AI products have yet to prove they are worth what users are being asked to pay. In discussing the US election and Big Tech’s performance, technology critic Nilay Patel directs his criticism at the products themselves: investment keeps rising, while the benefits delivered to users remain unclear.
What this is
This is an overarching assessment of the commercial reality of AI. Patel’s point is that companies cannot expand spending on data centers, model inference, and marketing while asking users to tolerate noise, synthetic content (low-quality material generated or manipulated by AI), and still largely impractical AI assistants (software capable of autonomously calling tools and completing tasks step by step).
He also considers the product experience alongside social costs: if the value provided by models does not justify what users pay, the power, equipment, and infrastructure expenses required to expand compute capacity lack a commercial foundation. This is not a rejection of technical capability; it is a demand to know whether users are getting sufficiently real and sufficiently frequent improvements.
Industry view
To his credit, this line of thinking forces AI companies back to basic business logic: revenue must cover continuously rising spending on inference, training, and infrastructure, rather than relying only on user scale, time spent, or financing to sustain growth. A genuinely useful product should eventually be one that users remain with without needing advertising to persuade them.
But the risks are equally clear. Many AI applications are still at the trial and paid-conversion stages, so pulling back investment too soon could mean missing the formation of new user habits. Conversely, if costs fall faster than willingness to pay, margins will not necessarily improve on their own. Data centers also involve siting, power grids, and local governance, which means they cannot be assessed solely through the ledger of a software product.
Moreover, Patel’s criticism of Big Tech’s political relationships carries a strong ideological position and cannot be treated as a direct financial conclusion. More reliable validation metrics include net revenue retention, frequency of active use, cost per task, and quantifiable savings generated by AI products.
Impact on regular people
For enterprise IT: budgets will shift from blanket adoption toward evaluating returns task by task, starting with use cases that clearly save time, reduce errors, or increase revenue.
For individual careers: tool selection will place greater weight on reliability and tangible output. Products with impressive demos but higher review costs will be eliminated more quickly.
For consumer markets: consumers are becoming less patient with AI features. Platforms that cannot continuously deliver obvious value will struggle to secure subscriptions or usage fees on the strength of concepts alone.