On May 4, 2026, the two biggest names in AI announced consulting-delivery companies on the same day, just hours apart. The author, a longtime AI industry observer, argues this marks the moment model companies realized the real money is not in the models themselves, but in rewiring client companies' code and processes.
Key points
The double announcement
- In the morning (US Eastern time), Bloomberg reported that OpenAI was secretly raising $4 billion to establish an independent company, "The Development Company," at a $10 billion valuation, with 19 investors already committed to most of the round.
- Hours later, Anthropic officially announced a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs focused on enterprise-grade AI services.
- Both moves point to the same business model: embedding engineers directly inside client companies, redesigning business processes, and billing per project — essentially the same "man-month" business Accenture, Deloitte, and IBM Consulting have run for thirty years, but with frontier models and internal toolchains in hand.
- No access to internal playbooks. Consultancies only get public APIs and documentation; Anthropic's own engineers get model cards, internal toolchains, and unreleased features. A project that takes a consultancy a year can be done in three months.
- A pricing model broken by AI. The classic model of a $2,000/day senior consultant and 50 man-months per project is overturned by "a few embedded engineers."
- 2023: research lab publishing papers and demos.
- 2024: API provider, selling tokens like a power plant sells electricity.
- 2025: consumer product company (ChatGPT, Claude.ai), like Apple selling phones.
- 2026 (May 4): systems integrator, like SAP and Oracle selling ERP plus implementation.
Not API, not SaaS — embedding AI into a company's DNA
The new ventures do not replace existing channels: token-based API billing and SaaS subscriptions continue. Anthropic itself disclosed ARR above $30 billion in 2026, up from $9 billion at the end of 2025. Instead, the joint venture dispatches small teams of top engineers to embed AI into core business processes, charging by project milestones.
The author's analogy: the model is a supercar, the API sells you fuel, and the new company provides the professional pit crew, the modified chassis, and coaching — until everything under the hood is your code and your processes. Enterprises that bought Claude, GPT, and Gemini APIs often stall at the hard part: security compliance, data privacy, legacy system integration, and employee training. The new ventures fill exactly those gaps.
Three structural weaknesses of the API business
1. Declining gross margins. Token prices halved repeatedly between 2024 and 2026, with models like Sonnet 4.6 and GPT-5.4 cutting prices round after round; volume-based revenue looks good but margins keep compressing. 2. Fragile lock-in. Switching providers can take hours and costs about as much as changing a food-delivery app. Today's Claude customer becomes tomorrow's Gemini customer after a price cut. 3. What enterprises actually want is a working solution, not a token pool. Salesforce proved the value of "implementation + subscription." Fortune 500 buyers care about KPI impact, not model benchmarks. Once your engineers rewrite a client's workflow, the model becomes irreplaceable DNA — that is the real lock.
The end of the good times for Accenture & co.
The author identifies two handicaps traditional consultancies cannot overcome:
Notably, Anthropic partnered with Wall Street private equity rather than other consultancies: the PEs bring the client pool and capital; Anthropic brings the models and engineers — no middleman needed.
Chinese cloud vendors already run the same playbook
This is not a Silicon Valley invention. In China, Alibaba Cloud + Qwen sells "large model all-in-one machines plus implementation services" to government and enterprise clients; Volcano Engine + Doubao offers industry solutions with on-site engineers for gaming, e-commerce, and content; Huawei Cloud + Pangu does end-to-end integration in industrial and telecom; Baidu AI Cloud + ERNIE pairs the Qianfan platform with industry consulting teams; Tencent Cloud + Hunyuan works through ecosystem system integrators.
The difference: Chinese vendors bundle model + sales + implementation inside one cloud company, while the Silicon Valley model is a cleaner alliance of model company + Wall Street capital + delivery partners. The lesson for Chinese peers: "selling APIs + SaaS" alone is not a deep enough business — the moat lies in the heavy, unglamorous step of embedding engineers until the client's workflow is rewritten.
The trajectory: 2023 → 2026
The post closes with a question for readers: by 2030, will the most profitable AI companies be model companies, a new generation of SaaS firms, or a new generation of consultancies?
References
1. Bloomberg. (May 4, 2026). OpenAI to Launch Independent "The Development Company" with $4B Funding Round. 2. Anthropic Official Press Release. (May 4, 2026). Anthropic Forms $1.5 Billion Joint Venture with Blackstone, Hellman & Friedman and Goldman Sachs to Deliver Enterprise AI Services. 3. TechCrunch. (May 4, 2026). Anthropic and OpenAI are both launching joint ventures for enterprise AI services. 4. Fortune. (May 5, 2026). Why the biggest AI labs are becoming the new Accenture. 5. Blackstone Press Release. (May 4, 2026). Blackstone Partners with Anthropic to Accelerate Responsible AI Adoption Across Portfolio Companies.