A robot folds a garment, artificial intelligence organizes a day, glasses translate a conversation. On a trade-show stand, the future always works. The real test begins elsewhere: when it is time to order, pay, connect the product to your tools and, above all, get it repaired. Amid the major technology shows of 2026, one question therefore deserves to replace the traditional “is it revolutionary?”: what remains of the promise when the spotlights go out?
A note on methodology: this analysis takes a September 2026 perspective but draws on facts documented through December 2025. It is not a verified account of announcements at CES or VivaTech 2026. The developments discussed are projections, not outcomes observed at those events.
A trade show demonstrates possibility, not maturity
CES in Las Vegas and VivaTech in Paris serve an essential purpose: bringing together manufacturers, startups, investors and buyers. They showcase technologies before they become widespread, sometimes before they enter industrial production. The problem begins when three different realities are grouped under the same innovation label: a laboratory experiment, a demonstration model designed to raise funds and a product actually on the market.
A prototype is not a deception. It is a way to test an idea, a use case or a process. But its demonstration may depend on precise lighting, a perfect connection, an unseen operator or a carefully prepared scenario. For a robot, ask whether its movements are autonomous, preprogrammed or remotely controlled. For AI, check whether it is handling a new request or replaying a previously validated workflow.
Availability: can you actually place an order?
The first test is almost trivial: look for the buy button. Then read what lies behind it. Joining a waiting list is not placing an order; a deposit does not guarantee delivery; a pilot program is not equivalent to general availability. Even the word “available” needs a location: only in the United States, through selected partners, or in France as well?
Apple’s Vision Pro offers an instructive precedent. Unveiled in 2023, it went on sale in the United States in February 2024 before expanding to other markets, including France. Between a public unveiling and a local purchase, timing matters. And a delivered product is not automatically an adopted product: comfort, available applications and frequency of use remain separate questions.
- Ask for a date: a stated delivery month rather than a vague “coming soon.”
- Check geographic coverage: supported languages, distribution and local support.
- Read the terms: deposit refunds, warranty, returns and the status of promised features.
In Europe, CE marking also warrants examination where applicable. Bear in mind, however, that it is neither an innovation award nor a performance guarantee. It indicates declared conformity with applicable requirements, under procedures that vary by product category.
Price: the bills often start after the purchase
An entry-level price is an excellent stage prop. It becomes less appealing when subscriptions, accessories, installation or AI service usage must be added. For businesses, there are also IT integration, training and the time spent correcting errors. The right measure is not just the purchase price, but the cost of reliable use.
Two launches in 2024 illustrate very different market positions: the Vision Pro had a US starting price of $3,499, while the Rabbit R1 AI device was announced at $199. Yet those initial prices alone say nothing about the value delivered. An inexpensive device can become redundant if a phone already performs its tasks; expensive equipment may be justified for a clearly defined professional use.
To compare, build a concrete scenario: how much does the service cost over three years, for how many users and at what volume of activity? Also ask what happens once a usage limit is reached. In AI, usage-based billing can make demonstrations seamless and budgets unpredictable. A credible estimate must state its assumptions, not just its best-case scenario.
Maintenance: who responds when the future breaks down?
This is the least photogenic part of the assessment, but often the most revealing. Can the battery be replaced? Are spare parts available? Is there an identifiable repair provider? Connected products add another layer of infrastructure: servers, updates, security patches and mobile applications. A physically intact device can become unusable if its remote service shuts down.
Humane’s AI Pin provided a stark illustration. After the February 2025 announcement that HP was acquiring certain Humane assets, the device’s consumer services were shut down at the end of the same month. This precedent is a reminder that buying cloud-dependent hardware also means betting on a service’s continued operation.
Before buying, request a stated support period, data export procedures and a list of features that work offline. For professional equipment, examine the contractual commitments: response times, replacement arrangements, liability and exit terms. “Our team is here to support you” is no substitute for a written obligation.
Early customers: look for use cases, not logos
A wall of major brands is reassuring without always being informative. Are these companies customers, partners, investors or merely participants in a trial? A free pilot demonstrates interest, not a willingness to pay. A first sale demonstrates a transaction, not yet lasting usefulness. Contract renewals and expansions generally provide stronger evidence.
The best testimonial describes a verifiable before and after: the task involved, the length of the trial, the conditions of use and the limitations encountered. For an AI assistant, time savings must account for human review. For a robot, throughput must include interruptions. Without these details, spectacular performance can conceal substantial work shifted onto operators.
Test on a small scale, decide on the evidence
The most robust approach is to organize a limited trial using your own data and criteria set in advance. Vary the conditions: a degraded network, an unusual request, a noisy environment. Measure failures too, along with how easy it is to take back control. A supplier willing to accept this protocol sends a more meaningful signal than another promotional video.
What next? For the rest of 2026, we can anticipate growing pressure on connected-device manufacturers and AI providers to turn demonstrations into repeated use. This does not offer certainty about the winners, but it is a reasonable expectation for buyers. The next decisive innovation may not be the one that attracts the most visitors. It will be the one you can order, finance, maintain and keep using once the excitement has faded.
