Better AI software is driving up spending on power, partners, and delivery

Salesforce told the market its Agentforce annual recurring revenue has reached $1.5 billion, and it said Slackbot passed 1 million active users within five months of launch. Those are product metrics, but they’re also procurement inputs. They’re the kind of numbers that turn “we’re piloting AI” into “we need to fund the rollout, staff it, and connect it to real systems.” StockStory tied a broader jump in enterprise software shares to this kind of upbeat AI monetization commentary, listing gains across Amplitude, GitLab, Doximity, Freshworks, and Sprinklr alongside bigger-platform readthroughs from Salesforce, CrowdStrike, and Okta.

For operators, the more consequential signal is what happens after those earnings calls: AI programs are moving from tool selection to delivery discipline. That shift shows up in three places at once, the channel partners who will implement the work, the data center and facilities constraints that set the pace, and the supply chain behind compute, power, and connectivity.

The new gating item: can you deliver AI, not just license it

Data Center Frontier’s Q1 2026 Executive Roundtable recap framed AI infrastructure as entering an “execution phase,” where success depends on power certainty, procurement discipline, coordinated delivery at scale, and credibility with local stakeholders. The emphasis is operational: meeting timelines over the next two years will hinge on coordination across developers, utilities, equipment suppliers, and contractors, and on designs that can flex with demand uncertainty.

That’s a direct counterweight to the narrative that AI is mainly a software feature race. If the infrastructure layer can’t be scheduled with confidence, AI roadmaps become sequences of partial launches, a copilot here, an identity module there, with the hard work of data gravity and latency pushed out another budget cycle.

Software vendors are publishing ROI clues operators can actually use

StockStory’s roundup highlighted a turning point for enterprise AI: vendors are increasingly tying AI features to concrete business results rather than broad adoption narratives. Alongside Salesforce’s Agentforce and Slack metrics, StockStory reported that Okta said its AI-driven identity products generated about 30% of new bookings in the quarter, and that deals including them lifted average contract values by roughly 40%. The roundup also pointed to CrowdStrike leadership linking greater AI use to an expanded attack surface, a shift that can speed adoption of additional security modules.

Those kinds of ratios create a practical benchmark for enterprise planning. If an IAM add-on is attached to deals often enough to represent 30% of new bookings, it suggests buyers are already treating AI as an identity and governance problem, not a chat interface. And a claimed 40% lift in average contract value when AI is in the bundle is a warning for sourcing teams: AI packaging can change the shape of renewals, even when the base platform is “already in place.”

CRN’s Solution Provider 500 is a reminder: most AI work lands on partners

CRN’s 2026 Solution Provider 500 isn’t about software features. It’s a map of who is positioned to deliver complex, multi-vendor projects across cloud, security, and managed services, the layer many enterprises lean on when internal capacity is thin. In the AI cycle, that layer matters because implementations touch data movement, IAM, security controls, observability, app modernization, and infrastructure, usually across more than one platform and contract.

The operational takeaway is to evaluate solution providers like production capacity, not like resellers. When AI initiatives shift from pilots to programs, partner selection starts to look like any other constrained resource planning problem: staffing, repeatable delivery methods, escalation paths, and commercial terms that survive a midstream change in model choice or hosting strategy.

Where this lands in 2026 budgets: power, identity, and implementation capacity

The four sources point to the same operational pattern even though they sit in different corners of the market. StockStory shows vendors putting real numbers behind AI adoption and monetization, which tends to pull forward deployment decisions. Data Center Frontier shows the infrastructure community warning that execution, especially power certainty and coordination, is now the determinant of schedule. CRN shows how much of the work will route through the solution provider ecosystem once deployments get real.

Semiconductor Digest’s issue page included in the source set is light on accessible detail at the provided link, but its presence in the reading list is a useful reminder of the physical layer behind all of this. Even when AI is “software-led,” it ends up as orders for compute, networking, and power equipment, and those categories bring their own lead times and qualification requirements.

Questions to take into the next AI refresh and partner RFP

• Power and site readiness: For any on-prem or colocation build tied to AI, what is the documented path to “power certainty,” including utility commitments, transformer strategy, and commissioning sequence, consistent with the execution risks highlighted by Data Center Frontier’s roundtable?
• Commercial packaging: If a platform vendor is bundling AI add-ons, ask for a renewal model that isolates the AI component, and test what happens to unit pricing and term flexibility when usage scales. Okta’s reported 40% contract-value lift when AI is included is a prompt to model the upside and the lock-in.
• Partner capacity: When comparing Solution Provider 500-type firms, require named delivery leaders, a staffing plan, and a governance model across identity, security, data, and app teams. Treat the integrator like a critical path supplier, because for many enterprises it is.
• Benchmarking success: Use public numbers as calibration points in your own KPIs. If a vendor cites a million active users in five months (Salesforce’s Slackbot, per StockStory), define what “active” means in your environment and what integration milestones must precede it.