The global technology services market just grew at its fastest pace ever. Inside that same quarter, managed services contract value in the Americas fell 12%, and traditional IT outsourcing fell 23%. Both numbers are true, and the gap between them is the most important thing happening in how companies buy engineering right now.
Two headlines ran within a week of each other in July 2026, and almost nobody put them side by side.
The first: ISG reported that the global market for technology services grew at its fastest pace ever in the second quarter of 2026, with combined annual contract value reaching $42.4 billion, up 43% year over year. Gartner raised its worldwide IT spending forecast for the fourth time in nine months, landing at $6.37 trillion for 2026, a 14.2% increase.
The second, from the same ISG data set: in the Americas, managed services annual contract value declined 12% to $5.2 billion, and within that, IT outsourcing fell 23% to $3.6 billion.
Record growth and a contracting outsourcing market in the same quarter, from the same source. That is not a contradiction, and it is not a rounding error. It is a structural reallocation, and if you are responsible for an engineering budget, it has already changed what a good deal looks like.
What Is Actually Happening to the Tech Services Market?
Money is not leaving technology services. It is leaving one specific shape of technology services. ISG's Q2 2026 data shows cloud and as-a-service consumption growing 65% globally to $31.5 billion, while managed services grew just 2.7% to $10.9 billion. In the Americas, infrastructure-as-a-service rose 86% to a record $13.2 billion even as IT outsourcing contracted 23%. The market is not shrinking. The long-duration, headcount-priced contract is.
Look more closely at the Americas breakdown, and the picture sharpens further. Business process outsourcing grew 37% to $1.3 billion in the same quarter that IT outsourcing fell 23%. Companies did not stop buying outside help. They stopped buying outside help to run their infrastructure, and started buying outside help to run business processes that AI can now partly automate.
Gartner's segment forecasts point in the same direction. IT services are projected to grow 5.3% in 2026, compared with 62.5% growth in data center systems and 14.2% in overall IT spending. Separately, Gartner forecasts worldwide AI spending of $2.59 trillion in 2026, a 47% year-over-year increase, of which roughly $585.5 billion is AI services and more than $1.43 trillion is AI infrastructure.
Put plainly: the money is going into compute, into platforms, and into AI-specific delivery work. It is coming out of the arrangement in which you paid a vendor a monthly rate to supply bodies to keep your existing systems running.
Why Is Managed Services Shrinking While AI Spending Explodes?
Because the traditional managed services contract was priced on the one input that AI most directly deflates. A conventional IT outsourcing agreement converts a scope of work into a headcount, converts the headcount into a monthly rate, and locks that rate in for three to five years. Every part of that structure assumes labor volume is a reasonable proxy for value delivered.
That assumption is now visibly wrong to the people signing the contracts. HFS Research, surveying 553 executives at Global 2000 enterprises in June 2026, found that 75% plan to replace people-run services with software-run services by 2028. When three-quarters of your buyers have publicly committed to a plan that reduces the labor content of the work, a multi-year contract priced on labor volume becomes very hard to sign.
There is a second, less flattering reason, and it comes from the buyers themselves. Everest Group surveyed more than 200 senior decision-makers at companies above $1 billion in revenue in April 2026. Eighty percent expect positive returns from AI. Sixty-seven percent name legacy infrastructure as their top barrier. And only 15% believe their service providers are leveraging AI extensively.
That last figure is the one worth sitting with. Buyers are not walking away from managed services because they no longer need help. They are walking away because they have concluded that the incumbent provider is not bringing anything to the engagement that they could not do themselves, and is charging them a multi-year commitment for the privilege. Eighty-five percent of large-enterprise buyers do not believe their provider is meaningfully using AI. That is a verdict on an entire delivery model.
Worth noting the honest counterweight: HFS also found that only 12% of those enterprises have reached true multi-agent systems in core operations, with 55% still classified as explorers. The buy side is confident about where it is going and considerably less far along than its own statements imply. The contracts are being restructured ahead of the capability, not behind it.
What Does This Mean for How You Buy Engineering?
The practical answer is that the unit you are purchasing has changed, and most procurement processes have not caught up. A few shifts are worth making deliberately.
Buy outcomes and capability, not seat-months. The reason ITO fell 23% while AI services grew is that buyers stopped believing headcount predicts value. If your next engineering agreement is still denominated primarily in full-time-equivalent months, you are pricing on the input that is deflating fastest. We wrote about what replaces it in our analysis of outcome-based pricing, including the uncomfortable part: outcome pricing requires you to define the outcome precisely enough to be measured.
Shorten the commitment horizon. A five-year managed services contract signed in 2026 is a bet that the cost and delivery capability will remain roughly the same through 2031. Given that Gartner revised its own IT spending forecast upward four times between October 2025 and July 2026, nobody should be confident enough in a five-year view to lock pricing to it. Shorter terms with defined expansion paths preserve the option value that long contracts destroy.
Make AI capability a scored requirement, not a slide. Only 15% of large enterprises believe their providers use AI extensively, which means the market is currently full of vendors claiming AI capability that their clients do not experience. Ask for specifics that are hard to fake: which parts of the delivery process are automated, what the evaluation harness looks like, how code review works when a model wrote the first draft. Our guide to choosing an AI development partner covers the questions that separate a real answer from a rehearsed one.
Separate the work that must be near your business from the work that must not. The Americas data, showing BPO up 37% while ITO fell 23%, reflects a real distinction. Undifferentiated infrastructure operations are increasingly a platform purchase. Work that touches your specific data, your specific workflows, and your specific customers is where engineering judgment still earns a premium, and that work benefits from being close, in your time zone, in your standups. That is the structural case for a dedicated nearshore engineering team over a distant managed services pool, and it is a different argument than the cost argument that dominated the last decade.
Fix the legacy problem before you buy more AI. Sixty-seven percent of large enterprises named legacy infrastructure as their top barrier to AI returns, ahead of talent and ahead of budget. Buying AI services on top of systems that cannot supply clean data is how organizations end up in the gap between 80% expecting returns and a much smaller share reporting them. We covered where those returns actually surface, where AI ROI shows up, and the honest answer usually starts with data and integration work that nobody wants to fund.
Price the transition, not just the steady state. The 12% decline in Americas managed services is partly due to companies letting contracts lapse without a replacement structure. That creates a capability gap right when the AI workload is increasing. Decide what your engineering function looks like on the other side of the transition before you unwind what you have, and staff the bridge deliberately. Our work on IT management and modernization generally starts at exactly this point.
Is the Traditional Outsourcing Model Finished?
Not finished, but repriced and rescoped. Managed services still grew 2.7% globally in Q2 2026, with $10.9 billion in annual contract value, indicating a large and functioning market. What has broken is the specific proposition that a provider can win a multi-year, labor-priced contract on the strength of scale and process maturity alone.
The providers gaining share are the ones whose pitch survives the Everest Group question. If a buyer asks what you are doing with AI that they could not do themselves, and the honest answer is "scale and lower hourly rates," then that engagement is now competing against a category of work that is being automated. If the answer involves specific engineering judgment applied to the client's specific systems, it holds.
For buyers, the practical consequence is that vendor selection now rewards technical diligence far more than commercial diligence. The cheapest rate card on a five-year term was a defensible choice in 2018. In 2026, it is how you end up in the 85% who do not think their provider brings anything to the table. Our guide to AI staffing models walks through the engagement structures that have replaced it.
Common Questions About the Managed Services Decline
Is the IT services market actually shrinking in 2026?
No. Total technology services spending is at a record. ISG reported a global annual contract value of $42.4 billion in Q2 2026, up 43%, the fastest pace on record, and Gartner forecasts $6.37 trillion in worldwide IT spending for the year. What is shrinking is a segment. In the Americas, managed services contract value fell by 12%, and IT outsourcing fell by 23%, while cloud and as-a-service consumption grew by 69%.
Why is IT outsourcing declining while AI spending grows?
Because traditional outsourcing pricing is based on headcount, and AI reduces the labor content of exactly the work those contracts covered. HFS Research found 75% of Global 2000 enterprises plan to replace people-run services with software-run services by 2028. Buyers are reluctant to lock multi-year labor-priced commitments against a cost base they expect to fall.
What should replace a traditional managed services contract?
Shorter terms, outcome-linked or capability-linked pricing, and a clear split between undifferentiated operations that can be bought as a platform and differentiated engineering that stays close to your business. The failure mode is unwinding the old contract without deciding what the engineering function looks like afterward.
How do I tell whether a provider is genuinely using AI?
Ask process questions rather than capability questions. Which delivery steps are automated, what the evaluation and review process looks like when a model produces the first draft, how they measure quality on AI-assisted work, and what they will show you from a live engagement. Only 15% of large-enterprise buyers believe their providers use AI extensively, so assume the claim is unverified until it is demonstrated.
If your current engineering agreements were structured for a market that no longer exists, the repricing is an opportunity rather than a problem. Talk to the Golabs team about what your engineering spend should look like on the other side of this shift, and how to staff the transition without losing capability along the way.