In my early days as an industry analyst, I did what most did and issued market forecasts. It didn’t take me long to find out that there was minimal correlation with what enterprises or operators said they would be doing in the future, and what they actually did. To get around that, I built a “decision model” to forecast not behavior but the drivers of behavior, then used the model to move to the creation of a forecast. I don’t survey users these days, or produce forecasts, but I thought it would be interesting to fiddle with the model using the things enterprises tell me. Not to produce long-term forecasts, but to predict near-term tech decisions. The results, I think, are interesting.
The biggest question vendors, and the tech markets, face today is probably one that’s almost never asked explicitly. That question is “How is buyer decision behavior changing in the next year or so?” If we expect the future of any technology to be different from the present, there has to be a pending change in decision outcomes. The model says that there’s a chain of truths that have been operative for decades, and it’s that chain we have to pull to get to some answers.
One thing clear from decades of history is that technology change is brought about by spending change. If you look at the residual value of a given class of technology (network gear, servers, etc.), you find that the pace of change we’ll actually see in a year depends on the ratio between the budget available and that residual value. When that ratio is large, enterprises entertain changes in technology direction, changes in vendor commitments, and so forth. When it’s small, they tend to stay the course.
Let me offer some specifics here. Over the two years of Andover Intel’s information-gathering, we’ve had a small ratio in play. We have that as 2025 opens, for three reasons. First, the number of new projects whose business case opens new budget contributions is at a historic low. Of the enterprises who offered comment here, the average contribution of new projects to IT and network spending is less than 10%, when over the last 30-plus years it’s averaged 34%. For fifteen years, it ran just around 50%, with four years when it went as high as 65%. Second, the asset base has grown over time, making it harder to displace stuff not yet fully depreciated, and finally the years and years of tech investment has leveled the peaks and valleys of asset depreciation, so there’s no single point where “modernization” is facilitated by having a lot of stuff aging out at once.
The result of this low-ratio condition is that any real opportunity to gain market share is generated by specialized places in the IT picture where new projects (that less-than-10% contribution) impact on a narrow front. Think AI and network equipment, for example. IT deployment in house depends largely on creation of “IT clusters” that generate more horizontal traffic, thus creating a specialized data center network that doesn’t have to conform to past technology choices or vendors. Why do network vendors love AI hype? Because it gives them a chance to gain market share, a chance that in 2025 isn’t offered by anything else.
Another interesting point is that the model says that new projects generating higher ratios and thus more transformational opportunity will take slightly over a year from approval to the delivery of a distinctively higher budget, and the budget/ratio good times will last (on the average) slightly less than 2 years, then trail back to baseline over another two to three years. That means that if we want 2026 to be a great year, a year of vendor opportunity, we’d need to identify something to drive it in 2025, and in early 2025 at that.
This truth creates another interesting model prediction. The value of startups is greatest during a period when a prospective driver has arisen, and projects that exploit it are being framed. This puts incumbent vendors under pressure to quickly prepare to exploit the gains in budget dollars, and thus facilitates an exit strategy for startups. Good exits rarely occur during periods when ratios are very low, absent that new driver, and also rarely occur during periods when the ratios are really high, largely because the budget largess has already driven users to make their product decisions. Only revenue will justify a good exit at that point.
A third point is that buyers, whether they’re enterprises or network operators or cloud providers, don’t actually believe the hype. Taking our current AI hype as an example, the tone of published material on AI was overwhelmingly positive in 2024 and is the same in 2025. It was like the cloud—everything is moving to it. Enterprise technologists never believed in the universal cloud, and they don’t believe in universal AI either. And here’s an interesting point; the cloud drove cloud spending, but didn’t drive a revolution in enterprise IT capex. It increased the expense budget only. Right now AI is doing the same thing.
Point four is that the greater the budget contribution, the greater the new-project benefit pool, the longer the positive impacts will sustain, but the longer the time it will take for the peak to be reached. What I’ve characterized as the “three cycles of IT” in the past (in the 1960s, the 1970s, and the 1980s, roughly) happened because of a transformational shift in technology application that had broad impact. Interestingly, each of these cycles got shorter and each contributed less to total spending than the one before. Since then, because of what my model identifies as the “ratio problem”, we’ve not had a major cycle. I submit that this is because it’s harder these days to transform.
Enterprises universally characterize the benefits of IT and networking as “improvements in productivity”. There’s only so much improvement possible in anything, and the most attractive targets are things that offer a lot of gain with minimal cost and risk. ROI, in short. We’ve done them, or at least the attractive and easy ones. What’s left is actually very large—the potential untapped productivity benefit pool is larger than what’s already been realized—but also very complex, and requires a larger “up front” investment. We’ve done what we’ve done by exploiting things familiar and in many cases already in place, but what remains will require breaking new ground.
The big take-away from all of this is that vendors who are looking for better profits had better starting to look for better business cases. Even stealing market share as a strategy will be difficult in a time when our ratios seem to be stuck at a low value; the safe choice and the easy choice is to stay the course. But that gets them to breaking new ground. The obvious questions are “where is the new ground” and “will it be broken”. I’ve pointed out in past blogs that roughly 40% of the workforce, the group not tied to desks, are obvious candidates. However, they’ve been that all along. The real question is whether we’ll uncover something to lower the barriers to reaching this non-empowered group. Since we’ve never been in ratio hell for as long as we have now, models can’t answer that one, but I did get some insights from enterprises that I’ll share in the next blog.
