‘The cost of compute is far beyond the costs of the employee’: Nvidia executive says right now AI is more expensive than paying human workers

Big Tech is still cutting jobs even as AI spending accelerates, with more than 118,000 tech layoffs in 2026 so far across nearly 100 companies. The article argues that AI remains more expensive than human labor in many use cases, citing examples of rising compute costs, higher software fees of 20% to 37%, and large AI budget overruns at firms like Uber and Microsoft. The key takeaway is that AI adoption is increasing, but the economics are still unfavorable and may pressure margins before any labor-saving benefits materialize.

Analysis

The market is still pricing AI as a labor substitute, but the near-term economics look more like a capital-intensity arms race that benefits infrastructure owners more than software adopters. That means the first-order “AI efficiency” story is likely overstated for the next 12-24 months: companies are cutting headcount to preserve margins while simultaneously stepping up compute spend, which can depress operating leverage and free cash flow before any productivity gains show up.

The cleaner second-order winner is the AI supply chain, especially vendors with pricing power in accelerators, networking, memory, power, and datacenter buildout. The loser set is trickier: SaaS and enterprise software names that sell “AI add-ons” on flat pricing can see margin pressure and customer pushback as usage spikes, while companies like UBER that are experimenting aggressively with AI tools may discover that adoption creates budget overruns before it creates measurable productivity.

The key risk catalyst is a pricing reset. If model providers move from flat subscriptions to metered usage, the apparent demand could slow abruptly, but unit economics for providers may improve, making the beneficiaries more durable than the current headline spending suggests. Conversely, if inference costs fall rapidly and reliability improves over the next 12-48 months, the labor-displacement narrative can re-accelerate and the current wave of layoffs becomes a lagging rather than leading indicator.

The contrarian read is that layoffs are not evidence of AI replacing humans yet; they are evidence that management is using AI as a budget justification while paying for optionality. That makes this a governance and capital-allocation story as much as a technology story: firms that force premature adoption are likely to destroy value in the near term, while those that monetize AI through infrastructure, not workflow theater, should compound through the cycle.

Key Decisions for Investors

• Long NVDA vs. short high-multiple enterprise software exposure for 3-6 months: favor the picks-and-shovels beneficiary of continued capex while avoiding names that must prove AI ROI before pricing power emerges.
• Initiate a short UBER / long MSFT pair for 2-4 months if the market starts discounting AI tooling budget blowouts as a broader enterprise issue; UBER has more visible near-term operating budget risk, while MSFT can reallocate within a larger ecosystem.
• Buy call spreads in AMZN and META on 6-12 month horizon, but only on pullbacks: both can absorb AI spend better than smaller peers, and a widening cost gap should ultimately consolidate share toward scaled buyers with internal distribution.
• Avoid chasing standalone AI software vendors on flat subscriptions; if holding, hedge with put spreads into earnings over the next 1-2 quarters because usage-driven cost overruns can compress gross margin before revenue recognition catches up.
• Watch for any disclosure of metered pricing or compute rationing from major AI providers; that would be the first credible catalyst to rotate from infrastructure longs into select application-layer beneficiaries with real productivity proof.