Reading old forecasts
This article was originally framed as a look ahead to 2023. That year is now well in the past, so the more useful exercise is to look back: what did people broadly expect from connected devices around that time, and how have things developed since? The answers here are general and hedged on purpose. Without specific figures that can be verified, it is better to describe directions than to claim precise outcomes.
A useful habit when reading any forecast is to separate three kinds of statements. Some describe a technical direction, which tends to age well. Some describe market size or adoption speed, which often age badly. And some describe a particular product or company, which is outside the scope of this site. The expectations below fall mostly into the first kind.
Expectation: devices will do more work locally
Around 2023, a common expectation was that more analysis would move from distant servers to the device or a nearby computer, a shift usually discussed under the name edge computing. Looking back, this direction has largely continued. Small chips capable of running simple models have become more common, and it is now ordinary for a camera, speaker or sensor to do some interpretation on the spot.
What has been less clear is how quickly this displaced central processing. Most real systems still combine both: quick local decisions and heavier work in the cloud. The forecast was broadly right in direction, while the idea of a clean break from the cloud was probably overstated.
Expectation: artificial intelligence will be built in everywhere
Expectations that artificial intelligence would be woven into connected products were widespread. In the years since, interest in this area has grown considerably, and more products advertise smart features. The honest caveat is that the benefit varies. Some uses, such as recognizing a spoken command or flagging an unusual vibration, are well established. Others add complexity without a clear gain for the user.
A lesson that seems to hold is that clever software cannot fix unreliable inputs. Sensors that drift, gaps in the data and poor placement still limit results, whatever method sits on top.
Expectation: security and rules will tighten
Many commentators expected pressure on security to increase, and regulators in several regions did move toward setting minimum requirements for connected products. The details differ by country and continue to evolve, so it would be unwise to describe any single rule as universal. Alongside that, public awareness of default passwords and unpatched devices has grown, and more makers now offer ways to deliver security fixes.
At the same time, the underlying problem has not gone away. Large numbers of older devices remain in use, some no longer receive updates, and new products still sometimes ship with avoidable weaknesses. Progress is real but incomplete.
Expectation: devices will finally work together
Compatibility was a recurring hope. Industry groups worked on shared specifications intended to let devices from different makers cooperate, and some of that effort has reached ordinary products. In practice, device interoperability is still uneven. Support varies by device type, and getting a mixed set of products to behave smoothly can still take effort. Expect improvement rather than a finished job.
Expectation: faster, wider connectivity
Forecasts also pointed to better networks: wider rollout of newer cellular generations, more options for long-range radios with tiny power needs, and improved home networking. The general pattern has been gradual improvement. Newer mobile networks have been built out at different speeds in different places, and many deployed IoT devices keep using older, simpler technology because it is cheap, available and good enough for sending small readings. That is a useful reminder that IoT hardware often stays in service for many years, so old and new generations coexist.
What was harder to predict
Several things tended to be underestimated in forecasts from that period.
- Supply and cost swings. Shortages and price changes affected electronics in the early 2020s and complicated projects that had been planned on neat assumptions.
- Slow replacement cycles. Industrial equipment and infrastructure last for years or decades, so change in those areas is gradual, and retrofitting old equipment with sensors is often more practical than replacing it.
- Ongoing costs. Subscriptions, support and maintenance often matter more over time than the price of the device itself.
- Fatigue and trust. Some buyers became more cautious after experiencing devices that stopped working when a service was discontinued.
Takeaways for reading future predictions
Three simple habits help. First, treat precise market numbers with suspicion unless a method is clearly explained. Second, trust technical direction more than timing: it is usually easier to say what will happen than when. Third, remember that adoption in homes, cities and factories moves at different speeds, so a statement that sounds true for one may be wrong for another.
Looking back at 2023-era expectations, the broad directions of local processing, smarter software, tighter security and better compatibility all remain relevant. The pace has generally been slower and messier than headlines suggested, which is a fair guide for how to read forecasts today.