Taken individually, five stories from the US tech press in the first week of November look unrelated: Apple tightening its internal AI spending, voters turning against data centres, mixed results for AI-industry political action committees, a hardware start-up teasing a mysterious "AI-first" device, and Uber revealing that almost nobody uses AI agents to book a ride. Read together, though, they tell the same story: the gap between how aggressively AI has been marketed and how it is actually performing, being paid for and being accepted once it leaves the slide deck.
Even the biggest spenders are counting the cost
According to a Bloomberg report covered by TechCrunch, Apple lowered the daily amount its engineers can spend on AI tools in October, from USD 300 to USD 200 per person, with further cuts to USD 100 reportedly being considered and limits as low as USD 25 a day outside engineering. This matters precisely because Apple can afford to spend freely if the returns justified it. When a company of that size starts rationing AI usage, it suggests that even sophisticated users are reassessing what they get for the money rather than assuming more spend means more productivity.
The infrastructure behind AI is becoming political
AI depends on physical infrastructure: power, water, land and grid capacity. In the US midterm elections on 3 November, that infrastructure became a voting issue. TechCrunch reported that opposition to data centres crossed party lines, citing an NBC exit poll in which almost 80% of respondents opposed new data centres in their neighbourhoods. At the same time, AI-industry political action committees that spent millions trying to shape Congress had mixed results, with several of their backed candidates losing. The regulatory and infrastructure environment around AI is far less predictable than many vendor roadmaps assume.
Adoption is lagging the narrative
Perhaps the clearest signal came from Uber. On its quarterly earnings call, the company said that during its busiest week of the quarter, around Halloween, only four rides were booked through an AI agent, and only about 500 Uber Eats orders, out of roughly 320 million trips and orders. Its chief executive described the traction as "very, very early and very limited", noting that booking a ride in the app already takes less than a minute.
Meanwhile, Nothing's chief executive says the company's first "AI-native" product will ship in December, but has not yet said what it is. Put side by side, these items capture the current state of applied AI rather well: plenty of ambition and announcement, and comparatively little proven, everyday use by real customers.
None of this means AI investment is misguided. It means the easy, hype-driven phase of adoption is ending and a more disciplined, evidence-based phase is beginning.
What this means for IT and security teams in the UAE
Treat AI tools like any other investment. Budget owners should measure actual usage, calculate the full cost including usage-based fees, and be willing to scale back deployments that are not delivering rather than renewing licences on momentum alone. Apple's reported limits are a useful reference point for any CIO building a 2027 AI budget.
Plan for power and capacity. Demand for AI-ready computing is growing fast in the region. Power availability, cooling and site planning will increasingly shape where and how organisations can expand, whether they run AI workloads on their own servers, in the cloud or in a mix of both. We looked at this in our article on the infrastructure behind Dubai's AI push.
Ask vendors for evidence, not promises. A feature existing is not the same as a feature being used. When you assess networking, security or productivity platforms that advertise built-in AI, ask for real usage and outcome data from customers like you, and run a proof of concept before committing budget.
What UAE and GCC organisations should do now
- Set usage limits and track AI spend per team, as you would for cloud resources.
- Define success measures before buying, such as hours saved, tickets resolved or errors reduced.
- Review AI licences before renewal and cut seats or features nobody uses.
- Include power, cooling and capacity in any plan to run AI workloads on your own infrastructure.
- Run a proof of concept with your own data before committing to an AI-enabled platform.
- Check data handling and security for every AI tool, including where data is processed and who can access it.
From hype to value
AI will keep changing how businesses work. The organisations that get the most from it will be the ones that buy on evidence, measure what they use and keep their infrastructure plans grounded in reality.
COMMPRO Technologies supplies servers, networking and security solutions from leading vendors across the UAE and GCC, and can help arrange proof-of-concept testing for AI-ready infrastructure. If you are planning your next AI investment, contact our team.
