IT Resilience Study: Costs Up 14%, Attacks Up, AI Readiness Down
A study commissioned by 11:11 Systems and conducted by Data Protection Matters, surveying more than 700 IT leaders across EMEA, the Americas and APAC, has reported its Australia and New Zealand findings on IT resilience. Ninety-three per cent of ANZ respondents saw total data centre infrastructure costs rise over the past twelve months, with an average increase of 14 per cent year over year. Forty-eight per cent name cyberattacks as their top data centre concern, 76 per cent experienced at least one attack in the past year and 25 per cent experienced three or more. Forty-three per cent say one hour of downtime from a cyberattack cost more than $250,000.
And while 63 per cent are investing in AI-enhanced IT or security toolsets, only 18 per cent have operational AI-enhanced backup or disaster recovery. 11:11 Systems sells managed infrastructure and recovery services, which is worth holding alongside the findings and the recommendation that closes them.
The number the release does not give
The survey covers more than 700 IT leaders globally. The findings reported here are ANZ-specific, and the release does not say how many of those 700 sit in Australia and New Zealand.
That matters more than it might appear. ANZ is a subset of the APAC portion of a three-region study, so the base for every percentage in this release is smaller than 700 and possibly considerably smaller. On a base of a hundred respondents a 93 per cent figure carries a margin of error of roughly five points; on a base of eighty it widens further. The direction of each finding is unlikely to be wrong. The precision implied by reporting 93 per cent and 14 per cent and $250,000 is more than an undisclosed sample can support.
The study also reports no margin of error and no sampling method. A reader should treat the figures as a well-constructed indication of what ANZ IT leaders are experiencing rather than as measurements, and the release would be stronger for stating the regional base plainly.
The AI readiness gap is the real finding
Sixty-three per cent are investing in AI-enhanced IT or security toolsets and 18 per cent have operational AI-enhanced backup or disaster recovery. The release describes this as a gap between AI ambition and recovery readiness, and that is the right reading.
The gap has a specific shape worth naming. Buying an AI-enhanced tool is a procurement decision; running one in a recovery path is an operational one. A backup or disaster recovery capability has to work when everything else has already broken, on data that may be corrupted, under time pressure, with a business waiting. Adding a model to that path means the model itself needs validating, its failure modes need understanding, and its outputs need a human who can overrule them at three in the morning.
That is why the 45-point gap exists and why it will not close quickly. The tools are available and the operational confidence is not, and the second takes longer to build than the first takes to buy. It is also the same pattern appearing across enterprise AI: a QualiZeal and HFS Research study this month found 57 per cent of enterprises running generative AI in production and 15 per cent with a purpose-built approach to testing it. Ambition outrunning operational discipline is the defining characteristic of this cycle.
What the study measures, and what it does not
The findings fall into two categories and the distinction is worth drawing. The cost increase, the attack frequency and the downtime cost describe experience. The budget expectation, the complexity ranking and the AI readiness gap describe intention and difficulty.
Both are useful and they answer different questions. The experience figures tell an enterprise whether it is typical. The intention figures tell it where the market is heading. An IT resilience strategy built only on the first would be reactive, and one built only on the second would be speculative. Read together, they describe a market that has decided the problem is serious and has not yet decided what to do about it.
One detail in the release deserves a note because it is easy to skim past. Identity platforms and production environments top the list of regional data centre concerns. Identity appearing alongside production is the more telling of the two, because identity is the attack path rather than the asset. An organisation that worries about its production estate worries about what it holds; one that worries about identity worries about how an attacker gets in, and the second concern implies a more mature reading of the threat.
Costs and attacks, converging
The two headline pressures connect, and the release treats them that way. Ninety-three per cent saw infrastructure costs rise, averaging 14 per cent, while 48 per cent name cyberattacks as their top data centre concern. Seventy-four per cent expect data protection, backup and recovery budgets to rise in 2027.
The budget expectation is the more interesting number, because it is the only forward-looking one and it suggests the market is responding rather than absorbing. A 14 per cent cost increase alongside a near-three-quarters expectation of rising protection budgets describes an estate where the bill is going up and the organisation has decided to keep paying.
The downtime figure gives that decision its justification. Forty-three per cent putting the cost of a single hour of cyberattack downtime above $250,000 makes a backup budget look inexpensive by comparison, and it is the number a finance committee will understand fastest.
Complexity as the barrier, and what it implies
Fifty-six per cent identify the complexity of planning for cyber incident recovery as their greatest challenge, followed by lack of in-house expertise and insufficient budgets. The ordering is informative for IT resilience planning: complexity ranks above both money and skills.
That ranking points at a specific problem. Planning a cyber recovery is not the same as planning a disaster recovery. A fire or a flood destroys infrastructure and the recovery target is a state you know. A cyberattack may leave the infrastructure intact and the data compromised, which means the recovery has to establish what is trustworthy before it can restore anything. That is a harder question than restoring from a backup, and it is why the complexity ranks first.
Marc Beder of 11:11 Systems makes the practical version of the point, that organisations need to understand which systems are most critical, how quickly they need to recover, and whether their recovery plans will work under pressure. He also notes the dependence on virtualised environments such as VMware for critical workloads, which is the detail that will resonate with most ANZ enterprises. Protecting a virtualised estate and recovering it quickly is a different exercise from protecting physical servers, and the licensing and support changes in that market over the past two years have made the question more pressing rather than less.
The recommendation that matches the vendor
The analyst quotation closes with a remedy: for many organisations, the way through is leaning on expertise as a service to reduce complexity while modernising how they protect and recover data.
That is a reasonable conclusion and it is also precisely the category 11:11 Systems sells. The vendor commissioned the study, an independent analyst firm conducted it, and its recommendation points at managed services. None of that makes the findings wrong, and a reader should know the shape of the arrangement before weighing the advice. The data is worth having; the prescription attached to it belongs to the company that paid for it.
Jason Buffington’s framing elsewhere in the quotation is sharper and more independent of the commercial interest: organisations now have to govern their own AI while defending against the attackers’ AI, all while rising hardware costs squeeze budgets that were already tight. That is a fair description of the current position and it does not require buying anything.
What would complete the picture
Three additions would turn this from a useful regional read into something an enterprise could plan against.
The ANZ sample size is the first, and it would change how a reader should treat every other number. A base of a hundred and a base of three hundred produce the same percentages and very different confidence.
A sector breakdown is the second. Infrastructure cost pressure and cyberattack frequency look different in a bank, a hospital and a mining company, and ANZ has all three in quantity. Knowing where the 14 per cent average concentrates would make it actionable.
And a definition of AI-enhanced backup or disaster recovery is the third. Eighteen per cent is a striking figure and a reader cannot assess it without knowing what counted. If the bar is a model in the recovery decision path, the number is unsurprising. If it is any tool with a model somewhere in the stack, the number is low enough to be alarming.

What to watch
Whether the 2028 edition reports the ANZ base and a sector split, because that would make the series genuinely useful rather than merely quotable. Whether the AI readiness gap narrows on the operational side rather than the investment side, since investment is the easy half. And, whether cyber recovery complexity falls as runbooks and testing practices mature, because it is the barrier that money cannot simply buy through. Whether the IT resilience conversation moves from tooling to testing, because testing is where the gap between investment and operational capability actually closes.
And whether the recovery testing that Beder describes becomes routine, because a recovery plan that has never run under pressure is a document rather than a capability.

