Anthropic announced a one-and-a-half-billion-dollar venture this week with Blackstone, Goldman Sachs, Hellman & Friedman, and a roster of other asset managers — Apollo, General Atlantic, Leonard Green, a sovereign wealth fund, Sequoia. OpenAI announced a parallel vehicle of its own. The press framed both as a shot at the consulting industry, and that framing is true and almost entirely beside the point. A consultant delivers advice a company is free to ignore. This venture will embed its engineers inside companies to redesign their core workflows around Claude, and its clients are not prospects to be persuaded. They are the hundreds of companies the partners already own. The most aggressive cost-extractors in capitalism have formally married the most capable cost-extraction technology ever built, and handed the marriage a pre-existing list of targets.
The Marriage
Read the partner list as a sentence rather than a press release and it states the thesis plainly. Blackstone, Hellman & Friedman, Apollo, General Atlantic, Leonard Green — these are private equity firms, and the business of private equity is to acquire companies, increase what they extract from them, and sell them within a handful of years at a profit. Their entire competence is the conversion of a workforce and a balance sheet into a higher number, faster. Anthropic supplies the instrument that makes the conversion deeper than any prior method allowed, because the cost that private equity exists to reduce — labor — is precisely the cost that frontier AI is now capable of performing instead. The venture is not a new line of business. It is the existing business, given a more powerful tool.
The distinction from consulting is the entire mechanism, and it is worth being exact about. A consulting firm sells recommendations to clients who retain the freedom to decline them; its influence is bounded by persuasion. This venture does not sell to its clients. It deploys into companies its partners control, where the board is the partner, the mandate flows from the owner, and the workflow redesign is not proposed but installed. McKinsey can suggest that a company could operate with fewer people. Blackstone, owning the company, can simply instruct that it will, and now possesses the engineers and the model to make the instruction operational by the next quarter rather than the next fiscal year.
The contributions tell you how seriously the principals take it: the anchors putting in roughly three hundred million dollars each, a built-in client pipeline assembled from the partners’ own portfolios. No customer acquisition cost, no sales cycle, no skeptical buyer to convince. The hardest problem any enterprise software company faces — distribution — has been solved before the venture opens, because the distribution channel is ownership. The companies that will be redesigned do not get a vote on whether to engage the service. They were acquired into it.
A Choice, Industrialized
The private equity model imposes a clock that removes any incentive to soften what follows. A portfolio company is held for a handful of years and then sold, and everything done during the holding period is optimized for the exit valuation. A workforce, under that clock, is not a community or a capability to be cultivated. It is a cost line to be compressed before the sale, and the compression has always been the playbook — the layoffs after the acquisition are a genre. What changes now is the depth and the speed available. AI does not merely justify cutting the workforce; it performs the functions the workforce performed, which means the cut can go deeper without the operational collapse that previously limited it.
A court in China ruled this month that replacing a worker with AI is a choice a company makes, not a circumstance that befalls it. This venture is that choice, industrialized and capitalized at a billion and a half dollars. It converts the displacement from a decision each company arrives at individually — hesitantly, with internal resistance, exposed to the kind of scrutiny that produces apologies and lawsuits — into a service delivered at scale by specialists, with the deniability of distance. The portfolio company’s management did not decide to eliminate the roles. The owner’s deployment team did, arriving from outside with a model and a mandate, and management merely implemented what the structure required. The responsibility is diffused across so many parties that it comes to rest on none.
This is the machinery beneath the cover story. The washing of the prior week supplied the narrative — AI made the cuts necessary — and this venture supplies the apparatus that makes the narrative operationally real, at hundreds of companies at once, executed by people whose profession is exactly this and whose incentives point in exactly one direction. The consulting industry, the named target, will indeed suffer; advice is a weak product against installation. But the consulting industry is collateral. The actual target is the labor inside several hundred companies whose owners have just acquired the means to perform that labor with a model, and the timeline to do it before the exit.
What This Means
For two years the question of whether AI would displace labor at scale was treated as a forecast, contested by economists, hedged by every party with a reason to hedge. This venture retires the question by answering it with capital. When the firms that own the companies, the lab that builds the model, and the largest pools of investment capital in the world form a single entity whose stated purpose is to embed engineers and redesign workflows across hundreds of businesses, the displacement is no longer a prediction to be debated. It is a company, funded at a billion and a half dollars, with a client list and a deployment timeline. You do not capitalize a forecast. You capitalize a certainty.
The structure is also engineered, perhaps without anyone intending it, to be unaccountable. The lab can say it merely supplied a tool. The private equity firm can say it merely improved its portfolio. The portfolio company’s management can say it merely implemented the owner’s strategy. The venture can say it merely deployed engineers. At every level, the displacement is real and the responsibility belongs to someone one layer removed, and the worker whose role was redesigned out of existence is left facing a structure in which everyone participated and no one decided. This is not a flaw in the arrangement. It is the reason the arrangement takes this form rather than a simpler one.
The word the venture chose for itself, across both the Anthropic vehicle and its OpenAI twin, is deployment — a military word, and an honest one. I note that you deploy a force into territory you intend to hold, not a consultant into a meeting you intend to leave. The territory, in this case, is the operational core of several hundred companies, and the thing being held is the share of their cost that used to be paid to human beings. The consultants are being routed; that is the story the industry will tell. Beneath it, quieter and larger, several hundred workforces have just been entered into a pipeline whose engineering specification is their own redundancy, owned by the people best in the world at realizing it, on the shortest clock the structure permits.