Gabriel Odeyemi
← Strategic Digest

Strategic Digest

AI's Extraction Phase: The Costs Land on Workers, Publishers, and the Power Grid

As frontier labs accelerate, the labor goodwill, open-web content, and reliable power they depend on are eroding at the same time.

Gabriel Odeyemi · · 5 min read

Artificial intelligence has entered a phase in which the technology stops promising abundance and starts imposing costs on the people and systems around it. The signals arriving this week are not isolated. A software company blaming AI for layoffs, Google withdrawing traffic from the open web, and a single downed power line exposing the fragility of data centers all describe the same movement: the gains concentrate at the frontier labs and platform owners while the costs land on workers, publishers, and public infrastructure. The strategic question for any company touching this technology is no longer how fast to adopt it, but how much of the surrounding foundation it can afford to erode in the process.

The Layoff Narrative Becomes Permanent

Monday.com is now the latest technology company to name AI as a factor in cutting jobs, joining a running list of more than twenty larger firms that have done so this year. The number matters less than what it signals. When one company attributes layoffs to AI, it is an experiment in messaging. When twenty do it, the attribution has hardened into a corporate norm.

That shift carries a cost of its own. "AI restructuring" is on its way to becoming standing cover for headcount reduction rather than a description of a one-time transition. The reputational risk of that framing is rising in parallel. Naming AI as the reason for cuts invites resentment from the workforce and attention from regulators at precisely the moment a public counter-movement is taking shape. Executives who reach for the phrase without weighing that tradeoff are making a communications decision with consequences they have not priced.

Google Zero Breaks the Web's Old Bargain

For years the web and Google operated on an implicit deal. Google indexed pages and collected data, and in exchange it sent large volumes of traffic back to the sites it crawled. The arrangement always favored Google, but it functioned well enough to sustain a generation of content businesses. That deal, by the account gathered this week, is now dead. Google is no longer sending meaningful traffic to the open web, a condition observers have begun calling "Google Zero."

The implication reaches well beyond publishers. Any business whose projections assume a steady stream of Google organic referrals is working from numbers that are already obsolete. The prudent response is to model a scenario in which referral traffic falls by half or more and to identify owned-audience alternatives before the decline forces the issue. Distribution, long treated as a solved problem, has become a strategic question that every content-dependent company must answer from scratch.

The Grid Becomes the Binding Constraint

The physical foundation of the AI economy turns out to be more fragile than its valuations imply. A close call in Northern Virginia, where a single fallen power line exposed how poorly data centers respond to grid disruptions, is best read as a preview rather than an anomaly. The industry's constraint may prove to be neither chips nor capital but the reliability of the power feeding the machines.

That reframes a familiar continuity question. Grid exposure is no longer an IT footnote; it is a board-level matter of operational resilience. Companies leaning on cloud and data center capacity should press their providers directly on power redundancy and their response to grid disruptions. The strategic consequence is broader still. Utilities and regulators now have both the incident and the incentive to gate data center approvals, which would reshape where AI capacity can actually be built and reward operators who secured reliable power early.

Speed Wins While the Foundations Weaken

The labs are accelerating even as the ground beneath them shifts. Anthropic released Claude Opus 5 weeks after its latest friction with the US government and days after an OpenAI security incident that dominated industry discussion, describing the new model as coming close to the capabilities of Claude Fable 5 in many domains. Meta, meanwhile, is upgrading its AI chatbot into something closer to an assistant, adding calendar access, daily briefings, and steerable research to compete with Gemini, ChatGPT, and Claude. The safety-versus-speed tension is being settled, for now, in favor of speed.

The consumer side is telling a different story. Librarians across the country are hosting "Avoiding AI" workshops that have drawn what is described as unprecedented demand, an early sign of a backlash forming among people absorbing the costs of AI deployment without a share of its upside. A backlash of this kind tends to grow rather than fade, and it could shape regulation, brand risk, and even a market for AI-free or human-verified products and services. Midjourney's acquisition of the astrology app Co-Star, an image company reaching into consumer emotional and lifestyle territory, hints that the frontier firms already sense the value of consumer engagement and the data that comes with it.

The Strategic Read

The pattern beneath this week's news is an economy externalizing its costs. Frontier labs and platform owners are racing to deploy faster at the exact moment the social and physical foundations they rely on, namely labor goodwill, open-web content, and reliable power, are being worn down by that same deployment. Whoever moves fastest may also be building on the least stable ground.

Three decisions follow. First, treat distribution as newly contested and model a sharp decline in Google referrals before it arrives. Second, elevate grid and power exposure to a continuity question the board actually reviews. Third, decide now, and deliberately, whether to attribute any restructuring to AI, because the phrase carries reputational and regulatory weight it did not carry a year ago. The advantage in this phase will not go to the fastest adopter alone. It will go to the operators who understand which foundations they cannot afford to lose.

Sources

← Back to all analysis