Hardware Outperforms Software Deals – Rethinking Capital Expenditure Priorities

15.07.2026

9 Min. read time

IBM reports a 7% decline in infrastructure for Q2, while distributed infrastructure surges by 37%. The contradiction is the real clue: customers pulled capex forward into servers, storage, and memory in the final weeks of June-delaying software and consulting deals.

Key Takeaways

Capex timing trumps roadmap slides. Companies securing hardware and storage now are pushing software renewals and consulting packages back by 30 to 60 days-intentionally or by default.
Two infrastructure worlds in one number. Overall infrastructure declined, but distributed systems and storage hit historic highs. The tight supply layer is the signal, not the average.
DACH leverage lies in H2 commitments. CFOs and CIOs need a shared list: which deals are pending, which capacity is secured, and who owns both tracks.

RelatedThe AI construction boom hits the cloud bill / Three AI budgets, no unified invoice

On July 14, 2026, Arvind Krishna released a letter to investors previewing select Q2 figures. Revenue grew by 1%. Software rose 5%-still below internal expectations. Consulting remained flat. Infrastructure fell by 7%. Yet IBM reported its strongest quarter ever for distributed infrastructure: up 37%, with a backlog in the high triple-digit millions by quarter-end.

This isn’t a conflict of investor relations storytelling. It’s a procurement and timing signal. In the final weeks of June, customers reallocated capex to servers, storage, and memory to lock in scarce infrastructure ahead of expected price hikes. Meanwhile, industry-wide cybersecurity concerns diverted attention. Several major deals missed their planned close windows. IBM openly acknowledges the teams didn’t adapt quickly enough.

+37%

Distributed infrastructure grew in Q2 2026-while overall infrastructure shrank by 7%.

Source: Arvind Krishna, Letter to IBM Investors, 14.07.2026

What the Capex Slide Really Measures

Traditional budget logic neatly separates hardware, software, and services into distinct buckets. But the market blends them during decision-making windows. When companies buy storage and servers now because lead times are rising, they drain the same Capex or cash buffer from the software close. This looks like “software demand is weakening”-but often, it’s just queue economics at play.

IBM identifies three drivers at once: weaker-than-expected z17 cycles and transactional software, Capex shifting toward constrained infrastructure, and distractions from security priorities. For decision-makers in DACH, the second driver matters most. It’s transferable-even without IBM in the stack. Hyperscalers and enterprises face the same bottleneck in memory and storage-adjacent components.

The report highlights another key point: Red Hat’s sequential growth accelerated to 11 percent, HashiCorp and Confluent performed strongly, and consulting signings with GenAI components rose. The portfolio isn’t breaking-it’s re-sequencing. Those who interpret this as a blanket “software is out” narrative risk making the wrong call for their H2.

Three Lists Instead of One Budget Slide

A workable solution isn’t a new framework. It’s three parallel lists that the CFO and CIO finalize in the same session.

List	Content in 10 Days	Ownership
Secure capacity	Servers, storage, memory, colocation or cloud reservations with delivery date and exit clause	CIO + Procurement
Software on hold	Renewals and new licenses that can be deferred 30–60 days without compliance breach	CIO + Business Unit
Consulting paused or sharpened	Packages without a clear capacity or ROI path; keep only what aligns with secured workloads	CFO + Transformation Owner

Digital Chiefs analysis based on IBM Investor Letter 14.07.2026 and typical DACH H2 commitment cycles

The lists deliberately separate concerns. Those who lump everything into a single “digitalization roadmap” end up measuring only the average-and miss the bottleneck. The IBM case proves it: the average (infrastructure −7 %) and the bottleneck layer (+37 % distributed) can move in opposite directions.

What breaks

Annual licenses assumed to renew “automatically” that compete with hardware during the close window
Consulting packages without workload anchors-costly, deferrable, politically tough to halt
Multi-vendor hardware without delivery SLAs: you buy scarce capacity yet still face delays

What holds

Secured memory/storage capacity with documented use case (AI training, HANA-adjacent, batch)
Software deals with real deferral clauses and no compliance deadline in the next 60 days
Consulting only where it makes secured capacity productive-not where it produces more slides

DACH Translation: Co-determination, Legacy, Procurement

US earnings calls talk about “client buying patterns.” In the DACH region, those same patterns go by different names: investment committees, works councils for system changes, SAP and mainframe legacies, and a procurement team that weighs supplier risk more heavily than feature roadmaps. The capex slide hits harder here because approval cycles are longer. If you secure hardware in June and software in September, you need two approvals-not one slide with a combined total.

The message lands especially clearly in SAP- and IBM-heavy environments. Transactional software and z-layer stacks depend on capex windows that now compete with AI infrastructure. Distributed power and storage keep running because the bottleneck is tangible there. This isn’t a case for IBM. It’s a case for sharpening your own procurement to match the urgency the market is forcing.

The second DACH-specific point is quarterly security distraction. NIS2 and KRITIS-related topics tie up budget and leadership attention. Treating security and AI infrastructure as separate “special projects” multiplies close risks. Better: a joint risk and capacity board with clear stop rules.

Ninety Days: Sequence Over Parallel Noise

H2 Roadmap After the Capex Signal

Days 1–14

Audit open software and consulting closes. Flag everything with hard deadlines (compliance, contract, audit) versus anything that can be deferred.

Days 15–45

Secure capacity: memory/storage/servers or equivalent cloud reservations. Every reservation needs a use case and an exit plan.

Days 46–90

Follow up with software and services-but only for workloads where capacity is already locked in. No “hope-driven” closes.

IBM has scheduled its earnings call for July 22. Between now and then, other vendors will echo the same bottleneck in their own stories. Those who wait until every earnings letter sings the same tune will have already missed the narrow window. Those who sequence early turn the noise into an approval argument instead of a surprise.

The first concrete step is unglamorous: a single page with three columns-secure, wait, stop-and a signature from both the CIO and CFO. Without that page, every roadmap remains a wish list that the market will reshuffle when the close window opens.

Frequently Asked Questions

Is this just an IBM issue or a market signal?

The letter describes customer behavior: Capex is shifting toward scarce infrastructure. This pattern holds even without IBM in the stack, as soon as memory, storage, or GPU-adjacent capacity becomes the bottleneck layer.

Should software renewals be paused across the board?

No. Only deals without hard deadlines and no direct link to secured capacity should be put on hold. Compliance- and audit-critical licenses remain a priority.

How does this align with separate AI budgets in business units?

Separate budgets accelerate the shift: each unit buys capacity or tools in isolation. A shared sequencing board prevents three budgets from tripling down on the same bottleneck.

Which figure in the letter is the key decision metric?

Not total revenue. The contrast between infrastructure at −7 % and distributed infrastructure at +37 % reveals where Capex is truly flowing.

What’s the minimum 90-day action step?

Three lists (secure, wait, stop) with clear ownership and deadlines. Without this, every H2 roadmap will lag behind the next vendor letter.