Starbucks builds AI-assisted software, investors reward with stock surge

Buying software is now the expensive option. Starbucks spends 400 million dollars a year on it. This week, Bloomberg revealed the company is building its own custom systems in-house, replacing tools it rents from IBM and Microsoft. IBM dropped 3 percent on the news. Salesforce dropped 4. One internal memo from a coffee company, and the market repriced the software industry. We have made one argument to enterprises since 2017: do more yourselves, own more yourselves. When we founded paterhn, this was the entire thesis: companies would need to own their intelligence layer instead of renting it. The systems, the models, the software that encodes how they actually operate. I wrote about the hidden costs of rented AI in September 2024, when that was an unfashionable position. We wrote in January that code is cheap now and software isn't. We wrote in February, when 800 billion dollars came off software stocks in one week, that the seat was the wrong unit of value. Each time, the response was some version of: too early, too hard, nobody serious builds their own. Starbucks builds its own now. The headline leaves out that companies never rented everything because renting was better. They rented because building cost too much: eight to twelve engineers, nine to twelve months, seven figures. That constraint is dead. The same scope ships today with three to five engineers in eight to twelve weeks, at a fraction of the cost. Almost 70% less.. When building costs that, the misfit tax on generic software becomes the most expensive line in your budget. And the order matters: rebuild the work first, then put the AI under it. Intelligence layered on a broken process automates the breakage. And here is the part Starbucks cannot show you: you do not need their army. That is what a development studio is for. We build it with you, on your infrastructure, and when we leave, you own everything. Code, weights, documentation, IP. The question was never in-house versus vendor. It is owned versus rented, and it always was. I write about this shift in my newsletter every other week in: Agent Labor in Plain English: The C-Suite Edition. The last issue covered exactly this question: what to own and what to rent. Welcome to software development in 2026: the coffee company builds, and the software companies watch... https://lnkd.in/evmBEVxP

SaaS implementations can be costly, distracting, and force your unique processes into a vendor's generalized framework. Now Starbucks is building their own POS, inventory, and maintenance software — in-house replacements for Microsoft Dynamics 365, Oracle Simphony, and IBM Tririga, targeted for end of 2027, with $10M of the CTO's $30M enterprise-tech budget cut riding on it. I watch for signs that AI has flipped the build-vs-buy math for ordinary operating companies, and if this is successful, it will be a strong data point: a coffee company (admittedly, a massive one with a 100Bn+ market cap, not your corner shop) deciding it's now cheaper to write its own enterprise software than to keep renting it. https://lnkd.in/gVWQpaUC (sorry, paywall) This will be an interesting case study to watch.

-- Big changes coming to the software industry if they pull this off... Starbucks is reportedly building in-house, AI-powered software tools to replace some Microsoft and IBM systems and cut an estimated $400 million annual software bill, with a gradual rollout expected through 2027.[fortune] What Starbucks is doing Bloomberg, via multiple outlets, reports that Starbucks is developing internal software using AI to take over tasks currently handled by third‑party applications from Microsoft and IBM. Examples mentioned include a Microsoft system used for inventory tracking and an IBM tool used for maintenance management across stores. Starbucks’ CTO Anand Varadarajan has said internally that the company spends about $400 million a year on software and sees “clear opportunities” to reduce that spend.[fortune] Goals and expected timeline The primary goal is to lower recurring license and support fees paid to large vendors by swapping some of their products for homegrown systems. Reports say Starbucks is targeting a phased transition, with some of these AI‑driven tools potentially rolling out by the end of 2027, rather than an abrupt cut‑over that could disrupt store operations.[finimize] Financial and operational implications Analysts note that fees paid to Microsoft and IBM show up immediately in operating expenses, whereas parts of internal software development can be capitalized and amortized, which can help operating margins even if total cash outlay doesn’t fall one‑for‑one at first. At the same time, building and running critical software in‑house shifts more risk to Starbucks, which must now secure, scale, and maintain these AI tools across thousands of locations worldwide.[finimize] Market reaction Reports of Starbucks reducing dependence on big software vendors have weighed on related stocks: IBM shares, and to a lesser extent other enterprise software names like ServiceNow and Salesforce, have traded lower on the news amid concerns about a broader trend of large enterprises using AI to build more of their own tools.[finance.yahoo]

I've been advocating this for a long time. Starbucks' decision to build AI-assisted internal solutions instead of spending hundreds of millions on enterprise software is another sign that the "build vs. buy" equation is changing. A timely validation for our newest clients who chose to build with Avonet Technologies rather than continue investing in expensive SaaS platforms. https://lnkd.in/gfkEazAa

𝗪𝗶𝗹𝗹 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗿𝗲𝘃𝗲𝗿𝘀𝗲 𝗱𝗲𝗰𝗮𝗱𝗲𝘀 𝗼𝗳 𝗲𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝘀𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗼𝘂𝘁𝘀𝗼𝘂𝗿𝗰𝗶𝗻𝗴 𝗯𝘆 𝗲𝗻𝗮𝗯𝗹𝗶𝗻𝗴 𝗳𝗶𝗿𝗺𝘀 𝘁𝗼 𝗶𝗻𝘁𝗲𝗿𝗻𝗮𝗹𝗶𝘇𝗲 𝗱𝗶𝗴𝗶𝘁𝗮𝗹 𝗰𝗮𝗽𝗮𝗯𝗶𝗹𝗶𝘁𝗶𝗲𝘀? Starbucks Corporation has recently reported that it is investing in 𝗶𝗻-𝗵𝗼𝘂𝘀𝗲 𝗔𝗜-𝗯𝗮𝘀𝗲𝗱 software to reduce its dependence on external software vendors such as Microsoft and IBM. The company reportedly spends around $400 million annually on software to pay large recurring software licensing fees. The first target systems are: - #InventoryManagement (currently supported by Microsoft-related systems) - #MaintenanceManagement (currently supported by IBM-related systems) Internal #AI-enabled software systems, can be more cost-effective and offer opportunities to develop customized capabilities. However, in the long run, creating internal AI Models/Agents may lead to higher 𝗺𝗮𝗶𝗻𝘁𝗲𝗻𝗮𝗻𝗰𝗲 and 𝗹𝗮𝗯𝗼𝗿 𝗰𝗼𝘀𝘁𝘀 for a company. This shift changes the cost structure from operational expenses (OPEX) to ownership of capabilities. Additionally, it may fundamentally alter the technical profile of the workforce within the company. https://lnkd.in/e3trCFiF #sovereignAI

Last week I wrote that seat-based software pricing is quietly breaking. This week gave us the clearest sign yet of what comes next. According to Bloomberg, Starbucks is building its own AI-assisted software  to replace specific applications it currently licenses from Microsoft and IBM starting with an inventory system and a maintenance platform. The reason is simple. Starbucks spends around $400 million a year on software. And it now believes AI makes building some of that in-house cheaper than buying it. For twenty years, the logic of enterprise software was settled. Building is hard, slow, and expensive. So you buy. You pay Microsoft, IBM,  or SAP, and you accept their product as it comes. That logic held because building was too costly for most companies to  justify. AI is changing that math. When AI meaningfully lowers the cost of building custom software, the  build-versus-buy calculation that held for two decades starts to shift. It is worth being precise about what this is and is not. Starbucks is not abandoning Microsoft. It is still building on Microsoft's  cloud and AI infrastructure underneath. It is replacing specific applications  on top, not the whole stack. And building software is easy. Operating it reliably across thousands  of stores is brutally hard. Starbucks may succeed or may not. But the signal matters more than the outcome. For the first time in twenty years, one of the world's most recognisable  companies looked at its software bill and decided that building some of it  was now the better option. When a company like Starbucks makes that call, others start asking why  they are not. The assumption that buying beats building is genuinely in play again. Watch what the biggest companies build next. #EnterpriseSoftware #AI #SaaS #BusinessStrategy #FutureOfWork

Businesses can treat Starbucks’ shift away from #Microsoft and #IBM as a #blueprint for how to cut #software #costs while adding #AI tools.  Starbucks’ enterprise #technology team expects #budget savings of about $30 million this fiscal #year, including $10 million in software #savings. https://lnkd.in/dYjV3GWr

☕ Starbucks may have just delivered one of the clearest signals yet that AI is changing the enterprise software market. The company is reportedly exploring AI-powered, in-house alternatives to some software currently purchased from vendors including IBM and Microsoft, part of a broader effort to reduce costs and rethink its technology estate. Starbucks spends roughly $400 million annually on software, and the shift has already raised questions across the market about the future of traditional enterprise software economics. The bigger story isn't simply “build versus buy.” It's this: AI is lowering the cost and complexity of building software, and forcing every enterprise technology provider to prove the value of what they sell. The market reaction has been significant. IBM shares fell sharply after the Starbucks news, reflecting investor concerns that AI could enable large enterprises to develop more of their own applications and reduce reliance on traditional software vendors. That concern has since been amplified by IBM's separate Q2 warning, which sent the stock down roughly 25% in its largest-ever one-day decline. But I think there's an important lesson here for enterprise technology leaders: AI may make it easier to build. That doesn't necessarily make building the right answer. The real questions are: Is the data trusted? Is the solution secure and governed? Can it scale? Who owns the ongoing maintenance? Can it support auditability, accountability, and compliance? Does it genuinely improve business outcomes? The future of enterprise software may not be about buying everything or building everything. It may be about knowing what should be built, what should be bought, and where trusted platforms create the greatest strategic advantage. The build-versus-buy conversation has changed forever. And AI is only getting started. 🚀 Note: Starbucks' decision is one factor in broader investor concerns; the recent dramatic IBM sell-off was primarily tied to its own preliminary Q2 earnings miss and outlook, not Starbucks alone. 💫 Read more here: https://lnkd.in/epvtnh7q

Is this a glimpse into the future? Starbucks just walked away from IBM and Microsoft in favour of custom-built software. $400 million a year in software spend, and they're building their own systems in-house instead. Here's the part people will miss: Starbucks could always afford custom software. Money was never the barrier. What's changed is that it's now a no-brainer. IBM dropped 3%. Salesforce dropped 4%. The market already knows what most companies haven't figured out yet. The real story is due to the step change in coding models over the last 6 months. What used to cost millions and take years to build has been shredded to a fraction of that, in both time and cost. Custom software is not longer an enterprise-only privilege. It's on the table for everyone. And that matters because most companies have spent two decades doing the opposite of what makes sense: buying software that only half fits how they work, then wrapping their processes around the tool instead of the other way round. Hard to customise, workarounds stacked on workarounds, and the whole team stuck learning software built for everyone, not for them. After implementation, it's rarely perfect going in and almost never gets much better. Those days are over. Now you can design your own business operating system from scratch. Not just custom screens bolted onto the same old stack - built on AI-native infrastructure from day one. That means your data can live in one central company brain, and every process, workflow and tool gets built around it, instead of scattered across ten different platforms that don't talk to each other. This is extremely difficult for large enterprises with an enormous amount of data built up over decades, but a lot more achievable for SMBs — less data to move, easier to get it all in one place, and much faster to agree on what you actually want out of it. Starbucks just showed the rest of the market where this is going. The only question left is whether you move now, while it's still an edge, or wait until it's just table stakes. The only question left is whether you move now, or explain to your board in three years why you didn't. -- I'm Colin, founder of CXO Studio. We're a bench of experienced operators — people who've built and run real companies — now obsessed with using AI to make businesses dramatically faster, leaner, and more valuable. The kind of team you wish was sitting around your boardroom table.

☕ Starbucks is ramping up its in-house tech strategy to cut costs and reduce reliance on major software vendors. Key moves include: * Replacing parts of Microsoft’s inventory management and IBM’s maintenance software with internally built alternatives. * Continuing development of a new point-of-sale system to replace Oracle Simphony. * Using AI-assisted coding to accelerate software development, with AI adoption now influencing employee bonuses. * Reviewing every technology contract to identify savings. The company spends around US$400M annually on software and expects to save ~US$30M this fiscal year, including software and contractor cost reductions. These efforts support Starbucks’ broader goal of cutting US$2B in costs while improving speed, operations, and the in-store customer experience under CEO Brian Niccol. If AI-assisted software development continues to improve, enterprise software vendors—and eventually even frontier AI model providers—could face increasing competition from software built in-house. For decades, the prevailing trend has been for organisations to buy increasingly sophisticated off-the-shelf software rather than develop it themselves. If AI significantly lowers the cost and complexity of building and maintaining custom applications, that dynamic could begin to reverse. Whether this becomes widespread remains to be seen, but it’s a useful reminder that the impacts of a general-purpose technology like AI are rarely straightforward. The biggest winners and losers may not be where today’s consensus expects them to be.