DXC’s AI partner said this about DXC
That argument is terminal for DXC’s independent thesis, and worth following all the way down.
The disintermediation trap
DXC’s OASIS pitch is: “we use AI to deliver enterprise IT faster and cheaper.” But that sentence contains its own refutation. If AI is the delivery engine, the client’s next question is obvious — why is DXC in the middle? What margin am I paying for, exactly?
The historical DXC value proposition was: we have 130,000 people, global delivery centres, integration expertise, and we’ll manage your IT estate for you. That justified a premium. AI erodes every one of those pillars. Headcount is no longer a proxy for capability. Integration expertise becomes a commodity when AI can read legacy code, write connectors, and configure systems autonomously. The “rates times hours” model — which the CEO himself acknowledged is dying — was the moat. There isn’t a replacement moat, because they don’t own the AI.
They’re a reseller with a thin value layer
DXC has a multi-year global alliance with Anthropic , but that alliance confers no exclusivity. Accenture has the same AI partnerships. Capgemini has them. Infosys has them. IBM has had its own AI story for a decade (Watson — a cautionary tale). Every competitor accesses the same models through the same APIs at the same price. DXC’s differentiation in that stack is vanishingly thin.
The client who wants Claude in their enterprise IT has three routes: direct API, an Anthropic enterprise deal, or through a system integrator. Each iteration of AI capability that reduces the integration complexity — better native connectors, more autonomous agents, simpler deployment — removes another reason to pay DXC’s margin for route three.
The simplicity argument is the ki-ler
This is the part that doesn’t get enough airtime in the earnings calls. DXC’s implicit bet is that enterprise AI deployment remains complex enough to require a managed service layer indefinitely. That bet looks worse with every model generation.
Early enterprise AI genuinely required significant hand-holding: prompt engineering, reliability management, integration scaffolding, output validation. Those were billable. But the trajectory is one-way — more capable, more reliable, more self-configuring, lower total cost of ownership with less implementation overhead. The skills gap that justified outsourcing to DXC is closing from below, not above. AI improves; the implementation complexity shrinks; the billable wrap around it shrinks with it.
The historical parallel is brutal: travel agents started saying “we’ll use Expedia to book your travel.” The client saw the margin and removed the agent. DXC is the agent.
What’s actually left
There are two genuine residual moats and neither is scalable:
One: government and regulated-sector contracts with security clearance requirements, long procurement cycles, and high switching costs. MoD/BAE-type accounts — fall into this category. These have genuine stickiness but are also the most likely to be served by security-cleared AI providers within 5 years, cutting out the SI layer even there.
Two: deep legacy system knowledge. Nobody else knows a client’s 30-year-old mainframe architecture. But AI is already eroding this — large models can now read COBOL, reverse-engineer undocumented systems, and produce documentation. That moat has a measurable shelf life.
The terminal conclusion
DXC is not using AI to build a new business. It is describing, in its investor communications, the mechanism of its own further disintermediation. Every OASIS success story — “we deployed this 40% faster using AI” — is simultaneously an advertisement to the client for why they could do it without DXC next time.
The company generates real cash and has real contracts. That’s why PE is circling. The play is: take it private, cut 30,000 jobs, harvest the cash from the existing contract base over a 5–7 year run-off horizon, sell or shut the rest. That’s not a technology company. That’s a liquidation vehicle with good near-term cash characteristics.
As an independent, publicly listed, growth-oriented technology company: it’s over. The only question is the pace of the wind-down and whether someone buys it before the market grinds it to zero.