OpenAI just launched a $4 billion AI services company with Capgemini, Bain, and McKinsey as investors. Hours earlier, Anthropic announced a competing $1.5 billion venture backed by Blackstone and Goldman Sachs. For CIOs, this isn't just another funding round—it's the moment AI vendors decided to replace your systems integrators.
Within a 24-hour window in early May 2026, the two leading enterprise AI companies made nearly identical announcements: both are creating massive services arms to deploy their technology directly into customer operations. The timing wasn't coincidental. The playbook wasn't either.
This is how vendor lock-in gets weaponized at scale.
Why AI Companies Are Building Services Arms Now
The problem both companies are solving is simple: pilot purgatory. Enterprises have been experimenting with AI for two years. Few have moved beyond proof-of-concept. According to industry analysts, the gap between what models can do and what enterprises can actually deploy has become a bottleneck that's stalling billions in potential revenue.
OpenAI's "DeployCo" will operate as a standalone business unit majority-owned by OpenAI, valued at $10 billion on its $4 billion initial raise. Nineteen investors are participating, including TPG (leading the partnership), Advent, Brookfield, Bain Capital, Goldman Sachs, and SoftBank. System integrators Capgemini, Bain & Co., and McKinsey are also direct investors—a signal that traditional consulting is hedging against its own disruption.
Anthropic's competing venture was announced just hours before, valued at $1.5 billion with $300 million commitments each from Anthropic, Blackstone, and Hellman & Friedman. Goldman Sachs appears in both cap tables—the only named investor backing both sides of this war.
The stated goal for both ventures is the same: embed forward-deployed engineers (FDEs) into customer organizations to turn AI capability into production systems. The model, popularized by Palantir, puts vendor engineers directly into customer workflows instead of handing over software and walking away.
Translation: AI vendors are moving from selling you models to controlling how you deploy them.
What Forward-Deployed Engineers Actually Do
Both companies describe their FDE strategy in nearly identical terms: engineers will sit with business leaders, operators, and frontline teams to identify high-impact use cases, redesign organizational infrastructure, and build durable AI systems integrated into existing workflows.
For technical leaders, this sounds like exactly what you've been asking for. Integration has been the hard part. Most enterprises lack specialized AI engineering talent. Vendor-supplied engineers who understand the model's capabilities and limitations could accelerate deployment timelines by months.
For business leaders, this solves the ROI problem. Pilots are cheap. Production systems are not. If the vendor can guarantee deployment success through its own engineering team, the business case for enterprise AI becomes simpler to justify and faster to execute.
But there's a structural trade-off hiding in this convenience.
When a vendor's engineers build your AI infrastructure, you're not just licensing a model—you're outsourcing architectural decisions. Those decisions determine data pipelines, workflow integration, governance frameworks, and operational dependencies. Over time, this creates lock-in not just to the model, but to the entire stack the vendor built around it.
Traditional systems integrators have been screaming this warning for weeks. Russell Goodenough, senior VP and AI lead for CGI (a top-15 solution provider and partner to both OpenAI and Anthropic), told CRN that unlike born-in-AI services firms, CGI brings trust, security, and the ability to avoid vendor lock-in by working within a customer's existing IT estate.
"We want to be the first organization to prove that [AI-powered ERP replacement] can be done in a trustworthy, dependable way, not just practiced at a hackathon," Goodenough said.
The subtext: vendor-led deployment may be fast, but it's not neutral.
The Lock-In Economics CIOs Need to Understand
Here's the math that matters: if your AI infrastructure is built by the model vendor, switching costs become prohibitive.
Tulika Sheel, senior vice president at Kadence International, pointed out the obvious trade-off: buying AI services directly from model providers reduces short-term deployment risk through tighter integration and specialized expertise. But it creates deeper dependency across the stack—from models to data pipelines and workflows.
"Over time, this could increase lock-in, making it harder to switch vendors without significant disruption," Sheel said.
Neil Shah, VP for research at Counterpoint, framed it more bluntly: AI model providers are trying to become a "one-stop shop" by tying AI applications and services more closely to their usage-driven business models. Controlling the application and services layer allows them to lock in enterprises while optimizing models based on firsthand understanding of customer needs and workflows.
In other words: the vendor gets smarter about your business, and you get more dependent on the vendor.
Lock-in isn't inevitable, according to Deepika Giri, head of research for AI, analytics, and data at IDC. But avoiding it requires deliberate architecture choices early in the process.
"While the model layer can increasingly be abstracted through modular architectures, avoiding lock-in requires deliberate design choices," Giri said. "Without that, organizations risk becoming dependent not just on a model, but on the entire stack: data pipelines, workflows, and governance frameworks tied to a specific provider."
For CFOs, the question becomes: what's the long-term cost of convenience?
Why This Announcement Happened Now
The timing of these announcements reflects pressure from both sides of the market.
On the vendor side, both OpenAI and Anthropic are fundraising at unprecedented scale while circling potential IPOs. OpenAI announced $122 billion in new funding at the end of March 2026 against a valuation of $852 billion. TechCrunch reported that Anthropic is finalizing a funding round seeking $50 billion against a $900 billion valuation.
Building services arms allows both companies to capture more value from enterprise deals and prove that their models can drive real operational outcomes—not just impressive demos.
On the customer side, enterprises are stuck. Pilots have proliferated, but production deployments remain rare. The gap between AI capability and enterprise readiness is measured in integration complexity, governance risk, and talent scarcity. Forward-deployed engineers address all three.
The services ventures also solve a channel problem. Both companies have been aggressively expanding partner ecosystems this year. OpenAI hired Colleen Kapase, a channel executive who led aggressive partner programs at Google Cloud and Snowflake, to lead its go-to-market strategy. Anthropic launched the Claude Partner Network with an initial $100 million investment and a new certification program.
But partners want margin. Vendor-controlled services let OpenAI and Anthropic capture implementation revenue without splitting it with systems integrators.
The result: a new competitive dynamic where AI vendors and traditional SIs are simultaneously partners and rivals.
What CIOs Should Do Right Now
If you're planning an enterprise AI deployment in 2026, here's what this announcement changes:
First, assume vendor-led services will create lock-in unless you design against it. Insist on modular architecture from day one. Make sure data pipelines, governance frameworks, and workflow integrations are abstracted from the model layer. If you can't swap vendors without rearchitecting core systems, you've built lock-in by accident.
Second, compare vendor-led deployment against traditional SI partnerships on total cost, not just speed. Vendor FDEs may get you to production faster, but if that speed comes with multi-year dependencies on proprietary infrastructure, the long-term cost could exceed the short-term savings.
Third, negotiate SLAs and exit terms before engagement begins. If a vendor's engineers are building your AI infrastructure, document who owns the architecture, the data pipelines, and the operational playbooks. Make sure you have the right to port that infrastructure to a different model provider without starting from scratch.
Fourth, watch for conflicts between vendor incentives and your architectural goals. Forward-deployed engineers work for the vendor, not for you. Their job is to maximize usage of their employer's models. Your job is to build systems that serve your business—even if that eventually means switching vendors.
Finally, don't treat this as a binary choice. Hybrid approaches exist. You can use vendor FDEs for initial deployment while maintaining architectural oversight through your own engineering team or a neutral SI. The key is making sure you control the critical dependencies.
The Bottom Line for Enterprise Leaders
OpenAI and Anthropic just announced that they're coming for your systems integrators' revenue. They're betting that enterprises want speed more than they want independence.
For many companies, that bet will be correct. Pilot purgatory is real. The gap between AI capability and enterprise deployment is costly. If vendor-led services can bridge that gap faster than traditional SIs, the market will reward them.
But speed has a price, and in enterprise IT, that price is usually paid years later when you realize how expensive it is to change direction.
The question CIOs need to answer isn't whether to use vendor-led services. It's whether the architectural decisions those services bake into your infrastructure will still make sense when the next generation of models arrives—or when your business priorities shift.
Because once a vendor controls your AI infrastructure, switching isn't a migration. It's a rebuild.
And rebuilds are how lock-in becomes permanent.
