Recently, OpenAI officially announced the establishment of OpenAI Deployment Company (hereinafter referred to as 'Deploy Co'), with TPG leading the investment and a total of 19 investors investing over $40 billion, valuing the company at $140 billion. The core business of this company is to embed AI Frontline Deployment Engineers (FDEs) into client companies, integrating the models behind ChatGPT into enterprises' data, processes, and workflows. Also in May, Anthropic took the lead in announcing a joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs, with committed capital of approximately $15 billion, to do the same thing—sending engineers into clients' offices.
These two investments, totaling about $55 billion, are the two most structurally significant events in the global AI sector from 2026 to date. They jointly mark one thing: frontier model companies are beginning to admit that relying solely on selling APIs can no longer support valuations, and they must, following the 'frontline deployment' model defined by Palantir in the mid-2000s, turn themselves into half-consulting companies. The capital structure, motivations, and labor market implications of this transformation are what this article will analyze.
$40 Billion, 17.5% Guaranteed Return
According to OpenAI's official announcement, Deploy Co is controlled by OpenAI, with external investors collectively committing over $40 billion, led by TPG, with Advent International, Bain Capital, and Brookfield acting as co-sponsoring partners. The remaining 16 investors include PE and strategic capital firms such as SoftBank Corp., Goldman Sachs, Warburg Pincus, BBVA, B Capital, Emergence Capital, Goanna, and WCAS.
What's truly unusual are the details of the capital structure. According to Axios citing informed sources, external investors received preferred shares rather than common shares, with the structure containing two core terms: OpenAI guarantees investors a minimum 17.5% return and sets a cap on profits. In other words, this is not a conventional equity financing round but a structured transaction akin to subordinated debt, where investors have a floor on the downside and a cap on the upside.
This arrangement is not common in the private equity industry. In its April analysis, SaaStr pointed out, "PE firms typically target internal rates of return (IRR) above 20%, but it's almost never guaranteed contractually by the investee company." MarketWise's interpretation is that this structure implies PE investors are cautious about the valuation and cash burn of OpenAI's main entity; they are unwilling to hold OpenAI common stock and prefer guaranteed preferred shares in the subsidiary. Considering the main OpenAI entity's valuation has reached approximately $8520 billion (StartupHub.ai estimate), this arrangement of "unable to refinance the main entity, so using a subsidiary with structured terms to raise funds" is itself a signal.
Another detail disclosed by Axios around the same time is that Deploy Co's pre-money valuation was $10 billion, with a post-money valuation of about $14 billion. This means OpenAI simultaneously packaged "future enterprise AI service revenue" into a measurable cash flow asset and priced it for sale to 19 institutions.
The implementation side was addressed by acquiring Tomoro. Tomoro is an AI consulting and engineering company registered in London in 2023, born in an "alliance-like" manner with OpenAI, headquartered in London, with offices in Edinburgh and Manchester. In the past year, it established its APAC headquarters in Singapore, with additional branches in Sydney and Melbourne. Its client list includes Tesco, Virgin Atlantic (for whom it built an AI travel concierge), Supercell (launched an in-game support agent serving 110 million users within 12 weeks), Fidelity International, Red Bull, Mattel, and the NBA. Tomoro claims to have quadrupled its headcount in the past 12 months, with global monthly revenue growing over tenfold. This acquisition will bring Deploy Co approximately 150 "experienced frontline deployment engineers and deployment specialists."
Bain, McKinsey, Capgemini Simultaneously Invest
Among the list of 19 investors, the most unusual are not the PEs, but three consulting firms: Bain & Company (the twin consulting firm of Bain Capital), McKinsey & Company, and Capgemini.
Axios columnist Dan Primack offered two interpretations of this arrangement. The moderate interpretation is that these three consulting firms will use this to gain a deeper understanding of OpenAI's capabilities and roadmap, which they can then pass on to their own clients. The sharper interpretation is that OpenAI convinced these traditional consulting institutions to invest in funding a company that will eventually disintermediate them.
This dynamic reappears in a more implicit form in Anthropic's $15 billion joint venture. According to The Wall Street Journal, the capital structure of the JV is: Anthropic, Blackstone, and Hellman & Friedman each contribute approximately $3 billion, Goldman Sachs as a founding investor contributes about $1.5 billion, with the remaining capital filled by Apollo Global Management, General Atlantic, GIC, Leonard Green, and Sequoia Capital, totaling around $15 billion.
The JV's positioning was described by Blackstone's COO and President Jon Gray as "breaking one of the most critical bottlenecks in enterprise AI adoption" by "expanding the pool of engineers with practical implementation capabilities." Marc Nachmann, Global Head of Asset & Wealth Management at Goldman Sachs, stated in the announcement that this JV will "enable mid-sized companies to access Anthropic's solutions, democratizing access to highly scarce frontline deployment engineers."
Its target clientele is noteworthy. Both joint venture companies, Deploy Co and the Anthropic JV, have initially locked their first clients onto companies invested in by the PE firms involved. PEs like Blackstone, Apollo, TPG, Bain Capital, Brookfield, Advent, and Warburg Pincus collectively manage over 2,000 companies, constituting a vast, contractually bound internal distribution channel. Embedding models into the operations of these portfolio companies is both a source of returns for LPs and a tool for PE partners to reduce costs and improve profitability.
Anthropic's Overtaking is the Real Reason for OpenAI's Bet
Venture capital firm Menlo Ventures has been releasing a bi-annual report on enterprise LLM market share since 2023. Data from their year-end 2025 report shows that Anthropic currently holds 40% of the enterprise LLM API market share, a significant jump from 24% last year and 12% in 2023; OpenAI's share fell from 50% in 2023 to 27% over the same period, having lost nearly half of its enterprise share; Google rose from 7% in 2023 to 21%.
The gap is even more significant in the coding domain. Anthropic holds about 54% of the programming market share, while OpenAI holds 21%. Since releasing Claude Sonnet 3.5 in June 2024, Anthropic has held the top spot on programming evaluation leaderboards for 18 consecutive months. Menlo Ventures partner Deedy Das stated: "Anthropic is sweeping the enterprise market, and OpenAI has ceded nearly half its share."
This share reversal created direct pressure on OpenAI's management. In March of this year, OpenAI's CEO of Application Business, Fidji Simo, described Anthropic's progress as a "wake-up call" in an internal all-hands meeting and characterized OpenAI's response state as "code red." According to The Wall Street Journal citing meeting minutes, Simo told employees, "We must not miss this moment by being distracted by peripheral demands," and urged the company to "deliver results in productivity, especially on the enterprise side."
The timeline thus becomes clear. In March, Simo sounded the internal alarm. By April, OpenAI was already in advanced negotiations with TPG, Advent, Bain Capital, and Brookfield for the $10 billion joint venture company. On May 4th, Anthropic announced its $15 billion joint venture first. On May 11th, OpenAI announced Deploy Co and the Tomoro acquisition. The entire process was driven by Anthropic's market share data and forced into action by Claude Code's penetration speed.
The Internal Reflection of the White-Collar Flip: 800% Surge in FDE Demand Occurs Simultaneously with SWE Demand Shrinkage
Deploy Co and the Anthropic JV aim to solve a human resources problem. Specifically, the supply problem of FDEs.
According to Indeed public data, job postings for FDE positions in the US skyrocketed from 643 to 5,330 in the past 12 months, a year-over-year increase of 729%. LinkedIn data shows that US FDE job postings from January to September 2025 increased by over 800% year-over-year, among the fastest-growing categories in tech jobs. Geographically, New York has replaced San Francisco as the top hiring location for FDEs, accounting for about 35%, compared to San Francisco's 11%. This change is driven by the financial services and regulated industries in New York absorbing FDEs.
Salary bands are significantly higher than those for traditional software engineers. According to PitchMeAI citing Anthropic's public recruitment information, the base salary band for Applied AI FDE positions (Anthropic's internal naming for FDEs) in the US is $280,000 to $320,000. The total compensation (TC) for mid-to-senior level FDEs at OpenAI and Anthropic has stabilized in the range of $350,000 to $550,000, with some staff-level positions approaching $600,000. The average TC for FDEs at Palantir is about $238,000, with staff-level exceeding $630,000. New graduates typically start with a TC of $180,000 to $250,000.
In stark contrast to the explosion in FDE demand is the continued contraction of traditional software engineer positions. According to Indeed via FRED data, the number of software engineer job postings nationwide has fallen between 35% and 45% from the mid-2022 peak, hitting a five-year low by early 2025. Research by the Stanford Digital Economy Lab based on ADP payrolls shows that employment numbers for early-career software engineers aged 22 to 25 have declined by nearly 20% from the late 2022 peak. An industry observation by The Pragmatic Engineer notes that while AI-related infrastructure roles and engineering roles in regulated industries are still expanding, traditional SWE demand in most other sectors is in a retreat phase.
Apollo Global Management Co-Founder and CEO Rowan described the AI wave as "by far the largest tech cycle in his career" during Apollo's quarterly earnings call, predicting that "almost every job will be augmented or replaced, and we will see a complete flip: blue-collar rises, white-collar faces pressure." Concurrently, Blackstone Group President Jon Gray made a similar judgment at the Milken Conference, believing AI will drive blue-collar employment into a "huge boom."
The internal form of the FDE rise within Silicon Valley aligns with the macro logic described by Rowan, just more subtle. FDEs are "engineer blue-collarization" packaged with high salaries. They travel 50% of the time, are stationed in client offices, deal with clients' legacy systems and compliance audits, debug in data silos, and respond to client CIOs' political demands. It violates the Silicon Valley creed long held sacred: "zero marginal cost, pure software, remote work." The FDE model essentially pushes engineers back to the frontline, next to clients, and into specific businesses.
The Pragmatic Engineer newsletter editor-in-chief Gergely Orosz said in a May analysis: "In the arrangements of both OpenAI and Anthropic, FDEs are placed in independent subsidiary companies. This means newly hired FDEs likely receive equity in Deploy Co or the Anthropic JV, not the parent company's stock." In other words, the valuation premium at the model layer and the human capital premium at the deployment layer are structurally separated. The parent company is selling a "future revenue stream," the subsidiary is performing a "far more labor-intensive than SaaS" business, and the two are bound by structured terms.
The Model Layer is Commoditizing, the Deployment Layer is Being Capitalized
Pulling together the four threads above reveals a relatively complete narrative of an inflection point. Differentiation at the model layer is narrowing. OpenAI, Anthropic, and Google collectively hold 88% of the enterprise API market, with model quality scores increasingly converging. However, enterprise AI implementation success rates have long been estimated between 5% and 20% in the industry, with deployment difficulty constituting the real bottleneck for AI monetization. Anthropic proved in 18 months that in a commoditizing model landscape, more focused products and more solid enterprise implementation capabilities can overtake the first mover.
OpenAI's response is to use capital structure to catch up with time. The $40 billion Deploy Co is not a conventional financing round; it's a "securitization of future enterprise revenue" arrangement with guaranteed returns and a profit cap, sidestepping the awkwardness of the main entity's valuation making refinancing difficult. Anthropic's $15 billion joint venture turns the PE firms' investment company network into its distribution channel in one go. The two deals collectively expand the AI giants' boundaries from "model API" to "on-site deployment," bringing the core profit pools of the traditional consulting industry into competitive view.
The simultaneous investment by Bain & Company, McKinsey, and Capgemini gives this transformation special significance at the level of financial dynamics. Whether these three consulting firms are approaching it with a mindset of "understanding the competitor" or are already prepared to be partially disintermediated, they are, in fact, providing capital to their potential future competitors—a landscape extremely rare in the consulting industry's history over the past 20 years.
The explosive growth in FDE positions and the contraction of traditional software engineer roles are not contradictory. They are two sides of the same structural flip. After the intelligence spillover from the model layer, what enterprises are willing to pay for is no longer "writing another piece of software," but "making this AI actually work in my business." The former is increasingly becoming a commodity, while the latter is increasingly becoming a high-premium service. Borrowing Rowan's words, it's a complete flip, and the world isn't ready.
The next observation point is whether PE giants like Carlyle, KKR, and EQT, which have not yet entered the fray, will follow suit, and whether Meta's announced Enterprise Solutions division will follow with a similar structure. If they follow, this will completely define the capital narrative for enterprise AI from 2026 to 2027. If they don't, then this $55 billion bet is merely an emergency self-rescue by leading AI companies.
