in that time range is almost never budgeted, so it’s not real.
- The survey said 90% believed AI was helping to increase annual revenue; none tell me that. Exactly the same result for opex, from the survey and from comments to me.
- 99% in the survey said that AI had boosted productivity; less than 20% tell me that. In the survey, 26% say the impact is major or significant; none tell me that.
- 89% say they are increasing their AI budgets for 2026; this is also what I hear, but only 11 characterized their increase as significant.
- The survey said that 48% were using or assessing AI agents; all telcos tell me they are assessing it, and 48 said they had at least one trial/test or use in play.
In all, I can’t validate the results of the survey except in a few areas. I’m not saying that it’s wrong, but that it doesn’t conform to what I hear myself. Subjectively, I think it is likely wrong, for the bias reasons I’ve noted above. I’ve done surveys for decades, and audited survey results from others. It is incredibly difficult to get an accurate result, even if you’re trying, and surveys published by vendors are usually published because they paint a picture favorable to the vendor, which should be no surprise and isn’t an unreasonable thing to find. AI in telco has most of the same challenges as AI in any other vertical; that’s what I’m hearing. Attitudes are hopeful, exploratory, but not often fully committed, and I don’t think that they’re likely to become fully committed in 2026.
