Anthropic just raised $1.5 billion to do something no AI company has done before: abandon the software business model. The joint venture with Blackstone, Hellman & Friedman, Goldman Sachs, and five other alternative asset managers isn't a bet on better models or faster inference. It's a bet that enterprise AI success requires human engineers sitting next to your finance team, not self-service APIs.
OpenAI announced an identical play hours earlier—$10 billion from 19 investors for "The Development Company." Same model, larger scale, zero investor overlap. The simultaneous launches aren't a coincidence. They signal a fundamental shift in how AI companies capture enterprise value: services revenue beats software revenue when the product requires custom integration at every deployment.
For CFOs evaluating AI vendors and CTOs building deployment roadmaps, this matters more than any model benchmark. The era of "buy the API and figure it out yourself" just ended.
What Anthropic Actually Built
The joint venture operates as an independent services company, not a division of Anthropic. The founding partners—Anthropic, Blackstone, and Hellman & Friedman—each committed $300 million. Goldman Sachs joined as a founding partner (commitment undisclosed). A consortium of five additional investors—General Atlantic, Leonard Green, Apollo Global Management, GIC, and Sequoia Capital—backs the venture with additional capital.
The Wall Street Journal valued the entity at $1.5 billion at formation. That valuation assumes services margins, not software margins—a critical distinction for understanding the economic model.
Here's how engagements work: A mid-sized company (Anthropic specifically mentions community banks, manufacturers, and regional health systems) contracts with the joint venture. A small team—combining the JV's engineers and Anthropic's Applied AI staff—embeds with the customer to identify where Claude can deliver the biggest operational impact.
From there, they build custom Claude-powered systems tailored to the organization's workflows. This isn't a 90-day implementation. Anthropic describes "long-term support" as part of the standard engagement model.
The example Anthropic provided: A multi-site healthcare services network struggling with documentation overhead. Clinicians spend hours daily on medical coding, prior authorizations, and compliance reviews. The engagement team sits down with clinicians and IT staff to build tools that fit existing workflows—not generic SaaS products that force workflow changes.
The clinicians identify where time disappears. The engineers build around that knowledge. The result: more time for patient care, less administrative burden.
This is the Palantir forward-deployed engineer model, applied to AI. The difference: Palantir charges millions per engagement. Anthropic's venture targets mid-sized businesses that can't afford Palantir's price point but still need hands-on integration.
Why OpenAI Copied the Same Playbook (10x Larger)
OpenAI's "The Development Company" raised $4 billion against a $10 billion valuation. Named investors include TPG, Brookfield Asset Management, Advent, and Bain Capital. The investor composition tells you the target customer base: mid-market and lower-middle-market portfolio companies owned by private equity and alternative asset managers.
The logic is identical to Anthropic's venture. The investors get preferred sales access to their portfolio companies. The JV captures services revenue from resulting contracts. The AI lab (OpenAI or Anthropic) gets deployment scale without building a consulting arm internally.
The simultaneous announcements reveal a strategic consensus: Both companies believe enterprise AI revenue will flow primarily through services contracts, not pure software licenses. This shifts the competitive battlefield from "who has the best API" to "who can deploy fastest with the least customer engineering burden."
For enterprise buyers, this creates an interesting dynamic. OpenAI's larger scale ($10B vs $1.5B) suggests broader deployment capacity. Anthropic's smaller, more focused approach may deliver more customized solutions. Neither model guarantees success—both depend on execution quality from forward-deployed engineering teams.
The Economics Tell the Real Story
Software companies trade at 5-15x revenue multiples. Services companies trade at 1-3x. That valuation gap exists because software scales without linear headcount growth. Services revenue requires adding engineers for every new customer.
Anthropic and OpenAI both understand this. They're raising services capital through joint ventures—keeping the high-multiple software business (API revenue) separate from the lower-multiple services business (forward-deployed engineers).
Here's why that structure matters for enterprise buyers:
The joint ventures operate as independent P&Ls. They succeed or fail based on services delivery quality, not on whether Anthropic or OpenAI ships a better model next quarter. This separation creates clearer incentives: the JV's engineering team gets paid to make Claude work in your environment, not to upsell you on the next model tier.
For CFOs evaluating total cost of ownership, this changes the calculation. Instead of licensing fees + internal engineering costs + third-party consulting, you're buying a bundled services contract. The trade-off: higher upfront cost, but predictable spend and faster time-to-value.
The investor composition provides pricing clues. Blackstone, Hellman & Friedman, and Goldman Sachs don't invest in low-margin businesses. They expect EBITDA margins in the 20-30% range for services firms. That implies pricing significantly above pure software licenses but below traditional Big 4 consulting rates.
A mid-sized manufacturer with $500M-$2B in revenue might pay $2M-$5M annually for an engagement that includes custom integration, ongoing optimization, and model updates. Compare that to a self-service API contract ($100K-$500K annually) plus $1M-$3M in internal engineering costs plus $500K-$2M in third-party consulting. The services model eliminates the coordination overhead.
What This Means for Mid-Sized Businesses
Anthropic explicitly targets companies that "lack the in-house resources to build and run frontier deployments." That's most mid-sized organizations. A $1B manufacturer doesn't have a 20-person AI engineering team. A regional bank doesn't have ML infrastructure. A multi-site healthcare network doesn't have the budget for a Big 4 consulting engagement.
The services JV model fills that gap. You get access to frontier models (Claude 4.6, Claude Code) without building internal ML expertise. You get custom integration without hiring a consulting firm. You get long-term support without maintaining a specialized team.
The trade-off: You're betting on Anthropic's roadmap. If Claude falls behind OpenAI or Google in model quality, you're locked into a services contract with an inferior product. If Anthropic's Applied AI engineers can't deliver the customization your workflows require, you've paid for a failed deployment.
Risk mitigation strategies for buyers:
-
Pilot before committing to multi-year contracts. Anthropic describes engagements starting with "a small team working closely with the customer." Insist on a 90-day pilot with clear success metrics before expanding scope.
-
Negotiate model flexibility. If Claude's performance degrades vs. competitors, you should have contract language allowing model switching or early termination.
-
Define handoff criteria. At some point, your internal team should own the deployment. Negotiate a clear transition plan—don't become permanently dependent on Anthropic's engineers.
-
Compare against in-house builds. For companies with existing ML teams, the cost comparison isn't JV services vs. self-service APIs. It's JV services vs. hiring 3-5 engineers internally. Run the numbers for your headcount costs and opportunity cost of build time.
What This Means for CTOs and VPs of Engineering
If you're building an AI deployment roadmap, this shift changes your vendor evaluation criteria. Model benchmarks matter less. Integration speed and workflow customization matter more. Here's what to prioritize:
1. Deployment velocity beats model performance (within reason). A Claude-powered system deployed in 8 weeks delivers more business value than a GPT-5-powered system deployed in 6 months. The services JV model promises faster deployment by embedding engineers with your team from day one.
2. Evaluate integration depth, not just API quality. Ask prospective vendors: "How many engineers will work on-site with my team? For how long? What's the escalation path when we hit workflow edge cases?" The answers reveal whether they're selling software or services.
3. Compare total engineering burden. Calculate: (vendor services cost) vs. (self-service API cost + internal engineering time + third-party consulting). For most mid-sized companies, the services model costs less when you account for fully loaded internal engineering time.
4. Assess post-deployment support. Anthropic promises "long-term support" as part of standard engagements. Define what that means in your contract. Monthly optimization reviews? Quarterly model updates? 24/7 on-call support? The details determine whether the services model delivers ongoing value or becomes a vendor dependency trap.
5. Plan for eventual in-house ownership. Even with a services JV, your team should build internal AI literacy. Anthropic's engineers will embed knowledge transfer into the engagement—insist on documentation, training, and transition planning from day one.
The Competitive Landscape Just Shifted
Before this week, enterprise AI vendors competed on model quality and API pricing. After this week, they compete on deployment speed and integration depth. That's a fundamentally different game.
Who wins in this new landscape:
- Anthropic and OpenAI have first-mover advantage with capital-backed services JVs. They can underprice traditional consultants while delivering frontier models.
- Google (via Vertex AI) already has a professional services arm and cloud infrastructure. They can match the JV model by bundling GCP credits with services contracts.
- Microsoft (via Azure OpenAI + Consulting) can leverage its existing enterprise relationships and deploy GPT models through Microsoft Consulting Services.
- Traditional consultants (Accenture, Deloitte, PwC) face margin pressure. Anthropic already partners with them through the Claude Partner Network, but the JV competes for the same mid-market customers.
Who loses:
- Self-service AI platforms that assumed enterprises would build deployments themselves. The services JV model proves that assumption wrong for mid-sized businesses.
- Niche AI consultants without access to frontier models or capital to match Anthropic/OpenAI pricing.
- Enterprise software vendors selling "AI-powered" features as upsells. If buyers can get custom AI integration from Anthropic's JV for comparable spend, why pay for rigid SaaS features?
Bottom Line: Services Revenue Beats Software Revenue (For Now)
The simultaneous Anthropic and OpenAI announcements reveal a strategic consensus: enterprise AI revenue flows through services contracts, not pure software licenses. This doesn't mean software revenue disappears—API fees still matter. But the biggest deals, the longest contracts, and the deepest customer relationships will come from forward-deployed engineering teams.
For CFOs: Budget for services contracts, not just API spend. The total cost may be higher, but the deployment risk is lower and the time-to-value is faster.
For CTOs: Evaluate vendors on integration depth and deployment speed, not just model benchmarks. The best model doesn't matter if it takes 12 months to deploy.
For mid-sized businesses: You now have access to frontier AI without building internal ML teams. The trade-off is vendor dependency, but that's manageable with clear contracts and transition planning.
The AI-as-software era lasted 18 months. The AI-as-a-service era just began.
Continue Reading
- Stanford AI Playbook: Why 95% Fail Before Technology — Organizational readiness beats model selection
- The $670K Gap: Why 78% of AI Pilots Die Before Production — Governance requirements for production AI
- AI ROI Calculator — Quantify services vs. in-house build costs
Sources
- Anthropic: Building a new enterprise AI services company
- TechCrunch: Anthropic and OpenAI are both launching joint ventures for enterprise AI services
- Wall Street Journal: Anthropic Nears $1.5 Billion Joint Venture With Wall Street Firms
- Bloomberg: OpenAI Finalizes $10 Billion Joint Venture With PE Firms to Deploy AI
