--- title: "Anthropic and OpenAI launch their AI consulting arms" date: 2026-05-04T18:00:00+02:00 language: en slug: 2026-05-04-anthropic-openai-joint-ventures-conseil-ia url: https://mathieuhaye.fr/blog/en/2026-05-04-anthropic-openai-joint-ventures-conseil-ia alternate: https://mathieuhaye.fr/blog/2026-05-04-anthropic-openai-joint-ventures-conseil-ia category: Business & Growth description: "On May 4, 2026, Anthropic and OpenAI launched multibillion-dollar joint ventures with Blackstone, Goldman, TPG and Bain. Target: the AI consulting market." --- # Anthropic and OpenAI launch their AI consulting arms > On May 4, 2026, Anthropic and OpenAI launched multibillion-dollar joint ventures with Blackstone, Goldman, TPG and Bain. Target: the AI consulting market. ## What was announced on May 4 Two announcements, same Monday. Anthropic, with three founding partners (Blackstone, Hellman & Friedman, and Goldman Sachs), unveiled a $1.5bn vehicle to push Claude into the enterprise. Anthropic, Blackstone, and Hellman & Friedman are each putting in roughly $300m; Goldman Sachs adds about $150m; the rest comes from Apollo Global Management, General Atlantic, GIC (Singapore's sovereign wealth fund), Leonard Green, and Sequoia Capital, as detailed by [CNBC](https://www.cnbc.com/2026/05/04/anthropic-goldman-blackstone-ai-venture.html). The same day, OpenAI closed its own joint venture, presented depending on the source as The Development Company or The Deployment Company. The vehicle is sized at $10bn, with $4bn already committed by 19 investors. TPG is the anchor; Advent, Bain Capital, and Brookfield Asset Management are co-founders. Notable design choice: OpenAI keeps super-voting shares to retain strategic control, and guarantees the sponsors a 17.5% annual return over five years, as [TechCrunch](https://techcrunch.com/2026/05/04/anthropic-and-openai-are-both-launching-joint-ventures-for-enterprise-ai-services/) reports. Both vehicles target the same buyer: mid-sized companies, with priority given to firms owned by the participating funds. The delivery model is identical, lifted straight from Palantir's playbook: forward-deployed engineers (FDEs) sent to embed inside the client team for several weeks, wiring the model directly into internal workflows. Palantir, founded in 2003, reached roughly $5bn in annualized revenue after more than two decades. Anthropic and OpenAI want to compress that timeline by an order of magnitude. There is no overlap between the two investor lists. ## The application layer is where the money sits What is happening here is not a new distribution channel. It is an admission, by both model leaders, that the margin is no longer in the API. Anthropic went from $9bn in annualized revenue at the end of 2025 to more than $30bn by April 2026, nearly a fourfold jump in four months. The acceleration came from Claude Code, Claude Cowork, and direct enterprise deployments, not from the consumer service. The company is in talks to raise at a $900bn valuation, ahead of OpenAI on the latest figures reported by [CNBC in late April](https://www.cnbc.com/2026/04/29/anthropic-weighs-raising-funds-at-900b-valuation-topping-openai.html). The ratio quoted internally at every AI vendor right now: for every $1 spent on enterprise software, $6 is spent on the surrounding services (integration, training, change management, data work). That is the wallet Anthropic is going after, and the pitch on Monday was direct. Anthropic's CFO Krishna Rao summed it up in [Fortune](https://fortune.com/2026/05/04/anthropic-claude-consulting-industry-joint-venture-blackstone-goldman-sachs/): enterprise demand for Claude is significantly outpacing any single delivery model. Translation: we are selling models faster than we can implement them, so we are buying an external delivery channel. Jon Gray, Blackstone's president, confirmed the complementary angle: the shortage of engineers capable of deploying AI inside customers is one of the most significant bottlenecks to enterprise AI adoption. Blackstone owns more than 250 portfolio companies; each one becomes a potential client for the new entity's FDEs. Goldman Sachs, which brings in its own corporate client network, framed the operation as democratization: giving mid-sized companies access to Anthropic engineers they could not afford to hire directly. ## Why private equity, not Microsoft or Salesforce The choice of partners is a strategic statement. Anthropic and OpenAI did not bring Microsoft, Salesforce, or ServiceNow as co-founders. They brought funds that already own the companies they want to transform. That is a triple shift. For the funds: applied AI has become the leading value-creation lever in mid-cap LBOs over the past 18 months. A fund that cannot show a credible AI roadmap loses deals on the buy side and on exit. A partnership with Anthropic or OpenAI becomes a term-sheet selling point. For the AI labs: having Blackstone, TPG, Apollo, and GIC in the cap table buys a guaranteed pipeline of logos. No more chasing wins, the portfolios do the work. For mid-cap company management: the mandate no longer comes from a traditional consulting firm; it comes down from the shareholder, with contracted performance commitments. OpenAI's structure is the most telling. A guaranteed 17.5% annual return over five years is a near-fixed-income vehicle, not a venture bet. OpenAI is selling financial predictability in exchange for a captive distribution channel; the sponsors are buying technology exposure with a protected downside. The super-voting share OpenAI keeps ensures the technology roadmap stays in San Francisco, not in Boston or Fort Worth. ## What this displaces in consulting The real loser in this operation is named McKinsey, BCG, Bain on the strategy side, and Accenture, Deloitte, EY, and IBM Consulting on the integration side. The Big Three and the Big Four built five years of growth on the promise of in-house AI: internal labs, partnerships with OpenAI or Anthropic, mass hiring of ML engineers. The pitch from the two new joint ventures short-circuits that chain. Why buy an AI transformation engagement from a consulting firm that itself consumes Anthropic's API, when you can buy Anthropic directly, with engineers who speak to the model at the source code level? The competitive context makes the gap even sharper. On May 1, 2026, Accenture and Databricks announced a [deeper partnership](https://newsroom.accenture.com/news/2026/accenture-and-databricks-accelerate-enterprise-adoption-of-ai-applications-and-agents-at-scale) on enterprise AI agents. McKinsey is pushing QuantumBlack to the front. But these structures still run on day rates with margins of 25% to 35%. An Anthropic-Blackstone joint venture delivering directly through FDEs has a different cost base: lower commercial overhead, model access at internal cost, and a return aligned with deployed AI performance instead of hours sold. The second circle of victims is the vertical SaaS layer. If the Anthropic-built agent lives inside Salesforce, Workday, or SAP, it is still acceptable. If it proposes to replace part of the SaaS at the workflow orchestration layer, it is an existential risk. Salesforce shipped [Agentforce Operations on April 29, 2026](/blog/en/2026-05-04-salesforce-agentforce-operations-back-office-ia) to take back the back-office. SAP needs to accelerate Joule. Read in that order, the May 4 announcements look like a coordinated response: the AI vendors are leaving wholesale and entering retail. ## What this changes in my freelance practice When I work with clients on engagements, the value gap is never in the model itself. On the **Salesforce 3CX integration for the 3018 helpline** at e-Enfance, what matters is the call routing algorithm and the Apex object that creates the case file in under 500ms. On **Horus Condition Report**, where Pipedrive becomes useful is in the bilingual FR/EN lead scoring and the follow-up rotation. On the weekly intelligence brief I run for **Fromagerie Ermitage**, the 93 n8n nodes that orchestrate the automation deliver more value than the LLM itself. What Anthropic and OpenAI are now selling to Blackstone and TPG is exactly that integration layer. The barrier to entry for a freelancer is no longer the technology (the Claude API costs cents and n8n is free), it is the deep knowledge of the client workflow: which field to fill, which SLA to trigger, which external partner to chase, which Apex trigger to wire to which webhook. That same reading underpins the [MSc Data & AI for Finance program at Albert School x Mines Paris-PSL](/) that I am joining in September 2026, and the ALM apprenticeship at AFD running in parallel: the technology is becoming a commodity, the durable differentiator is the business plus delivery pairing. ## The real shift May 4, 2026 marks the end of pure model AI. When the two best-capitalized AI labs in the world decide to become their own service channel, it means the margin has definitively migrated to the application layer. The question for the next six months is not which model wins, but how many traditional consulting firms and how many SaaS vendors get to refit their P&L before Anthropic and OpenAI field their first thousands of FDEs. --- Source: [https://mathieuhaye.fr/blog/en/2026-05-04-anthropic-openai-joint-ventures-conseil-ia](https://mathieuhaye.fr/blog/en/2026-05-04-anthropic-openai-joint-ventures-conseil-ia) | Other language: [https://mathieuhaye.fr/blog/2026-05-04-anthropic-openai-joint-ventures-conseil-ia](https://mathieuhaye.fr/blog/2026-05-04-anthropic-openai-joint-ventures-conseil-ia)