Starbucks Corp. is building its own artificial intelligence tools to replace software it currently buys from Microsoft Corp. and International Business Machines Corp., aiming to chip away at its $400 million annual technology bill. The move sits inside chairman and CEO Brian Niccol’s broader $2 billion cost-reduction turnaround, and the company is also preparing to cut about $30 million from its enterprise technology budget in the fiscal year ending in late September 2026. Chief Technology Officer Anand Varadarajan told workers in an internal forum earlier this year that there are clear opportunities to reduce software spend, according to a recording reviewed by Bloomberg News.
The coffee chain is building alternatives to a Microsoft system that tracks inventory and an IBM tool that manages equipment maintenance, per an internal presentation Bloomberg reviewed. Some of the in-house replacements could roll out by the end of next year, pending test results. The market reaction was immediate: Microsoft and IBM both fell in premarket trading the morning the report surfaced, though Microsoft moved less because it also sells the cloud infrastructure Starbucks will build the replacements on.
The $400 Million Software Line Item Inside a $2 Billion Turnaround
The $400 million figure came from Varadarajan’s own internal forum. In-house software can be cheaper, and Starbucks is hunting for savings as part of a broader $2 billion cost-cut campaign that runs across the whole company. That campaign sits inside Niccol’s “Back to Starbucks” plan, the multi-year reset he has been running since taking the corner office.
The reset is already showing up in the numbers. For Q2 fiscal 2026, ended March 29, 2026, Starbucks reported non-GAAP EPS of $0.50 and GAAP EPS of $0.45, per the company’s earnings release. Global comparable store sales rose 6.2%, and consolidated net revenues climbed 9% to $9.5 billion. North America comparable store sales increased 7.1%, and U.S. comparable store sales also increased 7.1%. CFO Cathy Smith framed the quarter as the moment cost discipline started showing up in margins.
The $2 billion cost-reduction target gives the AI build a clear home on the income statement. Software savings flow directly into the same line item that has to absorb elevated coffee costs, labor investments under “Back to Starbucks,” and pressure from faster drive-thru competitors. The full Q2 fiscal 2026 results are laid out in Starbucks’ Q2 fiscal 2026 earnings release.
What Starbucks Is Actually Replacing
Two specific vendors are in the crosshairs. The first is Microsoft, whose inventory-tracking system Starbucks engineers have had to heavily customize anyway, according to the Bloomberg-sourced internal presentation. The second is IBM, whose equipment-maintenance platform is being rebuilt using AI-assisted coding, with the replacement platform described as the more technically advanced of the two projects.
The deployment window runs through the end of 2027, pending testing. AI usage has become a formal part of how Starbucks evaluates its engineering staff: Bloomberg has reported that tech-worker bonuses now factor in AI adoption. Beyond Microsoft and IBM, Starbucks has been working for several years on a point-of-sale system that would take the place of Oracle Simphony, per people familiar with the matter who were not authorized to speak publicly.
The AI build is not a wholesale exit from Microsoft. Starbucks still runs on Microsoft software, including the Azure cloud and AI infrastructure that the new tools will sit on top of. Its Green Dot Assist barista tool already runs on Azure OpenAI, per Forbes. The bet targets the application layer, the customized software sitting on top of the cloud, rather than the infrastructure underneath. The vendor-dependency risk that surfaced during the Blue Yonder breach last year underlines why that distinction matters: the Blue Yonder breach that hit Starbucks baristas showed how a single third-party failure can take scheduling and store-floor operations down with it.
| Vendor | System | Status |
|---|---|---|
| Microsoft | Inventory tracking | In-house replacement under development; target deployment late 2027 |
| IBM | Equipment maintenance | In-house replacement built with AI-assisted coding; target deployment late 2027 |
| Oracle Simphony | Point of sale | In-house POS in development for several years, per people familiar |
Two Vendors Took a Hit, Two Barely Moved
The market did not treat every vendor equally. On the morning of July 9, 2026, after Bloomberg’s report circulated, Microsoft was down about 1.5% in premarket trading and IBM was down 4%, per Fortune. Forbes separately reported IBM down about 3% in the same window, with ServiceNow down 3.5% and Salesforce down 4%. ServiceNow and Salesforce were never named in the report.
Microsoft’s relatively muted move has a clear explanation. Microsoft sells both the inventory application Starbucks is replacing and the Azure cloud and AI infrastructure the replacement will be built on. IBM, ServiceNow, and Salesforce live at the application layer, the exact layer a coffee company just demonstrated it can rebuild in-house. Microsoft sells the layer underneath. Investors appear to have read the news as a story about the application layer, not the infrastructure layer.
The reaction was a category trade, not a contract loss. None of those companies lost a customer on the morning of the report. What they lost was a piece of the story that has propped up enterprise software valuations for two decades: the assumption that big companies will always buy because building is too hard. For broader context on the AI-versus-software dynamic weighing on the sector, see AI’s rapid reset of software stocks.
Premarket moves, July 9, 2026
- IBM: down 4% (Fortune; Forbes reported 3%)
- Microsoft: down about 1.5%
- ServiceNow: down 3.5%
- Salesforce: down 4%
Why a Failed AI Inventory System Made the New Bet More Credible
Earlier in 2026, Starbucks pulled an AI-powered system to track store inventory after it produced inaccurate counts, reverting operations to manual counting. On the surface, that failure looks like a reason to slow down on AI. The internal Bloomberg-sourced reporting suggests the opposite: the failure has reshaped how the company sequences the work.
The new Microsoft and IBM replacements are being built on a different theory. Rather than layering AI onto existing processes, the engineering team is redesigning the underlying workflow first, then building the system around the corrected process, then letting AI-assisted coding accelerate the build. Aaron Levie, CEO of Box, made a similar case on LinkedIn, arguing that the highest-upside AI projects change the work being done rather than just replacing an existing process.
The sequence matters because it explains why Starbucks is not rolling out the new tools broadly first. Late-2027 deployment is the staging point for a much larger architecture decision, and the team has already learned what happens when AI is dropped onto a broken workflow. Stores are counting by hand for now.
How the AI inventory bet has evolved
- AI-powered store inventory system deployed across stores in early 2026.
- System produced inaccurate counts and was pulled.
- Operations reverted to manual counting while engineering team redesigned the workflow.
- Replacement platform, built with AI-assisted coding, targeted for end-of-2027 rollout.
The Cost Stack Behind the $30 Million Fiscal Year Cut
The enterprise technology team is on track to reduce its budget by about $30 million in the fiscal year ending in late September 2026, per the Bloomberg-sourced internal presentation. About $10 million of that cut comes from software spending directly. Another $13 million is being saved mostly by cutting contractors from professional services firms and backfilling some roles with internal staff. The remainder comes from other cost actions inside the technology organization.
The workforce shifts are already visible. Starbucks has cut about 2,300 jobs since February 2025, including many in tech. New offices in Nashville and India will house some of the displaced tech workers, while others remain at the Seattle headquarters. AI usage is now formally factored into tech-worker compensation, turning the build-versus-buy decision into a metric on every performance review.
The $30 million is a near-term slice of a $400 million recurring annual line item, not a one-time cut. Varadarajan’s framing of the opportunity, captured on the internal recording, signals that the company sees more room to move after the current fiscal year closes. CTO Anand Varadarajan said:
There’s clear opportunities to reduce the spend in software.
Execution risk is real and sits on the other side of the same ledger. Building and maintaining proprietary AI demands specialized engineering talent, which shifts payroll burden from the retail floor to the technology department. Material failures during the late-2027 testing phases could trigger regional inventory availability problems and equipment maintenance disruptions across stores. The plan’s credibility will depend as much on the testing window as on the build itself.
What This Means for the Rest of Enterprise Software
If a non-tech operator can use AI-assisted coding to eliminate hundreds of millions of dollars in vendor spend, every Fortune 500 technology budget is now in play. The Starbucks playbook does not have to replace commercial software overnight. It starts as targeted substitutes where the vendor fit is worst and the license cost is highest, the systems engineers were already heavily customizing anyway.
Vendors will fight back on the terrain hardest to replicate: integration depth, governance, security, and decades of accumulated domain knowledge. Mati Greenspan, founder and CEO of Quantum Economics, told Forbes that “Companies are realizing that AI isn’t just a feature. It’s becoming the core nervous system of their operations.” Expect Microsoft, IBM, and Salesforce to reposition toward infrastructure, trust, and the data layer. The application layer is where the contract risk now lives, and the broader sell-off across enterprise software reflects that read. For a sector-level view of the same dynamic, see the AI disruption risk to North American software firms.
The savings feed directly into margin defense. Starbucks’ fiscal 2026 guidance, raised in late April 2026, calls for non-GAAP EPS in the $2.25 to $2.45 range and global and U.S. comparable store sales growth of 5% or greater, per the company’s earnings release. Software cost reduction lines up alongside elevated coffee costs, labor investments under “Back to Starbucks,” and the broader the Boyu Capital deal and the broader turnaround reshaping the company’s international footprint. The AI build is the technology leg of the same multi-year margin reset that Niccol is executing across the rest of the business.
Frequently Asked Questions
What software is Starbucks replacing with AI?
Starbucks is building in-house replacements for a Microsoft system that tracks store inventory and an IBM platform that manages equipment maintenance, per an internal presentation reviewed by Bloomberg News (Fortune, July 9, 2026).
When will the AI-built replacements roll out?
Some of the new tools could deploy by the end of 2027, pending test results (Fortune, July 9, 2026).
How much will Starbucks save?
The enterprise technology team is on track to reduce its budget by about $30 million in the fiscal year ending in late September 2026, including about $10 million in software spending (Fortune, July 9, 2026).
What happened to Starbucks’ first AI inventory system?
Earlier in 2026, Starbucks pulled an AI-powered store inventory system after it produced inaccurate counts and reverted operations to manual counting (Fortune, July 9, 2026).
Does this affect Starbucks’ relationship with Microsoft?
Starbucks continues to use Microsoft software, including Azure cloud and AI infrastructure. Green Dot Assist, the barista AI tool, runs on Azure OpenAI (Forbes, July 12, 2026).
Disclaimer: This article is for informational purposes only and does not constitute financial, investment, or legal advice. Stock prices, valuation figures, and forward-looking statements are subject to change. The figures cited are accurate as of publication (July 13, 2026). Past performance is not indicative of future results. Readers should consult a qualified financial professional before making any investment decisions.
