Starbucks taps AI to cut reliance on Microsoft, IBM software

Starbucks is developing AI-assisted in-house software to replace some third-party tools (e.g., a Microsoft inventory-tracking system and an IBM maintenance tool), with potential rollout by end-2025 if testing succeeds. The plan targets major cost reductions—about $400M/year in software spending, with the tech org on track to cut roughly $30M in FY ending late September (including ~$10M in software) and additional savings from contractor reductions. This also pressured vendors in premarket trading, with Microsoft down ~1.5% and IBM down ~4%, reflecting risk to incumbent software franchises.

Analysis

The market is likely overstating the near-term revenue impact on software vendors and understating the long-term pricing pressure. The real mechanism is not lost revenue from one customer; it is that AI lowers the cost of writing and maintaining bespoke tools, which gives procurement teams a credible threat in renewals and vendor negotiations. That matters more for IBM than MSFT: IBM’s mix is more exposed to services/legacy tooling economics, while Microsoft’s platform moat and bundle economics make it harder to displace than niche workflow apps.

The second-order effect is a broader shift in enterprise buying behavior. If a high-visibility consumer brand can insource workflow software, other multi-location operators may push internal IT to rebuild customized modules instead of paying for heavily tailored legacy products, compressing net retention and elongating sales cycles for vendors like ORCL and other vertical software names. But this is a slow-burn catalyst: the first 1-3 months are about narrative and multiples, while the 6-18 month risk is margin erosion from hidden maintenance, governance, and talent costs if the build/own model proves more expensive than advertised.

Contrarian view: the consensus may be too eager to extrapolate a few AI-assisted internal builds into a secular replacement wave. Most companies can prototype fast, but production-grade reliability, security, and support are where vendor lock-in reasserts itself; that argues for a repricing of pricing power, not a wholesale demand collapse. If vendor renewal rates and remaining performance obligations stay intact over the next two quarters, the selloff in legacy software should fade.

SBUX is a modest operational winner only if internal builds actually stick; otherwise the effort becomes another maintenance burden and a distraction from the turnaround. The best risk/reward is relative value in software where AI-native productivity helps offset lost seat growth, not a blanket short across megacap tech.

Key Decisions for Investors

• Short IBM on any post-print rebound; use a 1-3 month horizon and stop if management quantifies accelerating software/services bookings or AI-driven productivity that offsets customer insourcing risk.
• Pair trade: long MSFT / short IBM for 4-8 weeks. Thesis: AI displacement pressure is more narrative risk to IBM’s multiple than a material threat to MSFT’s platform revenue base.
• Watch ORCL and other vertical-app vendors for sympathy weakness; only initiate a short if upcoming commentary shows longer sales cycles or customers referencing internal-build alternatives in renewal discussions.
• No outright long SBUX here yet; treat this as an operating expense story, not an immediate equity re-rate. Reassess only if tech-led cost savings translate into sustained margin upside without added IT overhead over the next 2 quarters.