Microsoft’s competitive, multi-billion-dollar push into synthetic intelligence used to be intended to be a flawless victory. The mixing of Copilot into Home windows 11, Microsoft 365, and GitHub used to be designed to bring in a brand new generation of agentic computing. But, underneath the polished keynote shows and large infrastructure investments, a dramatically other truth is what we noticed.
As first reported by means of Home windows Newest, in keeping with a extremely revered former Microsoft govt, the corporate’s AI technique is basically failing to connect to actual customers, spurring requires an enormous inner “manufacturing facility reset.”
The manager in query is Mat Velloso, who used to be the Vice President of Product for the Developer Platform at Meta’s Superintelligence Labs. He additionally led AI developer merchandise at Google DeepMind (together with the Gemini API and Google AI Studio). However ahead of his stints at Google and Meta, Velloso spent over 12 years at Microsoft, the place he served as a Spouse Director managing AI innovation in Home windows and, curiously, spent 4 years because the Technical Marketing consultant to Microsoft CEO Satya Nadella.
When anyone with Velloso’s resume, having noticed the AI palms race from the best ranges of Microsoft, Google, and Meta, says Microsoft has “overlooked the AI wave,” it could possibly rip the lid off the deep tensions inside Redmond.
Ex-Microsoft exec says Microsoft overlooked the AI wave
Microsoft’s conduct over the previous few months is little short of stunning. Each the Home windows and Xbox divisions abruptly began prioritizing person comments and imposing asked options after years of ignoring them.
It’s additionally no longer a small activity to collect and prepare OEMs, ODMs, and chipset distributors in an tournament like WinHEC that had its final incidence nearly a decade in the past (2018).
Explaining this unexpected pivot to taking note of shoppers, Velloso remarked that in spite of making Bing the corporate’s largest AI guess, it did not seize a unmarried proportion level of seek marketplace percentage from Google. Extra damning is the state of Copilot.
In line with Velloso, not up to 3% of paying customers actively use Copilot, even if Microsoft has pre-deployed it at once into the Home windows 11 taskbar and around the Place of business suite.
Out of Microsoft’s 450 million Microsoft 365 person base, the corporate has handiest controlled to transform more or less 15 million paid Copilot seats. This implies a staggering 96.7% of customers are rejecting the top class AI options, yielding only a 3.3% paid adoption charge. When seen towards Microsoft’s estimated $37.5 billion quarterly AI spending, that is an alarmingly low adoption charge.
Nevertheless it’s no longer simply tool; Velloso often known as out the present state of AI {hardware}. During the last 12 months, Microsoft has closely driven OEMs to incorporate Neural Processing Gadgets (NPUs) of their newest laptops to energy complex Home windows 11 functions. Now we have tracked Microsoft’s push for NPU-powered AI options in Home windows 11, however as Velloso famous, OEMs invested closely in NPUs handiest to determine that “no one cares as a result of no longer a unmarried precious usecase used to be constructed for the ones in Home windows/Place of business.”
Moreover, he highlighted that GitHub, a platform that are meant to be thriving as the center piece of the AI coding revolution, has noticed its Provider Stage Settlement (SLA) reliability drop under 90%. Mix this with emerging Price of Items Offered (COGS) and shareholders starting to ask tricky questions, Microsoft used to be pressured to begin taking note of shoppers since the AI guess isn’t paying off as easily as expected.
Microsoft’s govt exodus
All this friction seems to be taking a toll on Microsoft’s management. Just lately, information broke that Julia Liuson, the extremely revered head of Microsoft’s Developer Department (DevDiv), used to be resigning after 34 years with the tool large.
Whilst reliable channels framed this as an ordinary retirement, Velloso critiqued the scoop, pronouncing, “Looks as if Microsoft simply went from hit refresh to hit manufacturing facility reset.” He additionally indexed an enormous string of high-profile departures and reassignments around the corporate, together with leaders from Xbox, GitHub, AI Infrastructure, Groups, and OneNote.
This public observation drew the ire of Frank X. Shaw, Microsoft’s Lead Communications govt. Shaw responded to Velloso, protecting the departing executives and accusing Velloso of jamming a “destructive body” onto commonplace company retirements.
All Velloso needed to do used to be indicate the cruel monetary realities that the marketplace is recently digesting. He moved from Microsoft to Google in early 2024, and whilst Google’s stocks surged by means of more or less 230%, Microsoft’s inventory expansion remained necessarily flat at 0%.
“I guess the entire marketplace may be fallacious about this,” Velloso fired again. “However what do I do know? I handiest noticed each corporations, how they perform, their strengths and their flaws. Microsoft overlooked the web wave, the cell wave and now it overlooked the AI wave. It’s what occurs while you stay doing the similar issues anticipating other effects.”
He suggested Microsoft’s management to forestall “gaslighting,” deflecting blame, and burying skeletons, arguing that hitting an entire inner “manufacturing facility reset” is the one pragmatic method to repair the exhausting issues plaguing the working device and venture tool stack.
OpenAI is chopping out the intermediary (Microsoft)
With the exception of the interior demanding situations, Microsoft is more and more being suffering from their closest allies. The corporate has staked its whole generative AI long run on its multi-billion-dollar partnership with OpenAI. Then again, OpenAI is swiftly construction out its personal venture infrastructure, threatening Microsoft’s ancient dominance within the company sector.
Simply days in the past, OpenAI formally introduced the “OpenAI Deployment Corporate” (DeployCo), a brand new trade unit sponsored by means of over $4 billion in preliminary funding from international corporations. This new undertaking options 150 “Ahead Deployed Engineers” (FDEs) tasked with embedding at once into Fortune 500 corporations to lend a hand them construct and deploy customized AI answers.
Traditionally, this hands-on, enterprise-level consulting used to be Microsoft’s bread and butter. As Velloso reminisced about his early profession as a specialist, he famous that Microsoft’s unbelievable penetration into huge enterprises used to be constructed on “armies of folks spending time, listening, figuring out trade targets and fixing them with era in each business vertical.”
Now, best AI labs like OpenAI are replicating this identical playbook. They now not need to delegate venture deployment to hyperscalers like Microsoft Azure. By way of chopping out the intermediary and proudly owning the direct courting with companies, OpenAI is stepping at once into the extremely profitable services and products layer the place the true venture AI greenbacks are headed. This items an enormous structural danger to Microsoft’s long-term venture earnings, particularly as customers proceed to thrust back towards pressured Copilot internet app integrations.
Why Microsoft is a long way from useless: “The moat is unbreakable”
In spite of those critical criticisms, Velloso defended his former employer towards apocalyptic tech media narratives. When a outstanding tech newsletter just lately claimed that “AI is killing Microsoft” and in comparison their present trajectory to the disastrous 2008 generation, Velloso stepped in to close the narrative down.
“Nope, they aren’t demise,” Velloso mentioned. “I do know I criticize them so much and that’s as a result of I care, however boy in the event you assume Microsoft is demise you haven’t watched how repeatedly they recovered from issues.”
He identified that whilst AI startups and labs may well be construction flashy deployment corporations, totally changing legacy venture tool is extremely tricky. When requested about corporations claiming they may be able to totally automate tool companies, Velloso really useful chatting with Fortune 500 CIOs to peer how real looking that actually is.
“There’s a explanation why the entire best AI labs are hiring huge consulting groups,” he defined. “The final mile is the toughest and Microsoft has the most efficient distribution for that. Their moat is unbreakable.”
That is the final paradox of Microsoft in 2026. They’re concurrently suffering to persuade customers to undertake their new AI options, but their foundational grip at the venture ecosystem continues to be forged. Hundreds of thousands of companies rely on Energetic Listing, Azure, Place of business, and Home windows. Whilst it’s simple to spin up an AI agent in Python, integrating it securely right into a legacy company atmosphere calls for the type of huge distribution community that Microsoft has spent 4 a long time construction.
A important warning call
The entire “new” options in Home windows 11, together with the much-awaited movable taskbar and resizable Get started menu, are a results of the entirety that went fallacious within the corporate and management in the end doing one thing about it.
When an organization realizes that pushing a three.3% adoption-rate AI chatbot down customers’ throats isn’t translating to inventory expansion or person pleasure, they’re forced to rethink their technique.
We’re in the end seeing the result of this “manufacturing facility reset” in actual time. Microsoft is dismantling slow internet wrappers, committing to local UI efficiency for Home windows 11 apps, overhauling motive force high quality requirements, and explicitly prioritizing long-ignored person comments.
Microsoft’s pivot again to platform basics proves that whilst AI may well be the longer term, you can not construct that long run on a crumbling basis. The marketplace pressured Microsoft to concentrate, and for pissed off Home windows customers, this is arguably the most efficient factor that will have perhaps came about.
