Thirteen days ago, on May 10–11, 2026, OpenAI launched a $4 billion joint venture called the OpenAI Deployment Company — DeployCo for short — and committed to embedding its own Forward Deployed Engineers inside Fortune 500 customers to do the AI integration work the model alone couldn’t close. Within 48 hours, the Nifty IT index dropped 3.6% on the news, with TCS down 4%, Persistent down 5%, and Infosys / Wipro / HCL down 2.5–4%. Within 13 days, every CIO with an open Q3 AI services RFP has been pitched DeployCo at least once.
Nobody is writing the version of this story that an actual procurement team can act on. The SERP is OpenAI’s press release (Advent partner statement here), the partner press, a few “Lock-In Machine” critiques on YouTube, and Indian IT business-press takes on the stock drop. None of those gives you the side-by-side a buyer needs.
So that’s what this is. Seven vendors, nine procurement dimensions, scored honestly with the conflicts disclosed. Then three buyer-type verdicts, then the three questions you should be asking every vendor on the table before signing anything in Q3.
The Two Facts That Should Anchor Every Procurement Conversation
Before the table, two non-obvious facts that the popular coverage has buried.
One. DeployCo’s economics are pre-committed to its investors. Multiple secondary sources tracking Reuters and FT reporting — including Gotrade (April 21, 2026) and Futunn (April 21) — describe a structure that guarantees private-equity backers a 17.5% annualized return over a five-year commitment. That’s a hard contractual obligation sitting underneath DeployCo’s pricing. If your DeployCo SOW comes back at margins that surprise you, this is why. The discount you can negotiate is bounded by the floor TPG, Advent, Bain Capital and Brookfield were promised before your RFP ever existed.
Two. Three of your would-be neutral advisors are now equity holders in DeployCo. McKinsey, Bain & Company, and Capgemini are listed in the official OpenAI announcement as founding partners alongside the PE firms — meaning when McKinsey QuantumBlack, Bain’s AI consulting practice, or Capgemini Generative AI recommends DeployCo in your Q3 AI services review, they have direct equity exposure to that outcome. Deloitte AI&Data and Accenture’s AI&Data practice are not listed as investors. That is not a small procurement-governance fact. It is the central one.
With those two anchors in place, here is the table.
The 7-Vendor × 9-Dimension Decision Table
| Dimension | DeployCo | Accenture AI&Data | Capgemini GenAI (DeployCo investor) | McKinsey QuantumBlack (DeployCo investor) | Bain AI (DeployCo investor) | Deloitte AI&Data | In-House Build |
|---|---|---|---|---|---|---|---|
| 1. Day-1 staffing capacity (US) | ~150 FDEs (Tomoro), scaling | ~75K AI-capable | ~30K AI-capable | ~5K consultants | ~3K consultants | ~25K AI-capable | 0 — must hire |
| 2. AI-model lock-in risk | High — OpenAI-only | Low — model-agnostic | Medium — partial OpenAI tilt now | Medium — partial OpenAI tilt now | Medium — partial OpenAI tilt now | Low — model-agnostic | None |
| 3. Data residency / IP | OpenAI-controlled by default | Client-controlled (standard) | Client-controlled | Client-controlled | Client-controlled | Client-controlled | Client-controlled |
| 4. Margin structure | Pre-committed to 17.5% PE IRR over 5 years | Standard SI margins (~35-45%) | Standard | Standard | Standard | Standard | Internal cost |
| 5. Negotiating leverage on enterprise API pricing | Theoretically high (parent company) — but governance unclear | None directly | Possible — partner status | Possible — partner status | Possible — partner status | None directly | None |
| 6. Recommendation conflict-of-interest | Self-evident | None | Yes — equity holder | Yes — equity holder | Yes — equity holder | None | None |
| 7. Exit cost / off-ramp credibility | High — proprietary stack | Low — portable artifacts | Medium | Medium | Medium | Low — portable artifacts | None |
| 8. Regulatory / EU AI Act posture | Untested at this scale | Mature compliance practice | Mature | Mature | Mature | Mature | DIY |
| 9. Cultural fit (engineering-led team) | Highest — Tomoro is engineering-first | Variable | Variable | Strategy-led | Strategy-led | Variable | Native |
Three of the cells deserve a sentence of unpacking before the verdicts.
Cell 4 (margin structure). DeployCo’s pricing is not a black box; it is bounded below by the 17.5% return its investors were promised. That doesn’t mean you can’t get a competitive number — it means the negotiated floor is higher than for vendors whose owners are happy with 12–14%. Procurement teams should expect smaller discount windows than they’re used to from Accenture or Deloitte.
Cell 6 (conflict-of-interest). This is the single most important cell in the entire table. If your Q3 AI-vendor-selection process used a McKinsey, Bain, or Capgemini advisor — and recommended DeployCo — you have a documented governance issue that your audit committee may or may not flag depending on how rigorous they are. The fix is not necessarily to reject DeployCo; it is to require explicit conflict-of-interest disclosure from any of the three before accepting their recommendation, and to source a second opinion from a vendor that does not own equity in any of the seven.
Cell 7 (exit cost). Hacker News commenters on the launch were direct: “Good luck ever cancelling [a vendor] when they have people inside your organization.” That cuts both ways — Accenture and Deloitte deployments have the same gravity once they’re in — but DeployCo’s combination of embedded engineers + OpenAI-proprietary tooling is a more aggressive lock-in surface than the traditional SI engagement. Negotiate the off-ramp before signing, not after.
Three Buyer-Type Verdicts
Different industries should weight these dimensions very differently. Here are three honest cuts.
If you’re a regulated financial-services CIO
Deloitte AI&Data favored. Two reasons. First, the conflict-of-interest cell matters more in regulated industries — your model-risk management framework and your board audit committee will both ask the question, and Deloitte is the only Big-Four AI practice with no equity exposure to DeployCo. Second, the AI-model lock-in cell matters more when your existing compliance posture is multi-vendor by default — the cost of being unable to add an Anthropic-based model later is bigger than the cost of slightly slower deployment.
DeployCo can still win the work, but only after explicit board-level disclosure that you considered the OpenAI lock-in and judged it acceptable.
If you’re a high-growth-tech CTO with engineering depth
DeployCo or in-house, often a hybrid. Two reasons. The Tomoro engineering bench is genuinely strong (the team built Supercell’s 110M-user in-game support agent in 12 weeks) and a tech-native culture absorbs FDEs better than a non-tech enterprise. The model-lock-in risk is also lower for you because you can absorb a model switch faster than a bank can. The in-house path stays defensible because you already pay engineers who can be retrained on the OpenAI Cookbook in a quarter.
The configuration that usually wins here is DeployCo for the first two production deployments + in-house ownership of everything after — explicitly written into the SOW as a transition clause.
If you’re a legacy-industrial CIO
Accenture or Capgemini, with Capgemini’s conflict disclosed. Two reasons. Day-1 staffing capacity matters more when you’re standing up AI in 40 manufacturing plants across three continents — Accenture and Capgemini have the scale DeployCo doesn’t yet. EU AI Act compliance is also a much bigger workstream for industrial firms with EU operations, and Capgemini’s compliance practice is mature in a way DeployCo’s three-month-old governance posture cannot be.
If you pick Capgemini, the equity-in-DeployCo disclosure should be in the procurement file. That doesn’t make Capgemini wrong; it makes the file complete.
The Three Questions to Ask Every Vendor Before You Sign
This is the operational layer. Use these in the next vendor call, regardless of which row you’re leaning toward.
Question 1 — to McKinsey, Bain, and Capgemini specifically: “Does your firm hold direct or affiliated equity in any AI deployment vendor you are recommending to us? If so, what is the size of that exposure, and what governance review have you applied to the recommendation?” Watch how fast they answer.
Question 2 — to DeployCo: “What is the contractual mechanism by which a Forward Deployed Engineer placed inside our organization can be replaced by a non-DeployCo engineer if our needs change? Show me the clause.” If there isn’t one, that’s information.
Question 3 — to all seven: “Walk me through your data-isolation guarantee. Specifically: do model training, evaluation, or caching use our data, our outputs, or our prompts in any way that survives termination of this engagement? Identify the contract clauses that protect each of those three.” The three data-isolation guarantees you want named are no-train, no-eval, no-cache. Any vendor that can’t point to all three clauses in their template is selling you a weaker data posture than they’re claiming.
What This Table Can’t Fix
Four limits before you act on this.
- The full 19-investor list is not yet on the public record. Reuters and FT have it in primary reporting; the secondary press has pieces. Once the full roster is public (expect it within the SEC-filing cycle around any later OpenAI IPO), additional conflict cells may need updating.
- The 17.5% PE return is reported, not confirmed by OpenAI directly. The number is consistent across Gotrade, Futunn, and TNW’s reporting attributed to FT/Reuters, but OpenAI’s own announcement does not state it. If your procurement file requires direct attribution, get FT or Reuters’ primary text.
- Vendor capabilities will shift fast. Anthropic’s Code w/ Claude London (May 19–21) included a “Managed Agents” announcement that competing SIs will pick up; the Accenture and Deloitte rows could look meaningfully different in 60 days. Re-run this table at quarterly cadence.
- None of this addresses whether you need outside help at all. The in-house column exists for a reason. The OpenAI B2B Signals report showing frontier firms use 16× as many Codex messages per developer is evidence that the deepest AI value is being captured in-house by teams that just got good at using the tools, not by teams that hired the most expensive vendor. Read that one before you sign anything.
The Bottom Line
DeployCo is not a bad option. It is a specific option with specific tradeoffs that the public reporting has either flattened ("$4B AI consulting firm scares Big-Four") or dismissed (“Lock-In Machine”). The honest read is in the table above: high engineering quality, real model-lock-in risk, a margin floor that bounds your discount window, and a conflict-of-interest situation involving three Big-Four-adjacent advisors that your audit committee should know about before any Q3 signature.
The procurement governance work — the conflict disclosure, the data-isolation review, the FDE replacement clause, the EU AI Act posture, the exit-cost negotiation — is the same work you would do for any vendor on this row. DeployCo just makes the consequences of skipping that work bigger.
If you’re standing up an enterprise AI rollout in Q3 and want a structured framework for the buyer-side decisions in this article, the Enterprise AI Rollout Playbook course on FindSkill walks through the procurement, governance, and integration layers in order — the first two lessons are free.
Sources
- OpenAI launches the OpenAI Deployment Company — OpenAI (May 10, 2026)
- OpenAI launches AI consulting arm valued at $14 billion — Axios (May 11, 2026)
- What People Are Saying About OpenAI’s Big Investment in DeployCo — Business Insider (May 10, 2026)
- OpenAI creates new unit with $4 billion investment to aid corporate AI push — Reuters via SAHM Capital (May 11, 2026)
- Advent International — OpenAI Deployment Company announcement
- OpenAI $1.5B PE Joint Venture — Gotrade (April 21, 2026)
- 17.5% guaranteed annual return reporting — Futunn (April 21, 2026)
- OpenAI launches $4bn Deployment Company with TPG, Advent, Bain, and Brookfield — TNW
- OpenAI Deployment Company is a Lock-In Machine — YouTube (May 19, 2026)
- OpenAI’s Deployment Company — CIO Dive (May 11, 2026)
- Tomoro AI consulting firm
- How Frontier Firms are Pulling Ahead — OpenAI B2B Signals
