Simple Google search
Yes, AI can actually cost significantly more than human labor. A major reason for this is that running advanced AI requires massive amounts of expensive hardware and energy, making human workers the cheaper and more economically viable option for about (77\%) of roles.Why AI Can Cost More Than HumansSky-High Compute Fees: The processing power—measured in tokens—required for complex, multi-step "agentic" AI systems can outpace the costs of human salaries. Companies like Uber have reported blowing through their entire yearly AI budgets in just a few months due to heavy infrastructure usage.Expensive Hardware and Energy: Nvidia's vice president of applied deep learning has noted that for his team, AI compute costs far exceed the salaries of the employees utilizing the tools.Required Human Oversight: AI still makes frequent errors, forcing companies to pay human workers to monitor the models, review outputs, and fix breakdowns.The Economic RealityA landmark MIT study found that it is only economically viable to automate about (23\%) of jobs. In the remaining (77\%) of cases, employing a human is cheaper, more accurate, and more efficient than relying on artificial intelligence.While research firm Gartner projects that the unit cost of running large language models will drop substantially by 2030, overall enterprise costs will likely remain high. This is because businesses are utilizing much more complex models that require significantly more processing power per task.You can read more about the challenges of AI replacing human labor in this MIT Study or learn about enterprise budget shifts on Forbes.
5 replies (most recent on top)
There is a lot of wishful thinking at the C-Level. Expectations will come down soon enough though.
You have real improvements in coding speed, but that's only 10% of budgets on the IT side. It will lead to more customization, and to support that cost will go up. Price of tokens may go up or down, there are two schools of thought on this, so I just dont know if this will be an issue or a non issue.
On the business solutions side, AI is being crammed into all kind of solutions - solutions are fragile and there is is a lot of work that's happening to iron out edge cases, so I am not sure if this makes sense from the cost perspective - probably does not in 2026, maybe 2030 and forward when tools mature. It's too early to make the call.
The remainder is lukewarm. It transcribes, can summarize or write an email. Presentations are getting better, its starting to do some real Excel work - but still, there is a lot of work to do to prep the data, context, instructions, prompts - only to realize that it sometimes work.
So, I'd say the current impact is underwhelming and no impact on labor aside from some jobs that are doing tasks that can be automated and some wins with knowledge/info discovery, and a range of half assed copilot features. The gains on the coding side are real, but someone still needs to initiate, plan (OK, AI helps here a lot), get the requirements. Design/Build/Test are faster already, production support is still messy with some tool improvements.
So, I'd say things are real but gains are very limited and tend to be expensive.
Tell that to brainwashed corporate risk leadership. AI GIVES BOGUS REG info then we end up having to fix it and it takes even more time.
Everything dad or trend we do ends up costing 10x more . Why? Because no one here knows how to reduce costs, streamline processes, or do anything. Off shore, Agile, AI, etc all are dark holes here
https://www.forbes.com/sites/timbajarin/2026/04/29/ai-compute-surpasses-human-costs-enterprise-budgets-shift/
This only matters if you buy the outward narrative that execs are pushing, that they are doing layoffs because they're replacing people with AI. That might be the case in some instances, but for a lot of these companies, and definitely for Wells, the decision to do layoffs was already made and AI is just a convenient cover story. "AI will be more expensive actually" doesn't matter if the CEO's actual intent was just to shed resources period, not replace humans with computers.