Frontier Labs Turn Consultant: Inside Anthropic's and OpenAI's Services Ventures
· TexaHub · AI in Industry, Technology Trends, Risks & Opportunities
Within a week, Anthropic and OpenAI each launched a private-equity-backed services company to embed engineers in client firms. We examine the structures, the economics and the risks for buyers.
Two ventures in one week
On 4 May Anthropic announced that it was forming a new AI services company with Blackstone, Hellman & Friedman and Goldman Sachs, aimed at bringing Claude into the core operations of mid-sized companies [1]. Hours earlier, Bloomberg had reported that OpenAI was raising money for a similar venture called The Deployment Company [3]. OpenAI made its own launch official on 11 May, saying the new unit had $4 billion in funding from OpenAI and more than a dozen other backers [5].
Both companies are effectively setting up consultancies: firms that put engineers inside client organisations to redesign workflows around their models. The timing reflects pressure on both labs to turn model capability into recurring enterprise revenue. TechCrunch noted that OpenAI announced $122 billion in new funding at the end of March at a valuation of $852 billion, and that Anthropic was reported to be in the final stages of a round seeking $50 billion at a $900 billion valuation [3]. SiliconANGLE reported that Anthropic's annual revenue run rate passed $30 billion in March, up from about $9 billion at the end of 2025 [4].
This report sets out how the two ventures are built, what they are meant to fix, how they change the position of systems integrators, and what buyers should check before signing.
How the deals are structured
Anthropic's venture, which has not yet been named, is backed by a consortium that also includes General Atlantic, Leonard Green, Apollo Global Management, GIC and Sequoia Capital [1]. SiliconANGLE reported that Anthropic, Blackstone and Hellman & Friedman each plan to contribute $300 million, with Goldman Sachs adding $150 million as a founding investor, for total capital of $1.5 billion [4]. TechCrunch, citing The Wall Street Journal, described the venture as valued at $1.5 billion including the three $300 million commitments [3]. The two descriptions differ on whether $1.5 billion is capital raised or valuation.
OpenAI's structure is larger and more layered. When Bloomberg first reported the plan, TechCrunch relayed figures of $4 billion from 19 investors against a $10 billion valuation, naming TPG, Brookfield Asset Management, Advent and Bain Capital as backers [3]. SiliconANGLE separately reported that OpenAI and its partners planned to put as much as $10 billion into the unit [4]. At launch, SiliconANGLE reported that TPG is the largest external investor, that SoftBank, Bain Capital and Brookfield also participated, and that the round included consulting firms Bain & Co., Capgemini and McKinsey [5]. Citing Axios, it reported that OpenAI keeps a majority stake, has promised external backers a minimum return of 17.5% subject to a profit cap, and that the round values the unit at $14 billion [5]. The reported valuation therefore ranges from $10 billion to $14 billion depending on the source and date.
OpenAI also announced its first acquisition alongside the launch: Tomoro, a London-based AI services firm whose clients include Virgin Atlantic and Tesco, bringing about 150 forward-deployed engineers and other technical staff [5]. Terms were not disclosed [5]. CIO.com, citing Reuters, reported earlier that the OpenAI venture was in advanced talks on three acquisitions [6].
Capgemini confirmed its stake on 12 May. It said the Deployment Company is intended to "enable, train, and partner with alliance members" to expand enterprise deployment capacity [8]. Capgemini was already a founding member of OpenAI's Frontier Alliance, announced in February, under which it set up a delivery function to work alongside OpenAI's own forward-deployed engineering team [11].
The forward-deployed model
Both ventures rely on forward-deployed engineers, a model TechCrunch says was popularised by Palantir [3]. OpenAI describes a staged engagement: engineers first identify the use cases with the largest potential return, then implement a small number of priority workflows, then build production-grade systems and, where necessary, redesign organisational infrastructure [5]. Anthropic describes a similar pattern in which a small team works with the customer to find where Claude can have the most impact, after which the venture's engineers and Anthropic's Applied AI staff build tailored systems [1]. Its example is a network of physician practices where clinicians lose hours to documentation, coding, prior authorisations and compliance reviews [1].
Both ventures also start with a captive market. SiliconANGLE reported that each will initially target its backers' portfolio companies, with Bloomberg reporting that OpenAI's partners have access to more than 2,000 mid-sized companies [4]. OpenAI said at launch that its investment and consulting partners sponsor more than 2,000 businesses worldwide [5].
Goldman Sachs' Marc Nachmann told CNBC there is a "big shortage" of people who know how to integrate AI with existing business processes, and that "Having the model alone doesn't change your workflows or how you operate" [4]. Anthropic's chief financial officer Krishna Rao said enterprise demand for Claude "is significantly outpacing any single delivery model" [1].
Productised offerings sit alongside the services push. A day after the venture announcement, Anthropic released ten agent templates for financial services, covering tasks such as pitchbook drafting, KYC screening and month-end close, delivered as plugins and as cookbooks for its Managed Agents platform [2]. FIS chief executive Stephanie Ferris said in the same release that the companies are building an agent that "compresses AML investigations from days to minutes" [2]. These are vendor claims and have not been independently verified.
Where integrators fit
Neither lab is presenting the move as a break with its integrator partners. Anthropic said it continues to invest in the Claude Partner Network, naming Accenture, Deloitte and PwC, and that the new company will itself join that network [1]. OpenAI announced Codex Labs in April with Accenture, Capgemini, CGI, Cognizant, Infosys, PwC and Tata Consultancy Services as initial partners, and TechCrunch reported that Codex had more than 4 million weekly active users [9].
Integrators are under pressure all the same. TechCrunch reported that Infosys shares had fallen more than 22% this year amid concern that AI tools could automate parts of traditional outsourcing, and that the company's AI-related services generated about 5.5% of its December-quarter revenue [9]. Analysts quoted by CIO.com read the services push as model providers moving into roles that integrators held. "AI model providers are moving beyond being platform vendors to actively shaping the entire AI value chain," said IDC's Deepika Giri [6]. Techarc's Faisal Kawoosa said AI companies want to stay "in the driver's seat" rather than become another IT vendor [6].
Capgemini's choice to invest rather than compete shows one response. Its chief strategy officer Fernando Alvarez said the investment would help the group "capture a greater share of the growing market for AI-enabled business transformation" [8]. IDC's Frank Dickson told CIO.com he sees consulting partners such as Capgemini, McKinsey and Bain less as investors than as deployment talent, and that their experience may be critical where AI interactions are spread across an enterprise [7].
Competition is not limited to the two labs. Sierra, the customer-service agent company, said on 4 May it was raising $950 million at a valuation above $15 billion, after reporting $100 million in annual recurring revenue in late November and $150 million in early February [10]. Uber's chief technology officer said the company "blew through our [AI] budget" soon after adopting agentic tools, an indication of how costly the ramp-up can be [10].
Risks for buyers
The main risk is control. CIO.com noted that OpenAI described the venture as "majority-owned and controlled by OpenAI" without giving percentages, and that it referred to 19 partners while naming only 15 [7]. OpenAI also said its engineers will "build for where OpenAI's frontier capabilities are headed" [7], which suggests customers may get early exposure to unreleased capabilities, and also that their systems will be shaped around one vendor's roadmap.
Greyhound Research's Sanchit Vir Gogia told CIO.com that "Majority ownership and control give buyers a headline answer, not a complete trust answer," and urged buyers to examine board rights, information rights, partner access and data controls [7]. Kodezi chief executive Ishraq Khan said enterprises will need contractual limits on data access, model training, subcontractors, audit rights and "where operational knowledge goes after the engagement ends" [7].
Lock-in is the second risk. Kadence International's Tulika Sheel said the model "creates deeper dependency across the stack" and could make switching vendors harder [6]. Counterpoint's Neil Shah said controlling the services layer lets providers "lock in enterprises" while learning how clients work [6]. Giri said lock-in is avoidable, but only through deliberate architecture choices made early [6].
There is also a conflict question that neither announcement addresses directly. When the engineers who design a client's workflows are employed by an entity majority-owned by the model vendor, model neutrality is not their job. Buyers should treat these teams as vendor-aligned and plan their evaluation and procurement accordingly.
What to watch
Several facts are still missing. The announcements reviewed here give no name, leadership team or fee model for Anthropic's venture. SiliconANGLE reported that OpenAI's unit was expected to be led by chief operating officer Brad Lightcap [4], while Capgemini's release quoted chief revenue officer Denise Dresser on the venture's purpose [8]. Neither venture has disclosed revenue targets or the share of engagements that must use the parent's models.
In our view, three indicators will show whether this model works. The first is acquisition pace: OpenAI has bought Tomoro and was reported to be negotiating more [5][6], and staffing will set how many clients either venture can serve. The second is whether engagements move beyond the backers' portfolio companies to the wider mid-market. The third is how integrators respond, whether by investing alongside the labs, as Capgemini has, or by positioning themselves as model-neutral alternatives.
The broader signal is already clear. Both leading labs have concluded that model access alone is not turning into enterprise deployment fast enough, and both are willing to take on labour-intensive services to close that gap. Whether services margins and private-equity return guarantees fit alongside software-style valuations is a question their own investors will be asking.
Sources
[1] Building a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs — Anthropic, 4 May 2026
[2] Agents for financial services — Anthropic, 5 May 2026
[3] Anthropic and OpenAI are both launching joint ventures for enterprise AI services — TechCrunch, 4 May 2026
[4] Anthropic and OpenAI establish joint ventures on Wall Street to accelerate enterprise AI adoption — SiliconANGLE, 4 May 2026
[5] OpenAI launches professional services business with $4B investment — SiliconANGLE, 11 May 2026
[6] OpenAI, Anthropic expand services push, signaling new phase in enterprise AI race — CIO, 6 May 2026
[7] OpenAI's new AI consulting offering raises questions of trust, strategy — CIO, 11 May 2026
[8] Capgemini strengthens its position in enterprise AI with investment in the OpenAI Deployment Company — Capgemini, 12 May 2026
[9] OpenAI teams up with Infosys to bring AI tools to more businesses — TechCrunch, 22 Apr 2026
[10] Sierra raises $950M as the race to own enterprise AI gets serious — TechCrunch, 4 May 2026
[11] Capgemini joins forces with OpenAI to accelerate new era of AI-powered enterprise transformation with Frontier Alliance — Capgemini, 23 Feb 2026